Bone 44 (2009) 251–265

Contents lists available at ScienceDirect

Bone

j ourna l homepage: www.e lsev ie r.com/ locate /bone

Role of MT1-MMP in the osteogenic differentiation

Paola Manduca a,⁎, Alessia Castagnino a, Domenico Lombardini a, Stefania Marchisio a, Stefano Soldano a,Valentina Ulivi a, Stefano Zanotti a, Corrado Garbi b, Nicoletta Ferrari c, Daniela Palmieri a

a Genetics, DiBio, University of Genoa, 26, C. Europa, Genoa 16132, Italyb Department of Cellular and Molecular Biology and Pathology, University Federico II, Naples, Italyc IST, Genoa, Italy

⁎ Corresponding author. Fax: +39 10 353 8267.E-mail address: man-via@unige.it (P. Manduca).

8756-3282/$ – see front matter © 2008 Elsevier Inc. Aldoi:10.1016/j.bone.2008.10.046

a b s t r a c t

a r t i c l e i n f o

Article history:

Metalloproteinase MT1-MM Received 26 April 2008Revised 27 September 2008Accepted 16 October 2008Available online 5 November 2008

Edited by: J. Aubin

Keywords:Osteogenesis in vitroAlkaline phosphataseMembrane metalloproteinase 1Nodule formationMineralizationHomeostasis

P is induced and Pro-MMP-2 up modulated early in rat preosteoblasts (ROB) setto differentiate. We here show that the induction of MMPs, accompanied by activation of Pro-MMP-2, occursby 6 h of adhesion on endogenous extracellular matrix (ECM), Fibronectin (FN) and Collagen type I (CI). Theseevents do not occur after adhesion on Collagen III (CIII), Vitronectin (VN) or BSA. Within the first hour oninducing substrata or plastic, FAK is unchanged and ERK1,2, is activated, but this activation is not sufficient forMT1-MMP induction. The function of p38 MAPK and PTKs is not required for the induction by substrata ofMMPs. Six hours after plating preosteoblasts on MMP-inducing substrata, complexes of β1 integrin withMT1-MMP are formed, that contain integrin dimers specifically engaged by the substratum, α4 and α5chains for cells plated on FN, and α2 chain for cells plated on CI and ECM. Induction of MT1-MMP and itsexpression during osteogenesis pleiotropically regulate alkaline phosphatase (AP) expression. Duringdifferentiation, variant clones derived from preosteoblasts and MMPs-over-expressing osteoblasts show highMT1-MMP level associated with high AP level both persisting in time, while inhibition of MMPs isaccompanied by inhibition of AP. Up or down modulation of AP, transcriptionally or by inhibition of theenzyme activity, has no effect on level or timing of expression of MT1-MMP and Pro-MMP-2. The persistencein expression of MT1-MMP during differentiation, and the associated persistence in expression of AP, as wellas their inhibition, both impair the formation of nodules and mineral deposition. A transient pattern ofexpression of MT1-MMP is required for the establishment of nodules, and MT1-MMP decrease is permissivefor nodule mineralization. The expression of AP is required for nodule formation and its level modulates themineralization. MT1-MMP has multiple functions and is implicated in multiple steps of the differentiationprocess, acting to regulate homeostasis of the osteogenic differentiation.

© 2008 Elsevier Inc. All rights reserved.

Introduction

Osteogenic progression of determined bone precursor cells in vitrois widely regarded as a model for the homeostatic process thatinvolves gene modulation in a developmental sequence and with astochastic pattern [1–3]. In vitro differentiation of preosteoblasts inthe presence of a source of phosphate and of ascorbic acid,differentiation proceeds independently from additional exogenousstimuli and results in the formation localized multilayered areas, ornodules, that behave as preferential sites of mineral deposition.Different genes are modulated or transiently expressed duringosteoblast differentiation, which poses the question of the modalityand of the significance of their timely regulation for its homeostasis.Among the genes transiently expressed, we focus here on MT1-MMPand alkaline phosphatase (AP) which have a pattern of expression

l rights reserved.

correlated in timing during differentiation and are highly expressed inthe nodules [3]. Although it is known that in other cell types MT1-MMP can be induced by Collagen type I [4] and that collagen cross-linking can elicit AP expression [1], it is not yet certain what are theevents that trigger MT1-MMP up modulation at the beginning ofosteogenesis. Also it has not been investigated what are themodalitiesthat cause down modulation of both enzymes after nodules areformed, nor if there is co-regulation of their expression duringdifferentiation.

The enzyme MT1-MMP is a transmembrane metalloproteinase,that acts in pericellular proteolysis [5,6]. It shows gelatinase activitytowards multiple substrata, and is at the top of a proteolytic chaininvolving the activation of secreted gelatinases Pro-MMP-2 and Pro-MMP-9 [7,8]. MT1-MMP binds TIMP-2 and Pro-MMP-2 in a complexrequired for the activation of Pro-MMP-2 to forms of lower molecularweight and higher specific activity [9,10]. In the absence of alternativeactivation pathways for Pro-MMP-2 activation, as for osteogenic cells,and when TIMP-2 is not limiting, the amount of activated MMP-2depends upon and measures MT1-MMP activity. MT1-MMP has also

252 P. Manduca et al. / Bone 44 (2009) 251–265

collagenase activity [11–14], and produces fragments of Collagen ofunknown functional significance and themselves substrates for MMP-2 [11,12]. In various cell types MT1-MMP forms complexes withintegrins [15–17] that modulate MT1-MMP function and is involved inmorphogenetic events and in carcinogenesis [15–20]. No data areavailable on the formation of MT1-MMP/integrin complexes inosteogenic cells. MT1-MMP is susceptible to self-regulation byautocatalytic degradation and shedding of an inactive fragment thatinhibits enzyme activity and binding to Collagen [21,22]. MT1-MMPmRNA and enzyme are induced very early at the onset of differentia-tion in ROB, human osteoblasts and MC3T3 E1 mouse osteogenic cells,and MT1-MMP function is transiently up modulated during in vitroosteogenesis. Pro-MMP-2 and -9 are also up modulated at thebeginning of osteogenesis [3]. The expression of MT1-MMP is grosslyconcurrent in timing with activation of Pro-MMP-2, and with thesimilarly transient induction of AP [3]. Transmembrane MT1-MMP ishighly expressed in mature osteoblasts. Down modulation of MT1-MMP occurs in nodules, uncoupling the level of its transcripts fromthat of the enzyme, prior to mineralization of cultures [23]. Thedecrease in MT1-MMP enzyme is associated with the decrease in theactivated forms of MMP-2. MT1-MMP, Pro-MMP-2 and -9 mRNAs areslightly decreased at mineralization [3]. The enzyme activity for APalso declines after the formation of nodules [3]. Osteopenia isassociated with inadequate collagen turnover and defective develop-ment of osteocyte processes in MT1-MMP knockout mice [24,25], andwith lack of collagen proteolysis and to defects in osteoblasts growthand apoptosis in MMP-2 knockout mice [26,27]. These data show therelevance of MMPs to in vivo in bone formation.

Alkaline phosphatase localizes on the cell membrane by a GPI-PLD-mediated association with PI-G. In skeletal tissue, AP is shed intomatrix vesicles and it reduces the physiological level of PiPi, creatingthe condition for Pi incorporation in crystals, that are deposited on thescaffold of Collagen I fibers in the osteoblasts ECM [28,29]. Expressionof AP is developmentally regulated during in vitro and in vivoosteoblasts maturation. In particular, AP is induced in preosteoblastcultures at the onset of osteogenesis, it is down modulated after theformation of nodules [1–3,23]. AP expression is finely tuned by manysystemic and osteoblast-endogenous inducers or inhibitors, amongthem Endothelin-1 (ET-1) [30], IGFII [31] and the ECM proteins of theSIBLING family, Bonesialoprotein (BSP) and Osteopontin (OPN)[32,33]. Mineralization is dependent upon AP expression, and furthermodulated by nucleotide pyrophosphatase/phosphodiesterase 1(NPP1) and by the pyrophosphate (PPi) channeling function of ANK[29]. Mice null for AP present defects in mineralization of bone; theosteoblast cultures derived from AP knockout mice fail to mineralize,and are rescued by soluble exogenous active AP [34–36].

Signaling by FAK and ERK1,2 is associated with induction of AP andis required for osteogenic progression of mesenchymal stem cells[37,38]. Activation of ERK1,2 is required for osteoblasts growth,differentiation and integrin expression [39] and is associated withthe increase in AP induced by Fluorate derivatives [40].

Preosteoblasts synthesize Collagen type I (CI) and Fibronectin (FN),and, lesser amounts of Collagen type III (CIII) and type V [23]. Ratpreosteoblasts recognize FN viaα5β1 andα4β1 integrins, CI viaα2β1and CIII via α1β1, α2β1 and α3β1 integrins [41].

The ECM deposed during osteogenesis is a dynamic microenviron-ment, that changes continuously by proteolytic remodeling and byapposition of GF and of developmentally expressed proteins [1,2,23]and of novel fragments from pre-existing components produced byproteolysis [4,12,19–21]. These changes might produce changes inintegrin engagements and cell adhesiveness. In other cells types,engagement of different integrins by ECM components was shown tocontrol MMPs expression and to affect growth and differentiation[20,21,42–46].

We assume that transient gene expression during osteogenesismust result from an initial self-triggering event that, in such a “closed

system”, might then regulate through interactions and feedbacksthe expression of other genes. Considering its functional charac-teristics and early up modulation at onset of osteogenesis, MT1-MMPis a candidate to govern self-regulating and homeostatic eventswith morphogenetic significance in the context of osteogenicdifferentiation.

