Transgenic mice were generated by microinjection of a construct with the αA-crystallin promoter linked to a mouse Fgf9 cDNA (Fig. 1A) 29. Eleven founders carrying the transgene were identified. Five of them had cataracts (data not shown). In the transgenic line OVE1070, in addition to lens defects, there were defects in the development of the skull (Fig. 1B, C). Mice in this family that are heterozygous for the transgene are born with 'dome-shaped' heads (Fig. 1C). This phenotype is even more pronounced in homozygous transgenic mice (data not shown).

The skeletal defects were examined by Alcian blue and Alizarin red staining of E15, P3, P10 and P30 mice (Fig. 1D–K). The skeletal preparations show a dramatic expansion of cartilage in the head region by E15 (compare Fig. 1E to 1D). Parietal bone development was particularly affected (Fig. 1E and inset). At P3, the phenotypic differences between the nontransgenic and FGF9 transgenic mice were still pronounced. At this age, the calvarial skull in nontransgenic mice is mostly ossified except at the sutures (Fig. 1F). In contrast, there is an expanded cartilage territory (blue) in the FGF9 transgenic mice (Fig. 1G). The transgenic mice also show a central hole in the skull where there is neither cartilage nor bone (Fig. 1G, inset, arrow). Development of the occipital bones appeared normal (Fig. 1G and inset). At P10, the Alcian blue stained tissues in the transgenic cranium were interspersed with Alizarin red stained regions suggesting that the cartilage territory was being replaced by bony structures (Fig. 1I). By 1 month of age, the cartilage territory was completely replaced by bone (Fig. 1K and inset).

The alterations in skeletal development in the FGF9 transgenic mice were examined at the histological level by analyses of hematoxylin and eosin stained sections of the cranium. No distinctive differences in morphology were seen at E11 (data not shown). By E13, more cranial mesenchymal cells were seen in the skull region of the FGF9 transgenic mice (Fig. 2B, B') and these cells were organized differently from nontransgenic mice (compare Fig. 2A, B). By E15, the cranial mesenchymal cells in the nontransgenic mice had initiated intramembranous ossification (Fig. 2C, C', arrow). In contrast, the mesenchymal cells in the FGF9 transgenic skulls differentiated into cells resembling chondrocytes (Fig. 2D, D'). Within the next few days these cells increased in size and became hypertrophic (Fig. 2F, F', H, H'). By P7, perichondrial cells (including the blood vessels) had begun to invade the cartilage territory (Fig. 2J, J'). By P18, the cartilage territory was transformed into trabecular bone and bone marrow (Fig. 2L, L'). These results considered together suggest that the cranial mesenchymal cells in the parietal region in these mice recapitulate the sequence of events that occurs during endochondral ossification.

To test if the alterations seen in the skulls of the OVE1070 mice were due to ectopic expression of the transgene, in situ hybridizations were performed using an S³⁵-labelled riboprobe that recognizes the SV40 portion of the transgene (Fig. 1A). Transgene expression was seen in the lens 29, the dorsal portion of the retinal pigmented epithelium 30 and also in the cranial mesenchymal cells beginning at E11 (Fig. 3B', D', F'). Expression levels peak by E13 and start to decrease by E15 (Fig. 3D', F'). Transgene expression is extinguished by E17 (data not shown). Section and whole mount in situ hybridizations show that the extraocular transgene expression is restricted to the mesenchymal cells overlying the future mid and hind brain regions (Fig. 3G, arrows; 3H-M). These results argue that the defects in skeletal development in the OVE1070 line are attributable to ectopic and transient expression of the FGF9 transgene.

