Commitment of the teratocarcinoma-derived mesodermal clone c1 towards terminal osteogenic differentiation

During embryogenesis, osteogenic stem cells may arise from the sclerotomal cells of the somitic mesoderm, the somitomeres of the hindbrain, or the mesectodermal cells of the neural crest (Couly and Le Douarin, 1988; Noden, 1988; Thorogood, 1988). In postnatal life, osteoprogenitors are located within the periosteum, the endosteum and the adjacent stroma. The differentiation pathway followed by mesodermal cells in order to become committed osteo-progenitors, and subsequently osteoblasts, may therefore be determined by their origin and their interactions with sur-rounding cells and matrices. Furthermore, the first stages of bone formation involves either direct conversion of meso-dermal stem cells into osteoprogenitors (intramembranous ossification) or conversion via intermediate stages where the cartilagenous matrix is replaced by bone (endochondral bone formation).

Our understanding of the ontogenesis of osteoblasts is limited by the lack of markers specific for mesodermal stem cells and their osteogenic progeny. At present, several bone matrix proteins, which may contribute to the process of extracellular matrix mineralization, serve as markers of osteoblast function. These proteins are: (1) type I collagen (type I col), the major component of the bone matrix; (2) osteopontin (OP), a sialoprotein made by osteoblasts, osteo-cytes and their precursors (this protein is expressed in tissues such as bone, gut, kidney and placenta, and is also synthesized by chondrocytes in the area of cartilage to bone transition; Castagnola et al., 1991); (3) osteonectin (ON), a phosphoprotein synthesized by osteoblasts and many other cell types; (4) osteocalcin (OC), the osteoblast-specific pro-tein strongly bound to hydroxyapatite. Alkaline phosphatase (AlkP), a membrane-bound enzyme, is considered to be an early marker of osteoblastic differentiation (Turksen and Aubin, 1991). In situ hybridization and immunolocalization studies, indicated that the expression of these bone matrix markers varies with the type of fetal bone analysed (Mark et al., 1988; Weinreb et al., 1990). To investigate osteoge-nesis in vitro, a variety of osteoblastic-like culture systems were derived from marrow stroma (Benayahu et al., 1989; Maniatopoulos et al., 1988), endosteum (Lomri et al., 1988) or, most commonly, calvaria (Aronow et al., 1990; Bellows et al., 1986; Escarot-Charrier et al., 1983; Gerstenfeld et al., 1987; Grigoriadis et al., 1988; Heath et al., 1989; Nefussi et al., 1985; Sudo et al., 1983; Yamaguchi and Kahn, 1991). In primary cultures, quantification of the num-bers of bone nodules formed suggests that a limited number of progenitors are present. Clonal cell lines (Benayahu et al., 1989; Grigoriadis et al., 1988; Heath et al., 1989; Sudo et al., 1983; Yamaguchi and Kahn, 1991) express pheno-typic features that are likely to correspond to their stage of commitment at the time of their isolation. They differ in their capacity to differentiate and in their responses to hor-mones and growth factors. The exact nature of progenitors cells in these systems and the precise mechanisms that con-tribute to their induction are not known.

We have shown that mouse teratocarcinoma could be a new model for determining the early steps of bone differ-entiation. We transformed the multipotent embryonal car-cinoma cell line 1003 with the hybrid plasmid pK4 (Kellermann and Kelly, 1986) carrying the early genes of simian virus 40 (SV40) under the control of the adenovirus type 5 E1a promoter. Various immortalized immature mesodermal derivatives were obtained including clone C1, which can be committed to the osteogenic pathway (Kellermann et al., 1990). C1 cells produce osteosarcoma in vivo and although expressing the SV40 T antigen, they differentiate in vitro. When cultured as nodules in the presence of ascorbic acid (AA) and β-glycerophosphate (βGP), classical mediators of in vitro osteogenic differentiation, C1 cells deposit a col-lagenous extracellular matrix, which progressively miner-alizes and forms hydroxyapatite crystals (Kellermann et al., 1990). In a previous report (Chentoufi et al., 1993) we described, from histochemical studies, the kinetics of osteogenic differentiation of C1 cells by following matrix deposition, mineralization and the expression of AlkP enzymic activity. Calcium deposition was detectable in aggregates treated with AA and βGP as early as 2 days after their addition and increased linearly up to 30 days. The high frequency and synchrony of nodule mineraliza-tion shown by Von Kossa staining and tetracycline incor-poration allowed us to follow the temporal sequence of expression of the osteoblastic genes by monitoring the tran-sition from exponentially growing C1 precursor cells to fully differentiated cultures.

