FGF2 promotes skeletogenic differentiation of cranial neural crest cells

Cells of the cranial neural crest make a major contribution to the skeletal tissues of the skull, differentiating into both primary and secondary cartilage, endochondral bone and membrane bone (Hall and Horstadius, 1988; Couly et al., 1993). Whether or not there is an established skeletogenic lineage within the cranial neural crest before migration is unclear. Moreover, the developmental events determining the skeletogenic fate of a sub-population of neural crest and their derivatives have only been identified at the level of epithelial-mesenchymal interactions (Bee and Thorogood, 1980) and interactions with the extracellular matrix (Hall, 1982, Thorogood, 1993).

A significant clue to the molecular mechanisms that underlie cranial skeletogenesis has come from identification of mutations in the genes encoding fibroblast growth factor receptor (FGFRs) in individuals with craniosynostosis. In this group of diseases, the sutures of the cranial vault lose their proliferative function and display precocious bone differentiation and fusion (Gorlin et al., 1990; Wilkie, 1997; Burke et al., 1998). Several FGFR mutations create constitutive activation of FGF signalling, endowing the receptor with a degree of ligand-binding independence for signal transduction (Webster and Donoghue, 1997). These studies point to a role for FGF signalling in the skeletogenic differentiation of neural crest-derived tissues. The precise role of FGF-mediated signalling in cranial neural crest cell development is, however, unclear. FGF2 is known to stimulate neural crest proliferation in cultures of mouse neural tube (Murphy et al., 1994) and in micromass cultures of chick facial ectomesenchyme, where growth is usually accompanied by an increase in chondrogenesis (Richman and Crosby, 1990). Correspondingly, FGF2- and FGF4-soaked beads implanted in vivo into facial tissue promote growth locally and increase the size of cartilage elements at that site (Richman et al., 1997). The facial ectoderm is a local source of FGF2 and is required for expression of FGFR2 in the underlying mesenchyme (Matovinovic and Richman, 1997), which further suggests a role for FGF signalling in the growth and differentiation of facial processes. The possibility that these mechanisms may have a broader and more fundamental role in skull morphogenesis, perhaps underlying the tissue interactions that direct skeletogenic differentiation of cranial neural crest cells, has not been fully explored.

We aimed to answer three questions relating to the role of FGF signalling in the skeletogenic differentiation of cranial neural crest cells. First, does neural crest responsiveness to FGF exist before migration? Second, does FGF-mediated signalling play a causal role in determining the skeletogenic fate of neural crest cells? Third, can the relationship between FGF and neural crest development be related to the clinical phenotype of craniosynostosis, in which neural crest-derived cells of the suture differentiate precociously? We find that FGF2 serves as a survival factor for avian pre-migratory mesencephalic neural crest cells in culture and elicits a concentration-dependent response, with proliferation at low FGF2 concentrations and skeletogenic differentiation at high concentrations. These findings indicate that FGF signalling plays a key role in skeletogenic differentiation of cranial neural crest and suggest that neural crest differentiation may be disturbed in craniosynostosis.

Fertile eggs of the Japanese quail, Coturnix coturnix japonica (Rosedean Farm, Hertfordshire, UK) were incubated at 37.5°C for 30 hours to obtain embryos at HH stage 8/9 (Hamburger and Hamilton, 1951). The apices of the neural folds were excised from the mesencephalic region (midbrain) using tungsten needles and used for explant cultures. The average size of the explants was approximately 0.1 mm.

To establish whether the size of the explants affected the outcome of the experiments, half-size explants were examined. Explants taken from embryos at early stage 8 and late stage 9 were also studied, as these were the extremes of the range from which embryos were routinely used for experiments. In some experiments, forebrain (from stage 8/9) and hindbrain (from stage 10) neural crest explants were cultured in parallel with mesencephalic ones for comparison. Neural crest explants were cultured in alpha MEM (alpha modification of Eagle’s minimal essential medium with ribosides and deoxyribonucleosides, Gibco BRL, UK) supplemented with 10% fetal calf serum (FCS; Sigma, UK), 25 units/ml penicillin, 25 μg/ml streptomycin sulphate (pen-strep; Gibco BRL, UK). The medium was changed on alternate days.