The different stages of endochondral ossification are characterized by expression of different marker genes. Expression of Sox9, a member of the Sox family of transcription factors, has been shown to be essential for chondrocyte condensation 31. In particular, Sox9 is required for the expression of cartilage-specific extracellular matrix components such as Collagen II (Col2a1) 32. In the long bone growth plates, Col2a1 is expressed in the resting and proliferating chondrocytes, Ihh in prehypertropic chondrocytes and Col10a1 in hypertropic chondrocytes 3. Expression of the runt domain transcription factor CbfaI has been shown to be essential for differentiation of the cells in the osteoblastic lineage 33. Osteocalcin, one of the downstream targets of CbfaI, is expressed exclusively in the osteoblasts 34. Expression of these six marker genes was examined by in situ hybridization using S³⁵-labelled riboprobes (Figs. 4 and 5). No appreciable differences in expression of Sox9 between nontransgenic and FGF9 transgenic mice were seen at E11 (Fig. 4A, A', B, B'). However, significant induction of Sox9 expression was seen in the cranial mesenchymal cells of the transgenic mice by E13 (Fig. 4C, C', D and 4D'). Sox9 expression was seen to persist in the developing chondrocytes (Fig. 4F, F'). CbfaI and Osteocalcin are expressed during the normal program of intramembranous ossification in nontransgenic mice, but their expression was not seen in the corresponding regions of the embryonic transgenic heads (Fig. 4G–J').

Col2a1 expression was seen in the developing chondrocytes in the transgenic cranium but not the wildtype, at E15, E17 and P3 (Fig. 5A–F'). By P3 expression levels were found to be reduced in transgenic hypertropic chondrocytes (Fig. 5E–H'). In situ hybridizations at P7 showed that Col10a1 and Ihh are expressed in the hypertropic chondrocytes (Fig. 5I–L'). Expression of Col10a1 was not seen in the chondrocytes at P3 or before (data not shown). These results demonstrate that the transgenic cranial mesenchymal cells undergo endochondral ossification instead of intramembranous ossification.

Fgf9 has been reported to be expressed in cranial mesenchymal cells as well as in the dural mesoderm during embryonic development 16. Expression levels of the Fgf9 transgene and endogenous Fgf9 were compared by in situ hybridizations using an [³⁵S]-labelled Fgf9 riboprobe (Fig. 6A–D'). At E11 and E13, Fgf9 was not expressed at detectable levels in the cranial mesenchymal cells in nontransgenic mice. In contrast, Fgf9 expression was clearly present in the transgenic mice (Fig. 6A–D').

As FGF9 binds to and signals through FGFR2 and FGFR3, in situ hybridizations were performed to examine their expression levels (Fig. 6E-L'). Low levels of Fgfr2 expression can be seen in cranial mesenchymal cells at E11 in both the nontransgenic and transgenic mice (Fig. 6E, E', F and 6F'). Fgfr2 expression is maintained in the induced chondrocytes in the transgenic mice (Fig. 6H, H'). In contrast, Fgfr3 expression was not detected at E11 either in transgenic or in nontransgenic cranial mesenchyme (Fig. 6I, I', J, J'). Fgfr3 was expressed at low levels in the differentiating chondrocytes of the FGF9 transgenic mice (Fig. 6L, L'). In addition, crossing the OVE1070 mice with FGFR3 null mice 14 does not alter the "dome-head" phenotype (data not shown). These results taken together suggest that the developmental switch in the transgenic mice may initially be mediated by ectopic activation of FGFR2.

To test if the cranial mesenchymal cells proliferate in response to FGF9 expression, sections of E11 and E13 skulls were assayed for BrdU incorporation (Fig. 7). At E11 and at 13, the cranial mesenchymal cells in the parietal region of FGF9 transgenic mice show a significant increase in BrdU incorporation (Fig. 7A–D). These results suggest that FGF9 expression in the cranial mesenchymal cells results in increased proliferation.

Cell lineage tracing studies performed in mice and in chicks show that the cranial skeletogenic mesenchyme is derived from two sources, neural crest and mesoderm 353637. To determine if the cells in the parietal region that form ectopic cartilage in the OVE1070 mice are neural crest or mesodermal derivatives, we performed fate mapping experiments using Wnt1-Cre and R26R mice 3839. As has been described previously, Wnt1-Cre/R26R compound heterozygotes stably express β-galactosidase in neural crest cells and their descendants but not in the mesodermal derivatives 4041. OVE1070 mice were crossed with Wnt1-Cre/R26R mice and heads of E17 embryos were stained with X-Gal (Fig. 8). In the absence of Wnt1-Cre, lacZ positive cells were not detected in the R26R embryo (A, A') or in the FGF9/R26R embryo (B, B'). In the Wnt1-Cre/R26R embryo (C,C'), the parietal region (p) of the skull showed no evidence for lacZ activity, implying that neural crest derivatives do not contribute to this region of the skull. In the FGF9/Wnt1-Cre/R26R embryo (D, D'), there is also no blue staining in the (enlarged) parietal region. These results establish that the cells in the parietal region that form ectopic cartilage in the OVE1070 mice are not neural crest derived but are for the most part, derived from mesoderm. In summary, in family OVE1070, ectopic FGF9 expression switches the differentiation and proliferation properties of cranial mesodermal cells, resulting in inhibition of intramembranous ossification and conversion to a program of endochondral ossification.