We searched for the event initiating the transient expression ofMT1-MMP by investigating the role of endogenous ECM and itsconstituent proteins.

Our results show that ECM induces MT1-MMP expression and thatMT1-MMP forms complexes with β1 integrin dimers specificallyengaged by FN and CI. We investigated if the induction of MT1-MMPcould affect the induction and sustain the expression of AP and testedthe relationships between MT1-MMP and AP expression by the use ofinducers and inhibitors. We show that MT1-MMP acts as a positivepleiotropic regulator of AP expression during osteogenesis andinitiates and supports AP expression in pre-osteoblasts. MT1-MMPexpression also supports increase in Pro-MMP-2 and its activation.Westudied the effects of MT1-MMP and AP deregulation on thecompletion of the osteogenic process, the formation of nodules andtheir mineralization and we show that MT1-MMP transient modalityof expression is required for nodule formation. The level of expressionof AP depends upon that of MT1-MMP expression and, when nodulesare formed, the extentmineralization depends upon the level at whichAP was expressed.

Materials and methods

Cells and cell culture conditions

ROB preparation and passaging, clones isolation and characteriza-tion were described [23]. ROB were used between passage 60 and 70.Differentiation medium is Coon's modified F12, 10% FCS (Seromed,Italy), 100 μg/ml ascorbic acid and 10 mM βGP. Inhibitors and inducerswere from Sigma with exception of Furine Inhibitor I, fromCalbiochem. The inducing antibody to MT1-MMP (Biotrend) wasutilized at 10 μl/ml. The cultures in all conditions of treatment weremonitored for DNA content. In all experimental conditions presentedhere there was not decrease in DNA in the treated cells, compared tocontrols.

AP, calcium and von Kossa stain

One exemplary experiment of two or more repeats is presented.Triplicate aliquots from each cell lysate in 0.1% SDS were utilized forAlkaline phosphatase determination with the AP-Kit Sigma 104-LL.Lectures of samples were at 420 nm. Calcium determinations werewith Calcium-Kit Sigma 587 and reading of samples was at 575 nm.Standard deviationwas calculated and, being less than 5% at all points,it is not indicated graphically. von Kossa stain was as described in ref23.

Coating with substrates

Fibronectin (Sigma, from human plasma), Vitronectin (Sigma), BSA(Sigma) were dissolved in PBS. Collagen I (Sigma type VII, from rattail), Collagen III (Sigma type X, from human placenta) were dissolvedin 0.1 N acetic acid. All substrata were added to cell culture dishes at afinal concentration of 110 nM, incubated ON at 37 °C, the proteinsolutionwas removed, plates were rinsed twice in PBS and then addedwith PBS 0.5% BSA at 37 °C for 1 h and rinsed twice in PBS before use.

Extracellular matrix preparations

Petri dishes of confluent preosteoblast in standard propagationmedium (for ECM t0) or of ROB at day 7 of differentiation in medium

253P. Manduca et al. / Bone 44 (2009) 251–265

added with ascorbic and βGP (for ECM t7) were put on ice, washedtwice with cold PBS and added with 0.5% Triton X-100, 10 mMNaH2PO4, 140 mMNaCl pH 7.4 for 15 min, followed by 2 changes of icecold water, for 5 min each. ECM was fixed in two changes of ice cold75% Ethanol and stored dry at −20 °C. Before use they were re-hydrated 30 min in PBS at 37 °C. In both ECM t0 and t7 FN is deposed.The ECMs thus prepared do not retain cell DNA.

Adhesion assay

Subconfluent pre-osteoblasts were trypsinized, washed andresuspended in Coon's modified F12, 0,05% heat denatured BSA,100 μg/ml ascorbic acid and allowed to adhere for the indicated time.The conditioned medium (CM) produced during the adhesion wascollected, centrifuged and the supernatant was stored at −20 °C or

Fig. 1.MT1-MMP induction in preosteoblasts by osteogenic ECM, Collagen I and Fibronectin.(B) Zymogram of CM from preosteoblasts plated on ECM t0 or t7, collected after 6 or 12 h. (C)BSA and plastic (PL). (D) Corresponding cell lysates for preosteoblast plated on inducing substor 6 h. All gels were loaded with equiamounts by cell DNA of preosteoblast CM or cell lyMT1-MMP and MMP-2 RNAs, in total RNA from cells adhering for 6 h on FN, CI or PL. Sigp= 0.004 and PL/CI, p= 0.0017.

used directly for zymograms. Lysates of the cell layers for studies ofimmunodetection and zymograms were prepared as described in thesections below. DNA content was determined for each sample.

Phosphorylation of ERK 1,2 and FAK

Subconfluent pre-osteoblasts cultures in standard medium, werestarved for 2 h in serum freemedium, trypsinized and plated in serumfree medium, at 106 cells/Petri dish of 100 mm diameter, on substratacoated or plain cell culture plastic. At the time points indicated themedium was gently removed and the adhering cells were collected,lysed in ice cold NP-40 1%, NaCl 150 mM, Tris HCl 50 mM pH 8.0, EDTA5mM, NaF 10mM, Na4P2O7 10mM, Na3VO4 0.4mM.100 μg of proteinsfrom each cell lysate was loaded in reducing conditions on 10%acrylamide-SDS gels: the proteins were electro-transferred to

(A) Zymogram of media conditioned (CM) for 6 h from preosteoblasts plated on Plastic.Zymogram of CM collected after 6 h from preosteoblast plated on ECM t0, FN, CI, VN, CIII,rata. (E) Zymogram of CM from preosteoblast plated on FN or Collagen I, collected after 2sates and are exemplary of at least 2 experiments. (F) Q-PCR determined amounts ofnificance is for MT1-MMP: PL/FN, p= 0,0067 and PL/CI, p= 0,0083; for MMP-2: PL/FN,

254 P. Manduca et al. / Bone 44 (2009) 251–265

nitrocellulose (Amersham), processed for immunodetection witheither polyclonal phospho-pERK 1,2 (44-680, Biosource) or polyclonalp-FAK antibodies (44-624, Biosource), followed by an HRP-secondaryantibody and the bands were visualized with ECL detection kit(Amersham, UK). The membranes were stripped and re-blottedrespectively with either polyclonal ERK 1.2 (M5670, Sigma) orpolyclonal FAK antibodies (sc558, Santa Cruz).

Western blotting and immunoprecipitation

Lysates of preosteoblast adhering for 6 h on different substrates orplastic were prepared in RIPA buffer (150 M NaCl, 1% Nonidet P-40,0.1% SDS, 0.5% Deoxicolic acid, 50 mM Tris pH 8), or in integrin lysisbuffer (500 mM NaCl, 20 mM Tris HCl pH7.5, 0.5% TritonX100, 1 mMCaCl2, 1 mM MgCl2, 1 mM MnCl2, pepstatin 4 g/ml, leupeptin 10 g/ml,aprotinin 0.1 IU/ml). Samples were loaded on gels or utilized forimmunoprecipitation (IP) in equiamount/cell DNA. For WB, theproteins were electro-transferred to nitrocellulose membrane andprocessed for immunodetection with standard procedures with theantibody indicated in the figure legends: Ab rabbit polyclonal anti-integrin β1 (Chemicon AB1952) and rabbit polyclonal antibodies anti-MT1-MMP hinge region (5980-1410, Biotrend). IP was prepared asdescribed [3] with either of goat polyclonal to the terminal N region ofMT1-MMP (Santa Cruz sc-12366) or polyclonal rabbit anti-integrin β1(courtesy of Dr. I. De Curtis, S. Raffaele Scientific Institute, Milano,Italy) and transferred to nitrocellulose. IP for MT1-MMP was revealedwith one of the following Ab: rabbit polyclonal anti-integrin β1(Chemicon AB1952), anti-MMP-2 (Ab 45 kindly donated by Dr. StettlerStevenson), rabbit polyclonal anti-integrin α2 (courtesy Prof. Tarone,University of Turin, Italy), anti-integrin α5 (AB1928, Chemicon) oranti-integrin α4 (courtesy Prof. Tarone). IP for β1 integrin wasrevealed with goat polyclonal to the terminal N region of MT1-MMP(Santa Cruz sc-12366) or anti-MMP-2 (Ab 45). The membranes wereincubated 1 h with HRP-conjugated antiserum against rabbit IgG(Sigma) in TTBS-5%BSA and the bands were visualized by ECLchemiluminescence.

Zymograms

Samples of CM were always collected after 6 h of incubation of thecultures at 37 °C in serum free medium, concentrated by cold ethanolprecipitation, collected by centrifugation, and resuspended in samplebuffer for electrophoresis. Cell lysates (in RIPA buffer) were useddirectly. Gels, 12%, unless differently stated, were cast with 0.28% w/vgelatin (Type A), run at 6 °–8 °C in awater-cooled box, rinsed twice for39′ in 2.5% Triton X-100, incubated 16–18 h at 37 °C in 40 mM Tris–HCl, 0.2 M NaCl, 10 mM CaCl2. After electrophoresis, gels wereprocessed as described [3]. Exemplary gels of at least two experimentsare shown. Amounts of CM and cell lysates were loaded on gels byequivolume which, in our differentiating conditions where noproliferation occurs, corresponds also to equal amount by cell DNA.For all the zymograms of CM from clones, given the high amount ofMMP expressed, the amounts of CM loaded on gels were 1/2 of thatused for ROB. No quantization of bands in the zymograms wasattempted because the conditions to detect the full pattern of MMPsexpressed coincide with saturation in some of the lanes. Here areshown gels with a standard/cell DNA amount of sample per lane, toallow comparison.

