Inhibition of chondrocyte differentiation in vitro by constitutive and inducible overexpression of the c-fos proto-oncogene

The control of proliferation and differentiation of chondrogenic cells is central to the coordinated development of the vertebrate skeleton. Vertebrate long bones develop by the process of endochondral ossification, which is initiated in the embryo with the condensation of undifferentiated mesenchymal cells and progresses with their commitment and differentiation into chondrogenic cells. By late embryonic development the epiphyseal growth plate has developed with distinguishable, well-organized and spatially distinct zones of resting, proliferating and post-proliferative hypertrophic chondrocytes (for review see Erlebacher et al., 1995). Chondrocyte proliferation and differentiation continue at the growth plate through juvenile growth and are partly responsible for regulating the rate of expansion of the long bones. The consequences of failure of the regulation of chondrocyte growth and differentiation can be seen in human chondrodysplasias. Interestingly, specific genetic defects associated with several chondrodysplasias have been mapped to genes involved in endocrine/paracrine signalling. For example, the fibroblast growth factor (FGF) receptor 3 (FGFR3) is disrupted in achondroplasia (Shiang et al., 1994), the transforming growth factor β (TGF-β) superfamily member CDMP-1 in Hunter-Thomson-type chondrodysplasia (Thomas et al., 1996) and the common parathyroid hormone (PTH) and PTH-related peptide (PTHrP) receptor, PTH1R, in Jansen-type and Blomstrand-type chondrodysplasias (Schipani et al., 1995; Karaplis et al., 1998; Jobert et al., 1998). Moreover, analysis of chondrodysplastic mouse models allows more detailed definition of the roles of these pathways. Thus, genetic disruption of the PTH1R gene (Karaplis et al., 1994; Amizuka et al., 1994), overexpression of PTHrP or PTH1R (Weir et al., 1996; Schipani et al., 1997), disruption of FGFR3 (Deng et al., 1996), or mutations of members of the TGF-β superfamily (Kingsley et al., 1992; Storm et al., 1994) all result in altered chondrocyte proliferation and/or differentiation. However, less is known about the molecular events downstream of endocrine/paracrine signalling and thus it is of interest that genetic disruption of several nuclear proteins, such as (i) transcription factors like ATF-2 (Reimold et al., 1996), the proto-oncogene Ets-2 (Sumarsono et al., 1996), and the runt-domain protein cbfa-1 (Otto et al., 1997; Komori et al., 1997; Inada et al., 1999), (ii) cell cycle control proteins such as the retinoblastoma (Rb)-related proteins p107 and p130 (Cobrinik et al., 1996) and (iii) the cyclin dependent kinase inhibitor (CKI) p57 (Zhang et al., 1997), also result in skeletal abnormalities in mice, with specific effects on chondrogenic cells.

One additional nuclear transcription factor that has a proven role in chondrocyte differentiation is c-Fos. The proto-oncogene c-fos was first identified as the cellular homologue of the v-fos oncogene from the FBJ- and FBR-murine sarcoma viruses. The c-Fos gene product is a member of the AP-1 family of transcription factors, which includes the other Fos-related (FosB, Fra-1, Fra-2) and Jun-related (c-Jun, JunB, JunD) proteins (for review see Angel and Karin, 1991). The functional activity of c-Fos, and of the other Fos family members is dependent on the formation of heterodimers with proteins of the Jun family. Fos proteins are unable to form homodimers, but Jun proteins can additionally dimerise with each other and with the related ATF-2 transcription factor. Dimerised complexes can then modulate transcription by binding to AP-1 consensus binding sites in the promoter regions of target genes (Angel and Karin, 1991). c-fos and other AP-1 family members have been shown to display an immediate early gene pattern of expression being rapidly and transiently induced by external mitogenic signals such as serum stimulation, suggesting a role for c-Fos in signal transduction of mitogenic stimuli (see also Morgan and Curran, 1991). The downstream effects of c-Fos induction are numerous and in addition to roles in proliferation, transformation and apoptosis, c-Fos has been implicated in the differentiation of several cell types: Clear c-Fos-dependent differentiation effects have been demonstrated by the overexpression or inactivation of c-fos in teratocarcinoma stem cells (Müller and Wagner, 1984), adipocytes (Abbott and Holt, 1997) and cells of the osteoclast/macrophage lineage (Grigoriadis et al., 1994).

Although c-Fos expression can be induced in most tissues, gain-of-function and loss-of-function studies have demonstrated a specific role in bone/cartilage biology (Grigoriadis et al., 1995). The earliest embryonic expression of c-Fos is restricted specifically to the growth regions of fetal bones (Dony and Gruss, 1987; DeTogni et al., 1988) and post-natal expression has been detected in osteoblasts and growth plate chondrocytes (Grigoriadis et al., 1993; Lee et al., 1994; Sunters et al., 1998). In transgenic mice expressing exogenous c-fos from a ubiquitous promoter the primary phenotype is the transformation of osteoblasts leading to bone tumour formation (Grigoriadis et al., 1993), whilst inactivation of the c-fos gene causes severe osteopetrosis with a complete block in osteoclast differentiation (Johnson et al., 1992; Wang et al., 1992; Grigoriadis et al., 1994). A specific role for c-Fos in chondrocytes in vivo is highlighted by several further lines of evidence. Infection of embryonic chick limb buds with c-fos expressing retroviruses results in truncation of the long bones due to severe retardation of the differentiation of proliferating chondrocytes into mature hypertrophic chondrocytes (Watanabe et al., 1997). Furthermore, c-fos-overexpressing embryonic stem (ES) cell chimeric mice demonstrate the transformation of chondrocytes and the development of chondrosarcomas, and transformed, type II collagen (coll II) expressing cell lines derived from these tumours failed to progress to hypertrophy (Wang et al., 1991, 1993). Finally, c-fos knockout mice display shortened limbs and disrupted growth plate architecture, with a significantly depleted zone of proliferating cells and an expanded zone of hypertrophic cells (Wang et al., 1992).

