INTRODUCTION

Signaling through FGF receptors (FGFRs) plays a major role in several developmental processes, and evidence from human and mouse genetics has highlighted its role in bone development (Goldfarb, 1996; Webster and Donoghue, 1997; Naski and Ornitz, 1998; Colvin et al., 1996; Chen et al., 1999). Skeletal formation proceeds through two distinct mechanisms (Gilbert, 1994). Intramembranous ossification is involved in the formation of the flat bones of the skull, where mesenchymal cells condense and differentiate directly into osteoblasts that secrete bone matrix proteins, ultimately resulting in calcification and bone formation. However, endochondral ossification, which accounts for the formation of most of the skeletal bones, is characterized by the sequential formation of a cartilage anlage, which then becomes calcified cartilage, and provides the template for bone formation. Longitudinal bone growth is a strictly regulated process that requires proper proliferation and differentiation of chondrocytes at the epiphyseal growth plate. Unregulated FGF signaling has been shown to affect both processes of bone formation. Gain-of-function mutations in FGFR3 are responsible for several forms of human dwarfism (Shiang et al., 1994; Rousseau et al., 1994; Tavormina et al., 1995), of which the most common is achondroplasia, while activating mutations in FGFR1 and FGFR2 are generally associated with craniosynostosis syndromes (Gorlin, 1997).

In line with the genetic evidence indicating that excessive FGF signaling retards long bone growth, we previously showed (Sahni et al., 1999) that FGF signaling in chondrocytes resulted in inhibition of proliferation. Furthermore, we have also shown that treatment of chondrocytes with FGF induces phosphorylation of STAT1, a phenomenon that appears to be chondrocyte and perhaps FGFR3 specific. We have also shown, through the use of chondrocytes from Stat1-null mice, that STAT1 function is required to mediate the inhibitory effect of FGF on chondrocyte proliferation (Sahni et al., 1999). STAT1, originally identified as a signal transducing molecule in the interferon (IFN) pathway, is activated by tyrosine phosphorylation to dimerize and translocate into the nucleus where it then functions as a transcriptional regulator. STAT1 activity has previously been shown to be associated with antiviral and antiproliferative effects, most of which could be ascribed to its role as an essential mediator of IFN responses (Darnell, 1997; Levy, 1999). Thus, our findings were the first example of a physiologically relevant link between STAT1 function and a different signaling system.

As our studies indicated that STAT1 is a negative downstream regulator of chondrocyte proliferation (Sahni et al., 1999), it was expected that Stat1^−/− mice would display unregulated bone growth. However, previous studies on Stat1-null mice did not reveal any gross skeletal phenotype (Durbin et al., 1996; Meraz et al., 1996). Therefore, to address the question of whether the role of STAT1 in mediating the FGF response that we observed in culture was also manifested in vivo, we crossed Stat1^−/− mice with a mouse transgenic line overexpressing FGF2 (TgFGF). Adult TgFGF mice show major defects in the long bones and in the cranial bones (Coffin et al., 1995). The long bone phenotype is very similar to the chondrodysplasia described in humans with activating mutations in FGFR3. We, thus, studied the phenotype of these crosses to determine the relevance of STAT1 to FGF signaling in vivo.

We now show that the dwarfism phenotype of TgFGF mice is accompanied by impaired proliferation and increased cell death of chondrocytes in the growth plate. Both of these FGF-induced phenotypes are substantially corrected in a Stat1^−/− background. Analysis of the long bone developmental program, in wild-type and Stat1^−/− mice, revealed that the growth plate of Stat1^−/− mice is characterized by an expansion of the proliferation zone and reduced apoptosis. These phenotypes are confined to the early postnatal development, and attenuate at adult stages. Our data demonstrate that under both physiological and unregulated FGF signaling, STAT1 function plays a crucial role in regulating the proliferation and apoptosis of the chondrocytes of the growth plate during early bone development.

