www.elsevier.com/locate/ybbrc

Biochemical and Biophysical Research Communications 333 (2005) 1116–1122

BBRC

Use of green fluorescent fusion protein to track activation of thetranscription factor osterix during early osteoblast differentiation

Guangping Tai 1, Ioannis Christodoulou 1, Anne E. Bishop, Julia M. Polak *

Tissue Engineering and Regenerative Medicine Centre, Imperial College London, Chelsea and Westminster Campus, London, UK

Received 23 May 2005Available online 14 June 2005

Abstract

Osterix (Osx) is a transcription factor required for the differentiation of preosteoblasts into fully functioning osteoblasts. How-ever, the pattern of Osx activation during preosteoblast differentiation and maturation has not been clearly defined. Our aim was tostudy Osx activation during these processes in osteoblasts differentiating from murine and human embryonic stem cells (ESC). To dothis, we constructed an Osx-GFP fusion protein reporter system to track Osx translocation within the cells. The distribution of Osx-GFP at representative stages of differentiation was also investigated by screening primary osteoblasts, mesenchymal stem cells,synoviocytes, and pre-adipocytes. Our experiments revealed that Osx-GFP protein was detectable in the cytoplasm of cultured, dif-ferentiated ESC 4 days after plating of enzymatically dispersed embryoid bodies. Osterix-GFP protein became translocated into thenucleus on day 7 following transfer of differentiated ESC to osteogenic medium. After 14 days of differentiation, cells showing nucle-ar translocation of Osx-GFP formed rudimentary bone nodules that continued to increase in number over the following weeks(through day 21). We also found that Osx translocated into the nuclei of mesenchymal stem cells (C3H10T1/2) and pre-osteoblasts(MC3T3-E1) and showed partial activation in pre-adipocytes (MC3T3-L1). These data suggest that Osx activation occurs at a veryearly point in the differentiation of the mesenchymal-osteoblastic lineage.� 2005 Elsevier Inc. All rights reserved.

Keywords: Osterix; Embryonic stem cells; Osteoblasts; Differentiation

Multiple transcription factors control the differentia-tion of osteoblasts at various stages of their develop-ment. Cbfa1 has been identified as one of the keytranscription factors for osteoblast-specific geneexpression [1–3]. More recently, osterix (Osx), a zinc fin-ger-containing protein, was shown to be the secondosteoblast-specific transcription factor. Osx-deficientmice lack osteoblasts and do not form endochondralor intramembranous bones [4]. It is likely that Cbfa1and Osx control different stages of osteoblast differenti-ation and, as Osx is not expressed in Cbfa1-null mice but

0006-291X/$ - see front matter � 2005 Elsevier Inc. All rights reserved.

doi:10.1016/j.bbrc.2005.05.195

* Corresponding author. Fax: +44208746 5619.E-mail address: julia.polak@imperial.ac.uk (J.M. Polak).

1 Present address: Department of Gene Expression and DevelopmentRoslin Institute, Roslin, Midlothian, EH25 9PS Edinburgh, UK.

Cbfa1 is expressed normally in Osx-deficient mice, it hasbeen suggested that Osx acts downstream of Cbfa1 [5–7]. It is known that osteoblast differentiation and lineagecommitment involve closely coordinated, multistep, par-allel pathways of chondrocyte and osteoblast differenti-ation [8]. Osx plays a role in the segregation ofosteoblast and chondrocyte lineages from a commonosteochondroprogenitor cell during the formation ofendochondral bone [5–7]. Osx negatively regulates theexpression of Sox9 and Sox5, transcription factorsknown to provide signals for induction of chondrogene-sis, and preosteoblasts are blocked from differentiatinginto osteoblasts in Osx-null mice, with the cells express-ing marker genes characteristic of chondrocytes instead[8,3]. The molecular mechanisms underlying these pro-cesses have not yet been defined and there is no direct

G. Tai et al. / Biochemical and Biophysical Research Communications 333 (2005) 1116–1122 1117

visual evidence of Osx activation during osteoblast celldifferentiation due to the lack of an appropriate anti-body to study its localization.