Immunofluorescence analysis

Immunofluorescence studies were performed on differentiatingosteoblasts at day 1 or 3, plated on 12-mm diameter glass coverslips.The cell layers were fixed in 3.0% paraformaldehyde in PBS for 20 min.After cell permeabilization with 0.5% Triton X-100 in PBS (3 min), thecoverslips were washed and incubated with PBS/0.1 mg/ml BSA for

1 h at room temperature, followed by two rinsing in PBS andincubation for 1 h with primary antibodies, or 5% not immune serumin PBS/0.1 mg/ml BSA for control. After washing, coverslips wereincubated for 1 h with the appropriate rhodamine- or fluorescein-tagged secondary antibody. Coverslips were mounted in 50% glycerolin PBS. Immunofluorescence analysis was performed at a confocallaser scanner microscope Zeiss LSM 510. Primary antibodies were:mAb anti-Talin (Sigma Chemical Co. St. Louis MO/USA); rabbitpolyclonal Ab to chicken β1 integrin cytoplasmic domain (from Dr. I.De Curtis) and rabbit polyclonal Ab to MT1-MMP hinge region (5980-1410, Biotrend).

Isolation of total cytoplasmic RNA and QPCR

Total RNAs were prepared with the RNeasy kit (Quiagen) andreverse transcribed with oligo(dT) primers. mRNA expression wasanalyzed by quantitative real-time reverse transcription-PCR usingthe following primers: MMP-2 sense 5′-CTGATGTCCAGCAAGTAGACand antisense 5′-CGGGGTCCATTTTCTTCTTTAC; MT1-MMP sense5′-CGAATACCCCAAAAACATCAAA and antisense 5′-ACCTTCA-GCTTCTGGTTGTT; Integrin β1 sense 5′-CTTCAACTGCGATAGGTCCAAand antisense 5′-CTGCCAGTGTAGTTGGGATA. The relative expressionof each gene was assessed in comparison with the housekeepinggene GAPDH amplified with the following primers: sense 5′-TTCAACAGCAACTCCCATTCTT and antisense 5′-ACCAGGAAATGA-GCTTCACAAA. cDNAs were amplified for 50 cycles using iQ Supermix(Bio-Rad, Richmond, CA) containing the intercalating agent SYBRGreen in a two-step amplification scheme (95 °C, 15 s and 60 °C, 30 s).Fluorescence was measured during the annealing step on a Bio-RadiCycler iQ instrument. Blank controls that did not contain cDNA wererun in parallel. All samples were run in triplicate. Followingamplification, melting curves, with 80 steps of 15 s and a 0.5 °Ctemperature increase per step, were performed to control ampliconidentity. Experiments were repeated twice. Relative expression valueswith SEs and statistical comparison (unpaired two-tailed t test) wereobtained using Qgene software.

Results

“Early” events in osteogenesis: induction of MT1-MMP expression bysubstratum and association of MT1-MMP with integrin β1 dimers

MT1-MMP is not expressed at detectable levels in preosteoblastsand it is up modulated early when they are set in differentiationcondition. Pro-MMP-2 is expressed in preosteoblasts, and it is upmodulated during osteogenesis. MT1-MMP-activated forms of MMP-2(64–62 kDa) were not detected in the conditioned media (CM) ofpreosteoblasts where only the self-activated 69 kDa form of MMP-2was detected (Fig. 1A). Conditioned media (CM) collected 6 h afterplating preosteoblasts on endogenous ECM, prepared either frompreosteoblasts cultures at confluence (ECM t0, lacking cross linkedCollagen I) or from ROB in differentiation medium for 7 days (ECM t7,presence of cross-linked collagen and of any other collagen-depen-dent and developmentally expressed or associated ECM component)displayed a pattern of secreted MMPs that includes Pro-MMP-2,69 kDa and all the MT1-MMP-activated forms of MMP-2. Samples ofCM corresponding to equal amounts by cell DNA collected at 6 h and12 h of culture after plating on ECM t0 or ECM t7 (Fig. 1B) showed thatall the forms of MMP-2 accumulate in time at similar rate on bothECMs. We tested if individual ECM components would also induceMT1-MMP by plating preosteoblasts on various proteins and testingfor the presence of activated MMP-2 in the CMs collected after 6 h.Conditionedmedia frompreosteoblasts plated on FN and CI, containedPro-MMP-2, 69 kDa and all of its MT1-MMP-activated forms (Fig. 1C).In the same figure we show the MMPs in the CMs collected afterplating on VN, CIII, BSA or Plastic, where only the 69 kDa MMP-2 form

Fig. 2. β1 integrin expression, PTKs signaling and formation of complexes between MT1-MMP and β1 integrin dimers engaged with substrata and with MMP-2. QPCR with β1integrin specific primers of total RNA extracted after 6 h of adhesion from preosteoblasts plated on FN, CI, and PL. (B) Western blotting of reduced samples of 100 μg of protein fromlysates of preosteoblasts adhering for 19′ to 69′ PL, FN or CI, followed by detectionwith antibodies to P-FAK and FAK or to P-ERK1,2 and ERK1,2. (C) Preosteoblasts were plated for 6 h onECM t0, without (control) or with SB 203580 (10 μM), or with Genistein (20 μg/ml). Equiamounts per cell DNA of the collected CM were analyzed on zymograms. Exemplary gel of 2experiments. (D) Cell lysates from preosteoblasts plated for 6 h in serum-free medium on PL, CI, ECM t0 or FN are tested, from top to bottom, by Western blotting for MT1-MMP andβ1 integrin or by immunoprecipitation for β1 integrin, followed by Western blot for MT1-MMP. Lanes of the gels were loaded with equiamounts per cell DNA of lysates. Gels were10%, reduced. (E) From left to right: WB for integrin β1, integrin α4, integrin α5 and integrin α2 of proteins immunoprecipitated with antibody to MT1-MMP from preosteoblastsplated for 6 h on FN or on CI, as detailed below. Far right,WB for integrin β1 from lysate of preosteoblast plated on plastic after immunoprecipitationwith antibody toMT-1MMP. Gelswere 10%, reduced for the WB for integrin β1, not reduced the WB for integrin α5. Gels were 8%, reduced, the WB for α4 and α2 integrins. Gels are loaded with equiamounts per cellDNA of lysates and are exemplary of at least two experiments. (F) Equal amounts of cell lysates from preosteoblasts plated on FN for 6 h were immunoprecipitated with either Ab toMT1-MMP or Ab to integrin β1 and detected with Ab to MMP-2. Gels were 10%, reduced.

255P. Manduca et al. / Bone 44 (2009) 251–265

256 P. Manduca et al. / Bone 44 (2009) 251–265

257P. Manduca et al. / Bone 44 (2009) 251–265

was detected. MT1-MMP was detected in the cell lysates ofpreosteoblasts plated on ECM, FN and CI (Fig. 1D), and not in theequivalent per DNA amounts of the lysates of cells plated on the othersubstrata, by zymogram orWestern blot (not shown). Thus MT1-MMPand Pro-MMP-2 induction in preosteoblasts occurs by cell adhesiononly on specific substrata. Fig. 1E shows the accumulation of Pro-MMP-2 and of its MT1-MMP-activated forms in the CMs ofpreosteoblasts plated on FN or CI, collected after 2 h and 6 h afterplating. On all the protein substrata tested, after 2.5 h cell adhesionwas above 60% and after 6 h it reached about 90%, and the cells werewell spread (not shown). Induction of MT1-MMP and Pro-MMP-2 byplating on ECM, FN and CI was inhibited by Actinomicin D andCycloeximide, in conditions where adhesion and spreading wereunaffected (not shown), indicating requirement for both mRNA andprotein synthesis. Using QPCR with the specific primers on total RNAsprepared at 6 h after plating on plastic, FN or CI we detected asignificant amount of MT1-MMP RNA only in cells plated on FN and CI,and a significantly higher amount of MMP-2 mRNA in preosteoblastplated on the inducing substrata than on plastic (Fig. 1F). By PCR oftotal RNAs prepared after 2 h of preosteoblast adhesion, induction ofmRNA for MT1-MMP was already observed in the cells plated on CIand FN, and a slight increase in the mRNA for MMP-2 was detectedonly in cells plated on FN (not shown), suggesting a sequentialinduction of these MMPs by substratum.

Preosteoblasts interact with FN, CI and CIII via engagement ofdifferent dimers of the integrin β1 and utilize β3 integrin dimers toadhere and spread on VN. Using QPCR for β1 integrin (Fig. 2A), at 6 hafter cell plating we did not observe an increase in the level of mRNAassociated with the substratum-dependent induction of MT1-MMP.Considering that signaling by β1 integrin dimers occurs via FAK andERK1,2, we tested if there are differences associated with induction ofMT1-MMP in their respective activation during the first hour ofplating on FN, CI, or plastic (Fig. 2B). We observed than phosphoryla-tion of FAK was not significantly changed by plating on any of thesubstrata, and that of ERK1,2 was equally increased on all substrata,suggesting that the induction of MT1-MMP by ECM components is notmediated by activation of FAK or ERK1,2. We also tested the relevancefor induction of MMPs of other PTKs (Fig. 2C). Neither the addition topreosteoblasts during 6 h of adhesion on ECM t0 of Genistein or ofSB203580, that respectively inhibit PTKs or p38 MAPK, affected theinduction by substratum of Pro-MMP-2 and its processing. This resultsuggests the irrelevance of these signaling for the induction bysubstratum of MMPs. The inhibitors at the concentrations that weused do not affect percentage of adhesion and spreading of cells (notshown).

