The T-box transcription factor Brachyury mediates cartilage development in mesenchymal stem cell line C3H10T1/2

The murine mesenchymal stem cell line C3H10T1/2 line presents many features of mesenchymal stem cells (MSCs). It can differentiate into mesenchymal lineages such as muscle-forming myoblasts, fat-storing adipocytes,cartilage-forming chondrocytes and bone-forming osteoblasts. Different BMP(Bone Morphogenetic Protein) family members are able to mediate induction of the osteogenic, chondrogenic or adipogenic lineages, but obviously not the myogenic pathway (Reznikoff et al.,1973; Taylor and Jones,1979; Wang et al.,1993; Ahrens et al.,1993).

Different complex transcriptional control mechanisms regulate the initial commitment stage of mesenchymal stem cell formation through to the final manifestation of the mesenchymal tissue type. It has been demonstrated that varying types of transcription factors govern such mechanisms. For example basic helix-loop-helix factors (bHLH) (Myo-D, Myf-5, MRF4 and Myogenin)regulate the differentiation of skeletal muscle cells(Arnold and Winter, 1998). A member of the nuclear hormone receptor family (PPARγ2) determines adipocyte differentiation (Tontonoz et al., 1994), and the runt-family member CBFA1 is required for osteoblast determination and differentiation(Komori and Kishimoto, 1998). The exact transcriptional mechanisms that determine development into the chondrogenic lineage are unknown, although it has been documented that SOX9, a member of the high-mobility-group (HMG) box protein superfamily, is required(Lefebvre and De Crombrugghe,1998). Mutations in Sox9 cause abnormalities (Campomelic Dysplasia) in cartilage-derived skeletal structures(Wagner et al., 1994;Foster et al., 1994). Several investigations, also involving chimeric mice, demonstrate that SOX9 is a major regulator of cartilage-specific genes (collagen II, XI)(Bridgewater et al., 1998;Lefebvre et al., 1998) and is crucially involved in chondrocyte formation(Bi et al., 1999). Similarly,it could be also demonstrated that HLH-transcription factor Scleraxis^-/- cells in chimeric mice are excluded from regions of the embryo that are involved in the formation of skeletal structures(Brown et al., 1999).

Recently, we and others found that in MSCs, BMP-signaling regulates differentiation to the osteogenic and the chondrogenic lineages in quite different ways. BMP-mediated SMAD-signaling seems to be necessary during the entire osteoblast-developmental sequence. In the case of chondrogenesis, BMPs are necessary for induction; however, the BMP-mediated SMAD signaling is not sufficient to significantly induce or promote chondrogenic differentiation in either the mesenchymal stem cell line C3H10T1/2 or in the prechondrogenic cell line ATDC5 (Fujii et al.,1999; Ju et al.,2000). We now show that BMP2-mediated upregulation of Fibroblast Growth Factor (FGF) receptor 3 (FGFR3) seems to be involved in the induction of chondrogenic differentiation of MSCs. This finding is with agreement with evidence that FGF-signaling is intimately involved in skeletal development(Wilkie et al., 1995;Colvin et al., 1996;Deng et al., 1996). It is shown here that forced expression of FGFR3 in MSCs (C3H10T1/2) leads to the onset of the chondrogenic lineage. We have also found the T-box containing transcription factor Brachyury, which is capable of mediating the FGFR3-dependent onset of chondrogenesis in these MSCs, following screening for transcription factors exhibiting a chondrogenic capacity in C3H10T1/2.

Brachyury, or T, is the founder member of a family of transcription factors that share the T-box, a 200-amino-acid DNA-binding domain (reviewed in Smith,1997; Papaioannou,1997). The mouse Brachyury gene is expressed at high rates during gastrulation and is required for differentiation of the notochord and the formation of mesoderm during posterior development(Kispert et al., 1995). Then Brachyury expression is downregulated at mid to late gestation periods. Here we show that FGFR3-mediated signaling induces expression of Brachyury in mesenchymal stem cell line C3H10T1/2. Forced expression of Brachyury in MSCs in vitro and ectopically, in vivo, is sufficient to initiate chondrogenic development in these MSCs. Therefore, T-box family members such as Brachyury may be factor(s) required not only for patterning but also contributing to the determination of the chondrogenic lineage.

For assessment of transcriptional activity a dimer of the double-stranded oligonucleotide of the Brachyury binding element (BBE)AATTTCACACCTAGGTGTGAAATT (Kispert et al.,1995) was incorporated in the BamHI site before the HSV thymidine kinase minimal promoter fused to the cloramphenicol acetyltransferase (CAT)-reporter of pBLCAT5(Boshart et al., 1992) to give reporter plasmid pBBE-CAT5. 20 hours before transfection, human embryonic kidney HEK293T cells were plated at a density of 1×10⁴/cm² in 6-well plates and allowed to grow under normal culture conditions. For cotransfection experiments, we used (per well) 250 ng of Brachyury expression vector and 250, 500 or 750 ng of the expression vector encoding dnBrachyury. Empty vector was added to adjust the amount of expression plasmids to 1 μg/ml. 260 ng of BBE-CAT reporter (pBBE-CAT5) were added in the presence of 140 ng of RSV-lacZ vector using the DOSPER procedure (see below). Cells were allowed to incubate for 48 hours. Then, cells were collected and β-galactosidase assays performed using the chemiluminescent β-gal reporter gene assay (Roche Diagnostics,Mannheim, Germany) and CAT-assays using the CAT ELISA kit (Roche Diagnostics,Mannheim, Germany). β-gal assay results were used to normalize the CAT assay results for transfection efficiency. All DNA transfection experiments were repeated at least three times in triplicate.