In this study, we have constructed an Osx transcrip-tion factor-green fluorescent protein fusion protein(Osx-GFP), in order to track the location and activationof Osx protein during osteoblast differentiation. Our re-sults showed that Osx was activated and translocatedinto the nucleus at the mesenchymal stem cell and pre-osteoblast stages of osteoblastic lineage differentiation.

Materials and methods

Construction of Osx-GFP fusion protein plasmid DNA. To constructthe vector expressing an Osx-GFP fusion protein, a 1.5-kb fragment ofthe full-length mouse Osx cDNA was excised from pcDNA (kindlyprovided by Dr. K. Nakashima, Anderson Cancer Center, Universityof Texas, Houston, TX) and inserted into the expression vectorpEGFP-C3 (Clontech, Palo Alto, CA). Insert orientation was con-firmed by restriction enzyme digestion. An osteocalcin-lacZ reporterplasmid (Invivogene, San Diego, CA) was used to assess the activity ofthe transcription factor fusion protein by co-transfecting it, with orwithout the Osx-GFP fusion protein, into a rabbit synoviocyte line(HIG-82: fibroblast type B synoviocytes, ATCC, LGC Promochem,Middlesex, UK).

Culture of cell lines. To investigate the specificity and activationstate of the Osx-GFP fusion protein, a panel of osteogenic and non-osteogenic cell lines was used: (a) murine mesenchymal stem cells (D1,CRL-12424, and C3H10T1/2), (b) murine pre-osteoblasts (MC3T3-E1subclone 4), (c) murine pre-adipocytes (MC3T3-L1), and (d) rabbitsynoviocytes (HIG82) were obtained from ATCC (LGC Promochem,Middlesex, UK). Cells were grown in: (a) Dulbecco�s modified Eagle�smedium (DMEM; Invitrogen, Paisley, UK) with 10% (v/v) FBS, (b)10% FBS (v/v) Alpha minimum essential medium (a-MEM) (Invitro-gen) with ribonucleosides and deoxyribonucleosides and with 2 mM L-glutamine and 1 mM sodium pyruvate but without ascorbic acid, (c)10% (v/v) FBS DMEM with 4.5 g/L glucose, or (d) 10% (v/v) FBSHam�s F12 medium, according to the supplier�s instructions. Both non-osteogenic (murine fibroblasts; NIH3T3 and SNL lines) and osteogenic(human osteosarcoma cells; MG63, murine and human fetal primaryosteoblasts) cells were used to verify the cell-specific activation of Osx-GFP. The NIH3T3, SNL, and MG63 cell lines were cultured in 10%FBS DMEM. Primary osteoblasts from human and murine fetal longbone were derived as described previously [9]. Isolated primary oste-oblasts were cultured in F12-DMEM full medium (Invitrogen) with10% FBS, 2 mM L-glutamine, and 1% (v/v) penicillin/streptomycin(100 U/ml penicillin, and 100 mg/ml streptomycin) (Invitrogen). Col-lection of human fetal long bone and the entire study were approvedby the Riverside Research Ethics Committee (GMIC-00033).

Cell culture and osteoblast differentiation from embryonic and mes-

enchymal stem cells. To investigate the activation of the Osx-GFP, itsintracellular location was studied during osteogenic differentiation ofmurine and human ESC. For the murine ESC experiment, a feeder-freecell line (Tg2a E14) was used as described previously [10] and osteo-genic differentiation was achieved using an established protocol [11].Briefly, cells from dispersed embryoid bodies (EBs) were transferred togelatin-coated tissue culture plates in osteogenic medium (supple-mented with 50 lg/ml L-ascorbate phosphate, 10 mM b-glycerophos-phate, and 5 · 10�7 M dexamethasone) maintained for up to 21 days.The cells were then transfected with the Osx-GFP DNA construct ondays 4, 7, and 21. These time points were selected to represent pro-genitor pools of early mesenchymal, osteo-progenitor, and mature cellstages of osteoblast differentiation [10,11].