The skeletal defects in this transgenic line (OVE1070) occur in mice that are either heterozygous or homozygous for the FGF9 transgene. It is, therefore, unlikely that the skull defects in this line of mice are due to disruption of a gene essential for membranous ossification of the parietal bones. The transgene expression pattern provides compelling evidence that the skull defects in these mice are due to ectopic expression of the transgene. In situ hybridization analyses show that the transgene is expressed in an appropriate spatial and temporal pattern to induce the developmental changes. In addition, the transgene is not expressed in regions where membranous ossification still occurs (Fig. 3G). Furthermore, transgene expression precedes Sox9 expression and condensation of the mesenchymal cells. Also, our results are consistent with the results of other studies. For example, recent studies of a mouse model for Apert syndrome, with a single amino acid change in FGFR2, support the notion that enhanced activation of FGFR2 can cause some cranial mesenchyme to convert to a chondrocyte differentiation program 10. In addition, FGF9 expression in a chondrocytic cell line is sufficient to induce expression of Sox9 42. Addition of FGF-2 in vitro to cranial mesenchymal cells from quail embryos can induce cartilage differentiation at high doses 17. Taken together, these results suggest that initiation of chondrogenesis in the cranial mesenchymal cells in this line of mice is due to ectopic expression of the FGF9 transgene.

In the OVE1070 family, expression of the FGF9 transgene is not only ectopic but also transient. Expression in the cranial mesenchyme was not detectable after E15. This implies that stimulation by FGF9 is sufficient to induce (competent) cranial mesoderm to switch to a chondrocytic differentiation program, but sustained expression of FGF9 is not required for the later stages of differentiation. Therefore, the endochondral ossification program, once initiated by FGF9, appears to become autonomous.

The reason for the ectopic Fgf9 expression is not clear at present. It is possible the transgene array has integrated in the neighborhood of an endogenous enhancer that directs expression to the cranial mesenchymal cells. Some of the integrated copies of the transgene still retain the αA-crystallin promoter since the transgene is still expressed in the lens 30. The transgene is also expressed in the dorsal margins of the retinal pigmented epithelium (RPE) and this expression transforms the RPE into neuroretina 30.

Expression of FGF9 in growth plate chondrocytes using the Col2a1 promoter results in reduced proliferation and terminal differentiation of chondrocytes 43. In addition, targeted deletion of FGFR3, one of the receptors through which FGF9 signals, results in mice with overgrowth of the long bones. These results suggest that a primary role of FGF signaling in the long bones is to act as a negative regulator of chondrocyte proliferation. These results are in contrast to our results that show that ectopic FGF9 expression in embryonic cranial mesenchymal cells induces overproliferation followed by endochondral ossification. How can these disparate observations be reconciled? One possible explanation is the differential timing of FGF9 expression. The Col2a1 promoter is active in the differentiated chondrocytes, while our transgenic FGF9 is expressed in undifferentiated cranial mesenchymal cells. Second, there are differences in the responding tissues. Chondrocytes in the long bones originate from the lateral plate mesoderm while the parietal mesenchyme is derived from cranial mesoderm. Since the stimulated cells have different developmental origins, there is no reason to expect that the responses will be identical. For example, although lens and corneal epithelial cells are related to each other, both morphologically and developmentally, they respond differently to FGF stimulation 2944. The reduction in proliferation in the long bones in response to FGF stimulation may be a unique property of the growth plate chondrocytes 4. In contrast, the FGF9-induced proliferation of cranial mesenchymal cells is consistent with the notion that FGFs, in general, act as mitogens during normal development.

