Complementary interplay between matrix metalloproteinase-9 ... · endochondral ossification requires restricted VEGF bioavailability that might be regulated by MMPs through ECM remodeling

RESEARCH ARTICLE

dmm.biologists.org224

INTRODUCTIONThe long bones form by the process of endochondral ossification,during which a cartilage template is first laid down, thensubsequently removed and replaced by bone (Colnot, 2005;Karsenty and Wagner, 2002; Olsen et al., 2000; Ortega et al., 2004;Provot and Schipani, 2005). During embryogenesis, at the sites ofthe future skeletal elements, mesenchymal cell condensationsform, within which precursor cells differentiate into chondrocytesthat proceed through a defined series of proliferative andmorphogenetic steps to form a cartilage template. This cartilagetemplate or anlagen contains less mature cells at the two ends andprogressively more mature cells toward the middle. Aschondrocytes mature, they hypertrophy and secrete a specializedextracellular matrix (ECM) containing collagen X, a collagen thatis specific to hypertrophic cartilage (HC). The HC ECM is

subsequently mineralized and invaded by blood vessels from theperichondrium. Vascular invasion is immediately preceded byinvasion by preosteoclasts that remove the HC ECM. Concurrentwith vascular invasion, osteoblasts are recruited and produce boneECM. The chondrocytes at the ends of the cartilage templatecontinue to proliferate and mature to form more HC, which in turnis continually removed and replaced with bone. This process ofendochondral ossification results in longitudinal bone growth andin a thin layer of HC at the two ends of the bone whose size isrelatively constant. Dysregulation of this complex and highlycoordinated process result in abnormal bone formation seen inmany of the skeletal dysplasias, and its abnormal activation maylead to heterotopic ossification (Cohen, 2000a; Cohen, 2000b;Cohen, 2006).

Numerous studies have contributed to characterization of theprecise interplay of diverse factors that regulate HC differentiation,vascular invasion, ECM degradation, and bone formation andremodeling (Colnot, 2005). Among them, hypoxia induced factor1-a (HIF1-a) and one of its targets, vascular endothelial growthfactor (VEGF), are crucial in controlling chondrocyte survival,proliferation, angiogenesis and bone formation (Gerber et al., 1999;Schipani et al., 2001; Zelzer et al., 2004). Some ECM molecules arealso essential: Col2a1 null mice lack endochondral ossification (Liet al., 1995) and perlecan null mice display chondrodysplasia(Arikawa-Hirasawa et al., 1999; Gustafsson et al., 2003). Several

Disease Models & Mechanisms 3, 224-235 (2010) doi:10.1242/dmm.004226© 2010. Published by The Company of Biologists Ltd

Complementary interplay between matrixmetalloproteinase-9, vascular endothelial growth factorand osteoclast function drives endochondral boneformationNathalie Ortega1,5,*, Ke Wang2,3,6, Napoleone Ferrara4, Zena Werb1,* and Thiennu H. Vu2,3,*

SUMMARY

Long bone development depends on endochondral bone formation, a complex process requiring exquisite balance between hypertrophic cartilage(HC) formation and its ossification. Dysregulation of this process may result in skeletal dysplasias and heterotopic ossification. Endochondral ossificationrequires the precise orchestration of HC vascularization, extracellular matrix remodeling, and the recruitment of osteoclasts and osteoblasts. Matrixmetalloproteinase-9 (MMP-9), vascular endothelial growth factor (VEGF) and osteoclasts have all been shown to regulate endochondral ossification,but how their function interrelates is not known. We have investigated the functional relationship among these regulators of endochondral ossification,demonstrating that they have complementary but non-overlapping functions. MMP-9, VEGF and osteoclast deficiency all cause impaired growthplate ossification resulting in the accumulation of HC. VEGF mRNA and protein expression are increased at the MMP-9–/– growth plate, and VEGFactivity contributes to endochondral ossification since sequestration of VEGF by soluble receptors results in further inhibition of growth platevascularization and ossification. However, VEGF bioavailability is still limited in MMP-9 deficiency, as exogenous VEGF is able to rescue the MMP-9–/– phenotype, demonstrating that MMP-9 may partially, but not fully, regulate VEGF bioavailability. The organization of the HC extracellular matrixat the MMP-9–/– growth plate is altered, supporting a role for MMP-9 in HC remodeling. Inhibition of VEGF impairs osteoclast recruitment, whereasMMP-9 deficiency leads to an accumulation of osteoclasts at the chondro-osseous junction. Growth plate ossification in osteoclast-deficient miceis impaired in the presence of normal MMP-9 expression, indicating that other osteoclastic functions are also necessary. Our data delineate thecomplementary interplay between MMP-9, VEGF and osteoclast function that is necessary for normal endochondral bone formation and provide amolecular framework for investigating the molecular defects contributing to disorders of endochondral bone formation.

1Department of Anatomy, 2Department of Medicine and 3Lung Biology Center,University of California, San Francisco, CA 94143, USA4Department of Molecular Oncology, Genentech, 1 DNA Way, South San Francisco,CA 94080, USA5Present address: IPBS, CNRS UMR5089, Laboratoire de Biologie Vasculaire, 31077Toulouse Cedex, France6Present address: Department of Respiratory Medicine, The Second Hospital of JilinUniversity, Changchun, Jilin 130041, People’s Republic of China*Authors for correspondence ([email protected], [email protected],[email protected])

Dise

ase

Mod

els &

Mec

hani

sms

D

MM

Disease Models & Mechanisms 225

MMPs and VEGF in endochondral ossification RESEARCH ARTICLE

matrix metalloproteinases (MMPs) are also crucial forendochondral ossification and bone formation, notably MMP-9 (Vuet al., 1998), MMP-14 (Holmbeck et al., 1999; Holmbeck et al., 2003;Zhou et al., 2000) and MMP-13 (Inada et al., 2004; Stickens et al.,2004). In humans and mice, MMP-2 contributes to bonehomeostasis but not to endochondral ossification (Egeblad et al.,2007; Inoue et al., 2006; Martignetti et al., 2001; Mosig et al., 2007;Tuysuz et al., 2009).

MMP-9–/– mice show impeded vascularization and ossificationof HC at the growth plate, resulting in a lengthened HC zone withaccumulation of late HC (Ortega et al., 2005; Vu et al., 1998). Thedelayed ossification may be the result of an imbalance betweenangiogenic and anti-angiogenic activities at the MMP-9–/– growthplate, since MMP-9–/– HC explants show delayed release ofangiogenic activities (Vu et al., 1998). One of these angiogenicactivities may be VEGF function. Sequestration of VEGF activity bysoluble VEGF receptor results in inhibition of endochondralossification and accumulation of HC similar to, albeit more severethan, that seen in MMP-9 deficiency (Gerber et al., 1999). DecreasingVEGF expression in HC by gene targeting using Cre-lox technologyalso gives similar results (Haigh et al., 2000). Mice that express onlythe VEGF120 soluble isoform also show defects in cartilagevascularization and endochondral ossification (Maes et al., 2002;Zelzer et al., 2002). However, mice expressing only VEGF164, or onlyVEGF188, have normal endochondral ossification of metaphysealgrowth plates, although the VEGF188 mice have abnormal epiphyseswith increased cell apoptosis and delayed secondary ossification(Maes et al., 2004). These studies suggest a model whereby normalendochondral ossification requires restricted VEGF bioavailabilitythat might be regulated by MMPs through ECM remodeling.

Based on this model, we proposed that VEGF synthesized byhypertrophic chondrocytes would be sequestered in the HC ECM,and that MMP-9 degrades ECM molecules leading to the releaseof bound VEGF. Indeed, this seems to be the case in models ofpancreatic islet carcinogenesis and glioblastoma (Bergers et al.,2000; Du et al., 2008). However, despite the expected matrixremodeling function of MMP-9, there are no direct data definingthe relationship between VEGF and MMP-9 in the context of matrixremodeling in postnatal endochondral ossification. In this studywe explore the mechanisms whereby VEGF and MMPs synergizeto drive endochondral bone formation.