MATERIALS AND METHODS

FGF2 transgenic and Stat1^−/− cross-breeding

A mouse line with constitutive overexpression of FGF2 and deficient for Stat1 was generated by crossing mice with a null Stat1 allele on a C57BL/6J background (Durbin et al., 1996) with mice expressing the human FGF2 cDNA under the control of a constitutive phosphoglycerate kinase promoter on a FVB/N background (Coffin et al., 1995). The double mutant line, TgFGF214/Stat1^−/− , carried the human FGF2 gene as an hemizygotic allele. All animals were maintained under specific pathogen-free conditions, according to institutional guidelines. Analysis of the phenotypes exhibited by animals with different combinations of Stat1 alleles and the FGF2 transgene were performed with F₂ animals resulting from crosses of TgFGF/Stat1^+/− mice with Stat1^+/− mice. Only littermate mice were compared. The phenotypes of TgFGF/Stat1^+/− and TgFGF/Stat1^+/+ mice were indistinguishable. For all the experiments shown, TgFGF/Stat1^−/− and Stat1^−/− control littermates were produced from crosses between F2 TgFGF/Stat1^−/− and C57Bl/6J Stat1^−/− animals. Similarly, TgFGF/Stat1^+/+ and Stat1^+/+ control littermates were obtained by crossing F₂ TgFGF/Stat1^+/+ with wild type C57Bl/6J mice. For all the experiments presented we used males with similar bodyweight at each given age.

Genotyping

Stat1 genotyping was performed by allele-specific PCR assay on tail DNA, using the primers: 5′-TAA TGT TTC ATA GTT GGA TAT CAT-3′ (neo-resistance gene specific), 5′-CTG ATC CAG GCA GGC GTT G-3′, 5′-GAG ATA ATT CAC AAAATC AGA GAG-3′. Reaction conditions were 30 cycles of 1 minute of denaturation at 94°C, 1 minute of annealing at 52°C and 2 minutes of elongation at 72°C. The PCR products were 150 bp and 320 bp for wild-type and mutant Stat1 alleles, respectively. Presence of the FGF2 transgene was confirmed by genomic PCR as described previously (Coffin et al., 1995).

Organ and cell culture

Embryonic metatarsal rudiments from E15.5 wild type, Stat1^−/− , TgFGF and TgFGF/Stat1^−/− were isolated and cultured as previously described (Sahni et al., 1999). Metatarsals were maintained in organ-culture for 48 hours and labeled with 50 μg/ml of BrdU solution for the last 6 hours of the culture. Bone rudiments were fixed in 4% paraformaldehyde, paraffin embedded, and cut in 5 μm sections. Bone sections were stained with anti-BrdU monoclonal antibody (Boehringer Mannheim) and a Vectastain Elite ABC Kit was used to stain cells according the manufacturer’s manual. Primary chondrocytes were isolated from 10-day-old growth plates and cultured, at equal density, for 24 hours, then labeled with BrdU for the last 6 hours of the experiment. The cells were fixed and stained with Hoechst and with TUNEL using in situ cell death detection kit according the manufacturer’s methods (Boehringer Mannheim). Cells were examined under a fluorescence microscope.

Histology and immunohistochemistry

Mice from the four genotypes described above were injected with BrdU 100 μg/g bodyweight at various ages then sacrificed after 2 hours. Femurs, tibia and calvarial bones were dissected and fixed in 4% paraformaldehyde after decalcification. The bones were embedded in paraffin and 5 μm sections were cut. Bone sections were either stained with anti-BrdU antibody, anti-Collagen X antibody, with Toluidine Blue, or with Hematoxylin and Eosin. The staining with TUNEL was performed using a POD converter followed by DAB staining as described above. The sections were counterstained with Alcian Blue.

RESULTS

Absence of STAT1 corrects the chondrodysplasia phenotype of TgFGF mice

In order to address the role of STAT1 in mediating FGF effects on skeletal development in vivo, we crossed Stat1^−/− mice with TgFGF mice to generate littermates for four genotypes, i.e. Stat1^−/− , TgFGF/Stat1^+/+, TgFGF/Stat1^−/− and wild type as described in Materials and Methods, and analyzed their skeletal phenotype.

Radiographic analysis of the long bones of 3-month-old mice (Fig. 1A) showed that the length of both femurs and tibias were significantly reduced in TgFGF mice, consistent with previously published data (Coffin et al., 1995). Interestingly, although at 3 months of age no obvious differences were observed between wild-type and Stat1^−/− bones, the crosses between TgFGF and Stat1^−/− (TgFGF/Stat1^−/−) showed a significant correction of the dwarfism phenotype observed in TgFGF hemizygotes (Fig. 1A,B). These results suggest that the absence of STAT1 can attenuate the inhibitory effect on bone growth produced by the unregulated expression of the FGF2 transgene. At 3 months the epiphyseal growth plate of murine bones shows significantly decreased chondrocytic turnover, and therefore the reduction of the growth plate could mask any effect that STAT1 might have on chondrocyte proliferation and/or differentiation at earlier ages. Thus, to determine the role of STAT1 in regulating FGF signaling in developing long bone, we studied the effect of Stat1 deletion on the progression of chondrocyte proliferation and differentiation during early stages of development ranging from embryonic day 17 (E17) to postnatal day 20 (P20).