Bovine brain-derived FGF2 (R&D Systems, UK) at different final concentrations (0.1, 1.0 or 10 ng/ml) was added to the culture medium either for the whole time in culture (10 days, unless otherwise specified) or for discrete periods (0-24, 0-48, 0-72, 24-48 and 24-72 hours), as indicated in the Results. Control cultures without FGF were always grown in parallel. Growth and changes in morphology were monitored daily.

Growth of the explants was assessed by measuring the diameter of the outgrowths using an eyepiece graticule with a linear scale fitted to an Olympus IMT microscope, and by immunodetection of the proliferation marker PCNA (proliferating cell nuclear antigen). The percentage increase in explant diameter was calculated using the following formula: (culture diameter at day x – culture diameter at day 0/culture diameter at day 0)×100. Changes in cell morphology were recorded using an Olympus OM-2_N camera.

To quantify the occurrence of chondrogenic differentiation following different exposure times to 10 ng/ml FGF2, the presence of chondrogenic nodules was assessed in 10-day-old cultures; the incidence of chondrogenesis was expressed as the percentage of the total number of explants examined that contained chondrogenic nodules. Statistical analyses were performed using a one way analysis of variance (ANOVA) for comparison of multiple groups or by t-test for analysis of two groups. All statistical analyses were carried out using SigmaStat 1.0 (Jandel Scientific Software, Erkrath, Germany).

To induce osteogenic differentiation, mesencephalic neural crest cells that had been cultured for 10 days in the presence of 10 ng/ml FGF2 were maintained in culture for a further 18 days in medium additionally supplemented with 50 μl/ml ascorbate, 10 mM β-glycerophosphate and 10⁻⁷M dexamethasone.

In addition to morphological analysis, the presence of cartilage and bone was assessed by Alcian Blue and Alizarin S Red staining with minor modifications (Simons and Van Horn, 1971). Further analysis of differentiated bone was carried out by staining the mineralised matrix (Von Kossa, 1901). Areas of the culture that were mineralised appeared dark brown. For closer examination, specimens were resin-embedded, as for electron microscopy, and 1 μm sections cut. Sections were stained with Toluidine Blue in 1% (w/v) sodium borate and examined under the light microscope for histological features of bone.

Collagen type II antisense and sense probes were prepared and used as described (Devlin et al., 1988). Cartilage derived from neural crest exposed to 10 ng/ml FGF2 for 10 days were fixed in 4% (w/v) paraformaldehyde (PFA) and in situ hybridisation was carried out as described (Wilkinson and Green, 1990) with some modifications. Briefly, riboprobes were labelled with [³⁵S]UTP, mixed with hybridisation buffer and placed on pre-treated cultures. Cultures were hybridised overnight at 55°C, treated with 40 μg/ml RNAse and washed with 0.1×SSC for 20 minutes at 65°C. Slides were dipped in K12 emulsion (Ilford, UK) exposed in the dark for 7 days, then developed using Phenisol (Ilford,UK) for 5 minutes. The slides were analysed using bright field optics.

Neural crest explants cultured in chamber slides (Gibco BRL, UK) and transverse 6 μm paraffin-embedded sections of HH stage 9 quail embryos were used for immunohistochemical studies. Tissues were fixed either in methanol for detection of PCNA, using a polyclonal antibody to PCNA (1:50; Santa Cruz, USA), or in 4% PFA for detection of collagen type II, using the monoclonal antibody II-II6B3 (1:200), and for detection of FGFR1, FGFR2 and FGFR3 and osteonectin, using polyclonal antibodies (1:50; Santa Cruz, USA). PFA-fixed specimens were incubated for 30 minutes at 37°C with 1 mg/ml testicular hyaluronidase in phosphate-buffered saline (PBS) for antigen unmasking prior to incubation with primary antibodies. In all experiments, specimens incubated with non-immune serum were used as negative controls. The secondary antibodies used were: fluorescein-conjugated anti-goat IgG (immunoglobulin G), fluorescein-conjugated anti-mouse IgG, rhodamine-conjugated anti-rabbit IgG and biotin-conjugated anti-rabbit IgG (all from DAKO).