The H1 human ESC line H1, isolated and established at the Uni-versity of Wisconsin (Madison, WI), was cultured as described [12].Briefly, a feeder layer of murine embryonic fibroblasts was prepared bychemically inactivating subconfluent cultures with mitomycin C (2 mg/ml in culture medium) (Sigma–Aldrich, Poole, Dorset, UK) for150 min at 37 �C. Inactivated fibroblasts were then reseeded at2 · 105 per well onto six-well plates that had previously been coatedwith a 0.1% (v/v) gelatin solution (Sigma). Human ESC were main-tained in DMEM:F-12 (Invitrogen) containing 20% (v/v) KnockoutSerum Replacement (Invitrogen), 1 mM L-glutamine (Invitrogen),0.1 mM of 2-mercaptoethanol (Sigma), 1% (v/v) non-essential aminoacids, and basic fibroblast growth factor (4 ng/ml, bFGF, human re-combinant; Invitrogen). Medium was changed every day. Passagingwas achieved by treating cells for 5–10 min with a 1 mg/ml solution oftype IV collagenase (Invitrogen) followed by mechanical scraping ofthe colonies to remove them from the feeder layer.

To initiate spontaneous differentiation of human ESC, colonieswere removed from the feeders using collagenase and mechanicalmethods, as described previously [13]. The cells were then placed intosuspension culture for 5 days in basic differentiation medium (a-MEMand 15% [v/v] heat-inactivated fetal calf serum, without penicillin andstreptomycin). The embryoid bodies that formed were disrupted tocreate a single-cell suspension by washing twice with agitation inDulbecco�s phosphate-buffered saline (PBS) without calcium or mag-nesium (Invitrogen) and incubating the cells with agitation for 5 min in1· trypsin–EDTA solution (Sigma). The cells were then reseeded ontoconventional tissue culture plastic at 5 · 105 per well and grown for 21days in an osteogenic medium comprising a-MEM medium (Invitro-gen) supplemented with L-ascorbate phosphate (50 mg/ml) and b-glycerophosphate (10 mM) with dexamethasone (5 · 10�7 M) [12].Transfection experiments were carried out on days 4, 7, and 21 duringdifferentiation towards the osteoblast lineage.

Spontaneous differentiation of mesenchymal stem cells (cell lineC3H10T1/2) was initiated by low serum culturing in DMEM with 2%(v/v) FBS induction for 11 days. Beating cardiomyocyte-like andneuron-like cells could be identified in the cultures on days 11–15 onthe basis of their distinct morphology. The state of osterix activation inthese cells was investigated by transient transfection with Osx-GFP.

Transfection and visualization of subcellular localization of GFP

tagged protein in a panel of osteogenic and non-osteogenic cells. Osx-GFP was transfected into a panel of osteogenic cells in order toinvestigate its cell phenotype-dependent activation and translocationpattern. These comprised murine and human primary fetal and adultosteoblasts, human osteosarcoma cell line (MG63), murine mesen-chymal stem cell line (D1,CRL-12424), murine pre-adipocytes(MC3T3-L1), and murine pre-osteoblasts (MC3T3-E1). Non-osteo-genic cells were also transfected in order to exclude the non-specificactivation of the fusion protein, including murine fibroblasts (NIH3T3and SNL) and beating cardiomyocyte-like and neuron-like cells de-rived from a mesenchymal stem cell line (C3H10T1/2). Briefly, cellswere cultured on gelatin-coated six-well plastic plates, and the mediumwas changed to OPTI-MEM medium (Invitrogen) for 1 h beforetransfection. A plasmid cocktail containing 4 lg Osx-GFP plasmid in500 ll OPTI-MEM medium was mixed with 10 ll of Lipofectamine2000 (Invitrogen) in 500 ll OPTI-MEMmedium for 30 min at ambienttemperature and added to the culture. After 3 h incubation, themedium was replaced with osteogenic medium and cells were incu-bated overnight. The fluorescence of cells in six-well plates was mon-itored and photographed using an Olympus BI-70 inverted microscopeequipped with a Nikon Pix950 digital camera (Nikon, Tokyo, Japan)with an FITC filter. b-galactosidase activity of osteocalcin-promoterreporter (LacZ) was detected by X-gal staining [13]. The mean per-centage of activated cells (i.e., showing nuclear translocation of Osx-GFP) was assessed in all transiently transfected cells in random fieldsfrom three separate experiments. A total of 400 of transfected cells wascounted. Data are expressed as means ± the standard error of themean.