The inhibition of chondrocyte proliferation by FGF in vivo is dependent on functional STAT1

In order to determine whether the correction of the FGF-induced chondrodysplasias by the loss of Stat1 was due to changes in the dynamics of chondrocyte proliferation, we performed organ culture of metatarsal bone rudiments that were isolated from E15.5 wild type, TgFGF, Stat1^−/− and TgFGF/Stat1^−/− embryos. Metatarsal rudiments from each genotype were cultured for 48 hours and labeled with BrdU for the last 6 hours of culture. Histological analysis (Fig. 2A) showed that the length of the metatarsals was significantly increased in Stat1^−/− embryos compared with either wild-type or TgFGF metatarsals. TgFGF metatarsals were slightly shorter than the wild type, but significantly wider in their extremities. Crossing TgFGF and Stat1^−/− mice resulted in a significant increase in the length of the rudiments over that of the TgFGF, suggesting that the absence of STAT1 accelerates the proliferation and attenuates the inhibitory effect of the FGF2 transgene on the chondrocytes of the proliferating zone. To test this theory, we determined the number of DNA synthesizing cells in the proliferating zone of metatarsal growth plates from wild-type, Stat1^−/−, TgFGF and TgFGF/Stat1^−/− embryos. BrdU labeling indicated that the rate of chondrocyte proliferation was significantly reduced in TgFGF compared with wild-type and Stat1^−/− rudiments, while the rate of DNA synthesis in the proliferating zone of TgFGF/Stat1^−/− was increased compared with their TgFGF littermates (Fig. 2B). Thus the loss of STAT1 relieved the inhibitory effect of the FGF transgene on chondrocyte proliferation.

To determine whether the dwarfism phenotype observed in TgFGF mice was due to an alteration in the proliferation of chondrocytes, we performed a detailed analysis of the effect of the FGF2 transgene and Stat1 on the rate of cell proliferation in the reserve and the proliferating zone of long bones at various stages of mouse development. Mice representing the four genotypes described above were injected intraperitoneally with BrdU and sacrificed after 2 hours. Femurs and tibia were decalcified and longitudinal bone sections were immunostained with specific antibody against BrdU. Positive cells were counted in the reserve and the proliferation zones of four consecutive sections from each femur (Table 1). Both the reserve and the proliferating zones of TgFGF growth plates exhibited a significant decrease in cell proliferation compared with either wild-type or Stat1^−/− at E17 and P1. The absence of Stat1 resulted in increased cell proliferation in mice that expressed the FGF2 transgene (TgFGF/Stat1^−/−). Even at later stages of development, i.e. P10 and P20, the rate of DNA synthesis in the chondrocytes of the proliferating zone was decreased in TgFGF, and returned to wild-type levels in TgFGF/Stat1^−/− mice. Interestingly, the rate of DNA synthesis of chondrocytes in the proliferating zone of Stat1^−/− growth plates was higher than that determined in wild-type animals up until P10.

Constitutive FGF signaling induces chondrocyte apoptosis in vivo

The inhibitory effect of FGF signaling on bone growth and chondrocyte proliferation could arise from increased cell death as well as from inhibition of proliferation. As the presence of apoptotic cells in cultures of human chondrocytes from fetuses with thanatophoric dysplasia I (TDI) has been reported (Legeai-Mallet et al., 1998), we addressed the question of whether unregulated FGF signaling caused chondrocyte apoptosis in vivo and whether STAT1 was involved in this process. Femur sections from the same animals used for proliferation studies were analyzed for the effect of constitutive FGF signaling on the degree of apoptosis in growth plate chondrocytes. TUNEL staining of growth plate sections from the four genotypes showed that the FGF2 transgene caused increased apoptosis both in the cells of the reserve zone and of the proliferating zone as early as day 1. The apoptosis in TgFGF remained significantly increased at day 10 compared with both wild-type and Stat1^−/− mice, in which apoptosis was very low (Fig. 3). The levels of cell death observed in TgFGF were significantly decreased in the Stat1^−/− background, suggesting that activation of STAT1 by excessive FGF signaling leads not only to reduced proliferation but also to increased apoptosis in vivo. In addition to the apoptosis observed in the reserve and the proliferating zones of the growth plate, we also detected apoptosis in the hypertrophic zone. This is a normal event that is linked to the progression of chondrocytes towards a terminally differentiated stage characterized by cell death and the creation of lacunae, followed by vascular invasion and bone formation (Gibson, 1998). Considerable cell death in the hypertrophic zone was observed in all the four genotypes (Fig. 3).

