Conditional ablation of Pten in osteoprogenitors stimulates FGF signaling

Endochondral ossification is the process by which long bones in vertebrates develop. In mice, during this highly coordinated process mesenchymal cell condensations form at around 10.5 days post coitum (dpc) and differentiate into chondrocytes, thereby forming the template of the future bones. Thereafter, the chondrocytes undergo a series of highly coordinated processes leading to the formation of the growth plate. The future long bone at this stage is surrounded by a sheath of cells called the perichondrium (Kronenberg, 2003; Karsenty, 1999). The perichondrium is adjacent to and surrounds the epiphyseal growth plate. It contains osteoprogenitors required for appositional growth of the long bone (Colnot et al., 2004). Osteoprogenitors residing in the perichondrium supply the osteoblasts that migrate with blood vessels invading the hypertrophic chondrocytes, thereby establishing the primary ossification center with trabecular bone. The skeletal precursor cells in the perichondrium also supply osteoblasts that produce cortical bone (Colnot et al., 2004). Growth factors such as fibroblast growth factor (FGF) activate signaling pathways transduced by tyrosine kinase receptors. Data show that FGF receptors (FGFRs) play a key role in coordinating events during endochondral ossification; for example, activating mutations in FGFR1, FGFR2 and FGFR3 have been linked to various skeletal abnormalities that affect normal endochondral ossification (Muenke et al., 1994; Reardon et al., 1994; Rousseau et al., 1994). FGF ligands that are required for normal bone development include FGF l, FGF2, FGF7, FGF8, FGF9, FGF17 and FGF18. These ligands effect bone development via binding to the cognate FGF receptors, which are normally expressed in bone, leading to activation of cellular signaling processes mainly through the Ras/MAP Kinase (mitogen-activated protein kinase), PKC (protein kinase C) and PI3K (phosphatidylinositol 3 kinase) cascades. (Ornitz, 2005; Miraoui and Marie, 2010).

Recent studies have shown that the PI3K signaling pathway has an important role in regulating bone development. PTEN is a lipid phosphatase that negatively regulates the PI3K signaling pathway. PTEN expression has been shown in chondrocytes and osteoblasts but the role it plays in bone formation signaling events has not been completely elucidated. Several recent studies (Ford-Hutchinson et al., 2007; Hsieh et al., 2009) reported the use of collagen2a1 cre to delete Pten in the cartilage of developing mice and saw defects in growth plate organization along with an increase in chondrocyte differentiation and increased bone formation resulting in skeletal overgrowth. Similar experiments carried out by Yang et al. (Yang et al., 2008) showed that the growth plate defects in collagen2a1 cre Pten cko mice resulted from increased endoplasmic reticulum stress in Pten-null resting chondrocytes. Other investigators (Liu et al., 2007) used an osteocalcin cre to delete Pten in mature osteoblasts. These data showed increased bone mass that accumulated throughout the animal's life span. Also, deletion of Pten in cultured calvarial osteoblasts led to accelerated differentiation with a decrease in cell death.

To define the role of PTEN in osteoprogenitors, we deleted Pten in mesenchymal condensations of nascent bones using the Dermo1cre deleter strain. Dermo1 (Twist2 – Mouse Genome Informatics) expression is turned on at 9.5 dpc in mice, thereby allowing us to study the role of Pten in osteoprogenitors (Li et al., 1995; Yu et al., 2003). We observed robust knockout of PTEN in the perichondrium using the Dermo1cre deleter strain. Pten deletion led to increased bone formation. Significantly, osteoblast differentiation was geographically altered in the Pten conditional knockouts. In addition to bone formation in the usual distribution, we found osteoblasts in regions of the perichondrium away from the hypertrophic chondrocytes. This suggested a differentiation pathway for a subset of osteoblast progenitors that is autonomous of growth plate control. We discovered that deletion of Pten stimulates FGF signaling. Activation of FGF signaling occurs via a bipartite pathway. First, the expression of the ligand FGF18 is increased and second, the FGF antagonist SPRY2 is decreased. This increase in FGF signaling stimulates osteoprogenitor cell expansion. We queried whether the increase in FGF signaling contributes to the autonomous osteoblast differentiation. We discovered an increase in the hedgehog-dependent transcription factor GLI2 in Pten-deleted perichondrium. Additional results support a model whereby the increase in GLI2 is driven by signaling downstream of FGFRs. Finally, the Pten-null phenotype was partially rescued by removing one allele of the major perichondrial FGFR, FGFR2. This study identifies for the first time that Pten deletion leads to an increase in FGF signaling, which can stimulate both perichondrial cell proliferation and osteoblast differentiation.