We demonstrate here that: (i) at low density, committed C1 precursor cells do not express any of the above bone-related genes; (ii) as soon as cells come into contact, irre-spective of the presence of AA and βGP, the genes are ‘switched on’, raising the possibility that cell-cell interac-tions may control their transcription (this first step of com-mitment is reversible); (iii) the corresponding transcripts are translated, as shown by the immunodetection of type I col and OC; (iv) finally, upon addition of AA and βGP, matrix deposition and mineralization are observed, and overex-pression of OC and OP transcripts is observed in cells lining the mineralizing foci. Addition of AA and βGP is thus absolutely required to turn on the terminal phases of C1 cells osteogenic differentiation.

C1 cells were isolated and cultured as described (Kellermann et al., 1990). Exponential cultures were established by dissociation of confluent cells, which were seeded at low density (500 cells per cm²) and grown for 3 days without cell-cell contacts. To form three-dimensional (3D) aggregates, C1 cells were first seeded at 3×10⁵ cells per 10 ml of DME supplemented with 10% fetal calf serum on untreated plastic dishes. After 9 days (day 0), the 4000 to 5000 3D clusters formed per dish were refed with DME sup-plemented with 1% fetal calf serum and induced to differentiate by the addition of 50 μg/ml ascorbic acid and of 7 mM β-glyc-erophosphate. Cells were refed every three days with or without pharmacological effectors.

The following DNA probes were used for in situ hybridization and northern blot analysis: a rat OC cDNA EcoRI insert of 520 bp (Yoon et al., 1988), a rat OP cDNA EcoRI insert of 1300 bp (Yoon et al., 1987), a rat AlkP, cDNA EcoRI insert of 520 bp (Noda et al., 1987), a mouse type I col cDNA XhoI insert of 850 bp (Schmidt et al., 1984), a Xenopus boralis histone H4 genomic HindIII-XbaI fragment of 594 bp and a bovine ON cDNA EcoRI insert of 1500 bp (Young et al., 1989). A mouse glyceraldehyde phosphate deshydrogenase (GAPDH) cDNA insert of 0.6 kb was used for northern blot calibration (Piechaczyk et al., 1984).

Total RNA was isolated from various cell lines or from calvaria of 1-day-old mice, as described (Chomczynski and Sacchi, 1987). Northern blots were prepared according to the method of Thomas (1980), with 20 μg of RNA resolved on 1.2% agarose gel and transferred to Genescreen membranes. Probes were labelled with ³²P-labelled dCTP in an oligonucleotide-primed reaction (Feinberg and Vogelstein, 1983). Hybridizations were carried out at 65°C for 12 hours. Final washes were performed at 65°C in 0.5× SSPE 0.1% SDS for heterologous probes, or 0.1× SSPE, 0.1% SDS for homologous probes.

Histochemical analysis was performed as described (Kellermann et al., 1990). For in situ hybridization and immunocytochemical studies, aggregates were collected, rinsed twice with PBS and fixed overnight at 4°C in a solution of 4% paraformaldehyde in PBS. The aggregates were embedded in 100 μl of 4% low melt-ing point agarose, prior to inclusion in Tissue Tek OCT for cryo-stat sections or Paraplast plus for paraffin sections. Cryostat and paraffin sections were obtained essentially as described (Poliard et al., 1986; Sassoon et al., 1988). For in situ hybridization aggre-gate sections were pretreated according to the method of Sassoon et al. (1988) with two modifications: (1) immediately after post-fixation the slides were treated in 0.25 M HCl for 20 minutes. This decalcification treatment was done to prevent non-specific binding of the probe to the mineral. (2) The proteinase K treat-ment lasted 15 minutes. For the labelling of cDNA probes, specific inserts were labelled with [³⁵S]dCTP and [³⁵S]dATP by nick-translation. In situ hybridization was then performed as described (Poliard et al., 1986) using 10⁵ to 2 ×10⁵ cpm of probe per slide. lides were processed for standard autoradiography and exposed at 4°C for 2 weeks to 1 month. For immunocytochemistry, cells and nodule sections were fixed and permeabilized with 3% paraformaldehyde followed by cold ethanol. Indirect immunoflu-orescence on cells and nodule sections were carried out using: rabbit antibodies against type I col affinity-purified on type I col and adsorbed on type II, III and IV collagen; goat anti-mouse OC antibodies (BTI, Stoughton, MA, USA). Specific secondary anti-sera coupled to fluorophores were used to visualize sites of pri-mary antibody binding.