In order to distinguish the naturally occurring apoptosis that was due to the differentiation of chondrocytes to hypertrophy from premature apoptosis caused by unregulated FGF signaling, we determined the number of apoptotic cells in each individual zone of the growth plates (Table 2). Analysis of the reserve zone and the proliferating zone of the growth plate reveal that excessive FGF signaling in TgFGF caused increased cell death in these two zones even at early stages of development, i.e. E17. In wild-type and Stat1^−/− mice, the number of TUNEL-positive cells remained significantly lower than in TgFGF mice at all ages analyzed. The crosses between TgFGF and Stat1^−/− showed a significant decrease in cell death of chondrocytes from the reserve as well as the proliferating zones.

The data presented in Table 2 show that the number of apoptotic cells observed in the reserve and proliferation zones was generally high. It has been observed that clearance of apoptotic cells in the growth plate is extremely slow, owing to the absence of macrophages in this tissue (Roach and Clarke, 1999). Thus, the number of apoptotic chondrocytes determined at any given time is likely to be an overestimate of the rate of apoptosis, owing to the accumulation of TUNEL positive cells over time. Although it is well established that apoptosis in the hypertrophic zone is a process that is part of the normal endochondral ossification program (Gibson, 1998), we also found that the rate of cell death in this zone was higher in the TgFGF mice than in wild-type or Stat1^−/− mice. These data suggest that the effect of FGF signaling was not limited to the proliferating chondrocytes but it also influenced later stages of chondrocyte differentiation. However, while the absence of STAT1 repressed the apoptosis induced by the FGF transgene in the reserve and the proliferating zone, it had little or no effect on the excessive apoptosis observed in the hypertrophic zone of TgFGF mice (compare TgFGF with TgFGF/Stat1^−/−, Table 2), suggesting that Stat1 is not crucial in mediating FGF-induced apoptosis in the hypertrophic zone. It should also be noted that the proliferating zone of growth plates of Stat1^−/− animals contained fewer apoptotic cells than the wild-type animals, particularly at P1 and P10.

To address the role of Stat1 in mediating FGF-induced apoptosis in chondrocytes further, we performed studies in culture. Primary chondrocytes were isolated from femurs and tibias growth plates and, after 24 hours in culture, were labeled with BrdU for 6 hours, fixed and stained either with anti-BrdU antibody or with TUNEL and Hoechst to visualize the nuclei. Cells that were maintained in culture for 24 hours showed a decreased rate of DNA synthesis in TgFGF, and higher levels of proliferation in Stat1^−/− compared with wild type (data not shown). The expression of the FGF2 transgene induced a significant amount of apoptosis in proliferating chondrocytes compared with the wild-type and Stat1^−/− mice in which apoptosis was almost undetectable. The absence of STAT1 repressed the apoptotic effect of TgFGF (Fig. 4A,B). Accordingly, treatment of wild-type cells with exogenous FGF1 (2 ng/ml) for 24 hours led to a considerable increase in the number of apoptotic cells, but this effect was much less pronounced in Stat1^−/− chondrocytes (not shown).

In summary, excessive FGF signaling greatly increased the level of apoptosis in proliferating chondrocytes and this effect required STAT1 function. Apoptosis of hypertrophic chondrocytes was also increased in the FGF2 transgenics but this effect was not counteracted by the absence of STAT1. Stat1-null mice showed reduced apoptosis compared with wild type throughout the growth plate and the increased apoptosis observed in primary cultures of TgFGF chondrocytes was reduced by the absence of STAT1. These findings raise the question of whether the reduction in cell death observed in the proliferating zone of TgFGF/Stat1^−/− is a general consequence of the lack of STAT1 or reflects a specific requirement for STAT1 in FGF-mediated chondrocyte apoptosis.