RESULTSVEGF accumulates in late hypertrophic chondrocytes in MMP-9-deficient mice during postnatal endochondral ossificationSince there is defective growth plate ossification in both MMP-9 andVEGF deficiency, and an apparent imbalance in angiogenic activitiesin the MMP-9–/– HC (Gerber et al., 1999; Vu et al., 1998), we firstasked whether VEGF expression or function is limiting at the MMP-9–/– growth plates. We previously showed that Vegf (also known asVegfa) mRNA is expressed in the most mature HC in normal mice(Gerber et al., 1999; Vu et al., 1998). Here, we compared the VEGFmRNA and protein patterns in wild-type (WT) mice with MMP-9–/– mice. In situ hybridization showed strong expression of VegfmRNA by HC in both WT and MMP-9 null growth plates (Fig. 1A-D). However, the area of Vegf expression in MMP-9–/– growth platesis expanded compared with that in WT growth plates, consistentwith the expanded zone of HC. VEGF immunostaining showed that

the zone of VEGF protein expression is also expanded in the MMP-9–/– growth plate (Fig. 1E,F). Notably, the zone of the VEGF proteinexpression was larger compared with that of Vegf mRNA expression,and included the upper HC in both WT and MMP-9–/– growthplates. This is probably the result of persistence of the VEGF proteinin the surrounding ECM, even when the cells have decreased theirexpression of mRNA. The increase in VEGF protein expression inthe MMP-9 growth plates was confirmed by western blotting ofprotein extracts from WT and MMP-9–/– growth plate HC, which,in addition, showed that the main isoform of VEGF that wasexpressed was VEGF164 (Fig. 1G). Interestingly, we observed smallerforms of VEGF migrating around 28 kDa in HC from both WT andMMP-9–/– HC (Fig. 1G). Considering its molecular weight, this formmay be the plasmin-processed form of VEGF, since the plasminogensystem is expressed in fetal bones and plays a role in endochondralossification (Daci et al., 2003; Hackel et al., 1995). These smaller formsof VEGF were noted only in the samples with intact perichondrium,suggesting that this is where this processing takes place (Fig. 1G).

MMP-9 deficiency does not impair VEGF release and activityAlthough there is abundant expression of VEGF, the impairedossification at the MMP-9–/– growth plates may still be becauseof limiting VEGF activity if VEGF is sequestered and not releasedefficiently. To determine whether VEGF release from HC isimpaired in MMP-9 deficiency, we compared VEGF released fromthe ossification fronts (including the bone-cartilage interface andthe last few rows of hypertrophic chondrocytes) of WT and MMP-9–/– growth plate HC. We found that there was no difference inthe amount of VEGF released into growth medium in the absenceof MMP-9 (Fig. 1H). Addition of active, purified MMP-9 did notincrease VEGF release from the cartilage explants (data not shown).

We next determined whether VEGF activity contributes tovascularization and ossification of the MMP-9–/– growth plate.We inhibited VEGF activity in vivo using soluble VEGF receptors.Administration of mFlt(1-3)-IgG systemically to MMP-9–/– miceby daily intraperitoneal injection for 10 days resulted in furtheraccumulation of HC in the treated mice. The treated HC zone wasapproximately 20% longer than control HC (Fig. 2A-C). As has beenseen with mFlt(1-3)-IgG treatment of WT mice (Gerber et al.,1999), there was inhibition of vascular growth into HC.Platelet/endothelial cell adhesion molecule-1 (PECAM-1)immunostaining showed a reduced number of metaphysealcapillaries running parallel to hypertrophic chondrocyte columnsin the treated mice (Fig. 2D,E). Bone formation was also inhibited,with a shortened trabecular bone region. There were few primarytrabeculae, and the secondary trabeculae were shortened andthicker in the treated mice (Fig. 2F,G). These data show that activeVEGF is present and contributes to the vascularization andossification of HC in the absence of MMP-9.

Exogenous VEGF partially normalizes the MMP-9 deficiencyphenotypeWe next asked whether bioavailable VEGF could still be limitingin MMP-9 deficiency. We determined whether exogenous VEGFcould rescue the MMP-9–/– phenotype. We administeredrecombinant human VEGF165 systemically to 1-week-old MMP-9 null mice by daily intraperitoneal injection. After 1 week oftreatment, there was a 35% reduction in the length of the HC zone

Dise

ase

Mod

els &

Mec

hani

sms

D

MM


MMPs and VEGF in endochondral ossificationRESEARCH ARTICLE

in the treated mice (Fig. 3A,B,F). Continuing treatment withVEGF165 for 2 weeks did not further reduce the length of MMP-9–/– HC, which was still significantly larger than WT HC (Fig.3C-F). An increase in the size of the trabecular bone region in thetreated mice accompanied the reduction in the size of the HC zone(Fig. 3G,H). The decrease in HC zone size was not because of adecrease in chondrocyte proliferation. The populations ofproliferative chondrocytes that incorporated bromodeoxyuridine(BrdU) were similar in both untreated and treated mice (Fig. 3I,J).Thus, exogenous VEGF was able to partially overcome the effectsof MMP-9 deficiency, and enhance ossification of MMP-9–/– HC,thus stimulating HC exit from the growth plate, ECM remodelingand trabecular bone formation.

MMP-9 deficiency affects the organization of the hypertrophicchondrocyte matrixThe partial normalization obtained with VEGF treatment, andthe fact that we could not detect a defect in VEGF release fromthe MMP-9–/– growth plate suggests that MMP-9 contributes

an additional function besides regulating VEGF bioavailability.Angiogenesis at the growth plate requires both endothelial cellgrowth stimulated by angiogenic factors and ECM degradationto allow for vascular invasion. The activity of MMP-9 may benecessary to degrade the ECM for path clearing, as well as forregulation of sequestered VEGF. We therefore determinedwhether there are alterations in ECM remodeling at the MMP-9–/– growth plates. A detailed analysis of the hypertrophic areaof postnatal metatarsals from WT or MMP-9–/– mice stainedwith Safranin-O showed that accumulation of HC correlates witha decrease of Safranin-O staining in MMP-9–/– mice (Fig. 4A,B),indicating a loss of matrix proteoglycans. Semi-thin sectionsstained with Toluidine Blue showed disorganized columns andthickening of the ECM pericellular to the hypertrophicchondrocytes, and delimiting the lacuna, in MMP-9–/– growthplates (Fig. 4C,D). Electron microscopy studies confirmed thesedifferences: granular staining characteristic of proteoglycans wasless dense in MMP-9–/– samples, concomitant with a clearincrease in collagen fibril density (Fig. 4E,F). In the most distal

Fig. 1. VEGF is expressed in the MMP-9–/– mice. (A-D) Bright-field (A,B) and dark-field (C,D) photographs of WT (A,C) and MMP-9–/– (B,D) sections of 1-week-old metatarsal growth plates hybridized with a Vegf antisense probe. Vegf is expressed in a subpopulation of hypertrophic chondrocytes (hc).(E,F) Immunostaining of tissue sections of 1-week-old WT (E) and MMP-9–/– (F) growth plates with VEGF antibody. VEGF protein is found in the ECM surroundingthe hypertrophic chondrocytes (arrows). (G) Western blot of protein extracts from 2-week-old WT and MMP-9–/– growth plate HC probed with an anti-VEGFantibody showing that the major VEGF isoform expressed is VEGF164. Note the smaller forms of VEGF migrating around 28 kDa in both WT and MMP-9–/– HCwith intact perichondrium (arrows). (H) Quantification by ELISA of VEGF protein secreted into the medium by cultured 2-week-old WT and MMP-9–/– growthplate ossification fronts. Bars, 200 mm (A-F).

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



rows of hypertrophic chondrocytes, the transverse septa showedstriking differences between WT and MMP-9–/– samples. In WTsamples, a progressive lysis of the last transverse septa eventuallyallows vascular invasion and elimination of condensed orapoptotic chondrocytes (Fig. 4G,I). In MMP-9–/– samples,transverse septa were thicker and denser with accumulation oflarge collagen type I fibers right under the septa (Fig. 4H,J). Highermagnification of the last transverse septa showed that, in WTsamples, collagen fibrils are completely degraded on the bonemarrow side, whereas in MMP-9–/– samples, collagen fibrils werestill visible and dense (Fig. 4I,J). These results suggest that ECMremodeling is impaired in MMP-9 deficiency.