Total RNA was extracted from cultured primary osteoblasts or immortalized preosteoblasts following a protocol described previously (Kapadia et al., 2005). Primer sequences used were (5′-3′): 18s_Fwd, CATGTGGTGTTGAGGAAAGCA; 18s_Rev, GTCGTGGGTTCTGCATGATG; Pten_Fwd, GACCAGAGACAAAAAGGGAGTCA; Pten_Rev, GTGCCACGG GTCTGTAATCC; BGLAP2_e1-3_A_Fwd, ACCTTATTGCCC TCCTGCTT; BGLAP2_e1-3_A_Rev, CTTGGTGCACACCTAGCAGA; BGLAP2_e3-4_A_Fwd, TTTGTAGGCGGTCTTCAAGA; BGLAP2_e3-4_A_Rev, AAGCAGGAGGGCAATAAGGT; SPRY2_Fwd, TATT TGCACATCGCTGGAAG; SPRY2_Rev, CTCCATCAGGTCTTGG CAGT; FGF18_A/B_Fwd, ACTGCTGTGCTTCCAGGTTC; FGF18_A_Rev, CCCAGGACTTGCATGTGCTT; FGF18_B_Rev, CCCAGGACTTGAATGTGCTT; SPP1_e1-3_A_Fwd, TGAGATTGGCAGTGATTTGC; SPP1_e1-3_A_Rev, TGGCTATAGGATCTGGGTGC; Osterix_Fwd, CCACTGGCTCCTCGGTTCT; Osterix_Rev, GTCCCGCAGAGGGCTAGAG. The data was analyzed using the method described by Livak and Schmittgen (Livak and Schmittgen, 2001). All data represent expression relative to 18s.

Bromodeoxyuridine (BrdU) labeling and immunohistochemistry was carried out following protocols described previously (Naski et al., 1998; Kapadia et al., 2005). Apoptotic cells in the long bones were identified by using TUNEL labeling of nicked DNA. Labeling was carried out by utilizing biotin-16-dUTP (Boehringer Mannheim) and terminal deoxynucleotidyl transferase (Gibco) to label nicked DNA. A DAB substrate kit (Vector Laboratories) was utilized to develop and visualize apoptotic cells.

One-day-old Pten flox/flox pups were sacrificed and the calvaria were dissected and collected in 1× Hank's buffered saline solution with 10% penicillin and streptomycin. The collected whole calvaria were gently scraped to remove any adhered tissue and digested in a buffer containing 0.5% trypsin, 0.2% collagenase. The calvaria were digested for 15 minutes in a 6-well plate in 3 ml of buffer shaken continuously in an incubator maintained at 37°C, humidified 5% CO₂. The first two digests were discarded and the remaining six digests were collected and cultured in 10% FBS-alpha-MEM. The cells that were collected after counting were plated in 1 ml 10% FBS-alpha-MEM and cultured as required to examine osteoblast differentiation. This protocol was modified from a rat calvaria isolation protocol (Yeh et al., 1996).

Metatarsal rudiments were dissected from 15.5 dpc embryos obtained when Pten flox/flox females were mated with Pten flox/wt and Dermo1cre heterozygous males. The dissected metatarsal rudiments were cultured as described previously (Kapadia et al., 2005).

Radiolabeled riboprobes were synthesized by in vitro transcription reaction and used to label paraffin embedded tissue as described previously (Naski et al., 1996).

Immunohistochemistry was performed on paraffin-embedded tissue using a tyramide amplification kit (Perkin Elmer) following the manufacturer's instructions. For immunofluorescence on cells plated on cover slips for AdGFP-AdCRE and 40 μg/ml BrdU (Sigma) treatments we followed the protocols described previously (Kapadia et al., 2005; Geng et al., 2009).

Mice tail snips were processed overnight in a solution containing proteinase K (40 μg/ml), 1 M Tris (pH 8.0), 5 M sodium chloride, 0.5 M EDTA (pH 8.0) and 10% SDS, and digested at 55°C. 1 μl of the extracted and resuspended DNA was utilized in a PCR reaction. Pten flox/flox mice, Fgfr2 flox/flox and Dermo1cre mice were described previously (Suzuki et al., 2001; Yu et al., 2003). The animals were maintained in accordance with protocols approved by the animal care committee at the University of Texas Health Science Center at San Antonio.

Protein lysates were prepared and run on SDS-PAGE gels as reported previously (Reinhold and Naski, 2007).