Human embryonic kidney cells HEK293T and murine C3H10T1/2 progenitor cells were routinely cultured in tissue culture flasks in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum (FCS),0.2 mM L-glutamine and antibiotics (50 i.u./ml penicillin, 50 mg/ml streptomycin). Cells were transfected using DOSPER according to the manufacturer's protocol (Roche Diagnostics, Mannheim, Germany). C3H10T1/2 cells that recombinantly express BMP2 (C3H10T1/2-BMP2) cells were obtained by cotransfection with pSV2pac followed by selection with puromycin (2.5μg/ml). FGFR3, Brachyury and the T-box domain were polymerase chain reaction (PCR)-amplified and cloned into expression vectors pMT7T3 and pMT7T3-pgk, which are under the control of the LTR of the myeloproliferative virus or of the murine phosphoglycerate kinase promoter-1, respectively(Ahrens et al., 1993). The integrity of the constructs was confirmed by sequencing. HA-tags were added to the carboxy terminus of full-length Brachyury and its T-box domain by PCR with primers encoding the respective peptide sequence. Stable expression of the DNA binding T-box domain (amino acids (aa) 1-229) and of the dominant-negative human FGFR3 without the cytoplasmatic tyrosine kinase domains (aa 1-414) in the C3H10T1/2-BMP2 background was done by cotransfection with pAG60, conferring resistance to G418 (750 μg/ml). Individual clones were picked, propagated and tested for recombinant FGFR3, dnFGFR3, Brachyury or T-box domain (dnBrachyury) expression by reverse-transcription-coupled PCR (RT-PCR) (see below). Selected cell clones were subcultivated in the presence of puromycin or puromycin/G418 and the selective pressure was maintained during subsequent manipulations. C3H10T1/2 cells were cultured in DMEM containing 10% FCS. The features of C3H10T1/2-BMP2 cells have been described (Ahrens et al.,1993; Hollnagel et al.,1997;Bächner et al., 1998). For assessment of in vitro osteo/chondrogenic development, cells were plated at a density of 5-7.5×10³cells/cm² and after reaching confluence (arbitrarily termed day 0),ascorbic acid (50 μg/ml) and 10 mM β-glycerophosphate were added as specified by Owen et al. (Owen et al.,1990).

For BMP2-stimulation studies, C3H10T1/2 cells were plated at a density of 1×10⁴/cm² in a 9-cm culture dish. After 48 hours cells were washed 3× with phosphate-buffered saline (PBS) and then cells were starved for 24 hours in DMEM without serum. Before induction the medium was replaced with fresh DMEM without serum. Cells were then treated for the indicated times using recombinant BMP2 from E. coli (50 ng/ml). Cycloheximide (50 μg/ml) treatment started 30 minutes prior to the addition of BMP2.

Total cellular RNAs were prepared by TriReagent^LS according to the manufacturer's protocol (Molecular Research Center Inc.). 5 μg of total RNA was reverse-transcribed and cDNA samples were subjected to PCR. RT-PCR was normalized by the transcriptional levels of hypoxanthine guanine phosphoribosyl transferase (HPRT). The HPRT-specific 5′ and 3′primers were GCTGGTGAAAAGGACCTCT and AAGTAGATGGCCACAGGACT, respectively. The following 5′ and 3′ primers were used to evaluate osteo/chondrogenic differentiation: collagen 1a1: GCCCTGCCTGCTTCGTG,CGTAAGTTGGAATGGTTTTT; collagen 2a1: CCTGTCTGCTTCTTGTAAAAC,AGCATCTGTAGGGGTCTTCT; osteocalcin: GCAGACCTAGCAGACACCAT,GAGCTGCTGTGACATCCATAC; PTH/PTHrP-receptor: GTTGCCATCATATACTGTTTCTGC,GGCTTCTTGGTCCATCTGTCC; FGFR3: CCTGCGCAGTCCCCCAAAGAAG; CTGCAGGCATCAAAGGAGTAGT;FGFR2: TTGGAGGATGGGCCGGTGTGGTG, GCGCTTCATCTGCCTGGTCTTG. The primer pairs for Brachyury and Sox9 have been described(Johansson and Wiles, 1995;Zehentner et al., 1999),respectively. Vector-borne transcripts for Brachyury were evaluated with nested primer sets using either vector-specific 5′- or 3′-primers: TTAGTCTTTTTGTCTTTTATTTCA; GATCGAAGCTCAATTAACCCTCAC.

Recombinant cells from Petri dishes (13.6 cm diameter) were harvested at different time points before (day B2), at (day 0) and after (days 2, 4, 7)confluence. Lysis was in RIPA buffer (1% (v/v) Nonidet P-40, 0.1% SDS (w/v),0.5% sodium deoxycholate in PBS, containing 100 μg/ml phenyl methyl sulfonyl fluoride (PMSF), 2 μg/ml aprotinin and 1 mM Na₃VO₄). Lysates were centrifuged (30 minutes, 10,000 g, 4°C) and the supernatants were stored at -70°C until analysis. Protein concentration of the lysates was determined using Coomassie Brilliant Blue staining. Protein was precipitated with ethanol,resuspended in reducing (containing dithiothreitol (DTT)) or non-reducing sample buffer and subjected to SDS-gel electrophoresis in 12.5%T polyacrylamide gels (20 μg/lane). Proteins were transferred to nitrocellulose membranes by semidry-blotting. Protein transfer was checked by staining the membranes with Ponceau S. After blocking, membranes were incubated incubated overnight at 4°C with a polyclonal antibody to the HA-tag (SC-805, Santa Cruz Biotechnology, Santa Cruz, CA, USA) diluted 1:200(v/v) in blocking solution. FGFR3 and FGFR2 antibodies were from Santa Cruz Biotechnology (#SC-123, #SC-122; Santa Cruz, CA, USA). The secondary antibody(Dianova, Hamburg) was applied at 1:5000 (v/v) dilution in blocking solution for 2 hours at room temperature. Color development was performed with 4-chloro-1-naphthol and H₂O₂.