We tested if the induction of MT1-MMP by substrata is accom-panied by direct interaction of the enzyme with β1 integrin. Differentβ1 integrin dimers are involved in the recognition by preosteoblasts ofFN, CI and ECM. Using WB of lysates obtained from preosteoblastsafter 6 h from plating on inductive substrata and on plastic, wedetected integrin β1, while MT1-MMP was detected only in lysates ofpreosteoblasts plated on inductive substrata (Fig. 2D). When weimmunoprecipitated these lysates with an antibody for β1 integrinand then probed by WB with antibody to MT1-MMP, we detectedcomplexes of the two proteins in the preosteoblasts induced toexpress MT1-MMP by ECM, FN and CI. Immunoprecipitationwith antiMT1-MMP followed by WB with antibody to β1 integrin of a similarcell lysate from preosteoblasts plated on FN is shown in the extremeleft lane of Fig. 2E and confirmed that complexes were formedbetween the newly synthesized MT1-MMP and β1 integrin within 6 h

Fig. 3. Immunolocalization of MT1-MMP, β1 integrin, and Talin in differentiating osteoblasts.or β1 integrin (H, K), and corresponding merge (right column) in osteoblasts set to differentiaArrows point to Talin containing streaks in d to f, to Talin containing focal adhesions in g to isome regions of colocalization at day 3 between MT1-MMP and Talin (F) and the colocalizantibodies (M and N). Fields were chosen for photography to permit to appreciate the chan

from adhesion. As a control we tested an equal per cell DNA amount oflysate obtained from preosteoblasts after 6 h from plating on plastic,that we immunoprecipitated with antibody to MT1-MMP andrevealed by WB with antibody to integrin β1 (see far right of Fig.2E). The complexes formed by MT1-MMP and β1 integrin involved β1integrin dimers specifically engaged by each of the inducingsubstratum, which are respectively α4β1 and α5β1 for FN andα2β1 for CI (Fig. 2E). After probing with antibodies to the differentalpha integrin chains the immunoprecipitate with antibody to MT1-MMP obtained from lysates of preosteoblasts plated for 6 h on FN, wedetected in complexes with MT1-MMP the processed form of the α4integrin subunit that participates to the α4β1 dimer that is engagedby FN, and the α5 integrin subunit, but did not detect α2 integrin.Using equal amounts per cell DNA of lysates from preosteoblast platedon CI,α2 integrinwas detected in immunoprecipitate with MT1-MMPand not α5 integrin. Thus, the β1 integrin dimer specifically engagedby FN or CI associates in complexes withMT1-MMPwithin 6 h from itsinduction by the substratum. In immunoprecipitate with antibody toMT1-MMP of lysates of preosteoblasts plated on FN for 6 h probed byWB with antibody to MMP-2 we detected Pro-MMP2, 69 kDa MPP-2and the lower size activated forms of MMP-2, an association reportedfor functionally active MT1-MMP (Fig. 2F). In immunoprecipitate withantibody to β1 integrin of the same amount of the same sample oflysate probed by WB with antibody to MMP-2 we detected only the69 kDa form of MMP-2. The amounts of 69 kDaMPP-2 associated withMT1-MMP and β1 were similar. We did not pursue furtherquantitative study or sequential immunoprecipitation at this time.

To test if interaction of MT1-MMP with β1 dimers occurred at thefocal adhesion we used confocal microscopy to detect the immunolo-calization of MT1-MMP and β1 integrin during the osteogenicdifferentiation in double immunostaining with Talin, as marker offocal adhesion (Fig. 3). Due to lack of adequate antibodies to rat proteins,we could not performdirect double immunofluorescence forMT1-MMPand β1 integrin. The immunolocalization of MT1-MMP confirmed itsincrease in expression on the cell membrane from day 1 of differentia-tion to day 3. At day1 MT1-MMP was distributed over the cell plasmamembrane only in the formof granular deposits (Fig. 3A–C),while at day3 anti-MT1-MMP antibodies detected, in addition to granular, alsoshort linear aspects of co-localization of MT1-MMP with Talin streaks(Fig. 3D–F), evidenced in the merged panel. Co-immunolocalization forβ1 integrin and Talin showed that they co-localize in focal adhesion-like structures at day 1 (Fig. 3G–I) and also in elongated fibrillaradhesions at day 3 (Fig. 3J–L).

Relationship between AP and MT1-MMP expression duringosteogenic differentiation

We investigated if the induction of MT1-MMP has any effect on theexpression of AP, a gene induced at the onset of osteogenicdifferentiation with a pattern of timing grossly similar to that ofMT1-MMP and also transiently up modulated during osteogenesis.Our first clue to look into this issue came from the study of two clonesfrom propagated cultures of ROB. These two clones, expressed a highbasal level of AP that is not down modulated even after a long time indifferentiation conditions (Fig. 4A). These clones did not form nodulesand mineralize during long term culture in differentiation conditions(see below), while they express at high level and with persistentpattern MT1-MMP and Pro-MMP-2, and displayed in the CM theactivated forms of MMP-2 (Fig. 4B). Clones with this phenotypediffered from bulk osteoblast cultures from which derived and from

Double immunofluorescence with antibodies to Talin (A, D, G, J) and to MT1-MMP (B, E)te at 1 day (A–C and G–I) and 3 days (D–F and J–L) of culture in differentiation medium., and respectively to elongated fibrillar adhesions in J to L. The merged panel evidencesation of β1 integrin and Talin at day 1 and d 3. Negative controls with only secondaryges in the colocalization of the antibodies for MT1-MMP and β1 integrin with Talin.

Fig. 4.MT1-MMP and AP expression is co-regulated in clonal isolates from ROB. Clones V2 and V4 were analyzed during differentiation and, at time points, the media conditioned for6 h (CM) were collected in serum free medium and cell lysates were prepared. (A) Expression of AP, by quantitative enzymatic assay. (B) Zymogram of cell lysates (top) and CM(bottom) for clones V2 and V4. (C) Clone V4 was cultured in differentiating conditions without (−) or with addition of 20 μM of GM6001 (+) and secreted MMPs were analyzed byzymograms of serum free media conditioned for 6 h. (D) Western blotting for MT1-MMP of lysates from clone V4 at different time in differentiation conditions in the absence orpresence of 20 μM GM6001. (E) Expression of AP in clone V4 at different time in differentiation conditions in the absence or presence of 20 μM GM6001. (F) Expression of AP upontreatment with 300 μM Levamisole of clone V4 in differentiation conditions. (G) Zymogram of CM collected during treatment with 300 μM Levamisole of clone V4. All samples in gelswere loaded in equiamounts per cell DNA. WB samples are reduced. Gels are representative of at least two experiments. AP measures are in triplicate, standard deviation is lesser-equal to 5% (not shown). Exemplary experiment is presented of at least two repeats with similar results.

258 P. Manduca et al. / Bone 44 (2009) 251–265

other 38 clones that we concurrently isolated, that all showeddevelopmental modulation of MT-MMP and AP and that formednodules and mineralized. Since it is unlikely that we isolatedindependently two clones with simultaneously the same two alteredphenotypes, we explored the possibility that pleiotropism or co-regulation of the expression of AP and of MT1-MMPmight account forthe double phenotypic change.

We used inhibitors of one or the other enzyme in order tounderstand the relationship that determines their concomitant de-regulation. The data below are from clone V4, which gave resultssimilar to thosewe obtainedwith clone V2. Using the generic inhibitorof MMPs, GM6001, in continuous treatment in differentiationconditions and in the absence of cytotoxic effects (not shown), weobserved a decrease in the level of Pro-MMP-2 secreted in the

259P. Manduca et al. / Bone 44 (2009) 251–265

conditioned media and the inhibition of the processing of Pro-MMP-2(Fig. 4C). Only a slight decrease in the transmembrane form 66 kDa ofthe MT1-MMP protein was detected by WB, and evident by 15 days indifferentiationmedium (Fig. 4D), in agreementwith themechanism ofaction of the inhibitor used. The inhibition of MMPs was paralleled bya reduction in the level of AP (Fig. 4E), suggesting co-regulation forthese differentiation markers. Levamisole (a reversible inhibitor of APactivity) in continuous treatment in differentiation conditions and inthe absence of cytotoxic effects (not shown), inhibited AP to a levellower than 40% of untreated clones (Fig. 4F), having no effect on thelevel and pattern of expression of Pro-MMP-2 and of its activation,implying also that had no effect on MT1-MMP expression (Fig. 4G).This observation suggests that co-regulation between AP and MT1-MMP is not reciprocal, and that the level ofMT1-MMP is pleiotropic onthat of AP and of Pro-MMP-2. These observations could rationalize thedifferences between clones V2 and V4 and the bulk ROB culture as due

Fig. 5. Association of MT1-MMP and AP expression in ROB set to differentiate in the presencedifferentiate in the absence or presence of (A and B) GM6001 (20 μM), or (C and D) Furine inwere collected after 6 h in serum free medium and analyzed on zymogram (A, C and E) and Agels in equiamounts per cell DNA, and they are double than the amounts used for the clonshown). Exemplary experiments of at least two repeats with similar results are shown.

to only one mutational or epigenetic event causing two phenotypicchanges. We also considered if the pattern of positive pleiotropism ofMT1-MMP on AP observed in the clones could represent a generalmechanism of co-regulation active in the bulk population of ROB aswell. We investigated if co-regulation based on the key element of theinduction of MT1-MMP could be a modality to ensure the coordinatedgene expression required for the progression of the osteogenesis.

