ORIGINAL ARTICLE 537J o u r n a l o fJ o u r n a l o f

CellularPhysiologyCellularPhysiology

Evaluation of VEGF-Mediated

Signaling in Primary Human CellsReveals a Paracrine Action forVEGF in Osteoblast-MediatedCrosstalk to Endothelial Cells

CLAIRE E. CLARKIN,1* ROGER J. EMERY,2

ANDREW A. PITSILLIDES,1 AND CAROLINE P.D. WHEELER-JONES1

1Department of Veterinary Basic Sciences, Royal Veterinary College, University of London, London, UK2Division of Surgery, Oncology, Reproductive Biology and Anaesthetics, Imperial College London, Faculty of Medicine,

St. Mary’s Hospital, London, UK

Communication between endothelial and bone cells is crucial for controlling vascular supply during bone growth, remodeling, andrepair but themolecularmechanisms coordinating this intercellular crosstalk remain ill-defined.Wehave used primary human and rat longbone-derived osteoblast-like cells (HOB and LOB) and human umbilical vein endothelial cells (HUVEC) to interrogate the potentialautocrine/paracrine role of vascular endothelial cell growth factor (VEGF) in osteoblast:endothelial cell (OB:EC) communication andexamined whether prostaglandins (PG), known modulators of both OB and EC behavior, modify VEGF production. We found that thestable metabolite of PGI2, 6-keto-PGF1a and PGE2, induced a concentration-dependent increase in VEGF release by HOBs but not ECs. InECs, VEGF promoted early ERK1/2 activation, late cyclooxygenase-2 (COX-2) protein induction, and release of 6-keto-PGF1a. In markedcontrast, no significant modulation of these events was observed in HOBs exposed to VEGF, but LOBs clearly exhibited COX-dependentprostanoid release (10-fold less than EC) following VEGF treatment. A low level of osteoblast-like cell responsiveness to exogenous VEGFwas supported by VEGFR2/Flk-1 immunolabelling and by blockade of VEGF-mediated prostanoid generation by a VEGFR tyrosine kinaseinhibitor (TKI). HOB alkaline phosphatase (ALP) activity was increased following long-term non-contact co-culture with ECs andexposure of ECs to VEGF in this system further increased OB-like cell differentiation and markedly enhanced prostanoid release. Ourstudies confirm a paracrine EC-mediated effect of VEGF onOB-like cell behavior and are the first supporting amodel in which prostanoidsmay facilitate this unidirectional VEGF-driven OB:EC communication. These findings may offer novel regimes for modulating pathologicalbone remodeling anomalies through the control of the closely coupled vascular supply.J. Cell. Physiol. 214: 537–544, 2008. � 2007 Wiley-Liss, Inc.

AndrewA. Pitsillides and Caroline P.D.Wheeler-Jones contributedequally to this work.

Contract grant sponsor: Biotechnology and Biological SciencesResearch Council.Contract grant sponsor: The Wellcome Trust.Contract grant sponsor: Arthritis Research Campaign.Contract grant sponsor: British Heart Foundation.

*Correspondence to: Claire E. Clarkin, Department of VeterinaryBasic Sciences, Royal Veterinary College, University of London,Royal College Street, London NW1 0TU, UK.E-mail: cclarkin@rvc.ac.uk

Received 22 May 2007; Accepted 28 June 2007

DOI: 10.1002/jcp.21234

There is an intimate physical proximity between blood vesselsand osteoblasts involved in bone formation during bothphysiological and pathological changes in bone turnover (Truetaand Little, 1960; Trueta, 1963; Deckers et al., 1995; Hauge et al.,2001). This relationship is highlighted by the pre-requisite forvascularization during both intramembranous andendochondral ossification (Shapiro and Boyde, 1987; Brownet al., 1990; Carringdon and Reddi, 1991; Bittner et al., 1998;Gerber et al., 1999). The importance of this relationship isevidenced by a reduction in the number of arterial capillaries inthe bone marrow in osteoporosis (Burkhardt et al., 1987),endorsing a central role of the vasculature in bone formationand remodeling. Intracortical remodeling is also dependentupon the vascular delivery of bone cell precursors as well astheir direct interaction with endothelial cells (EC) (see Brandiand Collin-Osdoby (2006)). However, despite reliance onappropriate vascularity, it remains undefined whether primarycontrol of bone remodeling is intrinsic to the osteoblast orendothelial cell population. To understand the basis of thisrelationship, it is essential to resolve both the sites ofproduction and the routes and factors involved inosteoblast:endothelial cell (OB:EC) intercellularcommunication.

A key candidate regulator of OB:EC communication isvascular endothelial cell growth factor (VEGF), a knownpro-angiogenic factor with well-established actions on ECs,which is required for bone development (Gerber et al., 1999;Zelzer et al., 2002), and which may also have direct effects on

� 2 0 0 7 W I L E Y - L I S S , I N C .

osteoblast function (Midy and Plouet, 1994). The response ofECs to exogenous VEGF is indeed a phenotypic characteristicdefining this cell type, and involves intracellular signaling thatimpacts upon vasoactive mediator release and EC proliferation(Gliki et al., 2001; Zeng et al., 2001). Whether VEGF exerts itsosteogenic action through targeting of both osteoblasts andECs or whether it preferentially targets one of these cell typesto regulate OB:EC communication remains unclear.

In addition to selective targeting, it is possible thatVEGF-mediated crosstalk depends upon the cellular site of

538 C L A R K I N E T A L .

production. Osteoblasts produce pro-angiogenic factors,including VEGF (Fong et al., 2003) and many osteotrophicfactors, such as insulin-like growth factor-1, PTH, and 1,25-dihydroxyvitamin D3 are known to promote osteoblast VEGFproduction (Wang et al., 1997; Esbrit et al., 2000; Akeno et al.,2002) suggesting enhanced bone formation via indirect controlof EC behavior. The importance of VEGF in longitudinal bonegrowth in vivo has also been emphasized (Gerber and Ferrara,2000). However, there is some evidence that ECs may alsoproduce VEGF under certain experimental conditions (Haideret al., 2005; Milkiewicz et al., 2007), raising the possibility thatVEGF may derive from both cell types. Thus, the cellular andmolecular mechanisms responsible for controlling VEGF-dependent OB:EC crosstalk are undefined. It is clear thatidentification of these mechanisms will inform strategies thatwill allow for the development of novel and innovativetherapeutic approaches for regulating bone formation andremodeling. Herein, we test the hypothesis that osteogenesis isregulated through directional targeting of EC by osteoblastVEGF.

