Raf Kinases Are Essential for Phosphate Induction of ERK1/2 Phosphorylation in Hypertrophic Chondrocytes and Normal Endochondral Bone Development*

Hypophosphatemia causes rickets by impairing hypertrophic chondrocyte apoptosis. Phosphate induction of MEK1/2-ERK1/2 phosphorylation in hypertrophic chondrocytes is required for phosphate-mediated apoptosis and growth plate maturation. MEK1/2 can be activated by numerous molecules including Raf isoforms. A- and B-Raf ablation in chondrocytes does not alter skeletal development, whereas ablation of C-Raf decreases hypertrophic chondrocyte apoptosis and impairs vascularization of the growth plate. However, ablation of C-Raf does not impair phosphate-induced ERK1/2 phosphorylation in vitro, but leads to rickets by decreasing VEGF protein stability. To determine whether Raf isoforms are required for phosphate-induced hypertrophic chondrocyte apoptosis, mice lacking all three Raf isoforms in chondrocytes were generated. Raf deletion caused neonatal death and a significant expansion of the hypertrophic chondrocyte layer of the growth plate, accompanied by decreased cleaved caspase-9. This was associated with decreased phospho-ERK1/2 immunoreactivity in the hypertrophic chondrocyte layer and impaired vascular invasion. These data further demonstrated that Raf kinases are required for phosphate-induced ERK1/2 phosphorylation in cultured hypertrophic chondrocytes and perform essential, but partially redundant roles in growth plate maturation.

Longitudinal bone growth is dependent upon the tightly regulated proliferation and differentiation of growth plate chondrocytes. Resting chondrocytes differentiate into proliferating chondrocytes that further differentiate into pre-hypertrophic and hypertrophic chondrocytes (1). Although recent investigations demonstrate that 60% of osteocalcin-expressing osteoblasts in growing bone originate from Col10a1-Cre-expressing cells (2), it is not currently known what percentage of hyper-trophic chondrocytes gives rise to osteoblasts versus undergo apoptosis. Hypertrophic chondrocytes secrete angiogenic factors that promote vascular invasion, undergo apoptosis, and are replaced by mineralized bone (3,4). Studies in numerous mouse models demonstrate that hypophosphatemia impairs apoptosis of terminally differentiated hypertrophic chondrocytes, which results in rickets (5)(6)(7). These studies establish a critical role for serum phosphate in normal growth plate maturation.
Exposure of cultured hypertrophic chondrocytes to phosphate leads to ERK1/2 phosphorylation and activation of the mitochondrial apoptotic pathway. Blocking phosphate induction of ERK1/2 phosphorylation with the MEK1/2 inhibitor U0126 impairs hypertrophic chondrocyte apoptosis in vivo and in vitro, leading to expansion of the hypertrophic chondrocyte layer of the growth plate (8). MEK1/2 can be activated by numerous signaling pathways, including Raf kinases. In the growth plate, A-Raf and B-Raf are primarily expressed in proliferative chondrocytes. Ablation of either or both of these Raf isoforms does not significantly affect normal growth plate development (9). In contrast, C-Raf is predominantly expressed in the hypertrophic chondrocyte layer of the growth plate (10). Chondrocyte-specific ablation of C-Raf leads to expansion of the hypertrophic layer of the growth plate, associated with decreased hypertrophic chondrocyte apoptosis. However, C-Raf is not required for phosphate-induced ERK1/2 phosphorylation in cultured hypertrophic chondrocytes, but rather causes rickets by increasing ubiquitin-dependent VEGF protein degradation and impaired vascular invasion at the chondroosseous junction (11). To define a role for Raf isoforms in normal growth plate maturation and phosphate-induced hypertrophic chondrocyte apoptosis, mice with chondrocyte-specific ablation of all three Raf isoforms were generated.