Either MMP-9 deficiency or VEGF inhibition results in anaccumulation of cleaved collagen II at the cartilage-bone junctionBesides stimulating endothelial cell growth and morphogenesis,VEGF recruits osteoclasts, which in turn secrete ECM degradingenzymes (Gerber et al., 1999). Thus, the activities of VEGF andMMP-9 may complement each other. In previous work we showedthat MMP-9 and MMP-13 co-operate in the degradation ofcartilage type II collagen (Engsig et al., 2000; Stickens et al., 2004).To further determine the relationship between VEGF and MMP-9 function in ECM degradation during endochondral ossification,we analyzed samples for the presence of a degraded MMP substrate,cleaved type II collagen, at the ossification front. Immunostainingwith an antibody that specifically recognizes a neoepitope on theN-terminal 3/4-length collagen fragments resulting fromcollagenase cleavage (Billinghurst et al., 1997) demonstrated thepresence of the specific cleaved type II collagen fragments in a thinline at the cartilage-bone junction in the WT growth plate (Fig.5A). The zone of staining slightly increased in the MMP-9–/– mice,suggesting that cleaved collagen fragments accumulate as a resultof MMP-9 deficiency (Fig. 5B). Both MMP-13 and MMP-14 (MT1-MMP) can cleave type II collagen and both are expressed at thecartilage-bone junction. Thus, the accumulation may be secondaryto the lack of MMP-9 activity, which would further degrade andsolubilize the collagen fragments cleaved by MMP-13 and/orMMP-14. Alternatively, it may be secondary to the induction ofMMP-13 or other enzyme activities that can generate these cleavedtype II collagen fragments. Indeed, as shown below (Fig. 8), weobserved the increased expression of MMP-13 in the expanded HCzone in the MMP-9–/– growth plates, which may account for thisincrease in collagen II cleavage.

Inhibition of VEGF activity in MMP-9–/– growth plates bytreatment with mFlt(1-3)-IgG resulted in a slight furtherenlargement of the zone of cleaved type II collagen staining (Fig.5C,D). These data suggest that VEGF activity also modulates thelocal balance of ECM degrading activities at the ossification front.

MMP-9 is not sufficient for normal endochondral ossificationwithout osteoclastsMMP-9 is expressed not only by osteoclasts and their progenitors,but also by TRAP-negative cells at the chondro-osseous junction(Fig. 5E,F). We previously showed that reconstitution of MMP-9expression in the hematopoietic, and thus the osteoclast, lineageby bone marrow transplantation is sufficient to restore normalendochondral ossification of the MMP-9–/– growth plates (Vu etal., 1998). We therefore postulated that MMP-9 expression by non-osteoclasts is not crucial in this process. To test this hypothesis,we investigated mice with null mutations in c-Fos, Csf1, PU.1 (alsoknown as Sfpi1) and Rank (also known as Tnfrsf11a), which arerequired for osteoclast development. Mice with null mutations inany of these genes are osteopetrotic owing to the absence ofosteoclasts (Dougall et al., 1999; Grigoriadis et al., 1994; Li et al.,2000; Marks and Lane, 1976; McKercher et al., 1996; Tondravi etal., 1997; Yoshida et al., 1990). We found that mice that weredeficient in any of these factors displayed increased HC zones atthe growth plates, consistent with defective endochondralossification (Fig. 6A,C,E,G,I,K,M,O).

We next asked whether this phenotype was associated withreduced MMP-9 expression at the growth plates. Except for the c-

Fig. 2. Inhibition of VEGF activity causes further accumulation of HC inthe MMP-9–/– growth plate. (A,B) Hematoxylin and eosin (H&E)-stainedsections of 1-week-old MMP-9–/– mice treated for 10 days with either controlIgG (A) or mFlt(1-3)-IgG (B) showing lengthening of the HC zone (hc) in thetreated mice. (C) Bar graphs of the mean and standard deviation of the HCzone lengths of six control and mFlt-treated bones; the differences betweencontrol and treated samples were significant (P<0.05). (D,E) Immunostaining oftissue sections with PECAM-1 antibody showing metaphyseal vessels runningparallel to hypertrophic chondrocyte columns in the control-IgG-treated mice(D, arrows) and a reduced number of these vessels in the mFlt(1-3)-IgG-treatedmice (E, arrows). (F,G) H&E-stained sections showing primary (F, arrowheads)and secondary (F, arrows) trabeculae in the control-IgG-treated mice, andthickened and shortened secondary trabeculae (G, arrows) in the mFlt(1-3)-IgG-treated mice. Bars, 400 mm (A,B); 100 mm (D-G).

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



Fos null growth plates, which showed decreased MMP-9 expression(Fig. 6B,D), all the other osteoclast-deficient growth platesshowed normal levels of MMP-9 protein expression byimmunohistochemistry (Fig. 6F,H,J,L,N,P), suggesting that theremay be a compensatory increase in expression of MMP-9 by theTRAP-negative non-osteoclastic cells. The decreased expressionof MMP-9 in c-Fos–/– growth plates may reflect the regulation ofMMP-9 expression by c-Fos (Bischof et al., 2003). The normalexpression of MMP-9, coupled with impaired endochondralossification in the osteopetrotic mice, indicates that MMP-9expression by non-osteoclasts is not sufficient to drive theossification process normally and that another osteoclastic functionis required for normal endochondral ossification.

MMP-9 deficiency leads to accumulation of TRAP-positive cells atthe chondro-osseous junction, which is reversed by VEGFtreatmentThe requirement of osteoclast function suggests their recruitmentmight be an essential regulatory mechanism of endochondral boneformation. During primary ossification, osteoclast recruitment intothe cartilage anlagen requires MMP-9 activity (Engsig et al., 2000).We therefore determined whether impaired osteoclast recruitmentmight contribute to the defect in HC ossification at the MMP-9–/–growth plates. We compared the number of multinucleated (fullydifferentiated osteoclasts) and mononucleated (osteoclastprecursors) TRAP-positive cells along the chondro-osseousjunction between WT and MMP-9–/– mice. Surprisingly,compared with WT, MMP-9–/– mice showed significantlyincreased numbers of both multinucleated osteoclasts (WT,3.0±0.4, versus MMP-9–/–, 6.5±0.8; P<0.01) and mononucleatedcells (WT, 4.8±0.4, versus MMP-9–/–, 8.5±0.9; P<0.01) along thetotal length of the chondro-osseous junction (Fig. 7A-C). These

results show that MMP-9 is not necessary for osteoclastrecruitment during growth plate ossification, and that impairedosteoclast recruitment is not a cause of the defective ossificationin the MMP-9–/– growth plates. This is in contrast to the functionof VEGF, which is necessary for osteoclast recruitment in bothprimary ossification of the cartilage anlagen (Maes et al., 2002;Zelzer et al., 2002) and during growth plate ossification (Gerber etal., 1999).

We next determined the effect of exogenous VEGF on osteoclastrecruitment at the MMP-9–/– growth plate. We assessed thenumber of TRAP-positive cells at the chondro-osseous junction of1-week-old MMP-9–/– mice treated for 1 week with either vehiclealone or VEGF. Surprisingly, MMP-9–/– mice treated with VEGFhad a decreased number of mononuclear TRAP-positive cellscompared with mice treated with vehicle alone (control, 7.7±0.7,versus VEGF treated, 3.8±0.4; P<0.01) (Fig. 7D-F). Thus, theimprovement in vascularization and ossification of MMP-9–/– HCby exogenous VEGF is associated with normalization of osteoclastnumber at the chondro-osseous junction.

Other MMPs contribute to endochondral ossification in theabsence of MMP-9Since MMP-9 alone is not sufficient for normal endochondralossification without osteoclasts, which also express other MMPs,we asked whether the activity of other MMPs might be necessary.We found high expression of MMP-13 and MMP-14 at thecartilage-bone junction by in situ hybridization (Fig. 8A-D). MMP-13 expression was observed in terminal hypertrophic chondrocytesand in cells on the trabecular bone surfaces (Fig. 8A), consistentwith previously reported results (Inada et al., 2004; Stickens et al.,2004). MMP-14 expression was seen in cells at the cartilage-bonejunction, which may be (pre)osteoclasts, as well as in cells along

Fig. 3. Treatment with VEGF rescuesendochondral ossification defects inMMP-9 –/– mice. (A-D) H&E-stained sectionsof metatarsals of 1-week-old MMP-9–/– micetreated for 1 week (A,B) or 2 weeks (C,D) witheither vehicle or recombinant humanVEGF165, showing shortening of the HCzone (hc) in VEGF-treated mice (B,D)compared with vehicle-treated mice (A,C).(E) H&E-stained section of 3-week-old WTmetatarsals showing the size of the normalHC zone (hc). (F) Bar graphs of the mean andstandard deviation of the HC zone lengths ofsix vehicle- and VEGF-treated bones after 1and 2 weeks of treatment; the differencesbetween control and treated samples ateach time point were significant (P<0.01).(G,H) H&E-stained sections of the metatarsalsof MMP-9–/– mice treated for 2 weeks witheither vehicle (G) or VEGF (H) showinglengthening of the trabecular bone region(tb) in VEGF-treated mice. (I,J) BrdU labelingof proliferating cells in the metatarsal growthplates of 1-week-old MMP-9–/– mice treatedfor 1 week with VEGF. BrdU-positive cellsstain brown. Bars, 400 mm (A-D); 200 mm(G,H); 100 mm (I,J).