In addition to growth factor control of cell behavior, OB:ECcommunication may also be mediated, by virtue of theirproximity in vivo (Hauge et al., 2001), by short-lived autocoids,such as prostaglandins (PG). Indeed, a range of PGs areproduced by both osteoblasts and ECs and have well-documented anabolic effects on bone in vivo (Machwate et al.,2001). Production of PGs, including PGE2, is controlled bycyclooxygenase (COX) activities. The constitutive isoform,COX-1, produces PGs for homeostasis, while the expression ofinducible COX-2 is enhanced during inflammation, tumor-associated angiogenesis, and load-induced osteogenesis(Hwang et al., 1998; Rawlinson et al., 2000; Li et al., 2002; Syedaet al., 2006). The importance of COX-2 activity is furtherhighlighted by the demonstrations that mice lacking COX-2exhibit defective osteogenesis following fracture (Zhang et al.,2002), and that osteoblasts from COX-2 knockout mice showreduced differentiation in vitro that is rescued by exogenousPGE2 (Okada et al., 2000). These results suggest that PGE2derived from COX-2 may act as a coupling factor forosteoblasts and ECs and serve a facilitatory role in OB:ECcrosstalk.

It is, however, an assumption that PGs enhance boneformation via direct effects on osteoblasts and similarly that theeffects of VEGF are achieved solely through direct actions onECs. Using primary human osteoblast-like cells (HOBs) andhuman umbilical vein endothelial cells (HUVEC) in vitro, wehave compared their VEGF-induced responses to determinescope for preferential VEGF targeting and have sought toidentify the predominant cell type responsible for VEGF and PGproduction. Our findings indicate that VEGF is releasedpredominantly by human osteoblasts and is promoted byexogenous PGs. We have also established that VEGF canpromote PG release from human osteoblasts via specificVEGF-receptor activation, but that its primary direct action isvia ECs. Exploitation of a non-contact co-culture model systemestablished that EC-derived soluble factors promote osteoblastdifferentiation and that this crosstalk is augmented byexogenous VEGF. These studies are the first to address thesequestions in primary human bone and ECs and together supportthe hypothesis that osteoblast-derived VEGF stimulates ECs tofacilitate osteogenesis.

Materials and MethodsMaterials

Human recombinant VEGF165 and rat recombinant VEGF164, IL-1aand EGF, were all from R&D Systems (Oxford, UK). Bovine serumalbumin (BSA; fraction V), and polyvinylidene difluoridemembranes (Hybond-PTM), PMA and PGE2 were purchased from

JOURNAL OF CELLULAR PHYSIOLOGY DOI 10.1002/JCP

Sigma (Poole, UK). Reagents for SDS–PAGE were purchased fromBio-Rad (Hemel Hempstead, UK) and National Diagnostics (Hull,UK). 6-keto-PGF1a was obtained from Cayman Chemical (MI).VEGFR tyrosine kinase inhibitor (TKI) (4-[(40-chloro-20-fluoro)phenylamino]-6,7-dimethyl-oxyquinazoline) was fromCalbiochem (Nottingham, UK). Culture media were purchasedfrom Sigma. COX-1, COX-2, and VEGFR2/Flk-1(N-931)antibodies were purchased from Santa Cruz Biotechnology, Inc.(Santa Cruz, CA). Polyclonal anti-phospho-ERK1/2 antibodieswere from New England Biolabs (Beverly, MA). Fluoresceinisothiocyanate-conjugated goat anti-rabbit secondary antibodieswere from Sigma. Horseradish peroxidase-conjugated goatanti-rabbit and rabbit anti-goat immunoglobulins were fromPierce (Rockford, IL). All other reagents were obtained from Sigmaor BDH (Poole, UK) at the equivalent of AnalaR grade.

Cell culture

HUVEC were isolated and cultured as previously described(Houliston et al., 2001, 2002). Experiments were performed onconfluent passage two HUVEC grown in gelatin-coated culturedishes. Primary HOBs were cultured from bone segments takenduring routine shoulder operations at The Hospital of St John andSt Elizabeth, London, by methods previously described (Zamanet al., 1999) and patient consent was gained before each operation.Briefly, bone samples were obtained from non-pathologicalexplants from the acromion, which is removed routinely duringrotator cuff repair procedures. After vigorous washing, boneexplants (�1 mm2) were placed into 75 mm2 tissue culture flaskswith DMEM supplemented with 10% FCS, 2 mM L-glutamine,100 U/ml penicillin, and 100 mg/ml streptomycin, incubated toallow outgrowth and media replenished twice weekly. After21 days, the fragments were removed and used to establish asecond set of cultures (P1). Rat long bone OB were derived byoutgrowth from cortical explants of femora from female SpragueDawley rats (90–110 g) and phenotype confirmed by measuringalkaline phosphatase (ALP) activity, osteocalcin expression,collagen type I mRNA expression, and potential for mineralization(Zaman et al., 1999). In some studies, primary rat OBdifferentiation was promoted by culture in differentiation medium,consisting of phenol red-free DMEM (plus 10% FCS, 100 U/mlpenicillin, 100 mg/ml streptomycin, and 2 mM L-Glutamine)supplemented with 10 nM dexamethasone, 50mg/ml ascorbic acid,and 2 mg/ml b-glycerophosphate for 18 days. Dense clusters ofpolygonal cells were obvious after 7 days of culture and formedmineralized, collagenous bone nodules 2–3 days later.

Western blotting

HUVEC and OB monolayers (1� 106 cells/dish) were serum- andgrowth factor-deprived for 16 h and subjected to treatments asdetailed. Whole cell lysates were prepared, and immunoblottinganalyses performed as previously described (Houliston et al.,2002). Blots initially probed with antibodies against eitherCOX-2 or phospho-ERK1/2 (1:1,000) were stripped in 62.5 mMTris-HCl (pH 6.7), 2% (w/v) SDS, and 0.7% (v/v) mercaptoethanolfor 30 min at 508C. After washing, blots were re-probed withCOX-1 or total ERK1/2 (1:1,000) antibodies and visualized byenhanced chemiluminescence. Where indicated, densitometricanalyses of immunoblots were performed using a Bio-Rad scanningdensitometer and Quantity One analyzing software.

Immunofluorescence

HUVEC or OBs were plated on 12-mm glass coverslips and grownfor 24–48 h until 80% confluent. After treatment, cells were fixed in4% paraformaldehyde (pH 7.4) and washed with PBS containing0.5% BSA (PBS-BSA). Coverslips were then incubated for 1 hwith anti-VEGFR2 in PBS-BSA at room temperature, washedtwice in PBS-BSA, and incubated with fluoresceinisothiocyanate-conjugated secondary goat anti-rabbit antibodies

Fig. 1. Prostanoids increase VEGF release from human OB-likecells. Quiescent human OBs were challenged with PGE2 (A) or6-keto-PGF1a (B) at the indicated concentrations for 24 h andsupernatants assayed for VEGF by ELISA. Data shown are fromOB-like cells extracted from explants from three donors, withduplicate observations per treatment. Results are given asmeanWSEM (MP<0.05, MMP<0.01 versus untreated control, (C)).