When 3D aggregates of the immortalized mesodermal C1 clone are exposed continously to AA and βGP, they deposit a mineralized collagenous matrix. As judged by Von Kossa staining, mineralization starts 2-4 days after addition of AA and βGP, and expands progressively (Fig. 1B,C). The bone matrix forms in a roughly synchronous way in all aggre-gates with osteocyte-like cells being progressively embed-ded within the matrix (Chentoufi et al., 1993). Remarkably, more than 95% of the aggregates mineralize upon treat-ment, whereas no matrix deposition and mineralization is ever observed in untreated aggregates (Fig. 1A).

To evaluate when and how bone-specific gene expression is triggered during C1 osteogenic differentiation, steady-state levels of AlkP, type I col, ON, OP and OC transcripts were first evaluated by northern blot analysis (Fig. 2). The immature C1 cells are first grown exponentially at low den-sity in the absence of cell-cell contact. AlkP, type I col, OC and OP transcripts are not detectable. When C1 cells reach confluency or interact within 3D aggregates, however, all the above transcripts become evident (Fig. 2A), irrespec-tive of the presence or absence of AA and βGP. This expression is reversible, since the same transcripts are not detectable 72 hours after dissociation of the confluent cells, as long as the cells do not establish contacts (exponential cultures, see Materials and Methods). It is noteworthy that this transcription pattern is specific to osteogenic cells, since the same transcripts are detected in calvaria cells but not in any of the other teratoma-derived cell lines tested: embryonal carcinoma (Kellermann et al., 1990), neuroec-todermal (Buc-Caron et al., 1990) (Fig. 2B) and endoder-mal cells (not shown). The only exception is the ON tran-script, which is found in all cell lines tested (Fig. 2B). This phosphoprotein is known to be present in many cell types (Nomura et al., 1988).

Whereas the level of bone-specific gene transcripts does not vary significantly over the 14-day-period studied in the untreated aggregates, addition of AA and βGP leads to con-siderable changes in the steady-state levels of these tran-scripts (Fig. 2C). From day 4 onwards, AlkP and type I col transcript levels decline while the aggregates progress through mineralization. By day 14, type I col transcripts have become almost undetectable. Between days 4 and 14 the level of AlkP transcripts diminishes by a factor of 4, in good aggreement with what has been observed histochem-ically and biochemically for AlkP activity (Chentoufi et al., 1993). The steady-state level of OC and OP transcripts, already high in untreated aggregates, increases by a factor of at least 4 between days 4 and 14 of the mineralization kinetics. The level of ON transcripts does not appear to vary significantly in untreated or treated aggregates.

To gain a better understanding of the variation in the number of bone-related transcripts at the single-cell level, an in situ hybridization analysis was performed. In control aggregates prior to the addition of AA and βGP a similar pattern is observed with all the probes tested: the majority of the cells are labelled without any preferential localization, irrespective of the time in culture (Fig. 3A). In treated aggregates, AlkP and type I col transcripts remain visible in most of the aggregated cells, until at least day 7. By day 12, however, the signal has become almost indistinguish-able from the background (not shown). Such a decrease in the silver grain density during osteogenic differentiation confirms the quantitative evaluation of type I Col and AlkP transcripts obtained by nothern blot analysis. In contrast, the distribution and intensity of the signals corresponding to OC and OP transcripts change under the action of AA and βGP. While uniformly present in the aggregates at the beginning of the experiment (Fig. 3A), these transcripts appear to be over-expressed in cells adjacent to the miner-alizing foci (Fig. 3B and C).