Apoptosis in the calvarial bones of TgFGF mice is not corrected with the lack of STAT1

To address the specificity of the FGF/STAT1 pathway in regulating cell death, we studied the effect of excessive FGF signaling on cell death in osteoblasts during intramembranous ossification of calvarial bones. We have recently demonstrated that FGF signaling enhances apoptosis in the osteoblasts of the calvarial post-frontal suture (PF sutures) of TgFGF mice (Mansukani et al., 2000), but we could not detect STAT1 activation in this system. To address the role of STAT1 in FGF-induced apoptosis in the calvarial osteoblasts, calvaria were isolated from P1 and P10 mice that represented the four genotypes described above. Serial sections performed along the PF sutures were analyzed for the presence of apoptotic cells. At early stages (P1) no significant difference in the rate of cell death was detected between wild-type, Stat1^−/−, TgFGF and TgFGF/Stat1^−/− mice (not shown). However, P10 TgFGF mice exhibited high levels of apoptosis in the PF sutures compared with both wild-type and Stat1^−/− mice, where the levels of cell death were low (Fig. 5A). Interestingly, when TgFGF were crossed with Stat1^−/− (TgFGF/Stat1^−/−) the rate of cell death remains similar to that observed in TgFGF mice. This suggests that, in contrast to growth plate chondrocytes, FGF-mediated apoptosis in osteoblasts is independent of the STAT1 pathway. Histological sections of calvarial bone (Fig. 5B) showed that in contrast to wild-type and Stat1^−/− calvaria, where the PF sutures are near closure, the PF suture in TgFGF and TgFGF Stat1^−/− remained wide open. Thus, in addition to STAT1-independent apoptosis, the Stat1 deletion also failed to correct the phenotype observed in calvarial sutures of TgFGF mice.

Effect of the FGF/STAT1 pathway on bone development

Longitudinal bone growth occurs by endochondral ossification and takes place in the epiphyseal growth plates of developing skeleton. The growth plate is characterized as a series of morphologically discrete zones that progress from the epiphyseal to the metaphyseal regions. As the prechondrocytes of the reserve zone further differentiate, they become proliferating chondrocytes and form the columnar zone. Proliferating chondrocytes mature and terminally differentiate to form the hypertrophic zone of the growth plate. The rate of the proliferation and the differentiation of chondrocytes contribute to the overall growth in the length and shape of a given bone.

To determine the role of STAT1 in regulating bone development in the presence and absence of unregulated FGF signaling, we performed a detailed histological analysis of the development of the growth plate. This demonstrated a general expansion of the chondrocyte population in the reserve zone, as well as in the proliferating and hypertrophic zones in the developing bones of Stat1-null mice. At P5, the reserve zone of Stat1^−/− femurs was larger than that of the wild type and of the transgenics, a phenotype that was also observed in the TgFGF/Stat1^−/− femurs (Fig. 6A). The reserve zones of both Stat1^−/− and TgFGF/Stat1^−/− mice also showed some evidence of premature chondrocyte hypertrophy, a prelude to the formation of the secondary center of ossification (Fig. 6A). The progression of chondrocytes from the reserve zone towards the proliferating zone appeared to be accelerated in Stat1^−/− mice, which showed an increase in the length of the proliferating zone compared with the other three genotypes. In TgFGF mice, the reduced proliferating zone was restored to normal by the loss of STAT1. The differentiation of proliferating chondrocytes to hypertrophy leads to accumulation of hypertrophic chondrocytes, that is also enhanced in the Stat1^−/− growth plates. In contrast, in TgFGF mice the hypertrophic zone is dramatically reduced and this phenotype is not corrected by the absence of STAT1 (Fig. 6A).

At later stages of bone development (P15) the differences between the growth plates of wild-type and Stat1^−/− mice become attenuated, but the hypertrophic zone of Stat1^−/− mice remains larger (Fig. 6B). In the TgFGF/Stat1^−/− mice, the Stat1 mutation still attenuated the effect of the transgene in the proliferating zone, but not in the hypertrophic zone (Fig. 6B). Furthermore, more extensive secondary centers of ossification appear to be present in the epiphyseal region of Stat1^−/− and TgFGF/Stat1^−/− femurs.