O S T E O B L A S T - E N D O T H E L I A L C R O S S T A L K M E C H A N I S M S 539

(1:100) for 45 min. Cells were washed (twice) and mounted ineither Fluorsave (Dako; Cambridgeshire, UK) or propidiumiodide-containingmountingmedium (Vectashield; Burlingame,CA)(Bastow et al., 2005). Immunofluorescence was monitored byconfocal microscopy using a Zeiss LSM 510 inverted microscope.

Measurement of PGI2, PGE2, and VEGF165 release

OBs or HUVEC in 24-well plates were treated as described in thelegends and supernatants assayed for 6-keto-PGF1a (stable PGI2hydrolysis product) by radioimmunoassay as previously described(Houliston et al., 2001). PGE2 was quantified using a Correlate-EIAProstaglandin E2 kit (Assay Designs Inc., Northampton, UK)according to the manufacturers’ instructions and VEGF wasmeasured by sandwich ELISA, as previously described (Pufe et al.,2003).

ALP activity of individual cells

OBs cultured in 12- or 24-well plates were washed with PBS priorto fixation in ice-cold 50% methanol, 50% acetone for 2 min, andwashed twice in PBS before reacting for ALP activity in mediumprepared by mixing 1 mg/ml Fast Blue RR in 50 mM Tris, pH 9.2 to10 mg/ml napthol-AS-MX phosphate in NN-dimethyl formamide(filtered). Fixed cells were then incubatedwith the staining solutionfor 15 min at 378C.

ALP activity in eluted preparations

OBs in 24-well plates were washed with PBS and fixed for 1 minin 100% ethanol. p-Nitrophenol substrate (1 mg/ml) was thenadded to 20% 0.1M NaHCO3 and 10% 30 mM MgCl2 at pH 9.5;500ml added to each well and incubated for 30 min at 378C.Eluent (2� 200 ml) was subsequently removed from each welland transferred into a 96-well plate. Absorbance (450 nm) wasmeasured after 1 min and concentrations expressed innmols/ml/minute.

Non-contact HUVEC and HOB co-culture

HUVEC and HOBs were maintained in non-contact co-cultureseparated by a polycarbonate filter insert (0.4 mm, Millipore;Bedford, MA). HOBs and HUVEC were seeded into the outer(24-well plates) and inner (Millcell-CM) chambers, respectively andboth sections maintained in media and conditions describedpreviously. Control comprised either HUVEC or HOBs in bothchambers.

Statistical analysis

Student’s t-test or ANOVA, as appropriate, were used to comparemeans of groups of data. P< 0.05 was considered statisticallysignificant.

ResultsProstaglandin treatment increases osteoblast but notendothelial cell VEGF release

We first examined the effects of exogenous prostanoids onVEGF release from primary human and rat OBs (OB-like)cells compared to HUVEC. VEGF was detected in mediumfrom untreated human OBs and levels increased,concentration-dependently, after treatment with either PGE2(Fig. 1A) or the stable hydrolysis product of PGI2, 6-keto-PGF1a(Fig. 1B). In contrast, VEGF was undetectable basally or afterprostanoid exposure in HUVEC-conditioned media (data notshown), suggesting that OBs predominate as a source ofextracellular VEGF in the bone microenvironment.

JOURNAL OF CELLULAR PHYSIOLOGY DOI 10.1002/JCP

VEGFR2/Flk-1 is detectable by immunofluorescencebut VEGF fails to activate ERK1/2 or induce COX-2in primary OB-like cells

Controversy exists regarding the ability of VEGF to directlymodify OB behavior. We therefore sought evidence ofVEGFR2/Flk-1expression in primary human and rat OBs usingimmunofluorescence, with expression in HUVEC as a positivecontrol. Immunofluorescent labeling for VEGFR2/Flk-1 wasdetected in both rat and human OBs and as expected HUVECalso exhibited clear VEGFR2/Flk-1 immunolabelling (Fig. 2). Todeterminewhether VEGF activates similar signaling pathways inprimary OBs and ECs, we have compared responses in rat andhuman OB-like cells to those known to be associated withVEGF165-induced angiogenic behavior in ECs, includingextracellular regulated kinase 1/2 (ERK1/2). Consistent withprevious studies (Takahashi et al., 1999; Gliki et al., 2001),treatment of HUVEC with VEGF165 induced a rapidphosphorylation of ERK1/2, with activation most marked atconcentrations exceeding 10 ng/ml (Fig. 3A, D). After 8 h,ERK1/2 activation had diminished to levels seen in untreatedHUVEC (data not shown). Preferential activation of ERK1/2 byVEGF165 was supported by only weak p38mapk and JNK

Fig. 2. VEGF receptor expression is detectable in primary OB-likecells. VEGFR2 expression inHUVECand in rat or humanOB-like cellswas detected by immunofluorescence. Human OB-like cells werecounterstained with propidium iodide and the inset to each partshows fluorescence in cultures without primary antibody exposure.Immunofluorescence was visualized using a Leica 300 fluorescentmicroscope equipped with a 100T objective. Data in each part arerepresentative of three experiments with similar results.

Fig. 3. VEGF robustly augments ERK1/2 activation and COX-2protein expression in HUVEC. Quiescent HUVEC were challengedwith VEGF165 (ng/ml) for either 10min or 8 h.Whole cell lysateswereanalyzed by SDS–PAGE and immunoblotted with anti-phosphoERK1/2 (A) or COX-2 antibodies (B). COX-2 blots were stripped andre-probed with COX-1 antibody (C). Densitometric analyses ofWestern blots demonstrated that treatment with 25 ng/ml VEGF165

for 10 min significantly increased pERK1/2 (D) and treatment for 8 hsignificantly increased COX-2 protein induction (E). Pooled resultsfrom three separate experiments performed on cells derived fromthree different cords are presented as fold increase versus control(intensity/mm2) meanWSEM (P<0.01MM, P<0.001MMM).