We tested preosteoblasts from bulk ROB cultures set todifferentiate with inhibitors and inducers of MT1-MMP or AP. In allcases the concentrations of drugs used were not cytotoxic and theamounts of DNA per culture relative to the controls were unchangedthroughout differentiation (not shown). We set preosteoblasts todifferentiate in the presence of GM6001, or of a Furine inhibitor,which specifically impairs the processing of Pro-MT1-MMP. Undercontinuous treatment with GM6001, the processing of Pro-MMP-2was severely impaired, while the amounts secreted were comparable

of inhibitors of MMPs or of an inducing antibody to MT1-MMP. ROB cultures were set tohibitor I (20 μM), or (E and F) of a rabbit polyclonal inducing antibody to MT1-MMP. CMP quantity was determined in cell lysates (B, D and F). All samples of CMwere loaded ones in Fig. 4. AP measures are in triplicate. Standard deviation is lesser-equal to 5% (not

Fig. 6. Induction or inhibition of AP expression in ROB set to differentiate does not affect MT1-MMP timely regulated expression and MMP2 levels during osteogenesis. ROB cultureswere set to differentiate in the absence or presence of (A and B) ET-1 (0.1 mM), (C and D) of the inhibitor of AP expression, and inhibitor of endogenous ET-1, BQ123 (1 μM), (E and F) ofLevamisole (300 μM). CM were collected after 6 h in serum free medium and analyzed on zymogram (A, C and E) and AP quantity was determined in cell lysates (B, D and F). Allsamples of CMwere loaded on gels in equiamounts per cell DNA, and they are double than the amounts used for the clones in Fig. 4. AP measures are in triplicate. Standard deviationis lesser-equal to 5% (not shown). Exemplary experiments of more than two repeats with similar results are shown.

260 P. Manduca et al. / Bone 44 (2009) 251–265

to control (Fig. 5A) and the level of AP was inhibited to 50%, asaverage maximum value in four experiments (Fig. 5B). Addition ofthe inhibitor after the induction of MT1-MMP and AP, here shown ina time window between 12 and 24 days during differentiation, wasstill effective in inhibiting the processing of Pro-MMP-2 and causinga decrease in the level of AP (Fig. 5B). Inhibition of MT1-MMPfunction and AP expression were reversible at all times tested (up to12 days in differentiation conditions, not shown). These data showthat also in ROB the functionality of MT1-MMP regulates the level ofAP. Treatment with GM6001 also decreased by 25% (maximum valuein three experiments) the total protein content of the cultures duringdifferentiation. This value, compared to the decrease of 50% in thelevel of AP, suggests that the down modulation of AP is specific.Treatment of differentiating ROB with the Furine inhibitor causeddecrease in the level of cellular MT1-MMP (by WB, not shown), aslight decrease in Pro-MMP2 and a severe decrease in its processing(Fig. 5C). We observed in this treatment a decrease in the level of APexpression, that was much more pronounced that the overalldecrease in protein amounts, (Fig. 5D). An antibody to MT1-MMPthat acts as inducer of the enzyme, was used to verify if up

modulation of MT1-MMP also affects the expression of AP. Thecontinuous treatment with MT1-MMP-inducing antibody causedincrease in the amount of Pro-MMP-2 and of its activated forms,which persisted also at late times of osteogenesis (Fig. 5E).Associated with this treatment we detected almost a doubling ofthe level of AP persists during culture (Fig. 5F) with a pattern thatfollowed that of the expression of MT1-MMP. This observationconfirms that MT1-MMP acts as positive regulator on AP.

We also tested ROB for expression of MT1-MMP in conditionswhere we modified the level of AP with inducers or inhibitors. Wetreated differentiating ROB continuously with one of three agents:ET-1, an endogenous positive regulator of AP, (Fig. 6A, B); BQ123, aninhibitor of endogenous ET-1 and of AP (Fig. 6C, D); Levamisole(Fig. 6E, F). We observed that neither induction of AP, up to 200% bytranscriptional up modulation with ET-1, or its inhibition of functionwith Levamisole, to 10% of control, or the down modulation of APtranscription with BQ123, to about 60%, had any significant effect onthe level or the processing of Pro-MMP-2. This confirms that pleiotropyof MT1-MMP on AP is the modality of the co-regulated expression ofthese genes in osteogenesis.

261P. Manduca et al. / Bone 44 (2009) 251–265

We also searched for molecular interactions of AP with MT1-MMPor integrin β1. Using double immunostain and confocal microscopywe detected AP at cell plasma membrane with granular distributionand in increasing amounts at day 1and at day 3 of differentiation. Nocolocalization was detected with β1 integrin at either time point (notshown). AP was not coimmunoprecipitated with MT1-MMP or β1from cell lysates of preosteoblast adhering for 6 h on ECM or FN or CIor from lysates of differentiating mature osteoblast (not shown).These, although negative, results suggest that it is unlikely that APinteracts directly at themolecular level with MT1-MMP or integrin β1.

Down modulation of MT1-MMP and AP, nodule formationand mineralization

It is widely acknowledged that the expression of AP, MT1-MMPand MMP-2, along with the formation of multilayered structures (i.e.nodules), are prerequisites for progression of osteogenesis to localized

Fig. 7. Calcium deposition and von Kossa stain of ROB differentiating in the presence of inhibthe absence or presence of GM6001 (20 μM). (B) Clone V4 was set to differentiate in the ababsence or presence of Levamisole (300 μM), or (D) of ET-1 (0.1 μM) or (E) of MT1-MMP-indu(A–C and E). (D) von Kossa stain at the last time of the culture (24 days). Exemplary experimetriplicates and the standard deviation was lesser that 5% at all points and is not shown.

mineral deposition. Indeed it has been shown that the level ofexpression of AP affects the amount of mineralization and that in theabsence of MT1-MMP or MMP-2 expression cultures of osteogeniccells do not mineralize. Less investigated are significance, in context,and mechanism(s) that govern the down modulation of AP afternodule formation in vitro, or in vivo within osteocytes. Also, poorlyunderstood is the relevance of the expression of MMPs for theformation of nodules and the functional significance that the downmodulation of MT1-MMP might have for mineralization. We herereport the capability to form nodules of cultures of ROB and of clones,in culture in differentiation conditions upon different treatmentswhich affect the expression and the level of MT1-MMP and/or of AP.We also show, by measuring the incorporation of Calcium, or by vonKossa stain, the effects of the co-regulated inhibition or induction ofMT1-MMP and AP, or of the induction or inhibition only of AP, on themineralization of the ECM (Fig. 7 and Table 1). In cultures of ROBcontinuously treated with GM6001, showing a decrease in AP

itors or inducers of MMPs or inhibitor of AP. (A) ROB cultures were set to differentiate insence or presence of GM6001 (20 μM). (C) ROB cultures were set to differentiate in thecing antibody. Insoluble calcium incorporation in ECM during culture was measured innts of at least two repeats with similar results are shown. The dosage of Ca+2 was done in

Table 1Summary of the level of AP expression, of MT-MMP-1 expression and pattern ofmodulation, of the capability to form nodules and of the Ca+2 incorporation in the ECMor of the mineralization in cultures of ROB and of clone V4 differentiating in thepresence of inhibitors or inducers of MT1-MMP or of AP

AP%ROBcontrol

MT1-MMPmodality

Nodules Ca+2% ROB control ormineralization

ROBUntreated 100 Modulated Present 100GM6001 62 Inhibited Absent 42Levamisole 10 Modulated Reduced 57ET-1 200 Modulated Present Highera

BQ123 58 Modulated Present Lowera

Ab/MT1-MMP 200 High, persisting Absent 28Furine inhibitor 54 Inhibited Absent n.d.

Clone V4Untreated 180 High, persisting Absent 14GM6001 118 Inhibited Absent 10Levamisole 108 High, persisting Absent 7

Here are listed for ROB cultures and for the derived clone V4, the maximumvalues of APor of Ca incorporated in the ECM attained during differentiation. These are expressed as% of the maximum value of the control, ROB mass cultures. It is reported if theexpression of MT1-MMP is modulated or affected by treatments during differentiation.It is registered if nodules are formed. Absent or present for nodules refer to theinspection of at least 3, 60 mm diameter, Petri dishes.

a von Kossa stain, visual evaluation.

262 P. Manduca et al. / Bone 44 (2009) 251–265

consequent to the inhibition of MT1-MMP, nodule formation did notoccur, and calcium deposition was decreased (Fig. 7A). In ROB treatedwith Levamisole, showing about 10% of the level of AP found incontrol, and expressing MMPs in quantity and modulation similar tothe control (Fig. 7B), was observed decrease of mineralization(incorporation of Calcium reduced to 57% of control) but also ofnodule formation, suggesting that the inhibition of AP expressionmight influence also the formation of nodules. In ROB that expresscontinuously MT1-MMP and high levels of AP upon treatment withinducing antibody, no nodules formed and calcium incorporation was28% than control. Similarly, in cultures of clone V4 (Fig. 7C) whichexpressed stably high level of AP (about 200% compared to the bulkculture of ROB) and continued to express in time high level of MT1-MMP, no nodules formed and Calcium incorporationwas about 14% ofthat of ROB. These data suggest that there is not a univocal or directrelationship between high expression of MT1-MMP and formation ofnodules, or high expression of AP and mineralization. Instead,formation of nodules took place only when the developmentallymodulated expression of MT1-MMP occurred above a certain thresh-old level; too low or too high level of MT1-MMP expression, and alsovery low expression of AP, inhibited nodule formation. Only if noduleswere formed, the level of expression of AP influenced the amount ofmineralization (Fig. 7). Moreover, mineralization took place in thenodules only if MT1-MMP was down modulated after its transientexpression, even when AP was highly expressed. Table 1 summarizesthe observations above. Nodule formation thus emerges as aprerequisite or a checkpoint for the deposition of mineral dependingfrom an adequate level of transient expression of MT1-MMP.