In order to firmly establish a link between the kinetics of gene expression and bone matrix formation, we examined the synthesis and cellular localization of two important matrix proteins, type I col and OC. In monolayer, in the absence of cell contacts, type I col and OC are not detected (Fig. 4A; type I col; and not shown). When cellular con-tacts are established, both proteins become detectable. In zones of high cellular density close to 100% of the cells present a signal (mainly in the Golgi; Fig. 4B, type I col; and G, OC). In control aggregate sections type I col immunolabelling remains weak, and is distributed within the majority of the cells (Fig. 4C). In contrast, in sections of treated aggregates type I col appears to be secreted pro-fusely, and is assembled extracellularly (Fig. 4D-F; and Kellermann et al., 1990).

The extracellular matrix of type I col is formed, in the main, 2 days after the addition of AA and βGP (Fig. 4D). The OC localization superimposes on that of type I col (Fig. 4H and I). Finally, type II collagen, a protein found pre-dominantly in chondrocytes is not detected in the mineral-ized aggregates (not shown). These immunochemical observations are in agreement with the histochemical analysis performed during the time course of C1 cell osteogenic differentiation (Chentoufi et al., 1993). The staining of the matrix with Goldner trichrome and Toluidine blue, and the lack of staining with Alcian blue have shown that the extra-cellular matrix is composed of collagen and bone-type gly-cosaminoglycans and not of cartilage.

C1 cells express the SV40 large T antigen and have, at least theoretically, an unlimited potential for proliferation when cultured in monolayers. We therefore wished to evaluate the proliferative state of the C1 cells as they differentiate along the osteogenic pathway. To this end, histone H4 gene expression was taken as an index of DNA replication. As expected, exponentially growing C1 cells express a high level of histone H4 transcripts. When cell contacts form, at confluence or in aggregates, this level decreases by a factor of 3 (Fig. 5) and remains roughly constant thereafter, during the 14 days of the kinetic study. Importantly, this pattern is independent of the presence of AA and βGP during the osteogenic differentiation pathway. When examined at the single-cell level, histone H4 transcripts appear uniformly distributed over the aggregates: in particular the cells adja-cent to the mineralizing foci, where OC and OP are over-expressed, display a signal equivalent to that of the periph-eral cells (Fig. 5). Thus C1 cells proliferate during the course of their differentiation. This result is in agreement with a previous analysis (Chentoufi et al., 1993) showing that differentiating C1 cells still incorporate [³H]thymidine. Consequently, the variations in bone-related gene expression, as reported here, are likely to reflect the differ-entiation state of the cells rather than their proliferative rate.

This study presents the teratocarcinoma-derived mesoder-mal C1 cell line as a new model to identify the parameters governing osteogenic differentiation. Our results suggest that the commitment of the mesodermal clone C1 to the osteogenic pathway is controlled at two levels: (i) cell-cell interactions at confluency or in 3D aggregates, which allow overt osteoblast-specific gene transcription in the absence of concomittant matrix formation and mineralization. This initial step does not require the presence of AA and βGP; it is reversible, since transcription of the monitored genes can be shut down upon dissociation of the confluent cells. (ii) The addition of AA and βGP, which enables confluent aggregated cells to proceed through the well-characterized steps of osteogenic differentiation.