540 C L A R K I N E T A L .

activation in VEGF-stimulated HUVEC at the times andconcentrations studied (data not shown). In marked contrast,we found that ERK1/2 was not significantly activated inresponse to VEGF in either rat (Fig. 4A, D) or human primaryOB-like cells (Fig. 5A, D). Weak ERK1/2 activation wasobserved in three of nine primary human OB cell cultures, butphosphorylation was far less marked than in VEGF-treatedHUVEC and failed to reach levels of statistical significance. Asexpected, significant increases in ERK1/2 phosphorylationwereinduced by epidermal growth factor (EGF; 25 ng/ml) and byPGE2 (1 mM; not shown), verifying intact receptor-mediatedERK activation in OB-like cells under these experimentalconditions (Fig. 4A). These results suggest that VEGF does notsignificantly activate ERK1/2 in OB-like cells when compared toECs. Our current studies also confirm previous observations(Wheeler-Jones et al., 1997; Houliston et al., 2002) that VEGFincreases COX-2 in ECs without modifying COX-1 expression(Fig. 3B, C, E). In contrast, VEGF did not enhance COX-2expression (or COX-1) in either rat (Fig. 4B, C, E) or humanOB-like cells (Fig. 5B, C, E) despite clear COX-2 induction byIL-1a (Fig. 5B), EGF (Fig. 4B), and PGE2 (not shown). Together,these results indicate that VEGF promotes robust ERK1/2activation and COX-2 induction in ECs, but not in primaryOB-like cells.

Several reports in OB-derived cell lines support a directaction of VEGF on proliferation and differentiation of bone cells(Midy and Plouet, 1994; Deckers et al., 2000; Mayr-wohlfartet al., 2002; Zelzer et al., 2002). In addition, measurements ofBrdU incorporation and ALP activity showed that in contrast to

JOURNAL OF CELLULAR PHYSIOLOGY DOI 10.1002/JCP

OB-derived cell lines (Midy and Plouet, 1994; Deckers et al.,2000; Mayr-wohlfart et al., 2002; Zelzer et al., 2002), VEGF didnot influence proliferation or differentiation of either human orrat primary OB-like cells (data not shown). As VEGFR2/Flk-1expression has been reported to increase in differentiated celllines (Deckers et al., 2000), we treated primary OBs withdexamethasone and found that this forced differentiationneither increased VEGFR2/Flk-1 expression nor facilitatedCOX-2 expression or ERK1/2 activation by VEGF (data notshown).

VEGF induces 6-keto-PGF1a release fromrat OB-like cells

To determine functional roles for COX we measured6-keto-PGF1a (PGI2 breakdown product) and PGE2 release.Basal PGI2 production differed markedly, with HUVECexhibiting five-fold to 10-fold greater release than OB-like cellsunder non-stimulated conditions, while in contrast PGE2release was greater in primary rat OB (legend to Fig. 6).Significantly, VEGF treatment increased both PGI2 and PGE2release from rat OB-like cells, and elevated PGI2 synthesis byHUVEC (Fig. 6). This increased PG production by OBsoccurred despite the inability of VEGF to significantly modify

Fig. 4. EffectsofVEGFonERK1/2activationandCOX1/2expressionin rat OBs. Quiescent rat OB-like cells were challenged with VEGF164

(ng/ml) at the indicated concentrations or epidermal growth factor(EGF; 25 ng/ml) for either 10 min or 8 h. Whole cell lysates wereanalyzed by SDS–PAGE and immunoblotted with anti-phosphoERK1/2 (A) or COX-2 (B) antibodies. Blots probed for COX-2 werestripped and re-probed for COX-1 expression (C). Densitometricanalyses of Western blots demonstrated that treatment with25 ng/ml VEGF164 for 10 min had no effect on pERK1/2 (D) andtreatment for 8 h failed tomodifyCOX-2 protein expression (E).Dataare from four replicate experiments and are presented as foldincrease versus controls (intensity/mm2) meanWSEM (NS; notsignificant).

Fig. 5. VEGF does not significantly modulate ERK1/2phosphorylationorCOX-2protein expression inhumanOB-like cells.Quiescent human OB-like cells were challenged with VEGF165 (V;100 ng/ml), phorbol 12-myristate 13-acetate (PMA; 100 nM), orinterleukin-1a (IL-1a; 100 U/ml) for either 10 min or 8 h. Whole celllysates were analyzed by SDS–PAGE and immunoblotted withanti-phospho ERK1/2 (A) or anti COX-2 (B) antibodies. COX-2 blotswere stripped and re-probed for COX-1 (C). Densitometric analysesof Western blots demonstrated that treatment with 25 ng/mlVEGF164 for 10min or 8 h had no effect on pERK1/2 (D) or COX-2 (E)expression, respectively. Data are from OB-like cells extracted frombone explants derived from five different individual donors. Resultsare given as fold increase versus controls (intensity/mm2)meanWSEM (NS; not significant).

O S T E O B L A S T - E N D O T H E L I A L C R O S S T A L K M E C H A N I S M S 541

COX-2 expression or ERK1/2 activation (Figs. 4 and 5). Incontrast, primary human OB-like cells failed to showVEGF-induced PGI2 release (control 257� 120, VEGF217� 106 pg/ml 6-keto-PGF1a, mean� SEM). Thus, VEGFdoes not affect COX-2 expression or ERK1/2 activation, butpromotes modest increases in rat, but not human, OB-like cellPG release. This highlights species-specific differences in OBbehavior and emphasizes that in the context of human bone,predominant VEGF-induced increases in PGI2 synthesis likelyreflect a selective influence of VEGF on ECs. Moreover, ourstudies show that VEGF-induced PGI2 release from rat OB-likecells and HUVEC is abolished by a COX-2 selective inhibitor,NS-398 (% inhibition, mean� SEM, of VEGF-induced6-ketoPGF1a release from OBs¼ 99� 1, P< 0.001; and fromHUVEC¼ 99� 1, P< 0.001). These data suggest that COX-2activity contributes to VEGF-stimulated PGI2 synthesis.

VEGF may thus promote OB PGI2 release independently ofVEGFR2/Flk-1 activation. To clarify this possibility, weexamined whether a VEGFR TKI modified VEGF-induced6-keto-PGF1a release. Pre-exposure of rat OB-like cells to TKIabrogated basal and VEGF-induced 6-keto-PGF1a release(% inhibition of basal, mean� SEM¼ 61� 1, P< 0.05; andVEGF-stimulated¼ 63� 13; P< 0.001; n¼ 3). HUVECutilization of VEGFR2/Flk-1 phosphorylation was confirmed bythe concentration-dependent blockade of VEGF-inducedERK1/2 activation, COX-2 induction (data not shown), and6-keto-PGF1a release by TKI (% inhibition of VEGF-drivenrelease¼ 64� 5, P< 0.001, n¼ 3).