Raf Isoforms Have Non-redundant Actions in Normal Growth
Plate Maturation-To generate mice with chondrocyte-specific ablation of B-and C-Raf, mice expressing the Col2a1-Cre transgene (12) were bred to mice with floxed B-Raf and C-Raf (B-Raf fl/fl or C-Raf fl/fl ) alleles (13,14). Offspring were bred to A-Raf Ϫ/y (A-Raf KO mice) (15) to obtain mice lacking all three Raf isoforms in chondrocytes. A-Raf Ϫ/y mice exhibit growth retardation and neurologic abnormalities after the second week of life; thus, postnatal phenotypes were analyzed on day 8 (P8). 2 As reported previously, A-Raf Ϫ/y , B-Raf f/f;Col2Cre , and C-Raf f/f;Col2Cre mice were phenotypically normal at P8 (9, 11), as were mice lacking two of the three Raf isoforms (A/B-Raf, A/C-Raf, or B/C-Raf) in chondrocytes. Histological analyses revealed a normal tibial growth plate in A-Raf Ϫ/y mice, in B-Raf f/f;Col2Cre mice, and in A-Raf Ϫ/y ; B-Raf f/f;Col2Cre mice ( Fig.  1). As reported previously, C-Raf f/f;Col2Cre mice exhibited an expansion of the hypertrophic chondrocyte layer of the growth plate (11) (Fig. 1). Chondrocyte-specific ablation of B-Raf did not alter the growth plate of the C-Raf f/f;Col2Cre mice ( Fig. 1); however, A-Raf Ϫ/y ;C-Raf f/f;Col2Cre mice exhibited an increase in the number of hypertrophic chondrocytes per column relative to that observed in the C-Raf f/f;Col2Cre mice, suggesting nonredundant functions of A-and C-Raf in normal growth plate maturation (Fig. 1).
To determine whether the modest phenotype observed in the growth plate of mice lacking two of the three Raf isoforms was due to compensation by the remaining Raf isoform, mice lacking all three isoforms in chondrocytes were generated. Consistent with a critical role for Raf signaling in the growth plate, ablation of all three Raf isoforms led to a neonatal lethal phenotype, as did the presence of a single A-Raf allele in female  (Fig. 2D). The femurs of the triple KO were 24% longer than those of control littermates (p ϭ 0.02, n ϭ 5), whereas tibiae, humeri, and skull parameters were not significantly different.
As reported previously, C-Raf f/f;Col2Cre mice exhibit an expansion of the hypertrophic chondrocyte layer at E18.5 relative to Cre-negative littermates (11). A-Raf Ϫ/y growth plates were similar to those of control mice, whereas the growth plates of mice lacking B/C-Raf exhibited a significant expansion of the hypertrophic layer (Fig. 2E) similar to that observed in C-Raf KO mice at E18.5.
Vascular invasion is required for normal growth plate maturation and hypertrophic chondrocyte apoptosis (16). Although ablation of all three Raf isoforms did not significantly alter osteoclast number at the chondro-osseous junction (Fig. 2F), E18.5 embryos lacking all three Raf isoforms in chondrocytes exhibited a marked reduction in vascular invasion based on the dra- Raf Signaling in Chondrocytes Is Required for Hypertrophic Chondrocyte Apoptosis-Raf kinases are known to induce phosphorylation of MEK1/2-ERK1/2, which is required for normal growth plate maturation. At P8, pERK1/2 and cleaved caspase-9 immunoreactivity were preserved in A-Raf Ϫ/y , B-Raf f/f;Col2Cre , and A-Raf Ϫ/y ; B-Raf f/f;Col2Cre mice, correlating with lack of expansion of the hypertrophic chondrocyte layer of the growth plate (Fig. 3A). Consistent with the known effects of C-Raf ablation, pERK1/2 and cleaved caspase-9 immunoreactivity were decreased in growth plates of mice lacking C-Raf alleles (11). A similar decrease in pERK1/2 and caspase-9 immunoreactivity is seen in A-Raf Ϫ/y ; C-Raf f/f;Col2Cre and B-Raf fl/fl ; C-Raf fl/fl;Col2Cre mice, correlating with the expansion of the hypertrophic chondrocyte layer observed relative to A-Raf ϩ/ϩ and A-Raf ϩ/y Cre-negative littermates (Fig. 3A). The growth plates of E18.5 mice lacking all three Raf isoforms in chondrocytes exhibited a dramatic decrease in pERK1/2 and cleaved caspase-9 immunoreactivity, demonstrating that impaired hypertrophic chondrocyte apoptosis underlies the marked expansion of the hypertrophic chondrocyte layer observed in the growth plates of these mice (Fig. 3B).