These in vivo models indicate a clear role for c-Fos in chondrogenesis in the context of an intact animal, with overexpression of c-fos inhibiting differentiation and endochondral progression and the absence of c-fos apparently accelerating this process. However, further cellular and molecular dissection of such a role by in vitro analysis has proved more difficult due to the relative difficulty in establishing stable, continuously growing, non-transformed chondrogenic cell lines which differentiate with reproducible kinetics. Some useful cell lines have indeed been isolated which display varying degrees of chondrogenic potential (e.g. Grigoriadis et al., 1989, 1996; Atsumi et al., 1990; Bernier and Goltzman, 1993; Lefebvre et al., 1995). However, no systematic approach to analysing the effect of c-fos overexpression on the differentiation of in vitro chondrocyte cultures has yet been undertaken. We have used an established model of chondrocyte differentiation in which the ATDC5 embryonal carcinoma derived cell line can be induced to undergo a reproducible, time-dependent in vitro progression from early precursors to hypertrophic chondrocytes (Atsumi et al., 1990; Shukunami et al., 1997). Using this model we have shown that stable overexpression of c-fos, either constitutively or by using an inducible promoter clearly demonstrates an inhibition of chondrogenic progression and that phenotypic and molecular characterisation demonstrates that c-Fos directly regulates chondrocyte differentiation.

The expression vector pJMF2 was obtained as a gift from Dr J. Feingold (UCHC, Farmington, CT), generated as a one-component construct based on the Tet-off system of Gossen and Bujard (1992). Briefly, this vector had been constructed by the removal of the promoter region of the expression vector pREP9 (Invitrogen, Groningen, Netherlands) and its replacement with a tetracycline-repressed transactivator (tTA) expression cassette (P_CMVtTA), a minimal tTA responsive promoter (tetO), and a multiple cloning site upstream of the SV40 poly A site of pREP9 (Lang and Feingold, 1996). The c-fos cDNA was excised by BamHI as a 1.35 kb fragment from pX-c-fos (Superti-Furga et al., 1991) and ligated into the BamHI site upstream of the tetO promoter region of pJMF2. Clones were isolated and the orientation of the c-fos cDNA insert was determined using the BglII and NcoI sites within the c-fos cDNA sequence. The size of the c-fos cDNA-SV40 poly A transcript is estimated to be ∽ 1.8 kb (Fig. 1). The construction of the vector p76/21 containing MT-c-fosLTR has been previously described and gives rise to 2 exogenous transcripts of 3.0 kb and 2.0 kb (Rüther et al., 1985; Wang et al., 1991).

ATDC5 cells were obtained from the RIKEN cell bank (Japan) and maintained as described by Atsumi et al. (1990). Cells were cultured in a standard medium of DMEM/Hams’ F12 (1:1) (Gibco BRL, Paisley, UK) supplemented with 5% FCS (M. D. Meldrum, Hants, UK), 10 μg/ml bovine insulin, 10 μg/ml human transferrin and 3×10⁻⁸ M sodium selenite (ITS) (Sigma, Poole, UK) and antibiotics (50 U/ml penicillin and 50 μg/ml streptomycin; Gibco BRL). Stable transfections were performed using either SuperFect (Qiagen, Crawley, UK; DT8 and DT7.1) or Effectene (Qiagen; DT12.4) according to the manufacturer’s instructions. After transfection, cells were plated at varying densities and transfected clones selected in standard media supplemented with 0.5 mg/ml G418 (Gibco BRL) and isolated by ring-cloning. Additionally, in transfections with pJMF2-c-fos the media were supplemented with 1 μg/ml tetracycline (Tc) (Sigma) throughout the selection, expansion and maintenance of clones, in order to repress exogenous c-fos expression. Exogenous c-fos expression levels of all clones were determined by northern blot analyses following Tc withdrawal. In all, 6 stable clones transfected with p76/21 were analysed (DT8 series) and 22 pJMF-c-fos-transfected clones, of which 2 were strongly positive (DT7.1 and DT12.4), and one was weakly positive (DT12.5, data not shown).

For all cell biological analyses cells were plated at a density of 6×10³ cells/cm² in 6-well plates (Nunc, Roskilde, Denmark) unless otherwise stated and the standard media were replaced every two days for 21 or 30 days. For differentiation assays cultures were fixed in 4% paraformaldehyde in PBS and stained with 0.25% Alcian blue at Ph 0.75. For quantification, stain was solubilised in 4 M guanidine HCl (pH 1.5) and the optical density measured at 595 nm. For expression analyses cells were plated in 140 cm culture dishes (Nunc), and were harvested by trypsinisation at least 24 hours after feeding. For longer term cultures after 21 days, the media were changed to αMEM (Gibco BRL) with the same additives to induce hypertrophic differentiation (Shukunami et al., 1997).

Analyses of pJMF2-c-fos clones were performed in the presence or absence of 1 μg/ml Tc (± Tc) as stated. For pulse experiments, induction of c-fos was performed by washing cell cultures 3-4 times in PBS to remove Tc, prior to feeding with Tc-deficient media. Cultures were ‘pulsed’ for 4-day time periods as induction experiments suggested that maximal c-fos expression is not achieved until 48-72 hours after the removal of Tc (data not shown). Limiting dilution analysis was performed by plating DT12.4 cells in 96-well plates with densities per well as indicated. Cells were cultured for 21 days ± Tc, fixed and stained with Alcian blue as stated. The fraction of wells without cartilage was plotted against cell density. From the equation F₀=e−x, where x is the number of chondroprogenitors per well, the probability of no nodule at the 1/e level (1/e=0.37, as indicated in Fig. 6C) determines the frequency of chondroprogenitors in each population (see also Grigoriadis et al., 1996).