Osteoblasts exhibit stellate morphology and display high levels of alkaline phosphatase, which was visualized by cellular staining with Sigma Fast BCIP/NBT (Sigma, St Louis, MO, USA). Proteoglycan-secreting chondrocytes were identified by staining with Alcian Blue at pH 2.5 and staining with Safranin O(Sigma, St. Louis, MO, USA). For collagen-immunohistochemistry cells were washed with PBS and fixed with methanol for 15 minutes at -20°C. Primary antibodies were diluted with 1% goat serum in PBS. Monoclonal anti-collagen II antibodies (Quartett Immunodiagnostika, Berlin, Germany, # 031502101) were diluted 1:50 (v/v) and monoclonal anti-collagen X antibodies (Quartett Immunodiagnostika, Berlin, Germany, # 031501005) 1:10 (v/v), respectively. Incubation was for 1 hour at room temperature followed by staining with Zymed HistoStain SP kit (Zymed Laboratories Inc., San Francisco, CA, USA), applying the manufacturer's protocol. A positive signal is indicated by a red precipitate of aminoethylcarbazole (AEC).

Before in vivo transplantation, samples (2-3×10⁶ cells)were mounted on individual type I collagen sponges (Colastat⁷#CP-3n, Vitaphore Corp., 2×2×4 mm) and transplanted into the abdominal muscle of female nude mice (4-8 weeks old). Before transplantation animals were anaesthetized intraperitoneally (i.p.) with ketamine-xylazine mixture (30 μl/per mouse) and injected i.p. with 5 mg/mouse of Cefamzolin(Cefamezin⁷, TEVA). Skin was swabbed with chlorhexidine gluconate 0.5% and cut in the middle abdominal area; an intramuscular pocket was formed in a rectal abdominal muscle and filled with the collagen sponge containing cells. Skin was sutured with surgical clips. For the detection of engrafted C3H10T1/2 cells the mice were killed at 10 days and 20 days after transplantation. Operated transplants were fixed in 4% paraformaldehyde cryoprotected with 5% sucrose overnight, embedded and frozen. Sections were prepared with a cryostat (Bright, model OTF) and stained with Haematoxylin and Eosin (HE), Alcian Blue and Safranin O.

Embryos were isolated from pregnant NMRI mice at day 18.5 post conception(d.p.c.). The embryos were fixed overnight with 4% paraformaldehyde in PBS at 4°C. 10 μm cryosections were mounted on aminopropyltriethoxysilane-coated slides and non-radioactive RNA in situ hybridizations were done as described(Bächner et al., 1998) and by following the instructions of the manufacturer(Roche, Mannheim). Briefly for hybridization, sense- and antisense RNA probes from a 1.8 kb murine Brachyury cDNA were used. For the generation of collagen 1a1 or collagen 2a1 the vector pMT7T3 was used, harbouring specific probes(Metsäranta et al., 1991). Hybridization was performed with 0.5-2 μg denatured riboprobe/ml) overnight at 65°C in a humid chamber. For digoxygenin (DIG)-detection, slides were blocked in 5× SSC, 0.1% Triton,20% FCS for 30 minutes following two washes with DIG-buffer 1 (100 mM Tris,150 mM NaCl, pH 7.6) for 10 minutes. Slides were incubated in anti-DIG-alkaline-phosphatase-coupled antibodies diluted 1:500 (v/v) in DIG-buffer 1 overnight in a humid chamber. Slides were washed with 0.1% Triton in DIG-buffer 1 for 2 hours with several changes of the washing solution and equilibrated in DIG-buffer 2 (100 mM Tris, 100 mM NaCl, 50 mM MgCl₂). Detection was performed using BM-purple substrate (Roche,Mannheim) in DIG-buffer 2 with 1 mM levamisole for 1-6 hours, depending on the probe. The reaction was stopped in TE-buffer and slides were incubated in 3%paraformaldehyde in PBS for 3 minutes, followed by 0.1 M glycine in PBS for 3 minutes, and washed three times in PBS for 3 minutes. Slides were counterstained with 0.5% Methylene Green in PBS for 1 minute, dehydrated in a graded alcohol series, air dried and mounted with Eukitt.