1118 G. Tai et al. / Biochemical and Biophysical Research Communications 333 (2005) 1116–1122

Results

Construction and transcriptional activity of the osterix-

GFP fusion protein

The transcription factor activity of the Osx-GFP fu-sion protein was tested by co-transfecting with an oste-ocalcin-promoter reporter (LacZ) gene in asynoviocyte/fibroblast cell line (HIG82). No fluores-cence could be detected and there was a lack of tran-scription of reporter gene and b-galactosidase activityin cells transfected with osteocalcin-promoter reporter(LacZ) gene only (Figs. 1A and B). After co-transfection

Fig. 1. Osx-GFP fusion protein is transcriptionally active. Synovio-cytes were transfected with 2000 ng of each of the expression vectorsfor OCN-lacZ (A,B) or OCN-lacZ plus Osx-GFP (C,D) as indicatedand images were taken 18 h after transfection. LacZ was not expressedin fibroblast type synoviocytes (B), but was activated and expressed inOsx-GFP co-transfected cells (C), LacZ stained by X-gal and activatedOsx-GFP in nuclei (D). Bar 100 lm.

Fig. 2. Osx-GFP fusion protein is inactive in non-osteogenic cells. Murine adwith Osx-GFP. (A,D) Fluorescence microscopy revealed the expression patterwith phase contrast microscopy. (C,F) Percentages of cells with nuclear (Nuc)counted in the following days from 400 randomly selected transfected cells.

with Osx-GFP plasmid, the fusion protein itself could bedetected in the cells (Fig. 1C) where it activated the oste-ocalcin promoter resulting in the expression of b-galac-tosidase enzyme, which cleaved the substrate, X-gal, toform a visible blue compound (Fig. 1D).

Comparison of the subcellular localization of osterix-GFP

in non-osteogenic and osteogenic cells

The cell specificity of the Osx-GFP fusion protein wasconfirmed further in a range of non-osteogenic and oste-ogenic cells. Osx-GFP remained mostly inactive in thecytosol of non-osteogenic cells, with nuclear transloca-tion not being seen at all in NIH3T3 fibroblasts, andonly in a small percentage of fetal embryonic fibroblasts(SNL) (34.5 ± 10.2%; mean ± SEM) (Fig. 2). Nochange in the subcellular distribution of the fusion pro-tein was seen in these non-osteogenic cells during a fur-ther 7-day culture in osteogenic medium. In contrast,Osx-GFP was clearly present in virtually all nuclei ofall the different types of osteogenic cells tested, includinghuman osteosarcoma cells (MG63) (98.2 ± 4.2%), hu-man fetal primary osteoblasts (99.5 ± 1.8%), and murineprimary osteoblasts (99.5 ± 3.4%) (Figs. 3A–I).

Activation and subcellular localization of osterix-GFP in

cells differentiated from ESC

In an attempt to elucidate how the nuclear transloca-tion of Osx transcription factor is regulated during thedifferentiation of ESC, we used our established cultureprotocol to differentiate ESC into osteoblasts [11]. Forthe human ESC osteoblast differentiation model, theOsx-GFP fusion protein was found to be uniformly dis-tributed throughout the cells after 4 days in osteogenicmedium (Fig. 4A) and had become activated and trans-located to the nucleus by day 7 (Fig. 4B). By day 21,

ult (NIH3T3) (A–C) and fetal (D–F) fibroblasts (SNL) were transfectedn of Osx-GFP in both types of fibroblasts. (B,E) The same fields shown, cytoplasmic (Cyt), and pan-cellular (Pan) expression of Osx-GFP wereBar 100 lm.