For proliferation and apoptosis assays cells were plated at standard density and cultured for 24 hours in full medium (5% FCS), then washed in PBS and fed with low serum media ± Tc as indicated. For proliferation assays cells were trypsinised every 48 hours and counted using a haemocytometer. For apoptosis assays, after a further 24 hours, non-adherent and adherent cells (after trypsinisation) were cytospun onto TESPA-coated slides, fixed in 4% paraformaldehyde in PBS and stained with haematoxylin and eosin. The cells were viewed microscopically and the proportion of morphologically apoptotic cells (condensed or fragmented nuclei) was calculated by counting at least 300 cells from random fields, from each of triplicate cell cultures. The frequency of apoptosis was also measured by staining cytospin preparations with Acridine Orange and by TUNEL assays using standard protocols. Similar results were obtained by all methods (data not shown). All apoptosis experiments were carried out in the absence of ITS except in 5% FCS, in order to mimic differentiation assay conditions. However, c-fos expression caused no difference in apoptosis rates at 5% FCS in the absence of ITS (data not shown).

Cell cultures for expression analyses were harvested by trypsinisation either at confluence or at the day indicated and cell pellets were stored at −80°C prior to processing. Poly(A)⁺ RNA isolation, northern blot analysis and hybridisation in Church buffer were performed as previously described (Wang et al., 1991). Probes were labelled with [³²P]dCTP (NEN, Boston, MA) to a specific activity of 4×10⁸ cpm/ μg DNA using Ready-To-Go oligo-labelling beads (Amersham Pharmacia, St Albans, UK). The following probes were used: v-fos/ fox (0.8 kb), which hybridises to c-fos and additionally, to c-fox, an abundant RNA found in mouse tissues but unrelated to c-fos (see also Wang et al., 1991), which was used here as a loading control; murine probes for fosB (0.27 kb), fra-1 (0.23 kb), fra-2 (0.24 kb), c-jun (0.45 kb), junB (0.48 kb), and junD (0.3 kb) were all obtained as gifts from Dr R. Bravo (Bristol-Myers Squibb, Princeton, NJ); murine coll II (0.4 kb), coll X (1.2 kb) and aggrecan (0.47 kb) were gifts from Dr E.Vuorio (University of Turku, Finland); murine PTH1R (2 kb), from Dr B. Lanske (MPI, Martinsried, Germany); murine Sox-9 (0.52 kb), from Dr P. T. Sharpe (Guy’s Hospital, London, UK); and human GAPDH (1 kb), from Dr J. Beresford (University of Bath, UK).

Total cellular proteins were extracted for use in western blotting and electromobility shift analysis (EMSA) studies. Briefly, cells were washed twice in ice-cold PBS then lysed in Tween lysis buffer (50 mM HEPES, 1 mM EDTA, 2.5 mM EGTA, 150 mM NaCl, 1 Mm DTT, 0.1% Tween-20, 1 mM NaF, 0.1 mM NaVO_4, 100 μg/ml PMSF, 1 μg/ml aprotinin, and 1 μg/ml leupeptin, pH 8.0). For western blot analysis 50 μg of protein/lane was resolved on a 7.5% SDS-PAGE gel (National Diagnostics, Atlanta, GA) and proteins were transferred onto Immobilon P PVDF membranes (Millipore, Watford, UK), which were incubated in block buffer (5% low fat milk powder, 2% bovine serum albumin (BSA) in Tris-buffered saline (TBS)). Blocked filters were incubated with a 1:1000 dilution of a rabbit polyclonal anti c-Fos antibody (sc-52 Santa Cruz, Santa Cruz, CA) in block buffer for 1 hour, and subsequently with a 1:1000 dilution of a polyclonal goat anti-rabbit antibody conjugated to horseradish peroxidase (Dako, Denmark) in block buffer. c-Fos was visualised using enhanced chemiluminescence (ECL) (Amersham Pharmacia).

For electromobility shift assay (EMSA), 5 μg extracted protein was incubated with 1-2×10⁵ cpm (∽ 35 fmol) of [³²P]ATP end-labelled AP-1 specific oligo (5′-GCGTTGATGAGTCAGCCGGAA-3′ – Promega, Southampton, UK), in band shift buffer (50 mM NaCl, 5 mM DTT, 10 mM Tris-HCl (pH 7.5), 4% glycerol, 0.5 mM EDTA, 1 mM MgCl₂ and 10 μg/ml poly(dI-dC)) for 1 hour at 25°C. For supershift analyses, complexes were subsequently incubated with 10 μg of c-Fos specific antibody (sc-52X-Santa Cruz), pan-Fos specific antibody (sc-523X – Santa Cruz), or non-c-Fos reactive rabbit immunoglobulins (Dako) for 1 hour at 4°C. Complexes were resolved by electrophoresis on 4% non-denaturing polyacrylamide gels in 0.5× TBE. Gels were dried and exposed for autoradiography.