Though FGF9 appears to be sufficient to induce chondrogenesis in the skull, the role of FGF9 during normal development of cartilage in the cranium is unclear. Our in situ hybridizations did not show any detectable expression of Fgf9 in the calvarial mesenchyme at E11 or E13. Expression is seen later during embryonic development in the endocranial portions of the mesenchyme and is downregulated during postnatal development 16. This expression pattern suggests that FGF9 is not important for initial specification of the cells in the chondrocytic lineage in the cranium. Consistent with this model, targeted deletion of FGF9 does not result in any visible skeletal abnormalities in the skull 2425. In contrast, in the transgenic mice, the ectopic FGF9 plays an instructive role and initiates chondrogenesis. FGF9 is known to signal through FGFR2 and FGFR3 45. Signaling by FGF9 in the transgenic heads is likely to be mediated initially through FGFR2 rather than FGFR3 as the parietal fate switch in the FGF9 transgenic mice is not rescued by the loss of FGFR3. After Sox9 expression has been induced and the chondrocytic program has been initiated (by E13), expression of the FGF9 transgene is no longer required. In contrast to the normal chondrocytic differentiation program in long bones, ectopic chondrogenesis in the cranium proceeds at a much slower rate. Though the reasons for this are presently unclear, we speculate that this may be due to the multi-step nature of the endochondral differentiation program. Respecification and reprogramming of the parietal environment is required, and this is accomplished in a less timely fashion than at the normal sites of differentiation.

Our findings are in apparent conflict with some of the current models for the roles of FGF receptors in intramembranous ossification. Autosomal dominant mutations in FGFR2 or FGFR3 lead to premature differentiation and fusion of the skull sutures in humans and these effects are thought to be the consequence of enhanced receptor activity 9. Based upon such a model, our transgenic mice would be predicted to exhibit an analogous phenotype, i.e. craniosynostosis. Inappropriate activation of the receptors in the cranial mesenchyme would be predicted to lead to premature differentiation. However, our studies indicate that ligand-mediated activation of FGFR2 induces proliferation of embryonic cranial mesenchymal cells and subsequent differentiation into cartilage. By extrapolation then, our results imply that mutations in the FGFR2 gene in human patients do not result in ligand-independent constitutive activation of the receptor. This prediction is supported by the finding that targeted inactivation of FGFR2 in mice causes developmental alterations in cell types that are not affected by the putative gain-of-function mutations. An alternative explanation is that transient stimulation of FGFR2 (as seen in our FGF9 transgenic mice) may lead to a qualitatively different cellular response compared to sustained ligand-independent stimulation of FGFR2 (as seen in the gain-of-function mutations in human patients). In either case, our results show that a single FGF can function as an instructive signal, inducing a specific cell type to switch from one differentiation program to an autonomous, alternative differentiation pathway.

The construction of the FGF9 transgene and the generation of transgenic mice have been described previously 29. The coding region of mouse Fgf9 (a gift from Dr. David M. Ornitz, Washington University School of Medicine, St. Louis) was inserted between the αA-crystallin promoter 46 and the SV40 small t intron/polyadenylation sequences of the CPV2 vector 47. FGF9 transgenic mice were identified by isolating genomic DNA and screening by PCR, using primers specific to the SV40 portion of the transgene: 5'-GTGAAGGAACCTTACTTCTGTGGTG-3' (SV40A) and 5'-GTCCTTGGGGTCTTCTACCTTTCTC-3' (SV40B). The PCR cycle conditions were as follows: denaturation at 94°C for 30 seconds, annealing at 58°C for 30 seconds and extension at 72°C for 60 seconds, for 35 cycles. A final extension step of 72°C for 10 minutes was included.

Skeletal preparations were performed as described previously 48. Briefly, embryos were first skinned, and eviscerated, then fixed in 95% alcohol. After fixation, the samples were stained with Alcian blue, for 1–2 days. The samples were then destained in 95% ethanol for eight hours, and cleared in 2% potassium hydroxide from eight hours to overnight depending on the size of the specimen. After clearing, the samples were stained in Alizarin red/1% potassium hydroxide overnight. The samples were cleared in 1% potassium hydroxide and 20% glycerol/1% potassium hydroxide for 2–3 days. Subsequently, the samples were allowed to harden in 1:1 95% ethanol/glycerol for one day and then transferred to absolute glycerol for storage and photography.