Raf Isoforms Regulate Proliferation of Growth Plate Chondrocytes-Constitutive activation of MEK1 in chondrocytes impairs proliferation (17). Thus, to determine whether enhanced proliferation contributes to growth plate expansion in mice lacking all three Raf isoforms in chondrocytes, chondrocyte proliferation was evaluated by BrdU incorporation.
Mice lacking all three Raf isoforms in chondrocytes exhibited a 61% reduction in the percentage of BrdU-positive proliferative chondrocytes (Fig. 4A), in the setting of an 80 and 97% decrease in expression of B-Raf and C-Raf, respectively, evaluated by quantitative RT-PCR (A-Raf KO mice are a global knock-out). Thus, ablation of Raf isoforms decreases chondrocyte proliferation in the developing growth plate. Consistent with this, cyclin D1, c-Myc, and Foxm1 expression was decreased in primary proliferating chondrocytes from triple KO mice, whereas p27 was increased (Fig. 4B).
Raf Isoforms Are Essential for Phosphate-induced ERK1/2 Phosphorylation in Hypertrophic Chondrocytes-Because MEK1/2 activation of ERK1/2 phosphorylation is required for phosphate-mediated hypertrophic chondrocyte apoptosis (8), investigations were performed to determine whether phosphate activation of this signaling pathway requires Raf isoforms. Mice lacking C-Raf in chondrocytes exhibit a decrease in growth plate pERK1/2 immunoreactivity. However, ablation of C-Raf in chondrocytes does not abolish phosphate-mediated ERK1/2 phosphorylation in vitro, suggesting that C-Raf is not required for this process or that other Raf isoforms can compensate for the absence of C-Raf. Notable in this respect, the expression of A-Raf and B-Raf in hypertrophic chondrocytes is 7.6 and 42.7% that of C-Raf (11). To determine the contributions of Raf isoforms to phosphate-induced ERK1/2 phosphorylation, primary hypertrophic chondrocytes from mice lacking one or all three Raf isoforms, or from females with one A-Raf allele, were treated with sodium phosphate or sodium sulfate for 30 min. The absence of A-Raf, B-Raf, or C-Raf did not impair phosphate-in- duced ERK1/2 phosphorylation (Fig. 5). However, studies in chondrocytes isolated from mice lacking all three Raf isoforms demonstrated dramatic impairment in phosphate-induced ERK1/2 phosphorylation. Similar findings were observed in chondrocytes from females with a single A-Raf allele (supplemental Fig. 1), the latter of which exhibit a phenotype similar to mice lacking all Raf isoforms in chondrocytes (supplemental Fig. 1). These data define a critical role for Raf kinases in phosphate-induced ERK1/2 phosphorylation and normal growth plate maturation.

Discussion
Phosphate and the MEK1/2-ERK1/2 signaling pathway have been shown to play a critical role in normal growth plate maturation (8). The current studies demonstrate an essential role for Raf kinases in phosphate-induced ERK1/2 phosphorylation, hypertrophic chondrocyte apoptosis, and normal plate maturation. Global ablation of B-Raf or C-Raf isoforms causes embryonic lethality in mice (18,19); therefore, mice with chondrocyte-specific B-Raf and C-Raf ablation were used to investigate the role of B-Raf and C-Raf in phosphate-mediated growth plate maturation. Despite 20% residual mRNA expression of B-Raf and 3% residual expression of C-Raf in chondrocytes, these studies demonstrate a critical role for Raf isoforms in growth plate maturation, phosphate-induced ERK1/2 phosphorylation, and hypertrophic chondrocyte apoptosis.
Previous investigations have shown an important role for C-Raf in stabilizing VEGF protein in chondrocytes (11). Although mice with global ablation of A-Raf do not exhibit a  growth plate phenotype, mice lacking both A-and C-Raf in chondrocytes exhibit a more significant expansion of the hypertrophic chondrocyte layer of the growth plate than do mice with chondrocyte-specific C-Raf ablation, suggesting that A-and C-Raf exert non-redundant functions in chondrocytes. In contrast, the absence of B-Raf in chondrocytes does not alter the growth plate phenotype of mice lacking either A-Raf or C-Raf. However, a non-redundant function for B-Raf is evident from the severe phenotype observed in mice lacking all three Raf isoforms in chondrocytes. Consistent with the essential role of Raf isoforms in normal growth plate maturation, males lacking all three Raf isoforms in chondrocytes and females with only one A-Raf allele exhibit neonatal lethality, in the setting of significant, but incomplete ablation of B-and C-Raf. The inability of a single A-Raf allele to prevent neonatal lethality in female mice suggests that X-inactivation may decrease the expression of A-Raf from the remaining allele. Consistent with this, mRNA levels of A-Raf in the chondrocytes of these females is Ͻ30% of that seen in control.