We have analysed the expression of cartilage marker genes, and of c-fos and c-jun family members, throughout the in vitro differentiation of ATDC5 cells. Expression was analysed at the following stages of cartilage differentiation: pre-confluent cultures (day 3), confluent cultures (day 5), the appearance of visible multi-layered nodules (day 11), matrix accumulation (day 15) and the onset of chondrocyte hypertrophy (day 21). Additionally, some cultures were continued until day 30 to demonstrate the further accumulation of hypertrophic cells (Shukunami et al., 1997). Type II collagen (coll II) is the predominant collagenous protein of cartilage matrix and its expression is highly specific for chondrogenic cells. Coll II expression was detectable in early stage cultures but levels showed a significant increase from day 11, reached a maximum at day 21, and decreased slightly with hypertrophic differentiation (day 30; Fig. 2). Aggrecan, which is the major non-collagenous protein in cartilage matrix, showed an almost identical temporal expression profile to that of coll II (Fig. 2). Expression of type X collagen (coll X), a marker for hypertrophic chondrocytes, was undetectable in early stage ATDC5 cultures, but was observed at low levels from day 11, and increased to a maximum in day 30 cultures (Fig. 2). The HMG-domain gene Sox-9, a known marker of early chondrogenic commitment (Wright et al., 1995), was expressed at detectable levels in proliferating ATDC5 cells (day 3) and at similar levels throughout the culture period (Fig. 2). We also observed a continuous increase over time in levels of mRNA expression for PTH1R (Fig. 2) which has previously been shown to be a marker for chondrogenic progression of ATDC5 cells (Shukunami et al., 1996). Additionally, the visible development of cuboidal cells into multi-layered nodules, and the deposition of cartilage matrix, as assessed by Alcian blue staining (data not shown), accelerated significantly from day 11 in parallel with the expression of coll II and aggrecan. Thus, as assessed by both gene expression and morphological criteria, the ATDC5 cells used in this study recapitulate the well-established sequence of cartilage differentiation from chondroprogenitor cells to hypertrophic chondrocytes.

Within the context of this in vitro differentiation, levels of expression of mRNA for the c-fos and c-jun family members were assessed. c-fos expression was detectable at all time points albeit at low levels with a slight increase at later time points (Fig. 3A), and this was also seen at the protein level (Fig. 3C). Similarly, levels of fosB mRNA appeared generally absent throughout differentiation but became detectable at later stages (Fig. 3A). In contrast, fra-1 expression was detectable throughout differentiation at significant levels and fra-2 showed differentiation-dependent variation in expression, with levels elevated prior to confluence (day 3) and during matrix deposition (day 15; Fig. 3A). Members of the c-jun family of genes were readily expressed at significant levels throughout differentiation (Fig. 3B). Levels of junB mRNA were initially elevated (day 3), but in post-confluent cultures both c-jun and junB showed similar profiles with levels increasing until day 15, then decreasing during hypertrophic differentiation (Fig. 3B). In fact, c-jun mRNA levels were further decreased at day 30 (data not shown). Expression of junD was initially low (day 3) but maintained steady-state levels throughout the remainder of the time course. To investigate the functionality of the different Fos and Jun proteins expressed throughout ATDC5 cell differentiation AP-1 binding complexes were assessed by electro-mobility shift assay (EMSA). AP-1 complexes were detected at all time points (Fig. 3D), and the specificity of binding was determined by oligo competition studies (data not shown). Supershift analyses demonstrated that no c-Fos-specific supershift was detected at any of the time points, but was seen in a stable clone constitutively overexpressing c-Fos (clone 8.6; see also Fig. 4A). In contrast an antibody that recognises all Fos family members (pan-Fos) retarded all AP-1 complexes, whereas non-specific rabbit immunoglobulins failed to elicit any supershift (Fig. 3D). This demonstrates that all active AP-1 complexes act as Fos-family:Jun-family heterodimers and Jun:Jun dimers are either low or absent. Since c-Fos and FosB expression were apparently low, the AP-1 complexes of ATDC5 cells most likely consist of Fra proteins dimerised with specific Jun protein partners.

Previous evidence from in vivo studies has suggested that elevated c-fos expression may have an effect on the differentiation of chondrocytes (Wang et al., 1991; Watanabe et al., 1997). We initially investigated this in vitro by transfection of ATDC5 cells with the construct MT-c-fosLTR (Rüther et al., 1985), which has previously been shown to express high levels of exogenous c-fos in chondrocytes in vivo (Wang et al., 1991, 1993). Six stable clones were selected in G418 and analysed for gene expression and differentiation potential. Three of the clones (DT8.2, DT8.5, DT8.6) expressed the transgene at high levels, whilst three (DT8.1, DT8.4, DT8.8) expressed no detectable exogenous c-fos transcripts (Fig. 4). All of the c-fos expressing clones failed to demonstrate significant levels of cartilage nodule formation implying that c-fos expression is inhibitory to chondrogenic differentiation (Fig. 4). One c-fos-negative clone (DT8.4) did differentiate to levels comparable to wild-type ATDC5 cells, although others (DT8.1, DT8.8) did not (Fig. 4). This may reflect clonal variation and we investigated the possible molecular basis for this by analysing markers of chondrocyte differentiation: Sox-9, was expressed in all clones at similar levels, but coll II was variable with highest expression in clones DT8.4 and DT8.6 (Fig. 4). Interestingly, these clones were the only two that demonstrated significant nodule formation and they also displayed a clear negative correlation between levels of exogenous c-fos and the extent of differentiation.