Cultures which had been immunostained for collagen type II were further stained with propidium iodide by a modification of the technique of Coles et al. (Coles et al., 1993) involving treatment with 4 mg/l propidium iodide (Sigma) and 100 mg/l RNAse (DNAse free; Sigma) in PBS. Specimens immunostained for FGFRs were counterstained with Methyl Green.

Fluorescent specimens were examined with fluorescence optics on a Leica confocal laser-scanning microscope (CLSM Aristoplan-Leica, Heidelberg, Germany). Bright field images were captured electronically from an Olympus BH2 microscope using a Kontron ProgRes 3012 digital camera and stored as Adobe Photoshop v3.0 files.

Neural crest explants cultured on 35 mm petri dishes were fixed in 3% (w/v) glutaraldehyde in 0.1 M sodium cacodylate buffer containing 3.5 mM calcium chloride at 4°C for 1 hour. Cultures were washed in distilled water, then placed in 1% (w/v) osmium tetroxide in distilled water for 2 hours. The specimens were dehydrated in an ascending ethanol series, then placed in a 1:1 mixture of ethanol and Agar 100 resin at room temperature for 30 minutes, and in resin only for 3 hours before polymerisation at 30°C for 24 hours. The petri dish was snapped away from the resin block and the area of interest cut from the block for ultrathin sectioning. Sections (70 nm thickness) were cut with an Ultracut ‘E’ ultramicrotome and treated with 25% (w/v) uranyl acetate in absolute methanol for 10 minutes, followed by Reynolds lead citrate (Reynolds, 1963) for 10 minutes. The ultrastructure was examined using a Jeol 1200EX electron microscope.

Neural crest explants were cultured either in the absence or presence of FGF2 at different concentrations (0.1, 1 and 10 ng/ml) and their survival and growth monitored daily. After adhering to the dish cells started to migrate out radially and proliferate. Enlargement of each individual explant was monitored by measuring outgrowth diameter, while neural crest cell proliferation was assessed by immunocytochemistry using the cell proliferation marker PCNA. After 5 days in culture, all explants had grown in diameter by approximately 15-20-fold (Fig. 1), but whereas many floating dead cells were observed in cultures lacking FGF2, all FGF-treated cultures exhibited little cell debris, suggesting improved cell survival in the presence of FGF2. Significant morphological differences were apparent in 5-day cultures between untreated cells, which displayed a very flattened morphology (Fig. 2A), and cells grown in the presence of 1 ng/ml FGF2, which were compact and retained their fibroblastic appearance (Fig. 2B). Most interestingly, cultures exposed to 10 ng/ml FGF2 developed aggregates of phase-bright cells at the centre, reminiscent of cartilage nodules in micro-mass cultures (Fig. 2C). Cells in the area immediately surrounding the nodules were so densely packed that it was difficult to distinguish individual cells.

After 10 days, morphological analysis showed that cells in control cultures were even more flattened and spread than at 5 days, with intracellular spaces increasing (Fig. 2D), consistent with poor cell survival in these cultures. In contrast, cells cultured in 1 ng/ml FGF2 were more densely packed (Fig. 2E). Culture diameters were higher in 10 day cultures treated with 0.1 or 1 ng/ml FGF2 than in controls (Fig. 1). This was consistent with the dramatic difference in the number of PCNA-positive cells in untreated cultures, where only occasional PCNA-positive cells were observed, compared with those exposed to 1 ng/ml FGF2 (Fig. 3A-D). In cultures grown in the presence of 10 ng/ml FGF2 for 10 days, the cartilage-like nodules were larger than in 5 day cultures (Fig. 2F, arrowheads), and secondary nodules often appeared adjacent to them (not shown). These FGF2-treated explants, however, did not differ in size from untreated ones (Fig. 1); significant PCNA staining was observed only at the edges of the explant, and very few cells in the cartilage nodule were PCNA positive (Fig. 3E,F).

This analysis demonstrates that the survival of pre-migratory neural crest cells is enhanced by FGF2 and that the type of response elicited depends on FGF concentration, with proliferation at 0.1 and 1 ng/ml, but chondrogenesis at 10 ng/ml.