To create shRNAi plasmids, purified oligonucleotides were cloned into the pSiren Shuttle expression vector. For generating stable transformants, a neomycin resistance gene from pcDNA3.1 was cloned into the pSiren Shuttle vector using compatible XhoI and BglII sites. Sequences used were: shRNAi Pten top strand, GATCCAGTAGAGTTCTTCTTCCACAAATTCAAGAGATTTGTGGAACTCTACTTTTTTTACGCGTG; shRNAi Pten bottom strand, AATTCACGCGTAAAAAAAGTAGAGTTCTTCCACAAATCTCTTGAATTTGTGGAAGAACTCTACTG; shRNAi control top strand, GATCCAA TTCTCCG AACGTGTAAATTCAAGAGATTTACACGTTCGGAGAATTTTTTTTACGCGTG; shRNAi control bottom strand, AATTC ACGCGTA AAAAAAATTCTCCGAACGTGTAAATCTCTTGAATTTACACGTTCGGAGAATTG. The plasmid for generating the in situ hybridization probe for osterix (Sp7 – Mouse Genome Informatics) was made as follows: 1285 bp of the mouse osterix coding sequence was cloned into an pSP72 vector, BamHI was used to digest the plasmid and T7 RNA polymerase was utilized to create the antisense riboprobe. Col1a1, osteopontin (Spp1 – Mouse Genome Informatics), Fgf18, Pthr1 (Pth1r – Mouse Genome Informatics) and patched (Ptch1 – Mouse Genome Informatics) in situ hybridization probes have been reported elsewhere (Yu et al., 2003; Reinhold and Naski, 2007; Naski et al., 1998). Other plasmids used were: pECE_FoxO3aAAA [kindly provided by Dr Burgering (Brunet et al., 1999)], pGL3_8XGli_luc (Cooper et al., 2003), pFC_CaMEK1, pcDNAS45A_bcat, pGL3FGF18, 3.4kbLuc, pCMV_Runx2 and pCS_FGFR3K650E. (Reinhold and Naski, 2007; Naski et al., 1996). Antibodies used were: anti-sprouty2 (Sigma, S1819), goat anti-FGF18 (Santa Cruz Biotechnology, sc16830), anti-Pten (Cell Signaling, 9559), anti-pAktSer473 (Cell Signaling, immunohistochemistry specific 3787), anti-pAktSer473 (Cell Signaling, 9271), anti-Akt (Cell Signaling, 9272), anti-pErk (p44/42MAPK Thr202/Tyr204 Cell Signaling, 4370) or anti-pErk (Santa Cruz, sc7383), anti-Erk (Zymed Laboratories, mouse anti-MAP Kinase ERK1+ERK2), anti-Gli2 (Rockland 600-401-695) and anti-actin (mouse monoclonal, Chemicon International) (Kapadia et al., 2005; Reinhold et al., 2006).

Transient transfections were carried out using a calcium phosphate precipitation method as reported in Reinhold et al. (Reinhold et al., 2006). Nucleofection was used to transfect shRNAi expressing Pten and scrambled control into C3H10T1/2 cells to look at the effect of knockdown of Pten on FGF18 (Reinhold and Naski, 2007).

Newborn pups were dissected in 1× PBS and fixed overnight in 4% paraformaldehyde (PFA), followed by 14% EDTA solution for decalcification for 2 weeks. Standard methods were followed for embedding, sectioning 5-μm tissue sections and Hematoxylin and Eosin (H&E) staining (Naski et al., 1996). Skeletal preparations of newborn pups was carried out following the protocol by Mcleod (Mcleod, 1980). Von Kossa staining was carried out following a protocol described by Zhao et al. (Zhao et al., 2002).

The hind limbs and fore limbs of newborn mice were dissected and fixed in 4% PFA. The undecalcified limbs were wrapped in foam and put into 15-ml tubes. The samples were packed tightly so that there were no air bubbles. For 27 μm scans a GE eXplore machine was utilized. The data that was obtained was analyzed using MicroView (open source software from GE). Using this software the femurs were first separated from the total scans and the volumetric bone mineral density (BMD) of whole femurs from the Pten conditional knockouts were compared with the wild-type controls from postnatal day one (Vasquez et al., 2008).

All the experiments were repeated at least three times. P value was calculated using Student's t-test and the null hypothesis was rejected for P values <0.05.

PTEN, a lipid phosphatase and tumor suppressor protein (Li et al., 1997; Myers et al., 1998), is expressed in both chondrocytes and osteoblasts (Fig. 1A). To study the role of Pten during bone development, we deleted Pten in mesenchymal cell condensations prior to cell fate decisions using the Dermo1cre deleter strain. Pten was efficiently deleted in osteoblasts and perichondrial osteoprogenitor cells as observed by the loss of indirect immunofluorescence staining for PTEN (Fig. 1A) in mice inheriting the Dermo1cre allele. PTEN was also deleted in chondrocytes (Fig. 1A). PTEN negatively regulates phosphatidylinositol-3 kinase (PI3K) signaling (Stambolic et al., 1998) and, therefore, ablation of Pten should upregulate PI3K signaling, including phosphorylation of the PI3K dependent kinase AKT (AKT1 – Mouse Genome Informatics). As expected, Fig. 1A shows increased phosphorylation of AKT at serine 473 following PTEN loss. Additionally, osteoblasts isolated from clavaria of newborn Pten knockout pups showed a decrease in Pten at the mRNA level (Fig. 1B).