Fig. 3. Osx-GFP fusion protein is activated in the nuclei of osteogenic cells. Human osteosarcoma cells (MG63) (A–C), primary fetal osteoblast cells(FOB) (D–F), and murine osteoblasts from long bone (MOB) (G–I) were transfected with Osx-GFP. (A, D, and G) Fluorescence microscopyrevealed the expression pattern of Osx-GFP in the cells. (B, E, and H) The same fields shown with phase contrast microscopy. (C, F, and I)Percentages of cells with nuclear (Nuc), cytoplasmic (Cyt), and pan-cellular (Pan) expression of Osx-GFP were counted in the following days from400 randomly selected transfected cells. Bar 100 lm.

Fig. 4. Dynamic activation and subcellular localization of Osx-GFP fusion protein during osteoblast differentiation. Human (A–C) and murine (D–F) ESC were transfected with Osx-GFP on days 4, 7, and 21. The percentages of activated (showing nuclear translocation) hESC (G) and mESC (H)on days 4, 7, and 21 were counted in the following days after transient transfection from 400 randomly selected transfected cells. Bar 100 lm.

G. Tai et al. / Biochemical and Biophysical Research Communications 333 (2005) 1116–1122 1119

most of the Osx-GFP fusion protein had translocatedinto the nucleus (Fig. 4C). Similar results were obtainedwith the murine ESC model (Figs. 4D–F). The percent-

ages of human (Fig. 4G) and murine ESC (Fig. 4H)showing Osx-GFP activation and translocation in-creased significantly (P < 0.01) from day 7 to day 21.

Fig. 5. Osx-GFP fusion protein is activated in mesenchymal progenitor cells. Cells with nuclear translocation of Osx-GFP fusion protein in murinemultipotent cells (D1) (A), pre-osteoblast (MC3T3-E1) (D), and murine pre-adipocyte cells (MC3T3-L1) (G), corresponding phase contrastmicroscopy (B, E, and H), Percentages of cells with nuclear (Nuc), cytoplasmic (Cyt), and pan-cellular (Pan) expression of Osx-GFP were counted inthe following days from 400 randomly selected transfected cells. Respectively (C, F, and I). Bar 100 lm.

1120 G. Tai et al. / Biochemical and Biophysical Research Communications 333 (2005) 1116–1122

Subcellular localization of Osx-GFP fusion protein in

mesenchymal stem cells and committed lineages

To define the precise stages of Osx activation duringosteogenic differentiation, a mesenchymal stem cell lineand committed cells, representing different lineages aris-ing from mesenchymal stem cells, were studied. Interest-ingly, Osx was found to be fully activated in all murinemesenchymal stem cells (D1) (Figs. 5A and D) (100%cells showed nuclear translocation) and pre-osteoblast(MC3T3-E1) cell lines (100%). Whereas, in the mesen-chyme-derived, non-osteogenic pre-adipocytic MC3T3-L1 cell line, Osx-GFP remained partly active(75.4 ± 4.6%) (Figs. 5H–J). Although all Osx-GFP wasfound to be fully activated in the C3H10T1/2 mesenchy-mal stem cells (Figs. 6A and B), after 11 days spontane-ously differentiation in low serum medium, Osx-GFPwas inactive and evenly distributed within the neuron-like cells (Figs. 6C and D) and had moved out of nucleiin all beating C3H10T1/2-derived cardiomyocyte-likecells (Figs. 6E and F).

Discussion

The molecular mechanisms underlying the differenti-ation of osteoblasts from a common mesenchymal stemcell through progenitor stages to a mature, fully func-

tional, matrix-synthesizing cell are not clear. Definitiveevidence for the existence of multipotent mesenchymalprogenitor or stem cells has been obtained by analysisof the in vitro differentiation outcomes of clonal cells de-rived from stroma, bone marrow or embryonic cells [15].In our previous study, we speculated that the process ofosteogenic differentiation of embryonic stem cells (ESC)involves transition through a mesenchymal stem cell-likeor multipotent mesodermal progenitor cell population[11]. Expression of the gene coding for the transcriptionfactor osterix (Osx) was detected on day 7 of osteogenicdifferentiation of murine ESC in vitro [11], but the timecourse of Osx activation has not been investigated previ-ously. In this study, we show for the first time that Osx isactivated in mesenchymal stem cells, which is earlierthan the preosteoblast stage that would be expectedfrom previous work [4–6]. We also found that Osx re-mained partly active in the mesenchyme-derived earlycommitted progenitors of adipocytes (MC3T3-L1). Thisis in agreement with results obtained from Osx-null miceshowing that Osx is expressed by bipotential cells and avariety of mixed, committed bipotential progenitors forbone/fat-forming and bone/cartilage-forming cells mayexist [7,3,14,15].