To quantitate the histological observations of the developing growth plates shown in Fig. 6, we measured the height of the proliferating zone and the hypertrophic and prehypertrophic zone containing collagen type X-positive cells. As shown in Fig. 7A, between E17 and P5 there is a significant increase in the height of Stat1^−/− proliferating zone compared with the wild type. These differences attenuate, as the Stat1 mutant mice become older (P10-P20). However, in TgFGF mice, the height of the proliferating zone is significantly reduced at all ages analyzed. Absence of STAT1 in TgFGF mice (TgFGF/Stat1^−/−) leads to a significant increase of the proliferating zone.

To further analyze the apparent acceleration of early bone growth observed in Stat1^−/− mice, we measured the distance between opposing femoral growth plates in each of four mouse genotypes under study. This provides an indication of the rate of trabecular bone deposition resulting from growth plate activity. Fig. 7B shows that Stat1-null mice had significantly higher distances than the wild type between the proximal and distal growth plates up to P10, indicating accelerated bone growth. At earlier times (E17-P1), the distance between growth plates of TgFGF mice was similar to the wild type, yet the proliferative zone was reduced (Fig. 7A), possibly indicating an early onset of ossification. At P5-P10, the distance between the two growth plates became progressively shorter. The absence of STAT1 attenuated the effects seen in TgFGF mice.

Thus, the results presented in this section show that Stat1^−/− mice exhibit an early bone phenotype, characterized by an expansion of the growth plate proliferative zone and a temporary acceleration of bone growth, in line with the observation that Stat1-null growth plate chondrocytes exhibited slightly higher DNA synthesis and reduced apoptosis compared with the wild type. These alterations in bone development, however, disappeared in older animals. Conversely, TgFGF mice showed a reduction in the height of the proliferative zone, with no indication of a further block in differentiation, and shorter bones. TgFGF/Stat1^−/− mice showed a partial restoration of the height of the proliferative zone and complete restoration of bone length.

DISCUSSION

The experiments described were undertaken to further define the role of STAT1 in transducing FGF signaling during endochondral ossification. We had previously shown that FGF signaling inhibited chondrocyte proliferation, and that this inhibition required STAT1 function (Sahni et al., 1999). It would therefore have been expected that Stat1 null mice would display a bone phenotype, resulting from unregulated chondrocyte proliferation, and perhaps similar to that exhibited by Fgfr3-null mice, which display prolonged chondrocyte proliferation resulting in abnormally long and distorted bones (Colvin et al., 1996; Deng et al., 1996). Such a phenotype had however not been reported in Stat1^−/− mice (Durbin et al., 1996; Meraz et al., 1996). Therefore to address the question of whether STAT1 plays a role in regulating bone growth in vivo, we undertook experiments aimed at determining whether the chondrodysplasic phenotype exhibited by mice with unregulated FGF signaling was corrected in a Stat1^−/− background. Furthermore, we examined in detail the dynamics of early long bone development in Stat1-null mice, and compared it with that of wild-type and FGF-induced chondrodysplasic mice.

The chondrodysplasia observed in transgenic mice overexpressing FGF2 results from reduced proliferation and increased apoptosis of growth plate chondrocytes

For our studies we used a mouse model of achondroplasia that consists of mice overexpressing a human FGF2 transgene under the control of the PGK promoter. As previously described (Coffin et al., 1995), these mice overexpress FGF2 ubiquitously but the major phenotypic alterations consist of dwarfism, owing to chondrodysplasia and shortening of long bones and vertebrae, and also of cranial malformations that consist of delayed closing of the PF sutures and macrocephaly. Comparison of the long bones and chondrocytes of TgFGF mice with the wild type showed that chondrocyte proliferation was decreased in the reserve as well as in the proliferative zone of the growth plates. This is in line with what has been previously observed in other mouse models of FGF-related chondrodysplasias (Segev et al., 2000; Naski et al., 1998; Li et al., 1999). Unexpectedly, we also found that the chondrocytes of the growth plate of these mice exhibited a very high level of apoptosis that was detectable in vivo as well as in primary culture of TgFGF chondrocytes. Apoptosis was drastically increased in the reserve and proliferative zones of the growth plates, but was also somewhat elevated in the hypertrophic zone, where apoptosis is a normal prelude to mineralization and bone deposition. Thus, TgFGF chondrocytes show two distinct abnormalities: decreased proliferation capacity and increased apoptosis.