Discussion

Homeostatic and morphogenetic roles of MT1-MMP were illu-strated in various cell types and contexts [4,6–10,13–20,42–44], andwere not yet investigated in osteogenic cells. We here report about thehomeostatic and morphogenetic role of MT1-MMP in osteogenicdifferentiation.

We show that induction of MT1-MMP occurs in response topreosteoblast and osteoblast-endogenous ECM and to specificcomponents thereof: Collagen type I and Fibronectin, recognized byosteoblasts via β1 integrins. Adhesion on these substrata induces theexpression of MT1-MMP in preosteoblasts, and increases that of Pro-MMP-2, causing accumulation of each gene's mRNAs and proteins,

initiating pericellular proteolysis and, through the activation of Pro-MMP-2, enhancing general proteolysis. Vitronectin or Collagen type IIIis not able to induce MT1-MMP or Pro-MMP-2 expression. The factthat the ECM deposed by preosteoblast and by mature osteogeniccultures equally induce expression of MT1-MMP in preosteoblasts,suggests that the interaction established by the cells with ECM mayregulate during osteogenesis the expression of MT1-MMP. Theinduction of MT1-MMP does not involve immediate responses viaFAK and ERK1.2, p38 MAPK or Genistein-sensitive PTKs. We areinvestigating if the src and Akt pathways might mediate induction bysubstratum of MT1-MMP, or if G-proteins might be involved.

MT1-MMP participates in the sensing of microenvironmentthrough the binding of the β1 integrins engaged with ECM and thesubstratum-engaged integrin dimer(s) form complexes with MT1-MMP once this gets exposed on the cell membrane. After 6 h ofadhesion on inductive substrata MT1-MMP is associated in complexeswith β1 integrin dimers andwith all the forms of Pro-MMP-2. Integrinβ1 coimmunoprecipitates MT1-MMP and only 69 kDa MMP-2. Wehave not tested for the presence of MT1-MMP, β1 integrin and 69 kDain the same complex, and we do not know if such tri-molecularcomplex form, or if β1 binds 69 kDa MMP-2 in an alternative fashionto its binding to MT1-MMP. This point requires further investigationand quantitative studies. After 1 day of osteogenic differentiation,MT1-MMP does not localize in focal adhesions formed by β1 dimers.However, MT1-MMP co-localizes with Talin-rich streaks after 3 days,but only a minority of the MT1-MMPmolecules on the cell membraneis involved. These findings are compatible with the possibility that thecomplexes MT1-MMP/β1 integrin, formed after 6 h of adhesion oninducing substrata are transient, or do not involve integrins organizedin focal adhesion streaks, and may be different from complexesformed later and localizing MT1-MMP in Talin streaks. In addition, wedo not know if other colocalization for MT1-MMP and β1 integrin, notoverlapping with Talin, might exist at any time.

MT1-MMP regulates positively and pleiotropically the expressionof AP and is required for the induction and for the maintenance of APexpression. The coincident changes in MMPs and AP observed in thevariant clones are best interpreted as being consequential to thepleiotropy of MT1-MMP on AP. The use of inhibitors and inducers ofMT1-MMP or AP in clones and in bulk population of ROB demon-strates that a general mechanism ensures the co-regulated expressionof these genes during osteogenesis, under the control of MT1-MMP.The positive pleiotropy of MT1-MMP on AP is in agreement with theobservation that induction of AP occurs with a delay after that of MT1-MMP when preosteoblasts are set in differentiation medium andwhen they are plated on MT1-MMP-inducing substrata (our unpub-lished data).

MT1-MMP activates Pro-MMP-2 and is also responsible for Pro-MMP-2 up modulation. Accordingly, treatment of differentiatingosteoblasts with a Furine inhibitor which reduces the production ofmature MT1-MMP, also slightly decreases Pro-MMP-2. In addition,treatment with an inducing antibody to MT1-MMP increases MT1-MMP and Pro-MMP-2 expression.

Co-regulation of MT1-MMP and AP is not a reciprocal event, asshown by the fact that inhibitors (Levamisole or BQ 123) or inducers(ET-1) of AP do not change induction, amount and pattern ofexpression of MT1-MMP and Pro-MMP-2 during osteogenesis, inagreement with previous reports using other transcriptional inducersof AP, IGFII [31], and FMSCalciumfluor [40]. Inhibition or induction ofAP expression instead affects the level of mineral deposition.

We could not find evidence of a physical interaction between APand the transmembrane form of MT1-MMP in lysates obtained frompreosteoblast after 6 h adhesion on MT1-MMP-inducing substrata orfrom osteoblasts at day1 or 3 of differentiation (not shown). It is thuslikely that mechanism(s) other than direct molecular interactionoccurs in the pleiotropic regulation of AP by MT1-MMP. A molecularmediator could exist for the effect of MT1-MMP on AP. This mediator

263P. Manduca et al. / Bone 44 (2009) 251–265

should be a protein that stimulates AP expression, that is inhibited byinhibitors of MMPs and that is up modulated as consequence of thepersistence of the expression of MT1-MMP. Theoretically, themolecular mediator of the pleiotropy could be either a transcriptionfactor(s) or a secreted protein. This issue remains open.

Osteogenic cells cultured in monolayer form, with a stochasticmodality, foci where the expression of MT1-MMP and AP aremaximal.In these foci multilayering takes place and nodules are formed. Thenodule is the prospective site of mineral deposition; during theformation of nodules local conditions favorable for prospectivemineral deposition must be generated and we thus regard theformation of nodules in culture as a checkpoint before mineralizationcan occur.

MT1-MMP expression during osteogenesis in vitro is required forthe changes that establish permissive conditions for the formation ofnodules, as shown by lack of nodules in differentiating cultures inthe presence of MMPs inhibitors. In addition, we show here thatwhile on one hand transient expression of MT1-MMP is required fornodule formation, on the other hand down modulation of MT1-MMP is required for mineralization. We already reported that adecrease of the MT1-MMP enzyme activity is observed in thenodules formed in differentiating cultures when a steady level ofMT1-MMP mRNA is still detected [3]. We show here that thecontinuing expression of MT1-MMP, in variant clones and in ROBcultures in the presence of a MT1-MMP inducing antibody, inhibitsthe formation of nodules and mineralization. We also show thatinhibition of AP by Levamisole also reduces nodule formation,besides decreasing mineralization.

In order to multilayer and form nodules, the cells expressing APandMT1-MMP present in the foci must loosen their attachment to theECM and then migrate. Creation of localized conditions for cellmultilayering can be started through pericellular proteolysis by MT1-MMP. The consequent localized changes in the ECM could makeavailable new fragments and ECM-bound components and also allowfor novel interactions with other osteoblasts products present in themicroenvironment. In this novel context, fragments derived by theproteolytic activity of MT1-MMP and proteins secreted in a devel-opmentally regulated pattern [1] could associate into the ECM focallylysed and be instrumental to cell multilayering. For instance, TCa andTCb fragments that are produced by the digestion by MMP-2 of theproducts of MT1-MMP collagenolytic activity [12] and they could beaccumulated where they are produced, in pericellular localization.Their biological function and cell receptors are yet unknown and needto be investigated. Among the developmentally up modulatedsecreted proteins [47], BSP is an interesting candidate to promotecell multilayering because of its binding to the cells viaαvβ3 integrinsthat can interfere with cell adhesion mediated by β1 integrins.Moreover, BSP binds MMP-2 and Pro-MMP-2, bridging them to thecell membrane [46] and protecting their gelatinase activity fromendogenous inhibitors [48,49] and the BSP/MMP-2 complex inducesthemigration of osteogenic cells through collagen, in an in vitro assays[50]. In support for a role of BSP in nodule formation, in preliminarytests, we observed that BSP accumulation in the cell layer isspecifically decreased in differentiating ROB in the presence of MT1-MMP inhibitors, while its soluble form is present in steady amounts. Itseems worth investigating further this aspect. In addition, the lack ofexpression of MMP-2 also impairs mineralization and the defect inMMP-2−/− osteogenic cells is not rescued by addition of soluble MMP-2, suggesting that pericellular localization of MMP-2 in the ECM isrequired for mineralization [26]. We suggest that BSP in the foci ofECM remodeling could produce pericellular localization of theproteolytic activity of MMP-2 in the same spots where greatestexpression of MT1-MMP occurred, enhancing pericellular proteolysis,and that this event is also required for mineralization. Worth of noticeis the recent finding that Pro and activated forms of MMP-2are susceptible of function-inhibitory phosphorylation and that their

P-forms are, in vitro, dephosphorylated by AP [51]. AP is highlyexpressed by cells in the foci where nodules form and it could enhancethe pericellular proteolysis via activation of free or BSP-boundpericellular MMP-2.