The mesenchymal stem cell line C3H10T1/2 has the capacity to undergo differentiation into all mesenchymal lineages, including osteogenesis and chondrogenesis. The responsiveness of C3H10T1/2 progenitors towards treatment with TGF-β and BMPs is often used to investigate mesenchymal cell determination and differentiation (e.g.Wang et al., 1993;Gazit et al., 1993;Ahrens et al., 1993;Hollnagel et al., 1997;Bächner et al., 1998). During a substractive screen for BMP-regulated genes in recombinant BMP2-expressing C3H10T1/2 (C3H10T1/2-BMP2) cells, we noted the upregulation of the Fibroblast Growth Factor Receptors 3 and 2 (FGFR3, FGFR2)at both the transcriptional and protein levels(Fig. 1A,C, respectively). These two receptor types exhibit different induction kinetics. FGFR3 is upregulated during early stages of cultivation in the stable C3H10T1/2-BMP2 line while FGFR2 shows a delayed response(Fig. 1A,C). The fast upregulation of FGFR3 seems to be due to an immediate response to BMP2, since exogenously added BMP2 mediated FGFR3 transcription in wild-type C3H10T1/2 cells in the presence of cycloheximide(Fig. 1B). In contrast to FGFR3 and FGFR2, FGFR1 is constitutively expressed in wild-type and C3H10T1/2-BMP2 cells (Fig. 1A) while FGFR4 does not show any significant rates of expression (data not shown). Since FGFs and their receptors are crucial modulators of chondrogenic development, we investigated whether the immediate BMP2-dependent upregulation of FGFR3 in C3H10T1/2 is involved in the onset of chondrogenic differentiation. Indeed,forced expression of the wild-type FGFR3 (FGFR3^WT) was sufficient for the development of morphologically distinct chondrocytes in C3H10T1/2-FGFR3^WT cells (Fig. 1D,E). Moreover, the constitutively active mutant FGFR3(Ach, G380R) possesses the same capacity (data not shown). The forced expression of FGFR3^WT in MSCs stimulates MAPK signaling in these cells, as documented by enhanced levels of ERK1 and ERK2 phosphorylation(Fig. 1D), leads to the development of histologically distinct chondrocytes and induces or increases expression of chondrogenic marker genes such as collagen 2a1, the PTH/PTHrP receptor and transcription factor Sox9(Fig. 1E). In C3H10T1/2 cells Sox9 is already expressed at substantial levels that are further upregulated by BMP2- and FGFR3, which is consistent with recent observations(Fig. 1E)(Zehentner et al., 1999;Murakami et al., 2000).

The immediate BMP2-dependent upregulation of FGFR3 in MSCs (C3H10T1/2) and the inherent capacity of this receptor to initiate chondrogenic development in these cells, prompted the development of a screen for FGFR3-regulated transcription factors. Among the transcription factors tested we observed that the T-box transcription factor Brachyury was upregulated in FGFR3-expressing C3H10T1/2 cells (see alsoFig. 5A). Furthermore, we noticed that Brachyury possesses a chondrogenic potential after recombinant expression in wild-type C3H10T1/2 cells (see below).

Brachyury was originally described as the first member of a family of transcription factors that harbors a T-box as the DNA-binding domain. In addition, it has been reported that FGF and/or TGF-β ligand-induced expression levels of T-box genes appear to be critical for their biological effects (O'Reilly et al.,1995; Tada et al.,1997). To investigate whether the FGFR3-dependent upregulation of the T-box factor Brachyury in C3H10T1/2 might play a role in chondrogenesis, we expressed Brachyury cDNA under the control of the murine phosphoglycerate kinase-1 (PGK-1) in the mesenchymal stem cell line C3H10T1/2 to allow moderate expression levels of Brachyury(C3H10T1/2-Brachyury). The recombinant expression of Brachyury cDNA under the control of the murine phosphoglycerate kinase-1 (PGK-1) in MSCs (Fig. 2A) gave rise to efficient chondrogenic differentiation, resulting in alkaline phosphatase-positive cells (beginning at day 4) and Alcian Blue-positive chondrocyte-like cells (at day 10 post-confluence;Fig. 2B). Three individual C3H10T1/2 clones were investigated for their chondrogenic potential and gave similar results. Immunohistochemistry confirmed the presence of the chondrocyte-specific collagen 2 but not of collagen X, which is typical of late stages of chondrocytic differentiation (hypertophic chondrocytes)(Fig. 2B, left). Major marker genes of chondrogenic and osteogenic development show a transient (collagen 2a1, PTH/PTHrP-receptor) or permanent (osteocalcin gene and the chondrogenic transcription factor Sox9) upregulation in C3H10T1/2-Brachyury, in comparison with C3H10T1/2 cells, that were stably transfected with an empty expression vector(Fig. 2C). Although induction of the osteocalcin gene indicates an osteogenic potential for C3H10T1/2-Brachyury, ectopic transplantation of these cells in murine intramuscular sites results exclusively in the massive formation of proliferating chondrocytes and cartilage(Fig. 2D). These ectopic transplantations were performed three times and in all cases the transplants developed chondrocytes and cartilage. After both 10 and 20 days, transplants exhibit the histological presence of proteoglycans (Alcian Blue, Safranin O)while bony elements or mineralized particles are not observed(Fig. 2D). After 20 days the ectopic implants show areas of extensive extracellular matrix production as visualized by histological analyses (Fig. 2D). The use of stronger viral promoters such as the LTR of the myeloproliferative virus (see Materials and Methods) resulted in increased cellular proliferation without the apparent formation of histologically distinct mesenchymal cell types (not shown).