To overcome the problem of clonal variation we sought to transfect ATDC5 cells with a regulatable expression construct, where levels of c-fos could be varied within single clones and therefore the effect of high and low expression can be assessed on a stable background of known differentiation potential. To this end, we have cloned the murine c-fos cDNA into pJMF2 (Lang and Feingold, 1996), a single vector system based on the Tet-off system as originally derived by Gossen and Bujard (1992), where expression of a cloned gene is repressed in the presence of tetracycline (Tc), but can be induced to high levels upon withdrawal of Tc from the culture medium of stably transfected cells. We transfected this construct (pJMF2-c-fos (Fig. 1)) into ATDC5 cells and selected clones in G418 and in the continuous presence of Tc. Two clones (DT7.1 and DT12.4) derived from independent transfections showed high levels of exogenous c-Fos induction upon withdrawal of Tc: western blot analyses demonstrated strong induction of c-Fos protein in clone DT7.1 (∽ 10-fold) and clone DT12.4 (∽ 100-fold) (Fig. 5A) and similar levels of induction were observed at the RNA level (see Fig. 8-top).

In order to assess the effects of induced c-fos expression on chondrocyte differentiation we performed the differentiation assay on clones DT7.1 and DT12.4 in the presence and absence of Tc. Both clones formed significant numbers of Alcian blue-positive nodules with levels of exogenous c-fos repressed (1 μg/ml Tc). However, in both clones there was a significant decrease in the levels of differentiation upon induction of exogenous c-fos (withdrawal of Tc), with an estimated 30-50% decrease in the number of nodules in clone DT7.1, and an almost complete abolition of nodule formation in clone DT12.4 (Fig. 5B). As with the stably transfected clones, there appeared to be variation in the basal levels of nodule formation (+Tc), most probably due to clonal variation, however, the inhibition of nodule formation in the absence of Tc in these clones was independent of such variation and thus dependent on the levels of exogenous c-fos expression. These results clearly demonstrate that chondrocyte differentiation is inhibited by exogenous c-fos expression and this effect is independent of clonal variability.

We have used the tightly regulatable expression of exogenous c-fos in DT12.4 cells to define more completely the role of c-fos in ATDC5 cell differentiation in vitro. Initially we analysed the effect of c-fos overexpression at different stages of the differentiation process. Thus, DT12.4 cells were cultured either in the continuous presence of Tc (i.e. low c-fos), or in its absence (i.e. high c-fos) for 4 day periods as indicated. Elevated exogenous c-fos levels during the pre-confluence stage (day 0-4) had a significant inhibitory effect with an ∽ 50% decrease in nodule formation, whereas, induction of high c-fos levels after confluence (after day 4) failed to cause an effect (Fig. 6A,B).

This suggests that the inhibitory effects of c-fos are manifested primarily at the early stages of differentiation, possibly by affecting specific chondroprogenitor cell populations. To address this possibility, we have used limiting dilution analysis to define the effects of c-fos on the number of chondroprogenitors (Fig. 6C). The results showed that in the absence of c-fos (+Tc) approximately 1 in every 22 cells plated is able to form cartilage, however, in the presence of elevated c-fos (−Tc) only ∽ 1 in 65 cells forms cartilage, an approximately 3-fold decrease in the proportion of chondroprogenitors (Fig. 6C). These data therefore demonstrate that the inhibition of differentiation in vitro by exogenous c-fos is a direct effect on the frequency of chondroprogenitor cells within the cell population.

In addition to direct roles for c-fos in chondrocyte differentiation, the possibility remains that exogenous c-fos expression additionally perturbs other cellular processes, such as proliferation and apoptosis, and these may contribute, at least in part, to the observed inhibition of differentiation. Indeed, roles for c-fos in proliferation and apoptosis have been cited in a number of systems (see Angel and Karin, 1991; Smeyne et al., 1993; Pandey and Wang, 1995). We have initially investigated a role for c-fos in proliferation of DT12.4 cells by growth curve analysis at different serum concentrations. Interestingly, in full serum conditions (5% FCS), there is no difference in cell number in the presence or absence of c-fos expression (Fig. 7A). However, at lower serum concentrations (0.5% FCS and 0.1% FCS), c-fos expression does appear to be mitogenic (Fig. 7A). Likewise, the effect of elevated c-fos expression on apoptosis of pre-confluent DT12.4 cells was quantified at different serum concentrations. Again, in standard culture conditions (5% FCS), apoptosis rates were low and unaffected by the expression of c-fos. However, in low serum conditions, rates of apoptosis are decreased approximately twofold in the presence of exogenous c-fos expression (Fig. 7B). As the rates of both proliferation and apoptosis were unaffected by c-fos expression in standard culture conditions (5% FCS) then it seems unlikely that these processes are contributing towards the observed inhibition of differentiation. Nevertheless, taken together, the proliferation and apoptosis data do imply that ectopic c-fos expression can lead to decreased serum dependence in DT12.4 cells.

Finally, the high levels of c-fos expression induced in clone DT12.4 resulted in a clear change, from a polygonal to a spindle-shaped morphology exhibiting elongated processes possibly indicating transformation (data not shown). This morphological change was reversible upon readdition of Tc and appears consistent with the previously reported effects of c-Fos on fibroblast morphology (Miao and Curran, 1994) and, more importantly, with the demonstration that chondrocytes exogenous c-fos (−Tc; Fig. 8) and correlated well with the differentiation status (Fig. 5B).

With respect to AP-1-related genes mRNA levels of both fra-1 and c-jun were upregulated in the presence of high levels of exogenous c-fos (−Tc) in both DT7.1 and DT12.4 cell lines and at both early and late stages of chondrocyte differentiation (Fig. 8). The elevation of c-jun appeared to be most significant at day 5 in DT12.4 cells, whereas c-fos-dependent increases in fra-1 expression were comparable between both cell lines at both early and late time points. were target cells for c-Fos-induced transformation in ES cell chimeric mice (Wang et al., 1991, 1993).