Also interesting is the fact that binding to BSP could also subtractPro-MMP-2 from interaction with MT1-MMP [46]. This last, when notbound to MMP-2, is susceptible to loose enzymatic activity viaectodomain shedding [21,22]. In other cell types [19,20], uponexhaustive degradation of collagen matrix by MT1-MMP, a feedbackis instituted that converts collagenolytic activity into gelatinolyticactivity and brings to autolysis of MT1-MMP and loss of the Collagen I-β1-mediated induction of its expression [21,22]. The soluble 44 kDafragment of MT1-MMP, produced by autolysis of transmembraneMT1-MMP is known to inhibit the transmembrane form frominteraction with Collagen I [21]. Our observation that both theexpression above a threshold level and the down-modulation ofMT1-MMP are necessary for the formation of the nodules and arerequired for the mineral deposition, together with the fact that MT1-MMP is indeed downmodulated in the nodules before mineralization,agree with the possibility that shedding of MT1-MMP by feed back isactivated in the nodules before mineralization. BSP binding toremodeled ECM during nodule formation and subtracting MMP-2from interaction with MT1-MMP could contribute to the shedding ofMT1-MMP and would be compatible with the fact that BSP expressionis required for mineralization, as shown by the report that BSPknockout mice and the osteogenic cells derived from them [52] andosteogenic cells transfected with iRNA for BSP [32], have defects inmineralization. It is possible that the production of soluble self-inhibitory shed fragments of MT1-MMP in nodules, may influence notonly locally but also in the rest of the culture theoverall expressionof theMT1-MMP. The combined pericellular proteolytic actions, collagenolysisby MT1-MMP and enhanced gelatinolysis by MMP-2, the consequentdegradation of FN and C I and loss of β1 integrin-mediated cellattachment, and the novel engagement of cells to BSP viaαvβ3 integrinsmay foster cell multilayering and the formation of nodules. Reduced celladhesion and partial disengagement of the cell receptors for CI and FNoccurring during formation of the nodules could cause the decrease ofMT1-MMP transcriptional activity, and, as corollary of the pleiotropicregulation, also decrease of the expression of AP. The radical change inthe balance of proteolysis/deposition, will eventually allow for novelcollagen deposition in fibers.

The local conditions favorable for prospective mineral depositioninclude deposition of collagen fibers and nucleating events. Studiesusing TEM show that in nodules formed by cultured human, rat ormouse osteoblast mineral is deposed on collagen fibers ([53], and ourunpublished results). However, BSP induces HA nucleation whenembedded in ECM and is localized in clusters of needle like crystals inthe osteogenic ECM, but not on collagen fibers [32]. Other proteinsmight be involved in predisposing focal deposition of mineral afterremodeling of ECM, shedding of MT1-MMP, and deposition of novelcollagen fibers in the nodules. A candidate for this role is the COOHterminal fragment of Pro-Collagen I (C3), increased in amounts duringosteogenesis [54] through the combined effects of increased CIsynthesis and increased of the Pro-collagen-processing enzymeBMP1 (our unpublished results). C3 interacts with cells via α2β1integrins, and is detected in newly deposed osteoid in fetal long bones(Manduca et al., in preparation). When novel collagen depositionoccurs in the nodules, C3 could increase the possibility of the cells torestore β1 integrin-mediated interactions with ECM, and/or titer theshed form of MT1-MMP that inhibits these interactions.

Once a novel local ECM is formed in the nodules deposition ofmineral could occur “by default” on the novel collagen fibers.

The amount of mineral deposition in the nodules is roughlydependent on the level of AP expression. Thus, mineral deposition inosteogenic cultures is severely decreased under AP-inhibitory treat-ment with Levamisole, a finding that is consistent with the full or

264 P. Manduca et al. / Bone 44 (2009) 251–265

partial impairment of mineralization in osteoblasts of AP null or +/−mice [36]. An increase in the mineralization is detected in culturesunder AP-inductive treatments, with ET-1, IGFII and FMSCalciumfluor[31,40], and that shows a pattern of MT1-MMP down-modulationsimilar to control cultures.

Concomitant expression by the cells of the osteogenic lineage of APand Collagen I was proposed to be the unique and sufficientpermissive condition for mineralization to occur by default in bonetissue [29]. We show here that high expression of AP in itself is notsufficient for mineral deposition if it is not associated with the downregulation of MT1-MMP. Our data require adaptation of the viewproposed above and argue that there are dynamic processes which arerequired for mineralization, that these occur by integration of theexpression of genes during osteogenesis, and that developmentallymodulated proteolysis by MT1-MMP is prerequisite for themineralization.

In summary, our data show that the pattern of expression of MT1-MMP regulates multiple steps in the osteogenic progression in vitro.MT1-MMP transient expression integrates the cell phenotypewith thechanges in ECM, and it links between them the expression of AP andMMP-2, in the phenotypic progression of osteogenesis. MT1-MMPexpression is required for morphogenetic changes and localizedpredisposition for deposition of mineral. The data show thatosteogenic cells traverse checkpoints. One such checkpoint occurs atthe onset of differentiation, another at nodule formation, with MT1-MPP involved at both points. There is sufficient potential for furtherinvestigation that might also give us tools to prevent pathologicalbone metabolism in syndromes with local or systemic manifestationsor in altered environmental conditions, such as long time bedding orspatial flight.

Acknowledgments

This work was supported with public funding to Prof. P. Manducafrom the Agenzia Spaziale Italiana, within the project OSMA.

References

[1] Owen TA, Aronow M, Shalhoub V, Barone LM, Wilming L, Tassinari MS, et al.Progressive development of the rat osteoblast phenotype in vitro: reciprocalrelationships in expression of genes associated with osteoblast proliferation anddifferentiation during formation of the bone extracellular matrix. J Cell Physiol1990;143:420–30.

[2] Liu F, Malaval L, Aubin JE. Global amplification polymerase chain reaction revealsnovel transitional stages during osteoprogenitor differentiation. J Cell Sci 2003;116:1787–96.

[3] Filanti C, Dickson GR, Di Martino D, Ulivi V, Sanguineti C, Romano P, et al. Theexpression of metalloproteinase-2, -9, and -14 and of tissue inhibitors-1 and -2 isdevelopmentally modulated during osteogenesis in vitro, the mature osteoblasticphenotype expressing metalloproteinase-14. J Bone Miner Res 2000;15:2154–68.

[4] Tam EM, Wu YI, Butler GS, Stack MS, Overall CM. Collagen binding properties of themembrane type-1 matrix metalloproteinase (MT1-MMP) hemopexin C domain. Theectodomain of the 44-kDa autocatalytic product of MT1-MMP inhibits cell invasionby disrupting native type I collagen cleavage. J Biol Chem 2002;277:39005–14.

[5] Hiraoka N, Allen E, Apel IJ, Gyetko MR, Weiss SJ. Matrix metalloproteinasesregulate neovascularization by acting as pericellular fibrinolysins. Cell 1998;95:365–77.

[6] Itoh Y, Ito N, Nagase H, Evans RD, Bird SA, Seiki M. Cell surface collagenolysisrequires homodimerization of the membrane-bound collagenase MT1-MMP. MolBiol Cell 2006;17:5390–9.

[7] Toth M, Chvyrkova I, Margarida Bernardo M, Hernandez-Barrantes S, Rafael. Pro-MMP-9 activation by the MT1-MMP/MMP-2 axis and MMP-3: role of TIMP-2 andplasma membranes. Biochem Biophys Res Commun 2003;308:386–95.

[8] Cowell S, Knäuper V, Stewart ML, D'Ortho MP, Stanton H, Hembry RM, et al.Induction of matrix metalloproteinase activation cascades based on membrane-type 1matrix metalloproteinase: associated activation of gelatinase A, gelatinase Band collagenase 3. Biochem J 1998;331:453–8.

[9] Strongin AY, Collier I, Bannikov G, Marmer BL, Grant GA, Goldberg GI. Mechanismof cell surface activation of 72-kDa type IV collagenase. Isolation of the activatedform of the membrane metalloprotease. J Biol Chem 1995;270:5331–8.

[10] Hernandez-Barrantes S, Toth M, Bernardo MM, Yurkova M, Gervasi DC, Raz Y, et al.Binding of active (57 kDa) membrane type 1-matrix metalloproteinase (MT1-MMP) to tissue inhibitor of metalloproteinase (TIMP)-2 regulates MT1-MMPprocessing and pro-MMP-2 activation. J Biol Chem 2000;275:12080–9.

[11] Ohuchi E, Imai K, Fujii Y, Sato H, Seiki M, Okada Y. Membrane type 1 matrixmetalloproteinase digests interstitial collagens and other extracellular matrixmacromolecules. J Biol Chem 1997;272:2446–51.

[12] Gioia M, Monaco S, Fasciglione GF, Coletti A, Modesti A, Marini S, et al.Characterization of the mechanisms by which gelatinase A, neutrophil collage-nase, and membrane-type metalloproteinase MMP-14 recognize collagen I andenzymatically process the two a-chains. J Mol Biol 2007;368:1101–13.

[13] Holmbeck K, Bianco P, Yamada S, Birkedal-Hansen H. MT1-MMP: a tetheredcollagenase. J Cell Physiol 2004;200:11–9.

[14] Hiraoka N, Allen E, Apel IJ, Gyetko MR, Weiss SJ. Matrix metalloproteinasesregulate neovascularization by acting as pericellular fibrinolysins. Cell 1998;95:365–77.

[15] Deryugina EI, Ratnikov B, Monosov E, Postnova TI, DiScipio R, Smith JW, et al. MT1-MMP initiates activation of pro-MMP-2 and integrin alphavbeta3 promotesmaturation of MMP-2 in breast carcinoma cells. Exp Cell Res 2001;263:209–23.

[16] Gálvez BG, Matías-Román S, Yáñez-Mó M, Sánchez-Madrid F, Arroyo AG. ECMregulates MT1-MMP localization with beta1 or alphavbeta3 integrins at distinctcell compartments modulating its internalization and activity on humanendothelial cells. J Cell Biol 2002;159:509–21.

[17] Zigrino P, Drescher C, Mauch C. Collagen-induced proMMP-2 activation by MT1-MMP in human dermal fibroblasts and the possible role of alpha2beta1integrins.Eur J Cell Biol 2001;80:68–77.