JOURNAL OF CELLULAR PHYSIOLOGY DOI 10.1002/JCP

VEGF enhances EC-mediated increases in OB-like celldifferentiation in co-culture

It remains to be established whether ECs have direct effects onOBs which, in turn, enhance the sensitivity to VEGF. Toinvestigate this possibilitymore directly, and to address the roleof VEGF inOB:EC crosstalk, we used non-contact co-culture ofhuman OB-like cells with HUVEC (homotypic OB:OB andEC:EC co-cultures were used as controls) and assessed ALPactivity. We found that co-culture with HUVEC increased OBdifferentiation to levels significantly greater than those inhomotypic co-culture (EC:EC co-cultures had negligible ALP;Fig. 7). To addresswhether VEGF exerts indirect actions onOBdifferentiation via ECs, we exposed HUVEC to VEGF andmeasured ALP activity in co-culturedOBs and found that VEGFtreatment of ECs potentiated OB-like cell ALP activity to asignificantly greater extent than did co-culture with ECs alone(in the absence of VEGF). We also observed that exogenousVEGF failed to modify OB-like cell differentiation directly in

Fig. 6. VEGF stimulates prostaglandin production by HUVEC andOB-like cells. Comparison of 6-keto-PGF1a (A) and (B) and PGE2

(C) and (D) release fromunstimulated (shaded) andVEGF (25ng/ml)-stimulated (unshaded) HUVEC (A) and (C) and rat OB-like cells(B) and (D) after 8 h treatment. Supernatants were assayed induplicate for 6-keto-PGF1a by radioimmunoassay and PGE2 byELISA. Data are from three separate experiments, with fourobservations per treatment and are given as mean fold increase(WSEM) versus control (P<0.01MM). Basal 6-keto-PGF1a productionby HUVEC and rat OBs ranged between 1,000–7,000 pg/ml and0–200 pg/ml, respectively and basal PGE2 production between100–200 pg/ml and 2,000–3,000 pg/ml, respectively.

Fig. 7. VEGF indirectly influences human OB-like celldifferentiation following non-contact co-culture with HUVEC.Homotypic non-contact co-cultures of ECs and OB-like cells wereestablished as described in the Materials and Methods. Cells on theinner chambers were exposed to vehicle alone (C) or to VEGF165

(V: 25 ng/ml). After 5 days of treatment, cells in the outer chamberswere fixed, ALP was eluted, and its activity measured at an opticaldensity of 450 nm against controls of known concentration (seeMaterials and Methods). Data are expressed as meanWSEM(triplicate determinations per treatment) from four separateexperimentsusingOB-like cells extracted from four individual donors(MP<0.05,MMM P<0.001; NS, not significant).

542 C L A R K I N E T A L .

homotypic OB:OB co-cultures (Fig. 7), confirming thisEC-dependency and suggesting that VEGF modified OBfunction indirectly through actions on ECs. In keeping with ourobservation that prostanoids modulate OB VEGF productionand may contribute toOB:EC crosstalk (Fig. 1), levels of 6-ketoPGF1a were at least twofold higher in both inner(EC-conditioned) and outer (OB-conditioned) wells ofco-cultures treated with VEGF than in similarly treatedhomotypic OB:OB co-cultures (data not shown).

Discussion

This study examined bone-forming osteoblasts and EC with afocus on identifying the cellular source of VEGF, its ability toactivate specific signaling events that influence cell behavior andthe role of osteogenic prostanoids in regulating VEGF release.Our studies revealed: (i) that osteoblasts release more VEGFthan EC and, unlike EC, show increased VEGF release inresponse to prostanoids, (ii) that the intracellular eventscharacteristic of endothelial cell VEGF signaling are largelyabsent in osteoblasts, despite their ability to respond effectivelyto other growth factors, (iii) that VEGF promotes aCOX-2-dependent release of prostanoids from rat osteoblasts,consistent with the presence of functional VEGF receptors,which is further supported by, (iv) the abrogation ofVEGF-mediated osteoblast prostanoid release by selectiveVEGF receptor blockade. Finally, we show, (v) that VEGF doesnot affect osteoblast differentiation directly, but does so only in

JOURNAL OF CELLULAR PHYSIOLOGY DOI 10.1002/JCP

OB:EC co-cultures. Our findings support the hypothesis (seeFig. 8) that OB-derived VEGF stimulates EC to facilitateosteogenesis and that this crosstalk may involve the release ofprostanoids. This is the first study to focus on thedirectionality of VEGF signaling in primary human bone/ECs andemphasizes the similarity of this to well-defined modes ofcommunication with the endothelium in a range of othertissues (Varney et al., 2005; Chang et al., 2006; Gonzalez et al.,2006; Muir et al., 2006; Schneider et al., 2006). These datahighlight the potential for a novel mechanism couplingosteogenesis and endothelial cell behavior during normaladult bone remodeling.

Studies in the microvasculature have established thatvascular smooth muscle cells are a major source of VEGF andthat this controls endothelial cell function in blood vessels(Chang et al., 2006; Elbjeirami and West, 2006). Ourdemonstration that primary osteoblasts, but not EC releasesignificant quantities of VEGF suggests that osteoblastsconstitute the principal VEGF supply in the bone osteogenicmicroenvironment. Our studies also show that PGE2 and PGI2promote significant increases in VEGF release fromosteoblasts,but not from EC. Our finding that 6-keto-PGF1a, the stablehydrolysis product of PGI2, is an effective stimulator of VEGFrelease from osteoblast-like cells is in keeping with therecognition that 6-keto PGF1a has biological actions (Blumeet al. 1998; Colgan et al. 2002). This may, in part, explain theosteogenic actions of PGs, and may also highlight a mechanismby which prostanoids produced in response to osteogenicstimuli, such as mechanical loading, regulate remodeling(Pitsillides et al., 1999; Rawlinson et al., 2000). In addition, ourstudy is unique in examining this cell:cell communication usingprimary human (and rat) bone-derived osteoblast-like cells andhuman endothelial cell cultures andmay, therefore, help explainseveral areas of controversy concerning the role of VEGF inOB:EC crosstalk.

Fig. 8. Potential mechanism of EC:OB communication.VEGF-stimulated HUVEC produce factors that influenceosteoblast differentiation. Osteoblasts release VEGF in response toprostaglandins, and endothelial cells respond to VEGF withphosphorylation of VEGFR2/Flk-1, activation of ERK1/2, COX-2induction, and release of prostanoids. This establishes a primarilyparacrine action of osteoblast-derived VEGF which may driveunidirectional crosstalk to endothelial cells in a process facilitatedby the release of prostanoids, such as prostacyclin, fromendothelial cells.

O S T E O B L A S T - E N D O T H E L I A L C R O S S T A L K M E C H A N I S M S 543

One area of dispute is whether VEGF has direct actions onOBs via VEGFR2/Flk1. Our results confirm that ERK1/2phosphorylation is a rapid and robust response to VEGF in EC(Ferrara, 1996; Abedi and Zachary, 1997; Wheeler-Jones et al.,1997; Xia et al., 1997; Takahashi et al., 1999; Gliki et al., 2001),but that VEGF does not affect ERK1/2 activation in either rat orhuman primary OBs. This lack of VEGF-stimulated ERK1/2activation in OB is consistent with their failure to proliferate inresponse to VEGF and, since ALP increases in parallel withdifferentiation and commits cells to the OB lineage (Stein et al.,1990), also coincides with the inability of VEGF to enhance OBdifferentiation in our study. Thus, it is unlikely that VEGFdirectly modulates OB function through intracellularmechanisms similar to those triggered in VEGF-stimulated EC.