The pattern of gene expression of DT7.1 and DT12.4 cell lines in the presence and absence of induced exogenous c-fos expression was assessed by northern blot analysis at both early (day 5) and late (day 21) time points of differentiation. Expression of the early chondrocyte marker Sox-9 was unaffected by either the levels of exogenous c-fos expression or the stage of differentiation in either cell line (Fig. 8). Interestingly, in DT12.4 cells at day 5, there was a clear decrease in the levels of coll II expression with high c-fos (−Tc; Fig. 8), whilst levels of PTH1R appeared unaffected at this stage. At day 21 the expression of coll II and PTH1R were markedly decreased in both cell lines in the presence of

In this paper we have demonstrated that overexpression of c-Fos directly inhibits differentiation in ATDC5 chondrocytes. Initially this was achieved by assessing differentiation in ATDC5 clones stably overexpressing c-Fos from a constitutive promoter and subsequently, by expression from an inducible promoter we have shown that elevated c-Fos levels inhibit differentiation within individual clones independently of clonal variation in chondrogenic potential. No previous study has investigated the effects of exogenous c-Fos expression on chondrocyte differentiation in vitro, although negative effects on matrix deposition have previously been observed in HCS chondrocytes (Tsuji et al., 1996). This in vitro inhibition is in accordance with the previous in vivo evidence for a role of c-Fos. Thus, gain-of-function experiments have shown that the progression of proliferating chondrocytes to hypertrophy was severely retarded when chick limb buds were infected with c-fos expressing retroviruses (Watanabe et al., 1997), whilst chondrocytes isolated from c-fos-induced chondrosarcomas of ES cell chimeric mice likewise were unable to progress to hypertrophy (Wang et al., 1991, 1993). In addition, loss-of-function studies in c-fos knockout mice demonstrated a diminished zone of proliferating chondrocytes at the epiphyseal growth plate (Wang et al., 1992), which may be due to altered chondrocyte progression in the absence of c-Fos.

Initially, we correlated the phenotypic differentiation of wild-type ATDC5 cells with the pattern of expression of a number of chondrogenic markers. Both Sox-9 and coll II, at low levels, were expressed in pre-confluent cultures, indicative of an already committed chondroprogenitor population. Cartilage differentiation was evident from day 11 as indicated by the dramatic rise in coll II expression, and this was coincident with the deposition of Alcian blue-staining matrix and with the expression of aggrecan and PTH1R. In addition, we demonstrated hypertrophic differentiation of longer term cultures (day 30), by elevated expression of coll X and decreased levels of coll II and aggrecan.

The analysis of the expression levels of c-fos and c-jun family genes throughout ATDC5 chondrogenesis revealed some interesting expression patterns. The most prominent fos family members expressed during differentiation were the fra genes, whilst c-fos and fosB were expressed at low levels which increased only slightly with differentiation. Expression of the jun family members was approximately constitutive throughout, with c-jun and junB elevated during matrix deposition, but decreased during hypertrophic differentiation. Interestingly, a decrease in c-jun expression has been observed during hypertrophic differentiation of embryonic chick chondrocytes (Kameda et al., 1997). These expression data suggest that the majority of AP-1-specifc DNA binding complexes in wild-type ATDC5 cells consist of Fra:Jun heterodimers, and this is supported by supershift analyses. However, it is likely that AP-1 complex composition will change during the differentiation process, which could result in differential regulation of target genes. This possibility is supported by evidence from an analogous osteoblast differentiation model, where low endogenous c-fos expression, and stage-specific changes in AP-1 complex composition have been demonstrated (McCabe et al., 1995, 1996). Whether comparable changes in complex composition occur during ATDC5 chondrogenesis remains to be determined by further supershift experiments using antibodies specific for each AP-1 family member.

The low levels of c-fos expression observed do not necessarily imply that c-fos has no function in chondrogenesis. The earliest embryonic expression of c-Fos is seen in chondrogenic cells of the developing limbs (Dony and Gruss, 1987; DeTogni et al., 1988), and postnatal expression has been detected in proliferating and articular chondrocytes (Grigoriadis et al., 1993; Sunters et al., 1998) as well as in hypertrophic chondrocytes by reporter gene expression in c-fos-lacZ transgenic mice (Smeyne et al., 1993). More importantly, it seems that high level c-fos expression can be induced in chondrocytes when required, for example, in response to specific extracellular stimuli, such as PTH (Lee et al., 1994; see also below). Basal levels of endogenous c-fos may not be important in its role in differentiation, as elevated levels of c-fos appear to inhibit chondrogenesis even when higher basal levels are detectable, both in vivo (Wang et al., 1991, 1993; Watanabe et al., 1997) and in vitro in C5.18 chondrocytes (D. P. Thomas and A. E. Grigoriadis, unpublished results). In fact, lower steady state levels of c-Fos may be preferable if c-fos expression is indeed inhibitory to chondrocyte differentiation. Moreover, previous evidence in vivo demonstrated that elevated c-Fos expression on a background of high endogenous expression may initiate a cascade of oncogenic transformation (Wang et al., 1991, 1993). Thus, it is probable that there exists a threshold of c-Fos expression, above which chondrocytes are susceptible to transformation, but below which c-Fos functions to regulate chondrocyte differentiation. As c-Fos can only act as one component of AP-1 complexes, the mechanisms whereby c-Fos causes inhibition of differentiation are likely to depend on the availability of dimerisation partners (e.g. Jun proteins) and the extent to which c-Fos can compete with other Fos-related proteins for dimerisation, resulting in the formation of new complexes, and presumably differential regulation of specific target genes.