The observed pattern of dynamic translocation ofOsx-GFP suggests a sequence of events during osteo-blast differentiation from ESC. (Fig. 7) Osx is inactivein undifferentiated ESC and throughout the early phase

Fig. 6. Osx-GFP fusion protein is inactive in non-osteogenic lineagescells derived from mesenchymal stem cells. Osx-GFP was translocatedto the nucleus of murine mesenchymal stem cells (C3H10T1/2) showingits activation (A,B). After 11 days of spontaneous differentiation, Osxactivation was switched off and it moved out nuclei in the non-osteogenic cardiomyocyte-like beating cells (C, Osx-GFP; D, phasecontrast) and evenly distributed in neuron-like cells (E, Osx-GFP; F,phase contrast) derived from the C3H10T101/2 cell line. Bar 100 lm.

G. Tai et al. / Biochemical and Biophysical Research Communications 333 (2005) 1116–1122 1121

of spontaneous embryoid body formation (i.e., the firstfive days of ESC differentiation). No nuclear transloca-tion of Osx-GFP was found up to day 4 followingESC growth in osteogenic medium, but Osx-GFP fusion

Fig. 7. Activation and hypothetical role of osterix in osteoblast differentiaembryoid bodies, and in early differentiated ESC (dESC) up to day 4 followiinto nuclei by day 7, when mesenchymal progenitors are formed (mesenchymaconfirmed in mesenchymal progenitors (C3H10T101/2), pre-osteoblasts (Mcardiomyocyte-like beating cells and neuron-like cells derived from C3H10T

protein was activated and translocated into nuclei fromday 7 onwards. This time point corresponds to whenmultipotent mesenchymal progenitor cells are formed[10,11]. It seems that Osx is activated and translocatedinto nuclei at this stage, thereby activating downstreamgenes and acting in co-ordination with other transcrip-tion factors (e.g., Sox 9) in segregating osteoblast andchondrocyte lineages from the mixed pool of osteo-blast/chondrocyte/adipocyte progenitors. In support ofthis contention, it has been reported that Osx negativelyregulates the expression and function of the gene codingfor the chondrocyte-specific transcription factor Sox 9[4,6,7] and that overexpression of Osx inhibits expres-sion of the adipocyte-specific transcription factorPPAR-c gene [11,16]. We have provided visual evidence,therefore, that Osx regulates and commits precursorcells to the osteoblast lineage, preventing them fromadopting a chondrocyte or adipocyte phenotype. Inaddition, this study provides a tool for the investigationof the dynamics of Osx activation and its interactionwith other transcription factors in osteogenic lineagecommitment and has potential applications in drug dis-covery where Osx-targeted compounds could be identi-fied by simple screening of nuclear translocation.

Acknowledgments

This work was supported by a Medical ResearchCouncil Development Grant, The RoseTrees Trust,The Grand Charity, and The Julia Polak ResearchTrust. The authors thank Dr. Kazuhisa Nakashimaand Professor Benoit de Crombrugghe for the osterixexpression plasmid and Professor Richard Lubman,Dr. Robert Bielby for helpful discussions of these data.

tion. Osx is inactive in undifferentiated embryonic stem cells (ESC),ng cultivation in osteogenic medium. Osx is activated and translocatedl progenitor pool). Activation and nuclear translocation of Osx-GFP isC3T3-E1), synoviocytes, and pre-adipocytes (MC3T3-L1), but not in101/2 cells.

1122 G. Tai et al. / Biochemical and Biophysical Research Communications 333 (2005) 1116–1122

References

[1] P. Ducy, R. Zhang, V. Geoffroy, A.L. Ridall, G. Karsenty, Osf2/Cbfa1: a transcriptional activator of osteoblast differentiation,Cell 89 (1997) 747–754.