[18] van Hinsbergh VW, Engelse MA, Quax PH. Pericellular proteases in angiogenesisand vasculogenesis. Arterioscler Thromb Vasc Biol 2006;26:716–28.

[19] Baciu PC, Suleiman EA, Deryugina EI, Strongin AY. Membrane type-1 matrixmetalloproteinase (MT1-MMP) processing of pro-alphav integrin regulates cross-talk between alphavbeta3 and alpha2beta1 integrins in breast carcinoma cells. ExpCell Res 2003;291:167–75.

[20] Ellerbroek SM, Fishman DA, Kearns AS, Bafetti LM, Stack MS. Ovarian carcinomaregulation of matrix metalloproteinase-2 and membrane type 1 matrix metallo-proteinase through beta1 integrin. Cancer Res 1999;59:1635–41.

[21] Ellerbroek SM, Wu YI, Overall CM, Stack MS. Functional interplay between type Icollagen and cell surface matrix metalloproteinase activity. J Biol Chem 2001;276:24833–42.

[22] TothM, Hernandez-Barrantes S, Osenkowski P, BernardoMM, Gervasi DC, ShimuraY, et al. Complex pattern of membrane type 1 matrix metalloproteinase shedding.J Biol Chem 2002;277:26340–50.

[23] Manduca P, Palermo C, Caruso C, Brizzolara A, Sanguineti C, Filanti C, et al. Rat tibialosteoblasts III: propagation in vitro is accompanied by enhancement of osteoblastphenotype. Bone 1997;21:31–9.

[24] Holmbeck K, Bianco P, Caterina J, Yamada S, Kromer M, Kuznetsov SA, et al. MT1-MMP-deficient mice develop dwarfism, osteopenia, arthritis, and connectivetissue disease due to inadequate collagen turnover. Cell 1999;99:81–92.

[25] Holmbeck K, Bianco P, Pidoux I, Inoue S, Billinghurst RC, Wu W, et al. Themetalloproteinase MT1-MMP is required for normal development and main-tenance of osteocyte processes in bone. J Cell Sci 2005;118:147–56.

[26] Mosig RA, Dowling O, DiFeo A, Ramirez MC, Parker IC, Abe E, et al. Loss of MMP-2disrupts skeletal and craniofacial development and results in decreased bonemineralization, joint erosion and defects in osteoblast and osteoclast growth. HumMol Genet 2007;16:1113–23.

[27] Karsdal MA, Andersen TA, Bonewald L, Christiansen C. Matrix metalloproteinases(MMPs) safeguard osteoblasts from apoptosis during transdifferentiation intoosteocytes: MT1-MMPmaintains osteocyte viability. DNA Cell Biol 2004;23:155–65.

[28] Hessle L, Johnson KA, Anderson HC, Narisawa S, Sali A, Goding JW, et al. Tissue-nonspecific alkaline phosphatase and plasma cell membrane glycoprotein-1 are centralantagonistic regulatorsofbonemineralization. ProcNatlAcadSciUSA2002;99:9445–9.

[29] Murshed M, Harmey D, Millán JL, McKee MD, Karsenty G. Unique coexpression inosteoblasts of broadly expressed genes accounts for the spatial restriction of ECMmineralization to bone. Genes Dev 2005;19:1093–104.

[30] Veillette CJ, von Schroeder HP. Endothelin-1 down-regulates the expression ofvascular endothelial growth factor-A associatedwith osteoprogenitor proliferationand differentiation. Bone 2004;34:288–96.

[31] Palermo C, Manduca P, Gazzerro E, Foppiani L, Segat D, Barreca A. Potentiating roleof IGFBP-2 on IGF-II-stimulated alkaline phosphatase activity in differentiatingosteoblasts. Am J Physiol Endocrinol Metab 2004;286:648–57.

[32] Gordon JA, Tye CE, Sampaio AV, Underhill TM, Hunter GK, Goldberg HA. Bonesialoprotein expression enhances osteoblast differentiation and matrix miner-alization in vitro. Bone 2007;41:462–73.

[33] Harmey D, Johnson KA, Zelken J, Camacho NP, Hoylaerts MF, Noda M, et al.Elevated skeletal osteopontin levels contribute to the hypophosphatasia pheno-type in Akp2(−/−) mice. J Bone Miner Res 2006;21:1377–86.

[34] Anderson HC, Sipe JB, Hessle L, Dhanyamraju R, Atti E, Camacho NP, et al. Impairedcalcification around matrix vesicles of growth plate and bone in alkalinephosphatase-deficient mice. Am J Pathol 2004;164:841–7.

[35] Fedde KN, Blair L, Silverstein J, Coburn SP, Ryan LM, Weinstein RS, et al. Alkalinephosphatase knock-out mice recapitulate the metabolic and skeletal defects ofinfantile hypophosphatasia. J Bone Miner Res 1999;14:2015–26.

[36] Wennberg C, Hessle L, Lundberg P, Mauro S, Narisawa S, Lerner UH, et al.Functional characterization of osteoblasts and osteoclasts from alkaline phospha-tase knockout mice. J Bone Miner Res 2000;15:1879–88.

[37] Salasznyk RM, Klees RF, WilliamsWA, Boskey A, Plopper GE. Focal adhesion kinasesignaling pathways regulate the osteogenic differentiation of human mesenchy-mal stem cells. Exp Cell Res 2007;313:22–37.

[38] Salasznyk RM, Klees RF, Hughlock MK, Plopper GE. ERK signaling pathwaysregulate the osteogenic differentiation of human mesenchymal stem cells oncollagen. Cell Commun Adhes 2004;11:137–53.

265P. Manduca et al. / Bone 44 (2009) 251–265

[39] Lai CF, Chaudhary L, Fausto A, Halstead LR, Ory DS, Avioli LV, et al. Erk is essentialfor growth, differentiation, integrin expression, and cell function in humanosteoblastic cells. J Biol Chem 2001;276:14443–50.

[40] Manduca P, Marchisio S, Astigiano S, Zanotti S, Galmozzi F, Palermo C, et al.FMS⁎Calciumfluor specifically increases mRNA levels and induces signaling via MAPK42,44 and not FAK in differentiating rat osteoblasts. Cell Biol Int 2005;29:629–37.

[41] Castoldi M, Pistone M, Caruso C, Puddu A, Filanti C, Piccini D, et al. Osteoblasticcells from rat long bone. II: adhesion to substrata and integrin expression inprimary and propagated cultures. Cell Biol Int 1997;21:7–16.

[42] Palmieri D, Poggi S, Ulivi V, Casartelli G, Manduca P. Pro-collagen I COOH-terminaltrimer induces directional migration andmetalloproteinases in breast cancer cells.J Biol Chem 2003;278:3639–47.

[43] Koshikawa N, Schenk S, Moeckel G, Sharabi A, Miyazaki K, Gardner H, et al.Proteolytic processing of laminin-5 by MT1-MMP in tissues and its effects onepithelial cell morphology. FASEB J 2004;18:364–6.

[44] Seiki M, Yana I. Roles of pericellular proteolysis by membrane type-1 matrixmetalloproteinase in cancer invasion and angiogenesis. Cancer Sci 2003;94:569–74.

[45] Borrirukwanit K, LafleurMA,Mercuri FA, Blick T, Price JT, Fridman R, et al. The type Icollagen induction of MT1-MMP-mediated MMP-2 activation is repressed byalphaVbeta3 integrin in human breast cancer cells. Matrix Biol 2007;26:291–305.

[46] Dumin JA, Dickeson SK, Stricker TP, Bhattacharyya-Pakrasi M, Roby JD, Santoro SA,et al. Pro-collagenase-1 (matrix metalloproteinase-1) binds the alpha(2)beta(1)integrin upon release from keratinocytes migrating on type I collagen. J Biol Chem2001;276:29368–74.

[47] Nefussi JR, Brami G, Modrowski D, Oboeuf M, Forest N. Sequential expression ofbone matrix proteins during rat calvaria osteoblast differentiation and bonenodule formation in vitro. J Histochem Cytochem 1997;45:493–503.

[48] Karadag A, Ogbureke KU, Fedarko NS, Fisher LW. Bone sialoprotein, matrixmetalloproteinase 2, and alpha(v)beta3 integrin in osteotropic cancer cellinvasion. J Natl Cancer Inst 2004;96:956–65.

[49] Jain A, Fisher LW, Fedarko NS. Bone sialoprotein binding to matrix metallopro-teinase-2 alters enzyme inhibition kinetics. Biochemistry 2008;47:5986–95.

[50] Karadag A, Fisher LW. Bone sialoprotein enhances migration of bone marrowstromal cells through matrices by bridging MMP-2 to αvβ3 integrin. J BoneMineral Res 2006;21:1627–36.

[51] Ferrera D, Poggi S, Biassoni C, Dickson GR, Astigiano S, Barbieri O, et al. Three-dimensional cultures of normal human osteoblasts: proliferation and differentia-tion potential in vitro and upon ectopic implantation in nude mice. Bone2002;30:718–25.

[52] Tye CE, Hunter GK, Goldberg HA. Identification of the type I collagen-bindingdomain of bone sialoprotein and characterization of the mechanism of interaction.J Biol Chem 2005;280:13487–92.

[53] Malaval L, Wade-Guéye NM, Boudiffa M, Fei J, Zirngibl R, Chen F, et al. Bonesialoprotein plays a functional role in bone formation and osteoclastogenesis. J ExpMed 2008;205:1145–53.

[54] Sariahmetoglu M, Crawford BD, Leon H, Sawicka J, Li L, Ballermann BJ, et al.Regulation of matrix metalloproteinase-2 (MMP-2) activity by phosphorylation.FASEB J 2007;21:2486–95.