We had previously reported that we could not detect increased apoptosis in rat chondrosarcoma (RCS) cells after FGF treatment (Sahni et al., 1999). It is likely that the failure to observe apoptosis in these cell lines resulted from the fact that RCS derive from a chondrosarcoma, and thus, like many other tumor cells, may have developed an anti-apoptotic phenotype. Clearly it does not derive from a different effect of exogenous versus endogenous FGF, as we could easily detect increased apoptosis in FGF-treated wild-type primary chondrocytes. However, this finding implies that inhibition of proliferation is a distinct phenomenon from apoptosis, even if both the processes appear to require STAT1 (see below).

Whether the induction of apoptosis caused by unregulated FGF signaling is direct or indirect is not clear at the moment. Increased apoptosis has been reported to occur in human chondrocytes isolated from individuals with TDI in culture (Legeai-Mallet et al., 1998), which suggests that apoptosis results from the direct action of FGF on chondrocytes, rather than being dependent on the induction of expression of pro-apoptotic molecules by FGF in other cell types. However, it is possible that FGF treatment induces the production of pro-apoptotic molecules (e.g. FAS ligand) in chondrocytes, which would, in turn, be the real effector of apoptosis. It has been shown that excessive FGF signaling decreases the expression of Indian hedgehog (IHH) and parathyroid hormone-related peptide (PTHrP) receptor in the growth plate (Naski et al., 1998), and this downregulation of PTHrP signaling could also play a role in promoting apoptosis, by decreasing Bcl2 expression (Amling et al., 1997).

In conclusion, examination of various aspects of long bone development in the TgFGF mice shows that they have reduced chondrocyte proliferation and increased apoptosis, phenomena that could both contribute to defective bone growth. Apoptosis is likely to be induced directly by FGF in chondrocytes, but the major mediators of FGF-induced apoptosis have not been identified. We examined the levels of expression of various molecules that are known to regulate apoptosis in transgenic as well as in FGF-treated wild-type primary chondrocytes, but we could not detect any significant difference in the expression of Bcl2, Bax, Casp3 and Casp8. Expression of Bcl-X_L, an anti-apoptotic molecule, was somewhat decreased in TgFGF chondrocytes, but the significance of this observation is not clear at this time.

The TgFGF chondrodysplasia phenotype is corrected in a Stat1-null background

By crossing the TgFGF mice with Stat1^−/− mice, we obtained TgFGF/Stat1^−/− mice and compared their bone phenotype with that of transgenic, Stat1^+/+ or Stat1^−/− littermates. As shown in the Results, the chondrodysplasia observed in TgFGF mice is substantially corrected when the FGF2 transgene is expressed in a Stat1-null background. The femurs and tibia of adult TgFGF/Stat1^−/− mice are longer and less thick than those of their TgFGF littermates. Furthermore, the rate of chondrocyte proliferation in the growth plate is restored to normal or near normal levels in TgFGF/Stat1^−/−. The high levels of chondrocyte apoptosis observed in growth plate chondrocytes are also considerably reduced in the crosses at all ages examined. In the hypertrophic zone, where apoptosis is a normal event and the increase in TgFGF bones is much less dramatic than in the reserve and proliferative zone, the effect of STAT1 deficiency was not very pronounced, suggesting that STAT1 does not play an important role in ‘physiological’ chondrocyte apoptosis.

Stat1 was originally characterized as an essential mediator of responses to IFN, including inhibition of cell proliferation (Schindler and Darnell, 1995; Darnell, 1997). In addition, it has been implicated in the regulation of apoptosis, both in the IFN system and independently (Levy, 1999). Loss of STAT1 function in mice results in increased proliferation and decreased apoptosis of lymphocytes and correlates with decreased expression of Casp1 and Casp11, suggesting that STAT1 is required to regulate these functions (Lee et al., 2000). Stat1-mutant human fibroblasts display reduced susceptibility to tumor necrosis factor (TNF)-induced apoptosis and express reduced levels of pro-apoptotic proteins, and STAT1 has been suggested to play an adaptor function that links the TNF receptor to induction of apoptosis (Kumar et al., 1997; Wang et al., 2000). All of these mechanisms could play a role in determining the apoptotic response of chondrocytes to FGF signaling, and its modulation by STAT1. Although the precise mechanism by which STAT1 mediates FGF-induced apoptosis in chondrocytes remains to be determined, our results show that this is not a generalized effect of STAT1 deficiency on apoptosis. When we examined the calvarial bones of TgFGF and TgFGF/Stat1^−/− littermates, we found that, as we previously described, osteoblasts in and around the PF suture exhibited a higher degree of apoptosis in the TgFGF skulls (Mansukhani et al., 2000). This high level of apoptosis is substantially unchanged in a Stat1^−/− background. It is also noteworthy that the cranial deformities observed in the TgFGF mice are also not corrected. Thus STAT1 does not seem to mediate the FGF effects on intramembranous ossification, and indeed we could not detect STAT1 activation after FGF treatment of osteoblasts.