To date, osteogenic differentiation has been mainly char-acterized using calvaria primary cultures. Calvaria cells have a limited potential for division at confluence, and are not uniformly responsive to the effect of the inducers (Owen et al., 1990). The clonal origin of the C1 cells may explain their capacity to form aggregates that are 95% capable of transforming into calcifying nodules and in which most cells express bone-related genes. As early as 2-3 days after the addition of AA and βGP to the aggre-gates, foci of calcification develop with a frequency close to 100%. In the absence of treatment, no matrix accumu-lation and no mineralization occur (this study; and Chen-toufi et al., 1993). Another important difference between calvaria cells and the clonal C1 system is the capacity of the C1 cells to continue proliferating after entering the pro-gram of osteogenic differentiation. This capacity probably reflects the expression of the SV40 T antigen previously introduced into the parental embryonal carcinoma cells from which the C1 clone was isolated. It seems likely, how-ever, that, as soon as they are embedded within the calci-fied matrix, the differentiated osteocyte-like cells stop dividing. This conclusion is based on photonic microscopy observations showing that osteocytes do not have neigh-bouring cells within the matrix (Chentoufi et al., 1993). Two rat calvaria cell lines that retain many of the characteristics of osteoblasts have been immortalized with a recombinant retrovirus containing the cDNA for SV40 large T antigen (Heath et al., 1989). Compared to these cell lines, C1 cells have two additional features. Firstly, the cells are osteogenic in vitro, allowing the study of the mechanisms underlying matrix mineralization. Secondly, C1 cells might represent an earlier stage of commitment, since they are derived from an island of cells resembling the somitic mesoderm of the early embryo.

The potential for in vitro differentiation of the C1 cell line strongly resembles that of primary osteoblastic cell cul-tures (Aronow et al., 1990; Bellows et al., 1986; Escarot-Charrier et al., 1986; Lomri et al., 1988; Nefussi et al., 1985) and clearly distinguishes the C1 clone from osteosar-coma-derived cell lines, which cannot be induced to dif-ferentiate in vitro (Rodan et al., 1987). Calvaria and C1 cells are both capable of synthesizing and secreting an extracellular collagenous matrix that then mineralizes. This differentiation is actually reminiscent of the process of in vivo bone formation.

A number of laboratories have shown AA to be neces-sary for collagen matrix deposition and expression of the osteoblast phenotype in vitro (Aronow et al., 1990; Franceschi and Iyer, 1992; Gerstenfeld et al., 1987; Owen et al., 1990; Sudo et al., 1983). βGP is known to promote the calcification of the extracellular matrix produced by osteoblasts. Upon addition of AA to C1 cell aggregates, type I col protein molecules already present in the cells accumulate in the extracellular space. The mineralization process is then rapidly initiated, and crystalline hydroxya-patite appears, closely associated with the striated collagen fibrils (Kellermann et al., 1990). Concurrently, a decline in the level of type I col transcription is observed. A similar relationship between mineralization and collagen I accu-mulation or transcription has been established in the case of primary cultures of osteoblastic cells (Gerstenfeld et al., 1987; Ibaraki et al., 1992; Owen et al., 1990).

In C1 cultures, both AlkP transcripts and activity become measurable when the cells reach confluency. Maximal levels are found in untreated aggregates and remain con-stant during at least 14 days of culture (Chentoufi et al., 1993). Like rat calvaria (Owen et al., 1990) and foetal bovine (Ibaraki et al., 1992) osteoblasts, a decrease in AlkP transcripts and activity is observed upon addition of AA and βGP. Our histological (Chentoufi et al., 1993) and in situ hybridization experiments show that the cells adjacent to the mineralizing foci, as well as the peripheral cells in the aggregates, all express AlkP, although at a reduced level. This behaviour is reminiscent of in vivo observations revealing AlkP expression in osteoblasts adjacent to the mineralized matrix and in still dividing osteoblast precur-sors extending from the bone surface (Mark et al., 1988; Weinreb et al., 1990).

OC protein is already detectable within confluent or 3D aggregated C1 cells. Upon addition of AA and βGP it is rapidly accumulated in the extracellular space. The levels of OC in the differentiating aggregates, increase concomi-tantly with the spreading of the type I col matrix and its calcification. A selective overexpression of the OC tran-scripts is observed within cells lining the mineralizing foci. This is also the case for OP transcripts. This observation may account for the overall 4 to 5-fold enhancement of the two corresponding RNAs measured by northern blot analysis during the 14-day period of in vitro osteogenic differ-entiation. Such a pattern of RNA and protein expression is in agreement with in vitro and in vivo data showing an ele-vated level of expression of OC and OP in mineralizing pri-mary osteoblastic cultures (Ibaraki et al., 1992; Owen et al., 1990) and in osteoblasts in close proximity to the bone sur-face (Mark et al., 1988; Weinreb et al., 1990). In C1 aggre-gates, as in these systems, hydroxyapatite deposition may be required to signal expression of genes such as the gene for osteocalcin in mature osteoblasts.