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



the trabecular bone, and in the perichondrium (Fig. 8C). Thepattern of expression of MMP-13 and MMP-14 in the MMP-9 nullgrowth plate resembled that in the WT mice, except that theirexpression levels were increased slightly (Fig. 8B,D). In addition,there was a larger zone of expression of MMP-13 in the latehypertrophic chondrocytes in the expanded HC zone in the MMP-9–/– mice (Fig. 8D).

To determine whether inhibition of all MMP activity wouldfurther inhibit endochondral ossification, we treated MMP-9–/–mice systemically by daily intraperitoneal injection with GM6001,a broad-spectrum inhibitor of MMPs. After 1 week of treatment,the HC zone in the treated mice further lengthened, increasing byapproximately 25% in length (Fig. 9A,B,E). This became morepronounced after 2 weeks of treatment, with a 65% length increase(Fig. 9C-E). Vascularization was completely inhibited in the treatedmice. PECAM-1 immunostaining showed that there were nometaphyseal capillaries that ran parallel to the HC columns, as seenin the control mice. Instead, there were dilated vascular spacesrunning perpendicular to the HC (Fig. 9F,G). Bone formation wasalso affected. Primary trabeculae were not present, and thesecondary trabeculae became a large mass of bone (Fig. 9H,I).

Fig. 4. Altered hypertrophic cartilage ECM in MMP-9 null growth plates.(A,B) Safranin-O staining of 3-week-old metatarsals from WT (A) and MMP-9–/–(B) mice, showing a decrease in the staining of the lower portion of thehypertrophic chondrocyte area in MMP-9–/– mice (outlined by dotted lines).(C,D) Semi-thin sections of Epon-embedded ossification fronts from WT (C)and MMP-9–/– (D) mice, stained with Toluidine Blue, showing the presence ofhypertrophic chondrocytes in the last rows at the ossification fronts in MMP-9–/– mice compared with WT mice. The longitudinal septae of the lacunaesurrounding the last rows of hypertrophic chondrocytes are thicker in theMMP-9–/– mice (D, arrow) than in the WT mice (C, arrow). (E,F) Transmissionelectron micrographs of the areas indicated by arrows in C and D, showing theultrastructure of the ECM in these areas. The ECM in these septae in WT mice isrich in proteoglycans, as indicated by numerous black dots, and collagenfibers are sparse (E). By contrast, in MMP-9–/– mice, the ECM of these septaecontains dense collagen fibers and fewer black dots, indicating a decrease inproteoglycan content, which was also observed with Safranin-O staining in Aand B. (G,H) Transmission electron micrographs showing the ultrastructure ofthe ECM of the last transverse septa in WT (G) and MMP-9–/– (H) mice. Theboxed areas are shown under higher magnification in I and J. Transverseseptae in the MMP-9–/– samples (J) show more collagen fibers and there is anaccumulation of collagen type I fibers under the septae compared with thetransverse septae in WT samples (I). Bars, 500 mm (A,B); 50 mm (C,D); 20 nm(E,F); 16 mm (G,H); 2 mm (I,J).

Fig. 5. MMP-9 deficiency and inhibition of VEGF alter the accumulation ofcleaved collagen at the growth plate. (A-D) Immunostaining with anantibody against cleaved fragments of type II collagen on sections ofmetatarsals of 2-week-old WT (A) and MMP-9–/– (B) mice, and of sections of1-week-old MMP-9–/– mice treated with cont-IgG (C) or mFlt(1-3)-IgG (D) for1 week, showing the presence of cleaved collagen at a thin line along thecartilage-bone junction in WT mice (A, arrows), a larger zone of cleavedcollagen in MMP-9–/– mice (B, arrows), and a larger zone with an irregularborder in MMP-9–/– mice treated with mFlt(1-3)-IgG (D, arrows).(E,F) Immunostaining for MMP-9 and substrate staining for tartrate-resistantacidic phosphatase (TRAP) activity on the same tissue sections from WTgrowth plates showing MMP-9-positive cells that are both TRAP positive (F,arrows) and TRAP negative (F, arrowheads). Bars, 100 mm (A-E); 32 mm (F).

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



Since the trabecular bone phenotype resulting from theinhibition of all MMPs may be because of inhibition of boneremodeling, we next determined whether the treated mice showedan abnormal osteoclast number by staining for TRAP activity, amarker of osteoclasts. We observed large numbers of osteoclastsboth at the cartilage-bone junction and on the trabecular bonesurfaces in the treated mice (Fig. 9J,K), indicating that the abnormalbone remodeling in the treated mice was probably due to analteration in osteoclastic activity and not in osteoclast number.

DISCUSSIONIn this study we explored the functional relationships betweenMMP-9, VEGF and osteoclasts at the growth plate duringendochondral bone formation. Our results demonstrated that allof these components are necessary for normal endochondralossification and that their functions are interdependent (Fig. 10).MMP-9 may degrade ECM components to regulate angiogenesiseither by path clearing or by regulating the bioavailability ofsequestered growth factors such as VEGF. VEGF not only regulates

angiogenesis, but also regulates the recruitment of other crucialcells such as osteoclasts. Osteoclasts function by providing MMP-9 activity, as well as other MMPs, which contributes to HC turnoverand bone ECM remodeling.

The functions of VEGF and MMP-9 in endochondral ossification donot completely overlapPrevious studies have suggested that the function of MMP-9 inendochondral ossification is to release VEGF that is sequesteredin the ECM. The accumulation of HC is seen in mice that aredeficient in MMP-9, treated with a VEGF inhibitor, or withconditional deletion of the Vegf gene in cartilage (Gerber et al., 1999;Haigh et al., 2000; Vu et al., 1998). In addition, MMP-9-deficientHC shows a delayed recruitment of capillaries in vitro (Vu et al.,1998). Our data now show that the functions of MMP-9 and VEGF,while synergistic, do not completely overlap, and thus the MMP-9–/– phenotype results from more than a defect in VEGF release.VEGF mRNA and protein are abundant in the MMP-9–/– HC.VEGF release from the MMP-9–/– HC is not impaired in vitro,

Fig. 6. Osteopetrotic mice have growth plate abnormalities with normal expression of MMP-9. Histological sections of metatarsal growth plates from c-Fos+/+ (A), c-Fos–/– (C), Csf1+/+ (E), Csf1op/op (G), PU.1+/+ (I), PU.1–/– (K), Rank+/+ (M) and Rank–/– (O) mice showing enlarged HC zones in the mutant growthplates. Immunostaining for MMP-9 showed similar expression of MMP-9 in the Csf1+/+ (F), Csf1op/op (H), PU.1+/+ (J), PU.1–/– (L), Rank+/+ (N) and Rank –/– (P)growth plates. The c-Fos–/– growth plate showed decreased immunostaining for MMP-9 (D) compared with the control c-Fos+/+ growth plate (B). Bars, 100 mm.

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



and there is VEGF activity contributing to the vascularization andossification of the MMP-9–/– growth plates. In addition,administration of exogenous recombinant VEGF to MMP-9–/–mice only partially rescues the phenotype. Thus, although VEGFbioavailability may be limiting in MMP-9 deficiency, our datasuggest that MMP-9 is also necessary for other processes unrelatedto the release of VEGF.