The phenotype of mice lacking all three Raf isoforms in chondrocytes is analogous to that seen in mice lacking both Erk1 and Erk2 in chondrocytes (20,21). Col2a1-Cre-mediated ablation of Erk2 in Erk1 knock-out mice results in expansion of the terminally differentiated hypertrophic chondrocyte layer of the growth plate and neonatal lethality. Neither cleaved caspase-9 nor TUNEL staining was evaluated in these mice. However, analogous to the current studies, an increase in length of the long bones was observed, primarily due to expan-sion of the epiphysis (21). Similar to mice lacking all three Raf isoforms in chondrocytes, this was associated with a decrease in chondrocyte proliferation (20). Analogous to our findings with Raf isoform ablation, mice expressing a single Erk allele had a significantly milder phenotype than those with ablation of all four alleles. Osx-Cre-mediated Erk2 ablation in the hypertrophic chondrocytes and osteoblasts of Erk1 null mice also leads to a marked expansion of the hypertrophic chondrocyte layer of the growth plate and is accompanied by impaired terminal chondrocyte differentiation (22). Paradoxically, diffuse TUNEL staining in the hypertrophic chondrocyte layer of the growth plate is observed, in contrast to the TUNEL staining present in the most differentiated hypertrophic chondrocytes in normal mice. Of note, in these studies, mediated recombination resulted in a 30% reduction of Erk2, and did not lead to neonatal lethality.
The epidermal growth factor receptor (EGFR) also plays a critical role in growth plate development. Blocking EGFR signaling leads to expansion of the hypertrophic chondrocyte layer of the growth plate without altering chondrocyte proliferation, differentiation, or vascular invasion. Rather, a dramatic decrease in chondrocyte MMPs and osteoclast recruitment at the chondro-osseous junction is observed (23). A similar impairment in MMP expression and osteoclast recruitment is observed in mice lacking TGF␣ (24). However, ablation of Raf isoforms does not impair osteoclast recruitment (Fig. 2F) or chondrocyte MMP13 expression (data not shown). VEGF signaling is also required for normal growth plate maturation and hypertrophic chondrocyte apoptosis (16). C-Raf f/f;Col2Cre mice exhibit expansion of the hypertrophic chondrocyte layer associated with impaired vascular invasion due to increased ubiquitin-dependent degradation of VEGF protein (11). Impaired vascular invasion was also observed in mice lacking all three Raf isoforms in chondrocytes and in Erk1 null mice with chondrocyte-specific Erk2 ablation (21). However, in contrast to mice with chondrocyte-specific C-Raf ablation, phosphate induction of ERK1/2 phosphorylation was prevented in mice lacking all three Raf isoforms in chondrocytes as well as in chondrocytes of females with a single A-Raf allele, demonstrating that Raf isoforms contribute to normal growth plate maturation by stabilizing VEGF protein levels as well as promoting phosphate-mediated ERK1/2 phosphorylation. Interestingly, basal ERK1/2 phosphorylation is increased in the Raf triple knock-out mice. An increase in Ras activity is known to occur with Raf inhibition, and leads to an increase in pERK1/2. However, this requires the presence of active Raf isoforms (25). Thus, the 20% residual expression of B-Raf and/or 3% residual expression of C-Raf could contribute to the increase in basal pERK1/2 by Ras-dependent C-Raf/B-Raf signaling (26). The molecular basis for the increase in Ras activity observed with Raf inhibition is not established, but has been shown to be associated with activation of receptor tyrosine kinases (RTKs), including the EGFR and FGFR3 (27,28).