The data described herein provide some evidence of alimited direct effect of VEGF on OBs. Thus, despite failure ofVEGF to modify COX-2 induction, VEGF-treated rat OBsexhibited increased COX-derived PG production, a finding inkeeping with a recent report that OBs produce PGD2 inresponse to exogenous VEGF (Gallant et al., 2005). Therestriction of VEGF-induced PG release to rodent OBs in ourstudy, however, highlights a need to apply caution whenextrapolating between data acquired from different species. Itnevertheless only strengthens the notion that VEGF does notdirectly target OBs in human bone.

In this respect, an obvious area of debate is whether OBsexpress functional VEGF receptors. Although VEGFR2/Flk1mRNA has been detected in several OB cell types, convincingevidence for the presence of VEGFR2/Flk1 protein in OBs islacking (Deckers et al., 2000). However, VEGFR2/Flk1 appearsto serve an autocrine role in the development (Zelzer et al.,2002) and retention of VEGFR2/Flk1 expression inOB-like cellsderived from mature bone may well reflect this developmentalrole. Our findings that VEGF-induced PG release by primaryOBs is inhibited by a VEGFR TKI, coupled withimmunochemical detection of VEGFR2/Flk1 in human and ratOB-like cells would support this contention. However, itshould be stressed that this TKImay not be completely VEGFR2selective, and it cannot be ruled out that VEGF-induced PG

JOURNAL OF CELLULAR PHYSIOLOGY DOI 10.1002/JCP

release fromOBs is mediated by VEGFR1 or through receptorsother than VEGFR2/Flk1. Indeed, VEGF-driven PGI2 synthesisby bovine ECs requires activation of VEGFR1/2 heterodimers(Neagoe et al., 2005) and VEGF-stimulated neuronal PGgeneration may depend, at least in part, on binding toneuropilin-1 (NP-1) (Cheng et al., 2004). The potential roles ofVEGFR1 and NP-1 as mediators of osteoblast PG synthesis byVEGF remain to be determined.

To further investigateOB:EC crosstalkmechanisms,weuseda non-contact co-culture model and demonstrated that humanECs produce soluble mediators that enhance human OBosteogenic behavior. This significantly extends previousfindings in non-human cells and transformed cell lines (Midyand Plouet, 1994; Deckers et al., 1995; Fons et al., 2004;Tombran-Tink and Barnstable, 2004). The possible role ofVEGF as an indirect regulator of OBs operating via anendothelial cell-derived osteogenic factor has received littleattention. Our studies show that VEGF can indeed intensifyEC-dependent osteogenic behavior, sinceOB:EC crosstalk waspromoted by exogenous VEGF. Failure of exogenous VEGF toaffect OB:OB co-cultures, or OBs cultured alone is strongevidence that ECs produce an, as yet, undefined diffusiblefactor(s) that modifies OB differentiation. Our data are inkeeping with those of Wang et al., which suggest that VEGFpromotes IGF-I and endothelin release from ECs to modify OBdifferentiation (Wang et al., 1996).

The intimacy of the physical relationship between OBs andECs (Hauge et al., 2001) suggests that an understanding of thefactors involved in sustaining their inter-communication shouldassist in developing novel therapeutic targets for bonedisorders characterized by a breakdown of this angiogenic/osteogenic coupling, such as Paget’s disease of bone andosteoporosis (Burkhardt et al., 1987).Our study shows that theosteoblast is the predominant source of VEGF and that PGsstrongly promote such VEGF release. We provide evidenceboth for an autocrine effect of VEGF on osteoblast PG release,mediated by VEGFR tyrosine kinase activity, but a primarydirect paracrine action via EC. Non-confrontational co-culturestudies also confirmed that EC-derived soluble factorspromote osteoblast differentiation and that this is furtherenhanced by exogenous VEGF. Our results have importantimplications for understanding the mechanisms that facilitateand control communication between bone and EC duringremodeling. Our current investigations are aimed at identifyingmediators of this OB:EC coupling.

Acknowledgments

Wewould like to thank Elaine Shervill for HUVEC isolation andculture and the staff and midwives at the Great Portland StreetHospital, London for the collection of umbilical cords. Thanksalso toMr.Dominic Simon for his assistance in collecting humanbone explants from the Hospital of St John and St Elizabeth,London. CEC was funded by a Biotechnology and BiologicalSciences ResearchCouncil studentship and by grants (toAP andCW-J) from the Wellcome Trust, Arthritis ResearchCampaign, and British Heart Foundation.

Literature Cited

Abedi H, Zachary I. 1997. Vascular endothelial growth factor stimulates tyrosinephosphorylation and recruitment to new focal adhesions of focal adhesion kinase andpaxillin in endothelial cells. J Biol Chem 272:15442–15451.

Akeno N, Robins J, Zhang M, Czyzyk-Krzeska MF, Clemens TL. 2002. Induction of vascularendothelial growth factor by IGF-I in osteoblast-like cells is mediated by the PI3K signalingpathway through the hypoxia-inducible factor-2{alpha}. Endocrinology 143:420–425.

Bastow ER, Lamb KJ, Lewthwaite JC, Osborne AC, Kavanagh E, Wheeler-Jones CPD,Pitsillides AA. 2005. Selective activation of the MEK-ERK pathway is regulated bymechanical stimuli in forming joints and promotes pericellular matrix formation. J BiolChem 280:11749–11758.

Bittner K, Vischer P, Bartholmes P, Bruckner P. 1998. Role of subchondral vascular system inendochondral ossification: Endothelial cells specifically depress late differentiation inresting chondrocytes in vitro. Exp Cell Res 238:491–497.

544 C L A R K I N E T A L .

Blume ED, Taylor CT, Lennon PF, Stahl GL, Colgan SP. 1998. Activated endothelial cells elicitparacrine induction of epithelial chloride secretion. 6-Keto-PGF1alpha is an epithelialsecretagogue. J Clin Invest 102:1161–1172.

Brandi ML, Collin-Osbody P. 2006. Vascular biology and the skeleton. J Bone Miner Res21:182–192.

BrownRA,Rees JA,McfarlandCD, LewinsonD,Ali SY. 1990.Microvascular invasionof rabbitgrowth plate cartilage and the influence of dexamethasone. J Bone Miner Res9:35–47.