We expected that Brachyury's DNA-binding domain (T-box, aa 1-229)without the associated regulatory domains (aa 230-436) should dominant-negatively (dn) interfere with endogenous Brachyury-mediated events in C3H10T1/2-BMP2 cells. A partial nuclear localization signal (NLS),which has been attributed to the T-box domain, should allow a substantial nuclear accumulation (Kispert et al.,1995). We confirmed the dominant-negative nature of the T-box domain in DNA cotransfection assays performed in HEK293 T cells. We used this particular cell line because expression levels are, in general, considerably higher in these cells than in C3H10T1/2. This cell line does not express Brachyury (data not shown). Exogenous Brachyurytransactivated a construct containing two copies of the consensus Brachyury binding element (BBE) oligonucleotide fused to a minimal HSV thymidine kinase (TK)-minimal promoter-CAT chimeric gene,pBBE(2×)-CAT5 (Fig. 3A). Indeed, contransfection of pBBE(2×)-CAT5 with a recombinant Brachyury-expressing vector resulted in a 25-fold activation, whereas an empty expression vector had no effect(Fig. 3A). Cotransfection of full-length Brachyury (Brachyury wt) with increasing amounts of an expression vector expressing the T-box domain (dnBrachyury)(1:1, 1:2, 1:3) led to a clear decrease in CAT (sevenfold). Exogenous dnBrachyury alone transactivated pBBE(2×)-CAT5 (BBE) only threefold.

The forced expression of the HA-tagged T-box domain(dnBrachyury) is observed throughout in vitro cultivation(Fig. 3B) and strongly interfered with the BMP2-mediated formation of alkaline phosphatase-positive osteoblast-like and Alcian Blue-stained chondrocyte-like cells in vitro(Fig. 3E). In vivo, in ectopic transplantations of C3H10T1/2-BMP2 in intramuscular sites,dnBrachyury allowed the development of connective tissue only(Fig. 3E). In addition, the chondrocyte-specific collagen 2al mRNA levels are more sensitive to the presence of dnBrachyury than mRNA levels of the distinct osteogenic marker osteocalcin. The latter is hardly affected, consistent with the idea that Brachyury possesses a predominantly chondrogenic capacity in this particular cell type. Interestingly, the BMP2-mediated transcriptional upregulation of FGFR3 in C3H10T1/2 is not obstructed by dnBrachyury,indicating that the immediate BMP2-mediated FGFR3 induction is independent of Brachyury or other T-Box factors(Fig. 3C,D). However,FGFR2-expression, which exhibits a delayed response in C3H10T1/2-BMP2 cells(Figs 1,3), displays a high sensitivity to dnBrachyury. BMP-mediated FGFR2 expression is almost completely suppressed by the dominant-negative acting T-box domain(Fig. 3C,D). This may indicate that that the presence of FGFR2 seems necessary for the osteo/chondrogenic differentiation in this mesenchymal progenitor line(Fig. 3C,D).

Furthermore, this suggests a hierarchy of FGFR-mediated signaling for chondrogenic development. FGFR3-dependent signaling is induced at first by BMP2 and, as a consequence, FGFR2-mediated signaling becomes active. Such a model is proposed in Fig. 7. This model predicts that a forced expression of dominant-negative FGFR3 would interfere with BMP2-mediated chondrogenesis and with FGFR2 and Brachyury expression. Indeed, an FGFR3-variant without the cytoplasmatic tyrosine-kinase domains downregulates BMP2-dependent mRNA expression levels of FGFR2 and Brachyury(Fig. 4B) and interferes with the histological manifestation of alkaline phosphatase- or Alcian Blue-positive chondrocyte-like cells (Fig. 4A).

During amphibian gastrulation, mesodermal Brachyury is involved in an autoregulatory loop with FGF that is present in the embryo(Kim et al., 1998). In C3H10T1/2 cells several FGF genes tested (FGF2, 4 and 9) were not Brachyury- or FGFR3- regulated (data not shown) and, therefore, are unlikely to be members of such a loop. However, a loop does seem to exist between FGFR3 and Brachyury, since forced expression of either one leads to the induction of the other in C3H10T1/2(Fig. 5A). These experiments indicate that after BMP2-mediated initiation of the chondrogenic lineage, the chondrogenic differentiation may advance for some time in a BMP2-independent fashion, maintained by the autoregulatory loop between FGFRs and FGF-regulated transcription factors such as the T-box factor Brachyury.

In an earlier study (Ju et al.,2000) we showed that BMP-mediated R-Smad signaling alone is not sufficient for cartilage development in C3H10T1/2 cells. Forced expression of Smad1 or the biologically active Smad1-MH2 domain is thereby able to mimic BMP2-mediated onset of osteogenic differentiation(Takeuchi et al., 2000). However, in contrast to osteogenic marker genes such as the osteocalcin gene,Smad1-MH2 domain-signaling is not sufficient to mimic BMP2-dependent FGFR3-and the concomitant Brachyury-gene induction(Fig. 5B). Other BMP-activated R-Smads such as Smad5 and Smad8 are also unable to mediate or to mimic BMP2-dependent FGFR3-induction in C3H10T1/2 cells (data not shown), indicating the existence of R-Smad-MH2-independent pathways for FGFR3 induction or,alternatively, cooperative activities of R-Smads with other transcription factors (Mazars et al.,2000).