To verify that chondrogenic differentiation was occurring in neural crest cells treated with 10 ng/ml FGF2, cultures were examined for several cartilage markers, including type II collagen, Alcian Blue, metachromasia after Toluidine Blue staining and ultrastructural appearance (Fig. 4). Radioactive in situ hybridisation showed collagen type II mRNA to be intensely expressed in the nodules developing in neural crest cultures exposed to 10 ng/ml FGF2 (Fig. 4A). The area surrounding the nodules, although densely packed with cells was largely negative. A few cells that expressed high levels of type II mRNA were also found in the vicinity, and these apparently became incorporated as the nodule enlarged (Fig. 4A). During the growth of the culture, isolated cells lying at a distance from the primary explant also underwent chondrogenic differentiation (Fig. 4B). Such cells were found to express type II mRNA at both intermediate and high levels (Fig. 4B), and probably gave rise to the secondary chondrogenic colonies, which accumulated as cultures aged.

Immunohistochemical evaluation in conjunction with confocal microscopy showed that the neural crest-derived nodules also contained type II collagen protein (Fig. 4C-F). Small clusters of cells first began to express type II collagen intracellularly (Fig. 4C), the granular nature of its appearance reflecting its location within the endoplasmic reticulum prior to secretion by the cell. As differentiation progressed with age of culture (Fig. 4D), type II collagen was found progressively in the extracellular matrix (Fig. 4E, arrowheads). In agreement with the in situ hybridisation findings, isolated cells undergoing chondrogenesis, as evidenced by the production of type II collagen protein, were also found scattered across the culture (Fig. 4F).

Neural crest cultures treated with 10 ng/ml FGF2 also stained positively with Alcian Blue, a histochemical marker for cartilaginous matrix (Fig. 4G). Alcian Blue-positive matrix was first identifiable at day 5 of culture, with staining intensity increasing as the nodules enlarged. Ultra-thin sections of the nodules, stained with Toluidine Blue, displayed the characteristic metachromasia seen in cartilage matrix (Fig. 4H) and revealed individual cells, rounded in shape and surrounded by large amounts of extracellular matrix, typical histological traits of cartilage.

Transmission electron microscopy also demonstrated that neural crest cells treated for 10 days with 10 ng/ml FGF2 had the typical ultrastructural characteristics of chondrogenic cells (Fig. 4I,J). Chondroblasts within the culture were surrounded by large areas of extracellular matrix, rounded in shape and with scalloped plasma membrane profiles (Fig. 4I). The endoplasmic reticulum was swollen with granular contents, presumably including collagen type II (Fig. 4I). Mature chondrocytes were embedded in lacunae within the extracellular matrix (Fig. 4J).

In order to establish whether there is a critical period of time during which FGF2 signalling is required for chondrogenic differentiation, explant cultures were exposed to 10 ng/ml FGF2 during several distinct ‘windows’ of time over 10 days (Fig. 5). As assessed by morphological analysis, 75% of neural crest explants differentiated into cartilage after continuous exposure to FGF2 for 10 days. Exposure during 0-48 and 0-72 hour periods caused chondrogenesis in 33% and 40% of the cultures respectively, while exposure of the cultures for the first 24 hours only was insufficient to drive chondrogenesis. However, when this period of exposure was omitted (i.e. 24-48 and 24-72 hour exposure periods), no cartilage differentiation took place (Fig. 5). This demonstrated that, exposure to FGF2 is necessary during the first 24 hours but is not required continuously, for neural crest explants to undergo chondrogenesis.

It has long been known that limb mesenchyme can form cartilage when cells are cultured at high density (Umansky, 1966), suggesting that cell density and critical mass are important factors in determining the expression of a chondrogenic phenotype. To evaluate this effect in neural crest chondrogenesis, we cultured mesencephalic fragments alongside half-sized fragments. No significant difference between the two differently sized populations was observed (Fig. 6A), demonstrating that the chondrogenic response of the mesencephalic neural crest is not dependent on a critical cell mass at explantation.

We also evaluated the chondrogenic potential of early and late migrating mesencephalic neural crest cells (early HH stage 8 or late stage 9, respectively). Both explant types were able to form cartilage at similar frequency (Fig. 6B) implying that, in terms of skeletogenic differentiation, there is no restriction in potency of early and late migrating mesencephalic neural crest cells.