Expression of conditionally active Raf isoforms in vitro increases proliferation of several cell types (29), and constitutive activation of B-Raf in vivo leads to tumors including melanomas and lung adenomas (30). Although B-Raf ablation in skin tumors reduces cell proliferation (31), the absence of a growth plate phenotype in mice with chondrocyte-specific B-Raf ablation suggests a redundant role for Raf isoforms in the regulation of chondrocyte proliferation. However, similar to mice lacking ERK1/2 in chondrocytes (20), chondrocyte proliferation was reduced in the growth plates of mice lacking all three Raf isoforms in chondrocytes. Cyclin D1, which is induced upon activation of C-Raf in a fibroblast cell line (32), was decreased in the triple KO mice, as was c-Myc, which is increased in the setting of increased B-Raf signaling (33). Foxm1, which mediates Raf-induced proliferation in fibroblast cell lines, (34) was also decreased in the chondrocytes from triple KO mice. P27, which negatively regulates proliferation of multiple cell types including chondrocytes (35), and is induced upon Raf inhibition (36), was increased in the triple KO chondrocytes, whereas p21, which mediates C-Raf regulation of chondrocyte proliferation (37), was unchanged.
The current investigations define an essential role for Raf isoforms in phosphate-induced ERK1/2 phosphorylation and normal growth plate maturation in vivo. Ablating Raf signaling leads to expansion of the hypertrophic chondrocyte layer of the growth plate, impaired vascular invasion, impaired hypertrophic chondrocyte apoptosis, and neonatal lethality. Thus, these studies support a critical role for Raf-MEK1/2-ERK1/2 in phosphate signaling and growth plate maturation.

Experimental Procedures
Animal Studies-Animal studies were approved by the Massachusetts General Hospital Institutional Animal Care Committee. All mice were on a mixed C57BL/6J, 129T2/SvEmsJ, 129/OLA background, maintained in a virus-and parasite-free barrier facility, and exposed to a 12-h light/dark cycle. Mice with chondrocyte-specific ablation of C-Raf or B-Raf were generated by crossing mice expressing Cre recombinase driven by the collagen type II (Col2a1) promoter with mice bearing C-Raf alleles floxed at exon 3 (C-Raf f/f;Col2Cre ) (13) or B-Raf alleles floxed at exon 12 (B-Raf f/f;Col2Cre ) (14). Germline A-Raf knockout mice (A-Raf X KO Y) have been characterized previously (15). These mice were bred to create mice with deletion of A-/C-Raf, B-/C-Raf, or A-/B-Raf, or deletion of all three Raf kinases in chondrocytes. Growth plates were analyzed at P8 for postnatal studies in single Raf mutant mice. Mice lacking three Raf isoforms in chondrocytes, or females with a single A-Raf allele, were analyzed at E18.5 because of their neonatal lethal phenotype. Both female and male mice were studied. The phenotype of mice with chondrocyte-specific Raf ablation (homozygous for the floxed B-or/and C-Raf allele and heterozygous for the Col2-Cre transgene (B-or C-Raf f/f;Col2Cre )) was compared with that of Cre-negative littermates homozygous for the Raf floxed allele (B-or C-Raf f/f ).
Evaluation of Chondrocyte Proliferation-Pregnant females were injected at 18.5 days post-coitum with 0.25 mg/g BrdU with 30 g/g of 5-fluoro-2Ј-deoxyuridine (FdU) (to decrease background) 2.5 h before harvesting embryos. BrdU was detected on paraffin sections using the BrdU staining kit (Invitrogen). Total and BrdU-positive cells in the proliferating chondrocyte layer were counted.
Cell Culture-Primary chondrocytes were isolated from the rib cages of P2 pups or E18.5 embryos by sequential collagenase II digestions and then plated at a density of 3 ϫ 10 5/cm2 (5,41). Cells were cultured in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin, and 25 g/ml ascorbic acid at 37°C, 5% CO 2 for 7-10 days. To evaluate phosphate induction of ERK1/2 phosphorylation, chondrocytes were serum-restricted (0.5% FBS) overnight prior to treatment. Cells were treated with sodium sulfate or sodium phosphate at a concentration of 7 mM for 30 min.
Evaluation of Gene Expression-Total RNA was isolated from proliferating primary chondrocytes using the RNeasy Mini Kit (Qiagen) and reverse-transcribed with SuperScript II (Roche Applied Science). Quantitative real-time PCR was performed using the QuantiTect SYBR Green RT-PCR Kit (Qiagen) on an Opticon DNA Engine (MJ Research). Gene expression was normalized to that of actin (Actb) for each sample (42).
Statistical Analysis-Student's t test was used to analyze significance between two groups. p Ͻ 0.05 was considered significant.