From the results in mesenchymal stem cell line C3H10T1/2 we concluded that Brachyury might also play a role in skeletogenesis in vivo. Brachyury is expressed at high levels early in vertebrate embryonic development and is involved in gastrulation and in the dose-dependent determination of mesodermal cell fates (see Introduction and Discussion). After gastrulation, Brachyury expression is downregulated and persists in the notochord to the end of embryogenesis(Kispert and Herrmann, 1994). Comparative mRNA expression analysis of murine Brachyury (Bra),collagen 1a1 (col 1a1) and collagen 2a1 (col 2a1) in skeletal development(18.5 d.p.c.) indicates that Brachyury is expressed at significant levels in cartilage-forming cells of the intervertebral disks and in limb bud development (Fig. 6). Expression of Brachyury is enhanced in intervertebral disc development in the nucleus pulposus in 18.5 d.p.c. mouse embryos(Fig. 6Aa,d) confirming earlier reports (Wilkinson et al.,1990). Collagen 1a1 is expressed in the outer annulus (arrowheads in Fig. 6Ab,e), and collagen 2a1 in the cartilage primordium of the vertebrae(Fig. 6Ac,f). In transverse sections made at the level of the upper lumbar vertebra, expression of Brachyury is also detectable in distinct chondrogenic cells of the neural arch (Fig. 6Ah) whereas collagen 1a1 expression is maintained in the outer annulus(Fig. 6Ai), as is collagen 2a1 in the cartilage primordium (Fig. 6Aj). In murine limb bud development (18.5 d.p.c.; hind limb)expression of Brachyury is evident in distinct chondrogenic cells of the forming metatarsal bones (Fig. 6Ba-c). In contrast, collagen 1a1 is expressed in the outer periosteal layer (Fig. 6Bd-f)and collagen 2a1 expression is enhanced in differentiating chondrocytes(Fig. 6Bg-i). Interestingly,like in intervertebral disc formation, the expression of Brachyury is only evident in chondrocyte-like cells that do not express Col 2a1 indicating that Brachyury expression is upregulated in chondrogenic cells before or after collagen 2 expression.

As documented in this investigation, FGFR3-mediated signaling is sufficient for the onset of chondrogenesis in the mesenchymal stem cell line C3H10T1/2. This evidence is surprising since FGF signaling is better known for exerting negative influences on endochondral bone formation. In vivo, FGFR3 is expressed in the epiphyseal growth plate and in FGFR3^-/-mice, the zone of hypertrophic cartilage is enlarged by excessive bone elongation (Colvin et al.,1996; Deng et al.,1996). Similar studies with FGFR1, which in vivo is expressed in osteoblasts and hypertrophic chondrocytes, or with FGFR2, which in vivo is expressed in the perichondrium, have not been performed since targeted -/-mutations are embryonic lethal before bones form. Moreover, several investigations have demonstrated that FGFR3-mediated signaling interferes with chondrocytic proliferation and/or differentiation(Naski et al., 1998;Sahni et al., 1999;Henderson et al., 2000;Segev et al., 2000). These reports, however, are in contrast with investigations indicating that FGF treatment stimulates proliferation and/or differentiation in chondrocytes(Kato and Iwamoto, 1990;Hill et al., 1991;Wroblewski and Edwall-Arvidsson,1995; Legeai-Mallet et al.,1998). Consistent with the latter observations, overexpression of sprouty, an antagonist of FGF signaling in mice, is responsible for the development of chondrodysplasia and not for excessive bone growth, as would have been expected from the FGFR3^-/- phenotype(Minowada et al., 1999). This could be explained by the necessity of FGF-signaling for chondrocytic cell determination. Moreover, other studies indicate that inhibitory influences on chondrocytic proliferation/differentiation by FGFR3-dependent signaling predominantly occur in postnatal rather than in embryonic skeletal development(Naski et al., 1998). Also, a decisive involvement of FGFR3- or MAPK-signaling with osteo/chondrogenic development has been observed recently in other studies(Lou et al., 2000;Murakami et al., 2000). Therefore, the in vivo or in vitro developmental state of chondrocytic cells or cell lines may define whether FGF exerts a stimulatory or an inhibitory potential on chondrocyte proliferation and/or differentiation.

FGFR3-mediated signaling initiated chondrogenic differentiation in mesenchymal stem cell line C3H10T1/2, and the T-box containing transcription factor Brachyury fulfilled the requirements as a target for FGF-mediated chondrogenesis (Figs1,2,3,4,5). Although expression of Brachyury in C3H10T1/2 upregulates mRNA levels of osteocalcin, which is a specific marker for late osteogenesis, bony centers or mineralized particles have never been observed in ectopic implantations. This might indicate that Brachyury expression is sufficient for chondrogenesis rather than for osteogenesis. However, Brachyury could mediate differentiation of bipotential osteo/chondrogenic progenitors where the chondrogenic lineage prevails, eventually. In addition we noted the absence of enchondral bone formation in ectopic transplants, which is in contrast to transplanted C3H10T1/2 cells expressing BMPs. It is known that BMPs may contribute to enchondral bone formation by their ability to induce VEGF. This may stimulate angiogenesis and vasculogenesis, which are prerequisites for the presence of phagocytic and osteogenic cells to replace hypertrophic cartilage by bone (Yeh and Lee, 1999;Kozawa et al., 2001). It seems that Brachyury is not involved in these events.

The model of BMP2-dependent osteo/chondrogenic development in mesenchymal stem cells C3H10T1/2 suggests that BMP2 predominantly initiates and determines osteo/chondrogenic development via BMPR-IA(Fig. 7). In C3H10T1/2 cells BMPR-IB only exerts minor functions (our unpublished observations). Then, the BMP2-mediated R-Smad signaling regulates predominantly osteogenesis and not chondrogenesis (Ju et al.,2000). R-Smad-signaling seems to be required during the entire osteoblast developmental sequence, possibly also by recruiting Cbfa1 into a heteromeric complex with activated Smad proteins(Hanai et al., 1999). In contrast, BMP2-dependent determination of chondrogenesis in the mesenchymal stem cell line C3H10T1/2 seems to involve the immediate upregulation of FGFR3 by an R-Smad-independent mechanism.