We tested whether neural crest cells arising from different axial levels of the cranial neural folds have the same capacity to form skeletogenic tissue in the presence of FGF2. Neural crest explants from fore-, mid- and hindbrain regions were cultured in the presence of 10 ng/ml FGF2 and their chondrogenic response assessed after 10 days in culture. Although cartilage formation occurred in explants from all three categories, the incidence of chondrogenesis in neural crest cells from the fore- and hindbrain levels was significantly reduced compared with that from the mesencephalic region (Fig. 6C).

The majority of bones in the avian skull, whether membranous or endochondral, arise from the neural crest (Couly et al., 1993). In our experiments, when neural crest cells were cultured for a period of 28 days in the presence of 10 ng/ml FGF2 and supplemented (from days 11 to 28) with ascorbate, β-glycerophosphate and dexamethasone (to create a medium permissive for bone mineralisation), they developed an osteogenic phenotype (Fig. 7). As well as staining with Alcian Blue, which is indicative of a cartilaginous matrix (Fig. 7A), one third of these cultures (n=27) also stained brown with Von Kossa staining, an indication of calcium phosphate mineralisation (Fig. 7A). Sections through these cultures revealed the presence of hypertrophic chondrocytes (Fig. 7B), as well as areas of new mineralisation (Fig. 7B). Osteonectin, a non-collagenous bone matrix protein also found in hypertrophic chondrocytes, was detected immunocytochemically both in the extracellular compartment (Fig. 7C) and intracellularly (Fig. 7D). Ultrastructural analysis of five samples confirmed the presence of hypertrophic chondrocytes: large, pale cells, which filled their lacunae (Fig. 7E) and contained many swollen endoplasmic reticulum cisternae. The mineralised areas of these cultures were often associated with osteoid matrix (Fig. 7F, arrowheads) which stained with Toluidine Blue and were adjacent to areas of hypertrophic cartilage.

In some neural crest-derived cultures (n=10), there were sheets of Alizarin Red-positive tissue (Fig. 7G) in areas distinctly separate from Alcian Blue-stained cartilage, indicating the differentiation of membrane bone. Elsewhere nodules could be double stained with Alcian Blue and Alizarin Red, indicating the presence of associated bone and cartilage matrix, suggesting endochondral ossification (Fig. 7H).

Immunohistochemical staining showed FGFR1, FGFR2 and FGFR3 to be expressed ubiquitously in transverse sections of HH stage 9 embryos (Fig. 8A-C). FGFR1 (Fig. 8A) and FGFR2 (Fig. 8B) were expressed with particular intensity in the apices of the neural folds from which the neural crest arises. This was also the case for FGFR3, although staining intensity was more evenly distributed between the apices and the rest of the neurepithelium (Fig. 8C). FGFR1, FGFR2 and FGFR3 were also expressed by cultured neural crest cells (Fig. 8E-G), although expression was heterogeneous, with many negative cells also present in the cultures.

We have shown that pre-migratory neural crest cells can respond to FGF2 and that this is probably mediated by one or more of the FGFRs that we have shown to be expressed by these cells both in vivo and in vitro. Clearly, the responsiveness documented by Richman and colleagues for facial ectomesenchyme (Richman and Crosby, 1990; Richman et al., 1997) already exists in the precursor cells before or early in migration. This responsiveness differs depending on the concentration of ligand to which they are exposed and, therefore to the intensity of FGF-mediated signalling. Thus, the lower concentrations of 0.1 ng/ml and 1 ng/ml FGF2 elicited a dramatic proliferative response whereas a higher concentration, 10 ng/ml, produced skeletogenic differentiation.

In contrast to neural crest cells cultured in only 10% FCS, where cell death increased from 72 hours onwards, all concentrations of FGF2 permitted the survival and growth of the cultures. The skeletogenic response is not a feature of cell survival, as neural crest cells proliferated and became more packed, but never differentiated, in the presence of 0.1 and 1 ng/ml FGF2. This outcome demonstrates that the differential responses of proliferation and differentiation are distinct phenomena, dictated by different levels of FGF signalling.