Burkhardt R, Kettner G, Bohm W, Schmidmeier M, Shalag R, Frisch B, Mallmann B,Eisenmenger W, Gilg TH. 1987. Changes in trabecular bone hematopoiesis and bonemarrow vessels in aplastic anemia, primary osteoporosis and old age: A comparativehistomorphometric study. Bone 8:564–571.

Carringdon JL, Reddi AH. 1991. Parallels between development of embryonic and matrixinduced endochondral bone. Bioessays 13:403–406.

Chang HJ, Park JS, Kim MH, Hong MH, Kim KM, Kim SM, Shin BA, Ahn BW, Jung YD. 2006.Extracellular signal-regulated kinases and AP-1 mediate the up-regulation of vascularendothelial growth factor by PDGF in human vascular smooth muscle cells. Int J Oncol28:135–141.

Cheng L, Jia H, Lohr M, Bagherzadeh A, Holmes DIR, Selwood D, Zachary I. 2004.Anti-chemorepulsive effects of vascular endothelial growth factor and placental growthfactor-2 in dorsal root ganglion neurons aremediated by neuropilin-1 and cyclooxygenase-derived prostanoid production. J Biol Chem 279:30654–30661.

Colgan SP, Taylor CT, Narravula S, Synnestvedt K, Blume ED. 2002. Endothelial COX-2induction by hypoxia liberates 6-ketoPGF1a, a potent epithelial secretagogue. Adv ExpMed Biol 507:107–112.

Deckers B, BartelsH,Deckers S. 1995. Relationships between endothelial cells, pericytes andosteoblasts during bone formation in the sheep femur following inplantation oftricalciumphosphate ceramic. Anat Rec 242:310–320.

Deckers MML, Karperien M, van der Bent C, Yamashita T, Papapoulos SE, Lowik CW. 2000.Expression of vascular endothelial growth factors and their receptors during osteoblastdifferentiation. Endocrinology 141:1667–1674.

ElbjeiramiWM,West JL. 2006. Angiogenesis-like activity of endothelial cells co-culturedwithVEGF-producing smooth muscle cells. Tissue Eng 12:381–390.

Esbrit P, Alvarez-Arroyo M, De Miguel F, Martin O, Martinez M, Caramelo C. 2000.C-terminal parathyroid hormone-related protein increases vascular endothelial growthfactor in human osteoblastic cells. J Am Soc Nephrol 11:1085–1092.

Ferrara N. 1996. Vascular endothelial growth factor. Eur J Cancer 32A:2413–2422.Fong KD, Nacamuli RP, Loboa EG, Henderson JH, Fang TD, Song HM, Cowan CM, WarrenSM, Carter DR, Longaker MT. 2003. Equibiaxial tensile strain affects calvarial osteoblastbiology. J Craniofac Surg 14:48–55.

Fons P, Herault JP, Delesque N, Tuyaret J, Bono F, Herbert JM. 2004. VEGF-R2 andneuropilin-1 are involved in VEGF-A- induced differentiation of human bone marrowprogenitor cells. J Cell Physiol 200:351–359.

Gallant MA, Samadfam R, Hackett JA, Antoniou J, Parent JL, de Brum-Fernandes AJ. 2005.Production of prostaglandin D2 by human osteoblasts and modulation of osteoprotegerin,RANKL, and cellular migration by DP and CRTH2 receptors. J Bone MinerRes 20:681–682.

Gerber H-P, Ferrara N. 2000. Angiogenesis and bone growth. Trends Cardiovasc Med10:223–228.

Gerber H, Vu TH, Ryan A, Kowalski J,Werb Z, Ferrara N. 1999. VEGF couples hypertrophiccartilage remodeling, ossification and angiogenesis during endochondral bone formation.Nat Med 5:623–628.

Gliki G, Abu-ghazaleh R, Jezequel S, Wheeler-jones C, Zachary I. 2001. Vascular endothelialgrowth factor-induced prostacyclin production is mediated by a protein kinaseC (PKC)-dependent activation of extracellular signal-regulated protein kinases 1 and 2involving PKC- and by mobilization of intracellular Ca2þ. J Biochem 535:503–512.

Gonzalez RR, Cherfils S, Escobar M, Yoo JH, Carino C, Styer AK, Sullivan BT, Sakamoto H,Olawaiye A, Serikawa T, LynchMP, Rueda BR. 2006. Leptin signaling promotes the growthof mammary tumors and increases the expression of vascular endothelial growth factor(VEGF) and its receptor type two (VEGF-R2). J Biol Chem281:26320–26328.

Haider DG, Bucek RA, Giurgea AG, Maurer G, Glogar H, Minar E, Wolzt M, Mehrabi MR,Baghestanian M. PGE1 analog alprostadil induces VEGF and eNOS expression inendothelial cells. Am J Physiol Heart Circ Physiol 2005. 289:H2066–H2072.

Hauge EM, Qvesel D, Eriksen EF, Mosekilde L, Melsen F. 2001. Cancellous bone remodelingoccurs in specialized compartments lined by cells expressing osteoblastic markers. J BoneMiner Res 16:1575–1582.

Houliston RA, Pearson JD, Wheeler-Jones CPD. 2001. Agonist-specific cross talk betweenERKs and p38mapk regulates PGI2 synthesis in endothelium. Am J Physiol Cell Physiol281:C1266–C1276.

Houliston RA, Keogh RJ, Sugden D, Dudhia J, Carter TD, Wheeler-Jones CPD. 2002.Protease-activated receptors upregulate cyclooxygenase-2 expression in humanendothelial cells. J Thromb Haemost 88:321–328.

Hwang D, Scollard D, Byrne J, Levine E. 1998. Expression of cycoloxygenase 1 (COX-1) andcycoloxygenase 2 in human breast cancer. J Natl Cancer Inst 90:455–460.

Li J, Burr DB, Turner CH. 2002. Suppression of prostaglandin synthesis with NS-398 hasdifferent effects on endocortical and periosteal bone formation induced by mechanicalloading. Calcif Tissue Int 70:320–329.

Machwate M, Harada S, Leu CT, Seedor G, Labelle M, Gallant M, Hutchins S, Lachance N,Sawyer N, Slipetz D, Metters KM, Rodan SB, Young R, Rodan GA. 2001. Prostaglandinreceptor EP4 mediates the bone anabolic effects of PGE2. Mol Pharmacol 60:36–41.

JOURNAL OF CELLULAR PHYSIOLOGY DOI 10.1002/JCP

Mayr-wohlfart U, Waltenberger J, Hausser H, Kessler S, Gunther K-P, Dehio C, Puhl W,Brenner RE. 2002. Vascular endothelial growth factor stimulates chemotactic migration ofprimary human osteoblasts. Bone 30:472–477.