The signaling pathway for this BMP2-dependent immediate induction of FGFR3 is not clear; however, recently it has been demonstrated that TGF-βsignaling activates MAPK-pathways through the small GTP binding proteins Cdc42 and Rac1, which leads to a cooperative effect between Smad2/3 and AP1 for gene activation. Similar mechanisms may play a role for BMP2-mediated activation of FGFR3 (Mazars et al., 2000;Dennler et al., 2000).

As discussed above, after this triggering event, chondrogenesis seems to be predominantly controlled by BMP-independent mechanisms. The interference of dominant-negative Brachyury (i.e. the T-box domain) with BMP2-mediated FGFR2- but not with FGFR3- expression indicates that FGFR2 expression is dependent on Brachyury, and that an autoregulatory loop may be initiated by FGFR3, and seems to be maintained between FGFR3/FGFR2 and Brachyury (Fig. 3A). In contrast, mRNA-levels for the transcription factor Sox9 are not significantly altered by dnBrachyury expression. Since both factors, Sox9 and Brachyury, are regulated by FGF signaling(Fig. 1,Fig. 2C) it is conceivable that different FGF-mediated signalling pathways activate Brachyury or Sox9, since FGFR-mediated MAPK-signaling also upregulates mRNA levels of the chondrogenic transcription factor Sox9(Murakami et al., 2000). This suggests that FGFs and FGF-receptors, the MAPK pathway, Sox9 and T-box factor(s) are essential components for the BMP-dependent onset of chondrogenesis (Fig. 7). Consequently, a block of the FGFR3-mediated signaling cascade should interfere with BMP2-mediated FGFR2 and Brachyury induction as well as with cartilage formation. Forced expression of dominant-negative FGFR3 (dnFGFR3) in C3H10T1/2-BMP2 cells fulfils these criteria, indeed. Osteo/chondrogenic differentiation is severely reduced on a histological basis(Fig. 4A) as well as on expression levels of FGFR2 and Brachyury(Fig. 4B). Interestingly, a role for another T-box factor (Tbx2) in skeletal cell development was postulated recently (Chen et al.,2001).

Brachyury expression in vivo is induced and carefully controlled during gastrulation. Thereafter Brachyury expression is downregulated but persists in the tailbud and the notochord. Brachyury expression in vivo is controlled by FGF, Wnt- and activin signaling pathways. Here in C3H10T1/2 an FGFR3-dependent induction of Brachyury transcription is monitored. What could be the mechanism for FGF-dependent activation of the Brachyury gene in the C3H10T1/2 cellular system?

One way of Brachyury induction that has been discussed is a synergism between FGF-signaling cascades and SRF (Serum Response Factor),because SRF is downstream of the MAPK signaling cascade and mouse embryos lacking functional SRF protein do not form mesoderm and do not express Brachyury (Arsenian et al.,1998). However, a constitutively active form of SRF does not induce expression of Xbra (Panitz et al., 1998) and therefore SRF activity in expression of Brachyury seems to be indirect. Also, based on in vitro and in vivo studies it has been suggested that Brachyury induction may be mediated by Wnt-signaling (Arnold et al.,2000; Smith et al.,2000). However, it has recently been documented that Wnt-signaling is involved in the maintenance but not in the induction of Brachyuryand mesoderm synthesis (Galceran et al.,2001).

So we think that among the likely mechanisms leading to the early upregulation of Brachyury in mesenchymal progenitors, C3H10T1/2 may be the derepression of members of the δEF1 family of transcriptional repressors. It has been demonstrated that point mutations disrupting δEF1 binding-sites in the Xbra promoter(Remacle et al., 1999) change the mesodermal expression of reporter constructs and result in widespread ectodermal and endodermal misexpression(Lerchner et al., 2000). Of the δEF1 family members, SIP1 is able to interact with activated Smad proteins and to interfere with transcription of endogenous Xbra(Verschueren et al.,1999).

It seems highly conceivable that SIP1 is bound to its binding sites in the Brachyury promoter in C3H10T1/2 since SIP1 has high basal levels of expression (data not shown). Although not yet entirely clear, SIP-1 could dissociate from DNA when associated with an activated Smad molecule. This would call for two signaling cascades involved in the early induction of Brachyury, one of which is involved in BMP/Smad-mediated derepression of the Brachyury promoter and the other for the onset of Brachyury transcription. Since FGF-signaling alone seems to be sufficient for Brachyury induction in C3H10T1/2, one might envisage that endogenously expressed TGF-β family members might be sufficient to enable a significant derepression of the Brachyury promoter but that additional (i.e. FGF-)signaling cascades are needed to elicit a significant Brachyury synthesis.

Also not seen during the cultivation of recombinant C3H10T1/2-FGFR3 cells,a downregulation of Brachyury transcription in vivo could be mediated by an FGFR3-dependent upregulation of sprouty gene expression. The latter inhibits FGFR-mediated signalling(Minowada et al., 1999;Wakioka et al., 2001) and would interfere with FGFR3-mediated Brachyury gene expression,eventually.