Osteoclasts are essential for normal endochondral ossificationand are regulated differently by MMP-9 and VEGFOsteoclasts are well-known matrix-resorbing cells and theirpresence at the chondro-osseous junction suggests that theycontribute to endochondral ossification by secreting ECM-degrading enzymes. However, cells that express MMP-9, but that

are TRAP negative, are also present at the chondro-osseousjunction and may also contribute to ECM degradation. Wepredicted that, in the absence of osteoclasts, non-osteoclastic cellsexpressing MMP-9 would contribute to growth plate remodeling.Consistent with this model, we found that mice lacking osteoclastsshowed MMP-9 expression in non-osteoclastic cells and displaymilder defects in growth plate ossification than those seen in MMP-9 deficiency. These non-osteoclastic MMP-9-expressing cells mayfunction as the TRAP-negative ‘chondroclasts’. The four differentgenetically osteopetrotic mice point to primarily osteoblasts andtheir progenitors as the non-osteoclast MMP-9-expressing cells:Rank–/– mice have no osteoclasts but still have mononuclearphagocytes; Fos–/– mice have no osteoclasts but have an excessof macrophages and monocytes; Csf1op/op mice have no osteoclastsand very few monocytes and macrophages; and PU.1 –/– mice lackmyeloid cells (Dougall et al., 1999; Grigoriadis et al., 1994; Li et al.,2000; Marks and Lane, 1976; McKercher et al., 1996; Tondravi etal., 1997; Yoshida et al., 1990). From their location, the TRAP-negative cells that express MMP-9 are presumably osteoblasts, theirprogenitors, or other mesenchymal cells that can express MMP-9(Heissig et al., 2002). However, since HC does accumulate inosteopetrotic mice in the presence of MMP-9 expression, it is clearthat MMP-9 produced by these TRAP-negative ‘chondroclasts’ isnot sufficient for normal growth plate ossification. Whether thisowes to a defect in MMP-9 activation or to other products madeby osteoclasts remains to be determined. Clearly other MMPs,which are produced by osteoclasts, contribute, as further inhibitionof MMPs in MMP-9–/– growth plates resulted in completeinhibition of endochondral ossification.

Fig. 7. Osteoclast recruitment to the chondro-osseous junction. (A,B) TRAPstaining of tissue sections showing more TRAP+ cells at the chondro-osseousjunction in MMP-9–/– growth plates (B) compared with WT growth plates (A).(C) Quantification of mononuclear and multinuclear TRAP+ cells at the WT andMMP-9–/– growth plates. (D,E) TRAP staining of tissue sections showing areduced number of TRAP+ cells at the chondro-osseous junction in MMP-9–/–growth plates treated with VEGF (E) compared with untreated MMP-9–/–growth plates (D). (F) Quantification of mononuclear and multinuclear TRAP+cells at the untreated and VEGF-treated MMP-9–/– growth plates. For C and F,mononucleated and multinucleated TRAP+ cells were counted under themicroscope at a 40� magnification and expressed as the total number ofTRAP+ cells along the total length of the chondro-osseous junction. Thecounting was performed on a minimum of three mice per group, using bothhind limbs, and on three to six different sections, with three metatarsals persection, for each hind limb. The results are expressed as mean ± standard errorof the mean (S.E.M.). The results were statistically significant (P<0.01) betweenWT and MMP-9–/– samples for both cell populations in C, and between VEGFtreated and untreated samples for the mononuclear population in F. Theresults were not statistically significant between VEGF treated and untreatedsamples for the multinuclear population in F.

Fig. 8. MMP expression at the growth plate. Dark-field photographs oftissue sections of metatarsals from 1-week-old WT (A,C) and MMP-9–/– mice(B,D) hybridized with antisense probes against MMP-13 (A,B) and MMP-14(C,D). MMP-13 is expressed in cells at the cartilage-bone junction, consistentwith the previously reported expression in terminal hypertrophicchondrocytes (A,B, arrowheads) and in the metaphysis (A,B, asterisks), in bothWT and MMP-9–/– mice. In addition, MMP-13 is also expressed in some lowerhypertrophic chondrocytes in MMP-9–/– mice (B, arrows). MMP-14 isexpressed in cells at the cartilage-bone junction (C,D, arrowheads) and themetaphysis (C,D, asterisks). It is also expressed in the perichondrium (C,D,arrows). Bars, 200 mm.

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



VEGF regulates osteoclast recruitment during formation of theprimary ossification site, as well as during growth plate ossification(Engsig et al., 2000; Gerber et al., 1999; Maes et al., 2002; Zelzer etal., 2002). It also regulates osteoclastic differentiation and function(Aldridge et al., 2005; Niida et al., 2005; Zelzer and Olsen, 2005).Similar to VEGF, MMP-9 positively regulates osteoclast recruitmentduring primary ossification (Engsig et al., 2000). Surprisingly,MMP-9 deficiency resulted in an increased number of osteoclastsat the chondro-osseous junction. Whether this is due to acompensatory mechanism that increases the recruitment of cellswith ECM-degrading activities to overcome the MMP-9 deficiencyor to negative regulation of osteoclast recruitment during growthplate ossification by MMP-9 remains to be elucidated. It is alsopossible that MMP-9 is required for osteoclast invasion of HC,either by degrading the ECM or by regulating the bioavailability ofosteoclast chemotatic factors, and thus its deficiency causesosteoclasts to accumulate at the chondro-osseous junction.

Interestingly, administration of exogenous VEGF did not recruitmore osteoclasts, but rather resulted in a reduction in theirnumbers, at the MMP-9–/– chondro-osseous junction. Theadditional VEGF appeared to act differentially on the recruitmentof osteoclast progenitors, because exogenous VEGF decreased thenumber of mononuclear TRAP-positive cells to a greater extentthan the multinucleated population. The improved angiogenesisand ossification by exogenous VEGF may negate the need for

compensatory recruitment of ECM-resorbing cells. However, VEGFalso affects the formation and/or function of osteoblasts (Hiltunenet al., 2003; Maes et al., 2002; Street et al., 2002; Zelzer et al., 2002),which are known to modulate osteoclast formation. Thus,exogenous VEGF may also reduce osteoclast number through itseffects on osteoblasts. We conclude that the effect of VEGF onosteoclasts is not simply to increase recruitment, but variesdepending on the complex interplay of many different factors.

MMPs and VEGF have different effects on bone turnoverIn contrast to the synergistic control of cartilage vascularization,the functions of VEGF and MMPs differ in the regulation of bonehomeostasis. Inhibition of VEGF activity led to fewer primarytrabeculae, and thick and shortened secondary trabeculae,suggesting defects in bone formation. This is consistent withprevious studies showing that VEGF regulates osteoblastdifferentiation and function (Hiltunen et al., 2003; Maes et al., 2002;Midy and Plouet, 1994; Street et al., 2002; Zelzer et al., 2002).However, inhibition of all MMPs resulted in both the absence ofprimary trabeculae and a thick mass of secondary trabeculae thatseemed to have fused together; this suggests defects in both boneformation and resorption. MMP activity may be necessary for boneformation by having a direct effect on osteoblast function orthrough its ECM-degrading activity that may release factors suchas VEGF or transforming growth factor (TGF)-b, which in turn act

Fig. 9. Inhibition of MMP activity inhibitsresidual endochondral ossification at theMMP-9–/– growth plate. (A-D) H&E-stainedsections of metatarsals of 1-week-old MMP-9–/– mice treated for 1 week (A,B) or 2weeks (C,D) with either vehicle or the MMPinhibitor GM6001, showing furtherlengthening of the HC (hc) zone in theGM6001-treated mice (B,D) compared withthe vehicle-treated mice (A,C). (E) Bar graphsof the mean and standard deviation of theHC zone lengths of six control and GM6001-treated bones after 1 and 2 weeks oftreatment; the differences between controland treated samples at each time point werestatistically significant (P<0.01). (F,G)Immunostaining of tissue sections withPECAM-1 antibody showing metaphysealvessels running parallel to hypertrophicchondrocyte columns in the vehicle-treatedmice (F, arrows) and dilated vessels runningperpendicular to the hypertrophicchondrocyte columns in the GM6001-treated mice (G, arrows). (H,I) H&E-stainedsections showing primary (H, arrowheads)and secondary (H, arrows) trabeculae in thevehicle-treated mice, and a mass of boneinstead of secondary trabeculae in theGM6001-treated mice (I, arrows). (J,K) TRAPstaining of tissue sections showing thepresence of TRAP+ cells at the cartilage-bone junction (arrows) and on trabecularbone surfaces (arrowheads) in both vehicle-treated (J) and GM6001-treated (K) mice.Bars, 400 mm (A-D); 100 mm (F-K).

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



on osteoclasts and osteoblasts. Significant numbers of osteoclastsare present on the bone surfaces of GM6001-treated mice,suggesting that the defect in bone resorption is the result ofinhibition of secreted MMP activity and not because of a lack ofosteoclasts.