We regularly observed that approximately 75% of cultures differentiated into cartilage when cultured in the presence of 10 ng/ml of FGF2. The failure of 25% of cultures to differentiate cannot be attributed to slight variations in developmental stage, as we found no difference in the incidence of chondrogenesis in cultures from early stage 8 and stage 9 neural crest. Similarly, by using explants of varying sizes, we have been able to eliminate critical mass of the initial explant as an explanation. The fact that chondrogenesis does not occur in all the cultures may be due to the stochastic nature of the primary explant system.

The first 24 hours after explantation are crucial if neural crest cultures are to go on to differentiate chondrogenically. This is consistent with FGF2 being present at the mesencephalic axial level during the early migration of neural crest cells (Kubota and Ito, 2000). Moreover, FGFRs are expressed in pre-migratory neural crest cells (Walshe and Mason, 2000; present study). The proportion of cultures differentiating increased significantly when the period of exposure to FGF2 was extended to 48 or 72 hours, but 48 hours exposure per se (e.g. during the 24-72 hour post-explantation window) was not sufficient to elicit a chondrogenic response. The reason for this loss of ability to respond to FGF2 is not immediately apparent but may relate to a phenomenon previously observed in ectomesenchyme. FGFR expression in the ectomesenchyme is downregulated in the absence of either a normal source of endogenous FGF ligand, or an implanted FGF2- or FGF4-soaked bead (Richman et al., 1997). If a comparable downregulation occurs in neural crest cells grown in the absence of FGF2 during the initial crucial period of culture, then the ability to respond to FGF at later times in culture might be absent or reduced.

Explants of neural crest taken from prosencephalic, mesencephalic and rhombencephalic levels of the neural primordium displayed different capacity to differentiate in response to 10 ng/ml of FGF2. This does not seem to be due to their inherently different survival and growth responses, as cultures from all three regions survived and grew equally well. It is possible that neural crest from different cranial regions contains skeletogenic precursor cells in equivalent proportions, but possesses inherent differences in responsiveness to FGF2. High levels of FGFR1, FGFR2 and FGFR3 have been detected in the anterior neural plate as the neural folds close (this study; Walshe and Mason, 2000); however, by stage 10-11, although FGFR1 is still expressed throughout the neural tube, FGFR2 and FGFR3 expression in the nervous system has become more restricted.

An alternative possibility is that quantitatively different proportions of cells with skeletogenic potential may be present along the anteroposterior axis of the cranial neural crest. Hence, it may be significant that fate mapping has revealed that forebrain neural crest produces little skeletal tissue, whereas most of the viscerocranium is derived from the hindbrain (Couly et al., 1993; Köntges and Lumsden, 1996). It is possible that skeletogenic cells involved in building the component parts of the skull (neurocranium, viscerocranium and dermatocranium) differ inherently in terms of the cues to which they can respond. This question will require further study.

Our data constitute the first report of pre-migratory neural crest cells being induced to form cartilage and bone in the absence of an epithelial tissue or chick embryo extract. This effect of FGF on neural crest is highly specific, as transforming growth factor β, another growth factor shown to play important roles in skeletal development and suture morphogenesis (Hall and Miyake, 1995; Opperman et al., 2000), is unable to induce chondrogenic differentiation in pre-migratory neural crest (S. S. and P. T., unpublished). The fact that a single protein (FGF2 at 10 ng/ml) is sufficient to elicit this complex differentiative response is remarkable. Cultures maintained for up to a month in medium permissive for bone mineralisation displayed three-dimensional nodules of cartilage projecting up from the surface of the culture dish and sheets of bone spreading across the surface. Skeletogenic neural crest in vivo produces four principal types of skeletal tissue in the skull of amniote vertebrates: primary and secondary cartilage, endochondral bone and membrane bone. We have used morphological and molecular markers in an attempt to distinguish between these skeletal phenotypes. We identified cartilage on the basis of its staining with Alcian Blue, expression of collagen type II mRNA and cellular morphology. It seems likely that the cartilage formed in the cultures corresponds to primary rather than secondary cartilage, as induction of the latter depends on biomechanical factors (Murray and Drachman, 1969) that are absent from the culture system. At least some of the bone appeared to be membranous, as it was formed in two-dimensional sheets often spatially independent from the cartilage nodules. Where bone was seen in association with cartilage, we conclude that it was endochondral, and the finding of hypertrophic chondrocytes associated with mineralising cartilage matrix supports such a conclusion. We suggest therefore that FGF2 is sufficient to elicit the formation of three of the four principal cranial skeletal tissues, thereby substituting for the normal in vivo cues and signals that exist within the migration environment and at sites of skeletogenic tissue interactions.