Constitutive overexpression experiments using a c-fos construct which caused chondrosarcomas in ES cell chimeric mice (Wang et al., 1991, 1993) yielded three ATDC5 clones which expressed high levels of exogenous c-fos but failed to differentiate. These clones thus provide an initial indication that c-fos overexpression is inhibitory to chondrocyte differentiation. However, other clones did not express exogenous c-fos, and despite being committed to the chondrocyte lineage, as judged by Sox-9 and coll II expression, they also failed to differentiate. The reasons for this clonal variation are presumably due to subpopulation heterogeneity which is not uncommon when clones are derived from essentially non-homogeneous progenitor populations, for example, C3H10T1/2 (Taylor and Jones, 1979), RCJ 3.1 (Grigoriadis et al., 1988), and RCJ 3.1C5 clones (Grigoriadis et al., 1996).

We subsequently derived ATDC5 clones where high level expression of exogenous c-fos could be induced by withdrawal of Tc. Such a regulatable system permits differentiation assays and expression analyses to be carried out at both high and low levels of c-fos in the same clone allowing unambiguous assessment of the potential role of c-Fos in differentiation and cartilage-specific gene expression. Additionally, by adding or removing Tc at different time points it is possible to assess the effect of c-fos expression at various stages of differentiation.

Clones DT7.1 and DT12.4 demonstrated tight regulation of exogenous c-fos expression in the presence and absence of Tc and exogenous c-fos induction significantly inhibited chondrogenesis in both of these clones with a 30-50% decrease in nodule formation for clone DT7.1 and an almost complete abolition of differentiation in clone DT12.4. This analysis further confirms the observations of the constitutively c-fos overexpressing clones. In clone DT12.4, there was a significant decrease following c-fos induction in the expression levels of coll II at day 5, whilst both clones displayed downregulation of coll II and PTH1R at late stages of differentiation. As both of these genes are markers of chondrocyte differentiation, then their down-regulation at late stages may be secondary to the inhibition of differentiation. However, down-regulation of coll II at early stages of DT12.4 differentiation, may indicate more direct regulation by c-fos that may, at least in part, contribute towards the inhibited differentiation. Whether c-fos can directly modulate coll II expression remains to be determined (see also below). In both of the regulatable clones, fra-1 and c-jun expression were upregulated together with c-fos at both at early and late stages of differentiation. This suggests a potentially more complex pattern of AP-1 transcription factor activity after elevation of c-fos that may have implications for the control of downstream targets. Fra-1 is considered to be a c-Fos target gene as it has been associated with high c-Fos levels in fibroblasts and osteoblasts (Braselmann et al., 1992; Grigoriadis et al., 1993; Schreiber et al., 1997). Moreover, ectopic fra-1 expression in vitro can stimulate osteoclastogenesis (Owens et al., 1999) and fra-1 transgenic mice have specific osteoblastic defects and can rescue the block in osteoclast differentiation in c-fos knockout mice (Jochum et al., 1999; Matsuo et al., 1999). Based on our inducible overexpression system it appears that fra-1 also lies in the pathway induced by c-Fos in chondrocytes. In contrast, the correlation between c-jun and c-fos expression in different cell types is not so well established. Exogenous c-fos expression in osteoblasts in transgenic mice, does not result in enhanced c-jun expression (Grigoriadis et al., 1993), however, in ES cell chimeric mice there is a clear correlation between c-fos and c-jun expression: All chimeric tissues expressing exogenous c-fos, including chondrosarcomas and chondrogenic cell lines derived from these tumours, demonstrate high c-jun levels concomitantly with exogenous c-fos (Wang et al., 1991, 1993). In addition, like c-fos, c-jun expression in chick chondrocytes delays their differentiation (Kameda et al., 1997). Although fra-1 and c-jun appear to be regulated by c-Fos in chondrocytes, whether they have specific roles in chondrocyte biology remains to be determined, for example, by specific gain-of-function and loss-of-function analyses.

In order to define the specific time points during differentiation in which the inhibitory action of c-Fos is manifested, DT12.4 cells were ‘pulsed’ with elevated c-fos levels for 4 day periods. Interestingly, only when exogenous c-fos was expressed during the subconfluent phase (days 0-4) was a decrease in nodule formation observed, whilst elevation of c-fos at later time points had no significant effect. The reduction in the number of differentiated nodules and the putative window of c-Fos action suggests that c-Fos acts directly on chondroprogenitor cells to inhibit their progression. This was confirmed by limiting dilution analysis of DT12.4 cells, which demonstrated an ∽ 3-fold decrease in the number of cells competent to differentiate into cartilage. Although, c-Fos clearly affects the frequency of chondroprogenitor cells within ATDC5 cell populations, high levels are nevertheless not completely incompatible with chondrocyte differentiation, as some nodule formation persists in these clones. These data imply that there may exist a subpopulation of precursors that are not affected by elevated c-Fos. Alternatively, together with the pulse experiments demonstrating an early effect, our observations may point to the presence of a restriction point, prior to which c-Fos expression can inhibit the differentiation of chondroprogenitors, but once passed c-Fos has no apparent effect. Whilst the effects of c-Fos on early chondroprogenitors are clear, at this point we can not exclude the possibility that, under different experimental conditions, other effects may be uncovered, for example, alterations in proteoglycan synthesis or changes in rates of matrix deposition.