METHODSAnimal studiesMMP-9–/– mice have been described previously (Vu et al., 1998).The treatment of mice was by daily intraperitoneal injection. Thefollowing doses were used: recombinant human VEGF165(Genentech, South San Francisco, CA) in PBS at 10 mg/mouse;mFlt(1-3)-IgG (Genentech) or control IgG in PBS at 25 mg/g;GM6001, which was made as a slurry in 4% carboxymethyl cellulose,at 100 mg/g. The GM6001 was a gift from Dr Richard Galardy. Weare grateful to Dr Adrian Erlebacher for providing samples fromc-Fos–/– mice and their WT littermates (Erlebacher et al., 1998);Dr S. R. McKercher for providing PU.1 –/– mice; and Dr JacquesPeschon for providing Rank–/– mice (Dougall et al., 1999;McKercher et al., 1996). The Csf1op/op mice were purchased fromJackson Labs.

Histological analysis, TRAP staining and immunohistochemistryAfter euthanasia, the metatarsals were dissected free from the skinand fixed in 4% paraformaldehyde at 4°C overnight. The tissueswere then washed in PBS and decalcified in 0.5 M EDTA (pH 7.4)for 7-10 days at 4°C prior to processing and embedding in paraffin.Sections (5 mm) were stained with hematoxylin and eosin, or withhematoxylin/Fast Green/Safranin-O. For quantification, the HClength was measured from control and treated samples (n=6). Dataare presented as mean ± standard deviation (S.D.). Statisticalanalyses were performed using the Student’s t-test.

For tartrate-resistant acidic phosphatase (TRAP) staining,sections were deparaffinized and stained for TRAP activity using

a leukocyte acid phosphatase kit with Fast Red Violet as a substrate(Sigma, St Louis, MO). After incubation with the TRAP solutionat 37°C for 1 hour, sections were washed in distilled water andcounterstained with either hematoxylin or Methyl Green (Sigma).Mononucleated and multinucleated TRAP-positive cells werecounted under the microscope at a 40� magnification andexpressed as the total number of TRAP-positive cells along thechondro-osseous junction. The counting was performed on aminimum of three mice per group, using both hind limbs, and usingthree to six different sections, with three metatarsals per section,for each hind limb. The results are expressed as the mean ± theS.E.M. Statistical analyses were carried out using the Student’s t-test. GraphPad calculator was used to perform the tests.

For immunohistochemistry, the sections were deparaffinized,washed in PBS, quenched in either PBS and 3% H2O2 for 5 minutesor in 0.3% H2O2 in methanol for 30 minutes, then washed in PBS,and treated with 0.1% trypsin in PBS for 30 minutes. For MMP-9immunohistochemistry, the sections were treated with ficin (ZymedLaboratories, Inc., South San Francisco, CA) for 15 minutes atambient temperature. The slides were then treated with 0.1 Mglycine in PBS for 30 minutes, washed in PBS, and blocked in PBSwith 3% BSA for 30 minutes, then washed in PBS, and blocked inPBS with 5% normal serum for 2 hours. The sections wereincubated with primary antibody diluted in PBS with 1 mg/ml ofBSA at 4°C overnight. The slides were then washed in PBS, blockedin PBS with 5% normal serum for 30 minutes, then incubated withbiotinylated secondary antibody for 1 hour, washed in PBS, andincubated with Vector Elite ABC reagent (Vector Laboratory,Burlingame, CA) for 1 hour, washed in PBS, and developed usingDAB substrate. Rat monoclonal anti-mouse CD31 antibody (cloneMEC 13.3, Pharmingen, San Diego, CA) was used at a dilution of1:100; the rabbit anti-cleaved collagen II epitope antibody (rabbitanti-Col 3/4 polyclonal antibody, Hdm Diagnostics and Imaging,Toronto, Canada) was used at 1:800; the MMP-9 rabbit polyclonalantibody was used at 1:100 (Vu et al., 1998); and the rabbit anti-human VEGF antibody (Ab1, Neomarkers, Fremont, CA) was usedat 1:100 on frozen sections with tyramide amplification (NEN,Boston, MA) and True Blue HRP substrate (KPL, Gaithersburg,MD).

Labeling of proliferative cells in vivo with BrdUMice were injected intraperitoneally with 10 mg/ml of BrdU(Sigma), at doses of 0.01 ml/g, 1 hour prior to sacrifice. Bones werefixed and processed for paraffin sections as above. Sections wereimmunostained with an antibody against BrdU using a kit fromZymed.

In situ hybridizationSections were deparaffinized, treated with proteinase K (20 mg/ml)for 5 minutes at ambient temperature, and hybridized with 35S-labeled antisense riboprobes in hybridization buffer (50% deionizedformamide, 300 mM NaCl, 20 mM Tris-HCl pH 8.0, 5 mM EDTA,0.5 mg/ml yeast tRNA, 10% dextran sulfate and 1�Denhardt’ssolution) in a humidified chamber at 55°C overnight. Followinghybridization, the slides were treated with RNase A, washed to afinal stringency of 50% formamide, 2�SSC at 60°C, then dippedin emulsion, exposed for 1-2 weeks, developed, and counterstainedwith hematoxylin and eosin.

Fig. 10. Complementary functions of VEGF, MMP-9 and osteoclasts at thegrowth plate. At the HC-bone junction, MMP-9 is secreted by mononucleatedand multinucleated osteoclasts, as well as by other non-osteoclastic cells.MMP-9 can degrade the ECM leading to the release of VEGF sequestered in theHC. However, MMP-9 also has functions besides regulating VEGFbioavailability, such as ECM degradation and remodeling. VEGF can act onendothelial cells, as well as on mononucleated osteoclasts and osteoblasts.Besides MMP-9, osteoclasts also secrete other MMPs that can also cleave theECM.

Dise

ase

Mod

els &

Mec

hani

sms

D

MM



VEGF western blotting and ELISAGrowth plate HC was dissected, with or without perichondrium,from 14-day-old WT and MMP-9–/– mice and homogenized inRIPA buffer. The protein was quantified using a micro BCA assay(Pierce, Rockford, IL) and 40 mg of total protein of each sample wasseparated on a 10% Bis-Tris gel in MOPS running buffer (NovexInvitrogen, Carlsbad, CA) under non-reducing conditions,transferred onto a PVDF membrane, and blotted with an anti-mouseVEGF antibody recognizing all isoforms of VEGF (R&D Systems,Minneapolis, MN). For the measurement of VEGF by ELISA, growthplates from 14-day-old WT and MMP-9–/– mice were dissected,and the ossification fronts with the last rows of hypertrophicchondrocytes were isolated and incubated in DMEM/F12 under 5%CO2 for 48 hours. Cultures were kept in 24-well plates, with 20 piecesof cartilage per well in 500 ml of culture medium. After 48 hours, theconditioned media were centrifuged and the VEGF content wasquantified by using a murine VEGF ELISA kit (R&D Systems),according to the manufacturer’s instructions.

ACKNOWLEDGEMENTSWe thank Helen Capili for her outstanding efforts in sectioning of the bonespecimens, Andrew Tauscher for assistance with electron microscopy, and DaniBehonick for helping with the Rank–/– mouse experiments and staining of sections.This study was supported by grants from the National Institutes of Health (AR046238to T.H.V. and Z.W.) and from the UCSF Program for Breakthrough BiomedicalResearch (to T.H.V). N.O. was a Michael Geisman Research Fellow of the OsteogenesisImperfecta Foundation. Deposited in PMC for release after 12 months.

COMPETING INTERESTSThe authors declare no competing financial interests.

AUTHOR CONTRIBUTIONSN.O. conceived, designed and performed experiments; analyzed data; andprepared the manuscript. K.W. performed experiments and reviewed themanuscript. N.F. contributed reagents and reviewed the manuscript. Z.W.conceived and designed experiments and edited the manuscript. T.H.V. conceived,designed and performed experiments; analyzed data; and prepared and editedthe manuscript.

Received 24 July 2009; Accepted 22 September 2009.

REFERENCESAldridge, S. E., Lennard, T. W., Williams, J. R. and Birch, M. A. (2005). Vascular

endothelial growth factor receptors in osteoclast differentiation and function.Biochem. Biophys. Res. Commun. 335, 793-798.

Arikawa-Hirasawa, E., Watanabe, H., Takami, H., Hassell, J. R. and Yamada, Y.(1999). Perlecan is essential for cartilage and cephalic development. Nat. Genet. 23,354-358.