[2] X. Yang, G. Karsenty, Transcription factors in bone: develop-mental and pathological aspects, Trends Mol. Med. 8 (2002) 340–345.

[3] J.B. Lian, A. Javed, S.K. Zaidi, C. Lengner, M. Montecino, A.J.van Wijnen, J.L. Stein, G.S. Stein, Regulatory controls forosteoblast growth and differentiation: role of Runx/Cbfa/AMLfactors, Crit. Rev. Eukaryot. Gene Expr. 14 (2004) 1–41.

[4] K. Nakashima, X. Zhou, G. Kunkel, Z. Zhang, J.M. Deng, R.R.Behringer, C.B. de, The novel zinc finger-containing transcriptionfactor osterix is required for osteoblast differentiation and boneformation, Cell 108 (2002) 17–29.

[5] K. Nakashima, C.B. de, Transcriptional mechanisms in osteoblastdifferentiation and bone formation, Trends Genet. 19 (2003) 458–466.

[6] J.B. Lian, G.S. Stein, Runx2/Cbfa1: a multifunctional regulatorof bone formation, Curr. Pharm. Des. 9 (2003) 2677–2685.

[7] D.S. de Jong, B.L. Vaes, K.J. Dechering, A. Feijen, J.M.Hendriks, R. Wehrens, C.L. Mummery, E.J. van Zoelen, W.Olijve, W.T. Steegenga, Identification of novel regulators associ-ated with early-phase osteoblast differentiation, J. Bone Miner.Res. 19 (2004) 947–958.

[8] C.B. de, V. Lefebvre, K. Nakashima, Regulatory mechanisms inthe pathways of cartilage and bone formation, Curr. Opin. CellBiol. 13 (2001) 721–727.

[9] I.D. Xynos, A.J. Edgar, L.D. Buttery, L.L. Hench, J.M. Polak,Ionic products of bioactive glass dissolution increase proliferation

of human osteoblasts and induce insulin-like growth factor IImRNA expression and protein synthesis, Biochem. Biophys. Res.Commun. 276 (2000) 461–465.

[10] L.D. Buttery, S. Bourne, J.D. Xynos, H. Wood, F.J. Hughes,S.P. Hughes, V. Episkopou, J.M. Polak, Differentiation ofosteoblasts and in vitro bone formation from murine embryonicstem cells, Tissue Eng. 7 (2001) 89–99.

[11] G. Tai, J.M. Polak, A.E. Bishop, I. Christodoulou, L.D. Buttery,Differentiation of osteoblasts from murine embryonic stem cellsby overexpression of the transcriptional factor osterix, Tissue Eng.10 (2004) 1456–1466.

[12] R.C. Bielby, A.R. Boccaccini, J.M. Polak, L.D. Buttery, In vitrodifferentiation and in vivo mineralization of osteogenic cellsderived from human embryonic stem cells, Tissue Eng. 10 (2004)1518–1525.

[13] K. Sakai, L. Hiripi, V. Glumoff, O. Brandau, R. Eerola, E.Vuorio, Z. Bosze, R. Fassler, A. Aszodi, Stage- and tissue-specificexpression of a Col2a1-Cre fusion gene in transgenic mice, MatrixBiol. 19 (2001) 761–767.

[14] J.E. Kim, K. Nakashima, C.B. de, Transgenic mice expressing aligand-inducible cre recombinase in osteoblasts and odontoblasts:a new tool to examine physiology and disease of postnatal boneand tooth, Am. J. Pathol. 165 (2004) 1875–1882.

[15] J.E. Aubin, Bone stem cells, J. Cell. Biochem. Suppl. 30–31 (1998)73–82.

[16] J.Y. Choi, J. Pratap, A. Javed, S.K. Zaidi, L. Xing, E. Balint, S.Dalamangas, B. Boyce, A.J. van Wijnen, J.B. Lian, J.L. Stein,S.N. Jones, G.S. Stein, Subnuclear targeting of Runx/Cbfa/AML factors is essential for tissue-specific differentiation duringembryonic development, Proc. Natl. Acad. Sci. USA 98 (2001)8650–8655.