Does Brachyury, which has been extensively characterized in early embryonic development, also play a role later in the determination/differentiation of chondrogenic tissue in vivo? It has been shown that Brachyury is highly expressed during gastrulation, where it plays a decisive role in the generation of undifferentiated mesoderm, and,thereafter, Brachyury expression is downregulated (reviewed inPapaioannou and Silver, 1998;Smith, 1999). Here, we show that in situ hybridizations detect substantial levels of BrachyurymRNA in maturing chondrogenic tissue at late stages of embryonic development(18.5 d.p.c.) during spine formation and especially in the intervertebral discs (Fig. 6). The part of the disc structure that appears to play an important role in its function is the nucleus pulposus, which is an avascular gelatinous tissue located between the endplates of the vertebral bodies and the inner lamellae of the annulus fibrosus. The integrity of the gelatinous matrix of the nucleus pulposus seems to be essential for the load bearing of the discs. The nucleus pulposus cells generate only poorly developed cartilage, with low amounts of collagens 2a1 and 1a1 but a high concentration of aggrecan. In situ analyses showed that collagen 2a1 is expressed in chondrocytes surrounding the nucleus pulposus while collagen 1a1 is expressed in the flanking outer annulus(Fig. 6). Brachyury is highly expressed in the nucleus pulposus cells that are of notochordal origin,consistent with it being a marker gene for axial (notochordal) mesoderm. Notochordal expression of Brachyury is regulated by an enhancer that is not yet mapped (Lerchner et al.,2000). In the notochord-derived nucleus pulposus cells it cannot be excluded that a similar element is used for driving T-gene expression. The high level of Brachyury expression and its potential chondrogenic capacity might indicate that this ensures the cartilaginous character of the nucleus pulposus. In contrast to the fibrous nature of the inner annulus, Brachyury is also significantly upregulated in distinct chondrocytes,forming hyaline cartilage of the vertebral body(Fig. 6A). An upregulation of Brachyury expression in chondrocytes of the hind limb metatarsals is also observed. There, Brachyury seems to be upregulated in a set of chondrogenic cells at an early state of maturation, i.e. in cells where collagen 2a1 is not yet expressed (Fig. 6B). Alternatively, these cells could be pre-hypertrophic chondrocytes. Expression of Brachyury in pre-hypertrophic chondrocytes would be consistent with the suggestion that this gene is upregulated in response to FGFR3, whose gene is expressed at high levels before chondrocytes undergo hypertrophy. It would also be consistent with the observations that ectopic expression of Brachyury in C3H10T1/2 cells resulted in production of alkaline phosphatase, and activation of the genes for PTH/PTHrP-receptor and osteocalcin.

Are there then genetic indications that Brachyury is involved in the differentiation/patterning of skeletal elements? Interestingly, Bennett(Bennett, 1958) demonstrated that cartilage formation is not induced in somites from T/T embryos. In addition, Brachyury mutant alleles have been described that counteract the activity of the wild-type protein(Kispert, 1995). These mutations carry truncations in the C terminus and appear to act as dominant-negatives. Consistent with the in situ expression analyses here(Fig. 6A), heterozygous dominant-negative Brachyury-mutant mice exhibit skeletal phenotypes such as complete or near-complete absence of the tail, absence of the odontoid process of the axis, absence of the nuclei pulposi of the intervertebral discs, a tendency for rib and vertebral fusions, and an occasional absence of presacral vertebrae and ribs, suggesting a later role for the Brachyury gene in morphogenesis and function of notochord-derived tissue (Wilkinson et al.,1990; Kispert and Herrmann,1994).

A key step to understanding Brachyury function is the identification of target genes. Two major cloning strategies have led to the isolation of Brachyury-regulated downstream targets in Xenopus and ascidians involved in gastrulation and switching on the notochord (Tada et al., 1998;Takahashi et al., 1999;Saka et al., 2000). Interestingly, mRNAs have been identified that seem to be upregulated by ectopic Brachyury expression such as cartilage-specific collagen 2a1 and collagen 11a1 (Takahashi et al.,1999).

It has been suggested that a hypothetical competence factor (CF) acts to promote the chondrogenic response to BMP signaling during the generation of axial and appendicular cartilage (Murtaugh et al., 1999). In the absence of such a signal, presomitic mesoderm assumes a lateral plate fate as its response to BMPs. Sonic hedgehog(Shh) can provide the competence to respond to BMP by differentiating into chondrocytes. Also limb bud mesenchymal cells assume the competence to convert to chondrocytes upon BMP treatment, otherwise exogenous BMPs induce apoptosis. As in the somite, it is suggested that a hypothetical factor (CF) acts to promote chondrogenic response to BMP signalling. The nature of signal which induces competence in the limb bud is unclear.

Although the nature of the CF is unknown it might be temptative to speculate that transcription factors of the T-box family could exert a CF-like role on transcriptional level. Brachyury is expressed in axial/paraxial mesodermal structures during early embryonic development while others that could exert such a role are expressed later in the limb bud. The finding mentioned above, that in somite/notochord cocultivation experiments cartilage formation may be induced in normal but not in somites from T/T embryos (Bennett, 1958), could indicate a CF-like role for Brachyury. It would also be consistent with the skeletal phenotypes in Brachyury mutant mice described above.

The authors are indebted to Drs B. Herrmann (Max-Planck-Institute,Freiburg) for proving Brachyury cDNA and A. Yayon (Weizmann Institute, Rehovot, Israel) for FGFR3 cDNA. We would like to thank Drs D. Huylebroeck and P. Tylzanowski (Celgen, Leuven, Belgium) for intensive and stimulating discussions and for editing the manuscript.