Bergers, G., Brekken, R., McMahon, G., Vu, T. H., Itoh, T., Tamaki, K., Tanzawa, K.,Thorpe, P., Itohara, S., Werb, Z. et al. (2000). Matrix metalloproteinase-9 triggersthe angiogenic switch during carcinogenesis. Nat. Cell Biol. 2, 737-744.

Billinghurst, R. C., Dahlberg, L., Ionescu, M., Reiner, A., Bourne, R., Rorabeck, C.,Mitchell, P., Hambor, J., Diekmann, O., Tschesche, H. et al. (1997). Enhancedcleavage of type II collagen by collagenases in osteoarthritic articular cartilage. J.Clin. Invest 99, 1534-1545.

Bischof, P., Truong, K. and Campana, A. (2003). Regulation of trophoblasticgelatinases by proto-oncogenes. Placenta 24, 155-163.

Cohen, M. M., Jr (2000a). Merging the old skeletal biology with the new. I.Intramembranous ossification, endochondral ossification, ectopic bone, secondarycartilage, and pathologic considerations. J. Craniofac. Genet. Dev. Biol. 20, 84-93.

Cohen, M. M., Jr (2000b). Merging the old skeletal biology with the new. II. Molecularaspects of bone formation and bone growth. J. Craniofac. Genet. Dev. Biol. 20, 94-106.

Cohen, M. M., Jr (2006). The new bone biology: pathologic, molecular, and clinicalcorrelates. Am. J. Med. Genet. A 140, 2646-2706.

Colnot, C. (2005). Cellular and molecular interactions regulating skeletogenesis. J. CellBiochem. 95, 688-697.

Daci, E., Everts, V., Torrekens, S., Van Herck, E., Tigchelaar-Gutterr, W., Bouillon, R.and Carmeliet, G. (2003). Increased bone formation in mice lacking plasminogenactivators. J. Bone Miner. Res. 18, 1167-1176.

Dougall, W. C., Glaccum, M., Charrier, K., Rohrbach, K., Brasel, K., De Smedt, T.,Daro, E., Smith, J., Tometsko, M. E., Maliszewski, C. R. et al. (1999). RANK isessential for osteoclast and lymph node development. Genes Dev. 13, 2412-2424.

Du, R., Petritsch, C., Lu, K., Liu, P., Haller, A., Ganss, R., Song, H., Vandenberg, S.and Bergers, G. (2008). Matrix metalloproteinase-2 regulates vascular patterningand growth affecting tumor cell survival and invasion in GBM. Neuro Oncol. 10, 254-264.

Egeblad, M., Shen, H. C., Behonick, D. J., Wilmes, L., Eichten, A., Korets, L. V.,Kheradmand, F., Werb, Z. and Coussens, L. M. (2007). Type I collagen is a geneticmodifier of matrix metalloproteinase 2 in murine skeletal development. Dev. Dyn.236, 1683-1693.

Engsig, M. T., Chen, Q. J., Vu, T. H., Pedersen, A. C., Therkidsen, B., Lund, L. R.,Henriksen, K., Lenhard, T., Foged, N. T., Werb, Z. et al. (2000). Matrixmetalloproteinase 9 and vascular endothelial growth factor are essential forosteoclast recruitment into developing long bones. J. Cell Biol. 151, 879-890.

Erlebacher, A., Filvaroff, E. H., Ye, J. Q. and Derynck, R. (1998). Osteoblasticresponses to TGF-beta during bone remodeling. Mol. Biol. Cell 9, 1903-1918.

Gerber, H. P., Vu, T. H., Ryan, A. M., Kowalski, J., Werb, Z. and Ferrara, N. (1999).VEGF couples hypertrophic cartilage remodeling, ossification and angiogenesisduring endochondral bone formation. Nat. Med. 5, 623-628.

Grigoriadis, A. E., Wang, Z. Q., Cecchini, M. G., Hofstetter, W., Felix, R., Fleisch, H.A. and Wagner, E. F. (1994). c-Fos: a key regulator of osteoclast-macrophage lineagedetermination and bone remodeling. Science 266, 443-448.

Gustafsson, E., Aszodi, A., Ortega, N., Hunziker, E. B., Denker, H. W., Werb, Z. andFassler, R. (2003). Role of collagen type II and perlecan in skeletal development.Ann. N. Y. Acad. Sci. 995, 140-150.

TRANSLATIONAL IMPACT

Clinical issueBone formation, remodeling and repair are important for skeletal function.Bone remodeling relies on maintaining a balance between bone formationand resorption. The dynamic bone structure is regulated by many bone tissuecomponents, including the vasculature, the cells that deposite bone(osteoblasts), and the cells that resorb bone (osteoclasts). Bone repairrecapitulates many of the pathways involved in normal bone formation andremodeling.

Disruption of the careful balance between bone formation and depositionleads to disease. Abnormal bone formation can result in malformed bones orectopic bone formation. Too much bone formation causes osteopetrosis andtoo much bone resorption results in osteoporosis. Understanding the factorsthat regulate these processes would shed light on disorders of bonedevelopment, maintenance and repair, and may provide therapeutic targets.

ResultsMatrix metalloproteinases (MMPs), which degrade extracellular matrix (ECM)proteins, and a cytokine that induces blood vessel growth, vascular endothelialgrowth factor (VEGF), work together with osteoclasts in bone formation,remodeling and repair. However, the integration of these different factorsduring endochondral ossification, a main process in bone growth, is notknown. In this study, the authors analyze mice that lack MMP-9 and that alsocontain osteopetrotic mutations in genes that affect osteoclast development.MMP-9 targets the cartilage ECM and reduces VEGF bioavailability.Homeostasis of the osteoclast population at the growth plate depends on thelevel of both MMP-9 and VEGF. Osteoclasts are key regulators of growth platedynamics because they express MMP-9 and other factors that are necessary forcartilage and bone resorption. These results elucidate some of the functionalrelationships among the components that regulate vascularization duringendochondral ossification.

Implications and future directionsThis study shows that cartilage vascularization and endochondral ossificationdepend on the integrated functions of MMP-9 and VEGF, together withosteoclasts, to regulate cartilage ECM remodeling and vascular growth. It ispossible that disorders of bone formation involve dysregulation of theseprocesses. This knowledge can also help build models of fracture repair, whichutilize endochondral ossification to form new bone, to determine whethermodulation of these factors promotes new bone formation and facilitatesfracture repair with good functional restoration.

doi:10.1242/dmm.005116Dise

ase

Mod

els &

Mec

hani

sms

D

MM



Hackel, C., Radig, K., Rose, I. and Roessner, A. (1995). The urokinase plasminogenactivator (u-PA) and its inhibitor (PAI-1) in embryo-fetal bone formation in thehuman: an immunohistochemical study. Anat. Embryol. (Berl.) 192, 363-368.

Haigh, J. J., Gerber, H. P., Ferrara, N. and Wagner, E. F. (2000). Conditionalinactivation of VEGF-A in areas of collagen2a1 expression results in embryoniclethality in the heterozygous state. Development 127, 1445-1453.

Heissig, B., Hattori, K., Dias, S., Friedrich, M., Ferris, B., Hackett, N. R., Crystal, R. G.,Besmer, P., Lyden, D. and Moore, M. A. (2002). Recruitment of stem and progenitorcells from the bone marrow niche requires MMP-9 mediated release of kit-ligand.Cell 109, 625-637.

Hiltunen, M. O., Ruuskanen, M., Huuskonen, J., Mahonen, A. J., Ahonen, M.,Rutanen, J., Kosma, V. M., Mahonen, A., Kroger, H. and Yla-Herttuala, S. (2003).Adenovirus-mediated VEGF-A gene transfer induces bone formation in vivo. FASEB J.17, 1147-1149.

Holmbeck, K., Bianco, P., Caterina, J., Yamada, S., Kromer, M., Kuznetsov, S. A.,Mankani, M., Robey, P. G., Poole, A. R., Pidoux, I. et al. (1999). MT1-MMP-deficientmice develop dwarfism, osteopenia, arthritis, and connective tissue disease due toinadequate collagen turnover. Cell 99, 81-92.

Holmbeck, K., Bianco, P., Chrysovergis, K., Yamada, S. and Birkedal-Hansen, H.(2003). MT1-MMP-dependent, apoptotic remodeling of unmineralized cartilage: acritical process in skeletal growth. J. Cell Biol. 163, 661-671.