In conclusion, these results indicate that STAT1 is an essential downstream modulator of FGF signaling in chondrocytes in vivo and that it mediates the two major FGF effects observed in the transgenic FGF2 mice: inhibition of chondrocyte proliferation and induction of premature apoptosis. These results prompted us to study, in detail, the bone phenotype of Stat1^−/− mice.

Stat1^−/− mice show accelerated bone development in embryonic and early postnatal life

When we compared E17 metatarsal bone rudiments of Stat1^−/− embryos with those of wild-type embryos, we noticed that Stat1^−/− metatarsals were invariably slightly longer than those of their wild-type littermates. The number of DNA synthesizing cells in the growth plate was also higher, and a consistent reduction in the number of apoptotic cells was also detected. A similar trend was also apparent in 1-day-old to 10-day-old mice. The slightly increased proliferation and decreased apoptosis of Stat1-null chondrocytes produced bones in which the proliferative zone was expanded compared with that of wild-type mice, and the distance between the two opposing growth plates was also higher, thus resulting in longer bones. However, by P10-P15, those differences had essentially disappeared. This phenotype is reminiscent of that observed in Mmp9-null mice, which show a delayed apoptosis and ossification of the growth plate that is abnormally lengthened up to 3 weeks after birth. After this period, these abnormalities disappear and ultimately result in a normal axial skeleton (Vu et al., 1998).

The observation that Stat1^−/− bones grow faster than the wild type is in agreement with our findings showing a profound effect of STAT1 on FGF-induced inhibition of proliferation and apoptosis. Chondrocytes produce small but detectable levels of FGFs, notably FGF2 and FGF9 (R. Kamijo, M. S. and C. B., unpublished), and the absence of STAT1 could relieve the homeostatic response to these growth regulators. The disappearance of this phenotype at later ages is more difficult to explain. As Fgfr3-null mice show longer bones than the wild type, it is unlikely that cessation of FGF production after day 10-15 renders STAT1 irrelevant to the cellular response to FGF. Rather it appears that the potential for extended growth of Stat1-null chondrocytes is limited in time by other factors or by a built-in clock that regulates the extent of chondrocyte proliferation. Alternatively, the ‘window’ during which FGFR3 signaling acts through STAT1 may be limited in time, and other signal transduction pathways may produce, at later stages, a similar growth inhibition in chondrocytes. It is also possible that the effect of FGFR3 deletion results from the lack of FGF growth inhibitory signals, as well as from concomitant modulation of expression of other proteins, such as PTHrP or IHH which play a crucial role in the homeostasis of bone growth (Karp et al., 2000). Indeed PTHrP is known to stimulate the proliferation of chondrocytes and delay their differentiation, since ablation of the gene for PTHrP (Pthlh – Mouse Genome Informatics) causes accelerated endochondral ossification (Karaplis et al., 1994; Chung et al., 1998; Lanske et al., 1999), while PTHrP overexpression lead to the opposite phenotype (Weir et al., 1996). PTHrP upregulation has been observed in Fgfr3^−/− mice (Deng et al., 1996), and could produce an additive effect on chondrocyte proliferation. FGF induced downregulation of PTHrP expression may not require STAT1 function and thus the additive effect of PTHrP upregulation on chondrocyte proliferation would be absent in Stat1^−/− mice. Experiments are in progress to determine whether this hypothesis is correct.

In conclusion, the results presented in this report show that STAT1 functions as a downstream mediator of FGF signaling during endochondral ossification in vivo, modulating the increased apoptosis and reduced chondrocyte proliferation induced by unregulated FGF signaling. The role that STAT1 plays in bone development under physiological conditions is more subtle and further experiments will be necessary to understand why it appears to be limited to early stages of bone growth.