It has long been known that epithelial-mesenchymal interactions underlie the chondrogenic and osteogenic differentiation of cranial neural crest and of its ectomesenchymal progeny (reviewed by Thorogood, 1993). Indeed, it has been argued that the spatio-temporal distribution of such interactions during craniofacial development is a major epigenetic factor in skull morphogenesis (Thorogood, 1988). Regional differences in the composition of the extracellular matrix at the site of such interactions appear to be implicated and the transient epithelial expression of type II collagen at such locations has been reported (e.g. Wood et al., 1991; Cheah et al., 1991).

Although type II collagen alone is insufficient to elicit a full chondrogenic response from cultured neural crest cells (S. S. and P. T., unpublished), our present results suggest that FGF2 possesses this ability. Several lines of evidence support this idea: the regional distribution of FGFs and FGFRs in the embryonic head is thought to be responsible for the regional differences in ectomesenchymal proliferation that underlie the allometric growth of the facial processes (Wilke et al., 1997). Several epithelia that are known to elicit a skeletogenic response from pre-migratory neural crest (Bee and Thorogood, 1980) have been identified as a source of FGF ligand, including the facial ectoderm (Richman et al., 1997), the wall of the optic cup (Gao and Hollyfield, 1995) and the neuroepithelium (e.g. Crossley et al., 1996). Moreover, we have shown that neural crest cells that respond to epithelial signals during these interactions express a range of FGFRs. Finally, the known role of heparan sulfate proteoglycans in binding FGFs, as a prerequisite for receptor binding (reviewed by McKeehan et al., 1998), correlates with the earlier finding of an extracellular matrix involvement in the tissue interactions. We suggest, therefore, that the transient epithelial expression of type II collagen may not be the principal matrix contribution to skeletogenic interactions, and that other matrix components, such as the FGF-binding heparan sulphate proteoglycans, are more likely to play an important role in skeletogenesis. Differential localisation of FGF2, essential for chemotaxis of cranial neural crest cells migration, has indeed been found at the mesencephalic axial level (Kubota and Ito, 2000). At such restricted locations, migrating neural crest cells that express FGFRs encounter the bound FGF and respond either by proliferating or differentiating, depending upon the intensity of FGF signalling.

FGF2 is known to be expressed at low levels in the neural crest-derived cells at the margins of the membrane bones of the mouse developing skull, where FGFR2 is abundant (Iseki et al., 1997). These are areas where rapid proliferation of putative osteogenic stem cells occurs, perhaps in response to low levels of FGF signalling, as suggested by the present study. In craniosynostosis, however, most mutations in FGFRs are thought to be constitutively activating (Webster and Donoghue, 1997; Burke et al., 1998), leading to an upregulation of FGF-mediated intracellular signalling. Our findings about the skeletogenic effect of 10 ng/ml FGF2 suggest that this increased FGF signalling probably leads to a premature differentiative response from the sutural cells, as is observed in craniosynostosis. This interpretation is consistent with reports showing that implantation of FGF-soaked beads in mouse cranial sutures results in inhibition of cell proliferation and ectopic expression of osteopontin (Iseki et al., 1997; Iseki et al., 1999), and accelerated suture closure (Kim at al., 1998). Conversely, we find that reduction in levels of endogenous FGF2 in chick cranial vault switches neural crest-derived cells from a differentiative to a proliferative mode (R. Moore, P. F., A. C. and P. T., unpublished). Taken together, these data indicate that the concentration-dependent effects of FGF2 on cranial neural crest are consistent with the functional consequences of FGFR mutations in promoting craniosynostosis in humans.

We thank Paul Hunt, Rachel Moore, Chi-tsung Joseph Chan, Jonathan Britto and Ivor Mason for helpful discussions and for comments on the manuscript. Professor Paul Brickell provided the probe for type II collagen mRNA and Dr Larry Fisher provided the antibody LF-8 for osteonectin.