Midy V, Plouet J. 1994. Vasculotropin/vascular endothelial growth factor inducesdifferentiation cultured osteoblasts. Biochem Biophys Res Commun 199:380–386.

Milkiewicz M, Mohammadzadeh F, Ispanovi E, Gee E, Haas T. 2007. Static strain stimulatesexpression of matrix metalloproteinase-2 and VEGF in microvascular endothelium viaJNK- and ERK-dependent pathways. J Cell Biochem 100:750–761.

Muir C, Leland WK, Chung LW, Carson DD, Farach-Carson M. 2006. Hypoxia increasesVEGF-A production by prostate cancer and bone marrow stromal cells and initiatesparacrine activation of bone marrow endothelial cells. Clin Exp Metastasis 23:75–86.

Neagoe P-E, Lemieux C, Sirois MG. 2005. Vascular endothelial growth factor(VEGF)-A165-induced prostacyclin synthesis requires the activation of VEGF receptor-1and -2 heterodimer. J Biol Chem 280:9904–9912.

Okada Y, Tomita M, Gronowicz G, Kawaguchi H, Sohn J, Tanaka Y, Morimoto I. 2000.Effects of COX-2 gene disruption on osteoblastic function. J Bone Miner Res 15:S217(abstract).

Pitsillides AA, Rawlinson SC, Mosley JR, Lanyon LE. 1999. Bone’s early responses tomechanical loading differ in distinct genetic strains of chick: Selection for enhanced growthreduces skeletal adaptability. J Bone Miner Res 14:980–987.

Pufe T, Scholz-Ahrens KE, Franke ATM, Petersen W, Mentlein R, Varoga D, Tillmann B,Schrezenmeir J, Gluer CC. 2003. The role of vascular endothelial growth factor inglucocorticoid-induced bone loss: Evaluation in a minipig model. Bone 33:869–876.

Rawlinson SCF,Wheeler-Jones CPD, Lanyon LE. 2000. Arachidonic acid for loading inducedprostacyclin and prostaglandin E2 release from osteoblasts and osteocytes is derived fromthe activities of different forms of phospholipase A2. Bone 27:241–247.

Schneider M, Buchler P, Giese N, Giese T, Wilting J, Buchler MW, Friess H. 2006. Role oflymphangiogenesis and lymphangiogenic factors during pancreatic cancer progression andlymphatic spread. Int J Oncol 28:883–890.

Shapiro IM, Boyde A. 1987. Mineralisation of normal and rachitic chick growthcartilage: Vascular canals, cartilage calicification and osteogenenesis. Scanning Microsc1:599–606.

Stein GS, Lian JB, Owen TA. 1990. Relationship of cell growth to the regulation oftissue-specific gene expression during osteoblast differentiation. FASEB J 4:3111–3123.

Syeda F, Grosjean J, Houliston RA, Keogh RJ, Carter TD, Paleolog E, Wheeler-Jones CPD.2006. Cyclooxygenase-2 induction and prostacyclin release by protease-activatedreceptors in endothelial cells require cooperation between mitogen-activated proteinkinase and NF-{kappa}B pathways. J Biol Chem 281:11792–11804.

Takahashi T, Ueno H, Shibuya M. 1999. VEGF activates protein kinase C-dependent, butRas-independent Raf-MEK-MAP kinase for DNA synthesis in primary endothelial cells.Oncogene 18:2221–2230.

Tombran-Tink J, Barnstable CJ. 2004. Osteoblast and osteoclasts express PEDF, VEGF-Aisoforms, and VEGF receptors: Possible mediators of angiogenesis andmatrix remodellingin bone. Biochem Biophys Res Commun 316:573–579.

Trueta J. 1963. The role of the vessels in osteogenesis. J Bone Joint Surg 45-B:402–418.Trueta J, Little K. 1960. The vascular contribution to osteogenesis. II. Studies with theelectron microscope. J Bone Joint Surg 42:367–376.

VarneyML, Olsen KJ, Mosley RL, Singh RK. 2005. Paracrine regulation of vascular endothelialgrowth factor—a expression during macrophage-melanoma cell interaction: Role ofmonocyte chemotactic protein-1 and macrophage colony-stimulating factor. J InterferonCytokine Res 25:674–683.

Wang DS, Yamazazaki K, Nohtomi K, Shizume K, Ohsumi K, Shibuya M, Demura H, Sato K.1996. Increase of vascular endothelial growth factor mRNA expression produced byosteoblasts and by growth factors produced by endothelial cells. Endocrinology138:2953–2962.

Wang DS, Miura M, Demura H, Sato K. 1997. Anabolic effects of 1,25-dihydroxyvitamin D3on osteoblasts are enhanced by vascular endothelial growth factor produced byosteoblasts and by growth factors produced by endothelial cells. Endocrinology138:2953–2962.

Wheeler-JonesCPD, Abu-Ghazaleh R, Cospedal R. 1997. Vascular endothelial growth factorstimulates prostacylin production and activation of cytosolic phospholipase A inendothelial cells via p42/p44 mitogen activated kinase. FEBS Lett 353:503–512.

Xia G, Aiello LP, Ishii H, Jiang ZY, Park DJ, Robinson GS, Takagi H, NewsomeWP, JirousekMR, King GL. 1997. Characterisation of vascular endothelial growth factor’s effect on theactivation of protein kinase C, its isoforms and endothelial cell growth. J Clin Invest98:2018–2026.

Zaman G, Pitsillides AA, Rawlinson SC, Suswillo RF, Mosley JR, Cheng MZ, Platts LA,Hukkanen M, Polak JM, Lanyon LE. 1999. Mechanical strain stimulates nitric oxideproduction by rapid activation of endothelial nitric oxide synthase in osteocytes. J BoneMiner Res 14:1123–1131.

Zelzer E,McLeanW,NgY-S, FukaiN, ReginatoAM, Lovejoy S, D’Amore PA,OlsenBR. 2002.Skeletal defects in VEGF120/120 mice reveal multiple roles for VEGF in skeletogenesis.Development 129:1893–1904.

Zeng H, Dvorak HF, Mukhopadhyay D. 2001. Vascular permeability factor(VPF)/vascular endothelial growth factor (VEGF) receptor-1 down-modulatesVPF/VEGF receptor-2-mediated endothelial cell proliferation, but not migration,through phosphatidylinositol 3-kinase-dependent pathways. J Biol Chem276:26969–26979.

Zhang X, Schwartz EM, Young DA, Puzas E, Rosier RN,O’Keefe RJ. 2002. Cyclooxygenase-2regulates mesenchymal cell differentiation into the osteoblast lineage and is criticallyinvolved in bone repair. J Clin Invest 109:1405–1415.