Inada, M., Wang, Y., Byrne, M. H., Rahman, M. U., Miyaura, C., Lopez-Otin, C. andKrane, S. M. (2004). Critical roles for collagenase-3 (Mmp13) in development ofgrowth plate cartilage and in endochondral ossification. Proc. Natl. Acad. Sci. USA101, 17192-17197.

Inoue, K., Mikuni-Takagaki, Y., Oikawa, K., Itoh, T., Inada, M., Noguchi, T., Park, J.S., Onodera, T., Krane, S. M., Noda, M. et al. (2006). A crucial role for matrixmetalloproteinase 2 in osteocytic canalicular formation and bone metabolism. J. Biol.Chem. 281, 33814-33824.

Karsenty, G. and Wagner, E. F. (2002). Reaching a genetic and molecularunderstanding of skeletal development. Dev. Cell 2, 389-406.

Li, J., Sarosi, I., Yan, X. Q., Morony, S., Capparelli, C., Tan, H. L., McCabe, S., Elliott,R., Scully, S., Van, G. et al. (2000). RANK is the intrinsic hematopoietic cell surfacereceptor that controls osteoclastogenesis and regulation of bone mass and calciummetabolism. Proc. Natl. Acad. Sci. USA 97, 1566-1571.

Li, S. W., Prockop, D. J., Helminen, H., Fassler, R., Lapvetelainen, T., Kiraly, K.,Peltarri, A., Arokoski, J., Lui, H., Arita, M. et al. (1995). Transgenic mice withtargeted inactivation of the Col2 alpha 1 gene for collagen II develop a skeleton withmembranous and periosteal bone but no endochondral bone. Genes Dev. 9, 2821-2830.

Maes, C., Carmeliet, P., Moermans, K., Stockmans, I., Smets, N., Collen, D.,Bouillon, R. and Carmeliet, G. (2002). Impaired angiogenesis and endochondralbone formation in mice lacking the vascular endothelial growth factor isoformsVEGF164 and VEGF188. Mech. Dev. 111, 61-73.

Maes, C., Stockmans, I., Moermans, K., Van Looveren, R., Smets, N., Carmeliet, P.,Bouillon, R. and Carmeliet, G. (2004). Soluble VEGF isoforms are essential forestablishing epiphyseal vascularization and regulating chondrocyte developmentand survival. J. Clin. Invest. 113, 188-199.

Marks, S. C., Jr and Lane, P. W. (1976). Osteopetrosis, a new recessive skeletalmutation on chromosome 12 of the mouse. J. Hered. 67, 11-18.

Martignetti, J. A., Aqeel, A. A., Sewairi, W. A., Boumah, C. E., Kambouris, M.,Mayouf, S. A., Sheth, K. V., Eid, W. A., Dowling, O., Harris, J. et al. (2001). Mutationof the matrix metalloproteinase 2 gene (MMP2) causes a multicentric osteolysis andarthritis syndrome. Nat. Genet. 28, 261-265.

McKercher, S. R., Torbett, B. E., Anderson, K. L., Henkel, G. W., Vestal, D. J.,Baribault, H., Klemsz, M., Feeney, A. J., Wu, G. E., Paige, C. J. et al. (1996).Targeted disruption of the PU.1 gene results in multiple hematopoieticabnormalities. EMBO J. 15, 5647-5658.

Midy, V. and Plouet, J. (1994). Vasculotropin/vascular endothelial growth factorinduces differentiation in cultured osteoblasts. Biochem. Biophys. Res. Commun. 199,380-386.

Mosig, R. A., Dowling, O., DiFeo, A., Ramirez, M. C., Parker, I. C., Abe, E., Diouri, J.,Aqeel, A. A., Wylie, J. D., Oblander, S. A. et al. (2007). Loss of MMP-2 disruptsskeletal and craniofacial development and results in decreased bone mineralization,joint erosion and defects in osteoblast and osteoclast growth. Hum. Mol. Genet. 16,1113-1123.

Niida, S., Kondo, T., Hiratsuka, S., Hayashi, S., Amizuka, N., Noda, T., Ikeda, K. andShibuya, M. (2005). VEGF receptor 1 signaling is essential for osteoclastdevelopment and bone marrow formation in colony-stimulating factor 1-deficientmice. Proc. Natl. Acad. Sci. USA 102, 14016-14021.

Olsen, B. R., Reginato, A. M. and Wang, W. (2000). Bone development. Annu. Rev. CellDev. Biol. 16, 191-220.

Ortega, N., Behonick, D. J. and Werb, Z. (2004). Matrix remodeling duringendochondral ossification. Trends Cell Biol. 14, 86-93.

Ortega, N., Behonick, D. J., Colnot, C., Cooper, D. N. and Werb, Z. (2005). Galectin-3is a downstream regulator of matrix metalloproteinase-9 function duringendochondral bone formation. Mol. Biol. Cell 16, 3028-3039.

Provot, S. and Schipani, E. (2005). Molecular mechanisms of endochondral bonedevelopment. Biochem. Biophys. Res. Commun. 328, 658-665.

Schipani, E., Ryan, H. E., Didrickson, S., Kobayashi, T., Knight, M. and Johnson, R.S. (2001). Hypoxia in cartilage: HIF-1alpha is essential for chondrocyte growth arrestand survival. Genes Dev. 15, 2865-2876.

Stickens, D., Behonick, D. J., Ortega, N., Heyer, B., Hartenstein, B., Yu, Y., Fosang,A. J., Schorpp-Kistner, M., Angel, P. and Werb, Z. (2004). Altered endochondralbone development in matrix metalloproteinase 13-deficient mice. Development 131,5883-5895.

Street, J., Bao, M., deGuzman, L., Bunting, S., Peale, F. V., Jr, Ferrara, N., Steinmetz,H., Hoeffel, J., Cleland, J. L., Daugherty, A. et al. (2002). Vascular endothelialgrowth factor stimulates bone repair by promoting angiogenesis and bone turnover.Proc. Natl. Acad. Sci. USA 99, 9656-9661.

Tondravi, M. M., McKercher, S. R., Anderson, K., Erdmann, J. M., Quiroz, M., Maki,R. and Teitelbaum, S. L. (1997). Osteopetrosis in mice lacking haematopoietictranscription factor PU.1. Nature 386, 81-84.

Tuysuz, B., Mosig, R., Altun, G., Sancak, S., Glucksman, M. J. and Martignetti, J. A.(2009). A novel matrix metalloproteinase 2 (MMP2) terminal hemopexin domainmutation in a family with multicentric osteolysis with nodulosis and arthritis withcardiac defects. Eur. J. Hum. Genet. 17, 565-572.

Vu, T. H., Shipley, J. M., Bergers, G., Berger, J. E., Helms, J. A., Hanahan, D., Shapiro,S. D., Senior, R. M. and Werb, Z. (1998). MMP-9/gelatinase B is a key regulator ofgrowth plate angiogenesis and apoptosis of hypertrophic chondrocytes. Cell 93,411-422.

Yoshida, H., Hayashi, S., Kunisada, T., Ogawa, M., Nishikawa, S., Okamura, H., Sudo,T. and Shultz, L. D. (1990). The murine mutation osteopetrosis is in the coding regionof the macrophage colony stimulating factor gene. Nature 345, 442-444.

Zelzer, E. and Olsen, B. R. (2005). Multiple roles of vascular endothelial growth factor(VEGF) in skeletal development, growth, and repair. Curr. Top Dev. Biol. 65, 169-187.

Zelzer, E., McLean, W., Ng, Y. S., Fukai, N., Reginato, A. M., Lovejoy, S., D’Amore, P.A. and Olsen, B. R. (2002). Skeletal defects in VEGF(120/120) mice reveal multipleroles for VEGF in skeletogenesis. Development 129, 1893-1904.

Zelzer, E., Mamluk, R., Ferrara, N., Johnson, R. S., Schipani, E. and Olsen, B. R.(2004). VEGFA is necessary for chondrocyte survival during bone development.Development 131, 2161-2171.

Zhou, Z., Apte, S. S., Soininen, R., Cao, R., Baaklini, G. Y., Rauser, R. W., Wang, J.,Cao, Y. and Tryggvason, K. (2000). Impaired endochondral ossification andangiogenesis in mice deficient in membrane-type matrix metalloproteinase I. Proc.Natl. Acad. Sci. USA 97, 4052-4057.

Dise

ase

Mod

els &

Mec

hani

sms

D

MM

Documents

Complementary interplay between matrix metalloproteinase-9 ... · endochondral ossification requires restricted VEGF bioavailability that might be regulated by MMPs through ECM remodeling