Having defined a role for c-fos expression in the differentiation of chondrocytes in vitro, we have also sought to analyse whether or not c-fos affects other cellular processes such as proliferation and apoptosis as have been demonstrated in various other systems (Angel and Karin, 1991; Smeyne et al., 1993; Pandey and Wang, 1995). One important reason for doing so is that decreases in chondroprogenitor proliferation and/or increases in apoptosis may to some degree contribute towards the observed decrease in nodule formation. In analysing the rates of proliferation and apoptosis of DT12.4 cells in the presence and absence of c-fos expression, we demonstrated that under the conditions whereby c-Fos inhibited differentiation (5% FCS) neither the growth rate nor the apoptotic index were significantly affected. However, modulation of growth rates and apoptosis were nevertheless observed under conditions of reduced serum concentrations, with rates of proliferation increased and rates of apoptosis decreased in the presence of exogenous c-fos. Therefore, under these conditions, c-Fos both increases the mitogenicity of these chondrocytes, and protects them from apoptosis, indicative of decreased serum dependence. We have also observed similar effects in osteoblastic MC3T3-E1 cells overexpressing c-fos (A. Sunters, D. P. Thomas and A. E. Grigoriadis, unpublished), and the roles of c-fos expression on the molecular mechanisms of proliferation and apoptosis in chondrocytes in vitro are currently under investigation. Thus, although c-Fos has the potential to regulate the rates of cell growth and programmed cell death of ATDC5 chondroprogenitors, these processes apparently do not contribute to the inhibitory effect of c-Fos on chondrocyte differentiation.

The in vitro evidence presented here clearly defines a role for c-Fos in inhibiting the differentiation of ATDC5 chondrocytes, but how does this fit into previously described models of chondrocyte differentiation? One strong candidate for a physiologically relevant stimulus of c-Fos expression in chondrocytes would be PTHrP. In this regard, it is extremely interesting that the endochondral growth plates of PTHrP knockout mice (Karaplis et al., 1994) and c-fos knockout mice (Wang et al., 1992) look very similar, suggesting that the normal role of both of these molecules is to inhibit the differentiation of growth plate chondrocytes. Signalling via the PTH1R has been shown to upregulate c-Fos expression in osteoblasts in vitro (Pearman et al., 1996; McCauley et al., 1997) and in growth plate chondrocytes in vivo (Lee et al., 1994), and using transgenic and knockout mice the PTH1R signalling cascade has been shown to regulate normal chondrocyte differentiation in vivo (Weir et al., 1996; Vortkamp et al., 1996: Lanske et al., 1996; Schipani et al., 1997). Interestingly, in ATDC5 cells, our preliminary results indicate that PTH efficiently stimulates c-fos expression, and inhibits cartilage differentiation with similar kinetics to the effects demonstrated by pulsed elevation of c-fos in our Tc-regulatable clones (D. P. Thomas and A. E. Grigoriadis, data not shown). Thus, it is likely that c-fos represents a specific physiological target for PTHrP signalling and may mediate at least some of the phenotypic effects of PTHrP action in chondrocytes. In addition, it is possible that c-fos expression is also regulated by other signalling pathways important in chondrogenic cells, such as those induced by bone morphogenetic proteins (BMPs) and BMP receptors (Zou et al., 1997), hedgehog proteins (Vortkamp et al., 1996; Iwasaki et al., 1997), and FGF receptors (Peters et al., 1992; Naski et al., 1998). The identification of c-fos responsive genes in chondrocytes will be important in understanding the molecular roles of c-fos in mediating the observed phenotype. In particular, the early effects of c-Fos provide a useful time window for analysis of genes that are potentially direct transcriptional targets of c-Fos, and the tight regulation of c-fos expression in our clones provides an excellent system for the screening of such targets. Candidate c-Fos-regulated genes may be proposed on the grounds either that they have demonstrated important roles in chondrocyte biology, for example from gene deletion studies, or that they have been shown previously to be regulated by c-Fos. Such genes fall into several categories: Firstly, cellular transcription factors such as additional members of the Sox family (Sox-5 and Sox-6; Lefebvre et al., 1998), as well as ATF-2 (Reimold et al., 1996), Ets2 (Sumarsono et al., 1996), and cbfa-1 (Inada et al., 1999) affect cartilage differentiation and some of these have been shown to interact with AP-1 complexes and modulate gene expression (De Cesare et al., 1995; Basuyaux et al., 1997; Selvamurugan et al., 1998; Porte et al., 1999). Secondly, profound effects on chondrogenesis have been reported in response to autocrine or paracrine signalling mediated by BMPs (see Hogan, 1996), hedgehog proteins (Vortkamp et al, 1996) and FGFs (Naski et al., 1998; for review see Tickle and Eichele, 1994), as well as by inhibitors, e.g. Noggin (Brunet et al., 1998; Ito et al., 1999). Thirdly, several genes associated with cell cycle control have demonstrated specific roles in chondrocyte differentiation, such as the Rb-related genes p107 and p130 (Cobrinik et al., 1996), and the CKI p57 (Zhang et al., 1997), whilst cyclin D1 has been shown to be regulated by c-Fos and Fos-related genes in fibroblasts and chondrocytes in vitro (Brown et al., 1998; Beier et al., 1999), and in osteoblasts in vivo (Sunters et al., 1998). Finally, regulation of the apoptosis genes, Bcl-2 and Bax by PTH has been demonstrated in growth plate chondrocytes in vivo (Amling et al., 1997), and therefore represent potential c-Fos targets. Besides elucidating whether known genes are affected by altered c-Fos levels, this inducible expression model allows for the identification of novel targets by cDNA subtractive hybridisation techniques, specifically during early stages of differentiation and these studies are currently underway.

We thank Dr Peter Angel (DKFZ, Heidelberg, Germany) and Dr Bernd Baumann (University of Würzburg, Germany) for helpful advice with EMSA analyses and Dr Chris Healy for discussions and critical review of the manuscript. This work was generously supported by the Arthritis Research Campaign (G0519).