For routine histology, embryos were obtained from timed pregnancies using FVB/N females that were mated to heterozygous FGF9 transgenic males. Embryos were delivered by Caesarean section, fixed in 10% formalin, dehydrated, embedded in paraffin, sectioned (5 μm) and stained with hematoxylin and eosin.

To analyze the expression of different markers, in situ hybridizations were performed. Mouse cDNA clones for Col2a1, Col10a1, Ihh and Sox9 were obtained from Dr. Benoit DeCrombrugghe (University of Texas, MD Anderson Cancer Center, Houston). Mouse cDNA clones for CbfaI and osteocalcin were obtained from Dr. Gerard Karsenty (Baylor College of Medicine, Houston). cDNAs clones for Fgf9, Fgfr2 and Fgfr3 were obtained from Dr. David Ornitz (Washington University School of Medicine, St. Louis). To analyze expression of the FGF9 transgene, a [³⁵S]-UTP-labeled riboprobe specific to the SV40 sequences of the transgene was made (see Fig. 1). To assay for endogenous Sox9 expression, a Sox9 antisense probe was synthesized using a HindIII-digested mouse Sox9 cDNA and T7 RNA polymerase (Promega). For Col2a1, the antisense probe was synthesized using EcoRI-digested DNA and T3 RNA polymerase. The antisense probe for Col10aI was synthesized using a HindIII-digested mouse Col10a1 cDNA and T3 RNA polymerase. The antisense probe for Ihh was synthesized using a BamHI-digested mouse Ihh cDNA and T7 RNA polymerase. The antisense probe for Fgf9 was synthesized using a KpnI-digested mouse Fgf9 cDNA and SP6 RNA polymerase. The antisense probe for CbfaI was synthesized using an EcoRI-digested mouse CbfaI cDNA and T7 RNA polymerase. The antisense probe for osteocalcin was synthesized using BamHI-digested mouse osteocalcin cDNA and T3 RNA polymerase. The antisense probe for Fgfr2 was synthesized using an EcoRI-digested mouse Fgfr2 cDNA and T7 RNA polymerase. The antisense probe for Fgfr3 was synthesized using a HindIII-digested mouse Fgfr3 cDNA and T3 RNA polymerase. In situ hybridizations on tissue sections were done using hybridization and washing conditions described previously 49. The hybridized slides were soaked in Kodak NTB-2 emulsion, dried and exposed for 3–5 days at 4°C. Following development and fixation, the slides were lightly counterstained with hematoxylin.

DNA replication was examined by BrdU incorporation as described previously 50. Cell proliferation was analyzed by counting the number of BrdU positive nuclei from more than 100 cells in a defined area in the parietal region in three serial sections from four (E11) or six (E13) nontransgenic and FGF9 transgenic heads. The counts for BrdU-positive cells in nontransgenic and transgenic mice were compared using the t-test (p < 0.01).

For whole mount in situ hybridizations, tissue samples were washed thrice in PBS and fixed in 4% paraformaldehyde. Hybridizations were performed using digoxygenin-labeled sense or antisense SV40 riboprobes following standard procedures 51.

Embryos were obtained from timed pregnancies of R26R homozygote females mated to OVE1070/Wnt1cre double transgenic males 36. X-Gal staining was performed as described previously 44. Briefly, heads of E17 embryos were collected and fixed for 2 hours at 4°C in 0.1 M phosphate buffer (pH 7.3) containing 2% paraformaldehyde, 0.2% glutaraldehyde. Following fixation, the tissue samples were rinsed thrice at room temperature in 0.1 M phosphate buffer (pH 7.3) containing 0.01% sodium deoxycholate, 0.02% NP-40, 2 mM MgCl₂, then stained in an X-gal substrate solution (0.01% sodium deoxycholate, 0.02% NP-40, 2 mM MgCl₂, 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, 1 mg/ml X-gal in 0.1 M phosphate buffer (pH 7.3)).