Taken together in the C1 clonal in vitro system the initial role of cellular interactions (cell-cell and/or cell matrix) would be to provide a permissive environment for specific differentiation to occur rather than to act directly as inducer (e.g. it is permissive rather that instructive). Addition of AA and βGP, as also demonstrated in the cases of parietal, endosteal and calvaria primary cultures, induces the C1 cells to differentiate along the osteogenic pathway. As observed in osteoblastic cultures, as soon as type I col matrix is formed, the progressive development of mineral-izing foci is accompanied by temporal changes in gene expression. Development of the osteoblast phenotype from progenitor cells and bone formation are likely to depend partly on the major matrix components studied here. How-ever, osteoinductive proteins concentrated within the matrix and autocrine or serum factors are also likely to be involved (Rosen and Thies, 1992). Under AA and βGP culture con-ditions, C1 cells appear univocally committed towards the osteogenic pathway. The lack of staining for alcian blue (Chentoufi et al., 1993) and the absence of type II collagen (Kellermann et al., 1990; and this study) within the extra-cellular matrix indicate the absence of cartilage.

In this study the use of a clonal cell line has allowed us to emphasize the fact that the establishment of cellular inter-actions may have a critical function in the initial transcrip-tion of bone-related genes. Increasing numbers of reports have demonstrated the pre-eminent regulatory role of cell-cell or cell-matrix interactions in the control of cellular gene expression (Gurdon, 1988; Streuli et al., 1991; Takeichi, 1988). In the skeleton (Solursh, 1989), initiation of chon-drogenesis depends critically on cellular condensation, and osteogenesis appears to be partly mediated by cell-extra-cellular matrix interactions. Little is known about the type of cadherins and integrins in osteogenic cells that are involved in these interactions. To date, the expression of specific chains of integrins has been defined in human bone cells (Clover et al., 1992; Horton and Davies, 1989; Hughes et al., 1993). Further studies, in particular during bone development are required to gain an understanding of the roles of integrin-mediated cell-matrix interactions in the control of osteoblast differentiation and function. As the cell density increases, nearly 100% of C1 precursor cells are ‘switched on’, becoming competent to respond to induc-tive signals capable of triggering the terminal osteoblastic differentiation program. This unique property may help in deciphering the molecular and cellular mechanisms under-lying the initial stages of osteogenic differentiation and in assessing the possible role of cell adhesion receptors in mesodermal gene regulation.

Finally, it is noteworthy that the clonal C1 cells behave in a synchronous manner during their differentiation. The complete lack of OC and type I col in sparse cultures, com-pared to the positive labelling for both proteins in nearly 100% of the cells at confluence, supports the idea that C1 cells are committed mesodermal precursors. Further cellu-lar interactions would then be permissive for the expression of a ‘determined’ phenotype. In agreement with this, our current unpublished studies show that confluent C1 cells can also differentiate into either adipocyte-or chondrob-last-like cells, in a mutually exclusive manner, under the action of the appropriate inducers. This behaviour may open the way for the identification of epigenetic mechanisms (mediated by cell or cell-matrix interactions, or growth factors) and transcriptional factors involved in the restriction versus promotion of osteoblastic, chondroblastic or adipoblastic differentiation from a unique mesenchymal progenitor cell.

The authors thank Dr P. Avner, Dr B. Laoide and Dr A. Sobel for constructive comments on reading the manuscript. We also thank M. Hott for expert histological analysis, and J. Lobrot for excellent photographic services. We thank Dr M. Noda, Dr M. Young and Dr B. de Combrugghe for various cDNA probes and S. Guerret for the generous gift of affinity-purified antibodies. We are grateful to F. Petrou for her editorial assistance. This work was supported by grants from the Centre National de la Recherche Scientifique (UR 1148), the Fondation pour la Recherche Médi-cale, the Ligue Nationale contre le Cancer and the Institut National de la Santé et de la Recherche Médicale (grant CRE 900 403).