We examined the gross skeletal morphology of the Pten conditional knockouts and the wild-type controls by staining with Alcian blue and Alizarin red. Fig. 2A shows that the femurs, as well as other long bones of the conditional knockouts were shorter (Fig. 2G) and broader than those of the wild-type controls. To examine the architectural and cellular basis of the skeletal changes, we prepared histological sections stained with H&E. Light microscopic examination of the Pten conditional knockout and wild-type tibia at 16.5 dpc and one day postnatal (Fig. 2B,C) showed an increase in perichondral and trabecular bone in the conditional knockouts (ckos). Interestingly, the ckos displayed multiple layers of perichondrial osteoid bone. A single layer of osteoid was deposited directly on top of the hypertrophic chondrocytes in the wild-type mice; however, the ckos displayed layers of osteoid at increasing distance from the hypertrophic chondrocytes. To assess the mineralization of the bone, we performed Von Kossa staining on undecalcified tibial sections and, consistent with the increased osteoid, we saw increased bone mineralization in the ckos (Fig. 2D). These findings were further substantiated by microCT analysis of femurs, which showed increased bone mineral density (BMD) at 1 day old (Fig. 2E,F).

We found an increase in bone matrix and BMD in the ckos (Fig. 2C,E). Inspection of the histological sections showed a parallel increase in osteoblast numbers. To determine whether deletion of Pten leads to an increase in cell proliferation we performed BrdU labeling studies. Significantly, the ckos demonstrated increased BrdU labeling in the perichondrium compared with the wild-type control (Fig. 3A,B). An alternative explanation for the increased osteoblast number is a reduced rate of cell death. We, therefore, tested for apoptosis using TUNEL labeling and saw only scattered cells that were positive for TUNEL in the perichondrium of the humerus sections (see Fig. S1A in the supplementary material). There was no significant difference in the number of apoptotic cells (see Fig. S1B in the supplementary material).

We have shown augmented bone matrix in the Pten ckos and have morphologically coupled this matrix to increased numbers of osteoblasts. We next wished to determine whether these cells expressed normal markers of osteoblast differentiation. Assays of gene expression by in situ hybridization showed normal levels of markers of osteoblast differentiation in the Pten ckos (Fig. 4A-D) but significantly increased levels in the perichondral region. To determine whether expression is upregulated developmentally and in other skeletal elements, we looked at Col1a1 at 16.5 dpc tibial and 18.5 dpc humerus sections and observed a similar increase in expression in the Pten conditional knockouts compared with the wild-type controls (data not shown). Significantly, the expression of the gene encoding the PTH/PTHrP receptor (Pthr1; Pth1r – Mouse Genome Informatics) was unchanged in the chondrocytes (Fig. 4E).

We found increased cell proliferation in the Pten ckos. Given that fibroblast growth factors are essential regulators of skeletal growth, we hypothesized that FGF signaling might contribute to the increase in cell proliferation. Because Fgf18 is expressed in the perichondrium and is an important osteoblast mitogen, we examined Fgf18 expression in the perichondrium of osteoblast markers using in situ hybridization. Significantly, we found increased Fgf18 expression in the Pten ckos compared with the wild-type control (Fig. 5A). To confirm this, we examined FGF18 protein using immunohistochemistry and, similarly, we observed increased levels in the periosteum of Pten ckos (Fig. 5B). Because we observed increased FGF18 expression we also queried whether there is altered antagonism of FGF. The sprouty proteins inhibit FGF signaling by decoupling intracellular connections of the FGF receptor with the Ras-Raf signaling axis (Hacohen et al., 1998; Gross et al., 2001). Recently, data show that Spry2 expression is inhibited by PI3K activity (Paik et al., 2007). We examined SPRY2 protein in the Pten ckos by immunohistochemistry. Significantly, deletion of Pten led to reduced SPRY2 in the perichondrium compared with the wild-type control (Fig. 5C).

To evaluate more closely the pathways leading to increased Fgf18 expression, we investigated whether the FOXO transcription factors regulate Fgf18 expression. The FOXO transcription factors are phosphorylated in response to PI3K activity and this phosphorylation leads to export from the nucleus and repression of FOXO transcription (Brunet et al., 1999). We first asked whether Fgf18 expression can be regulated by FOXOs. Overexpression of FOXO3A (constitutively active; FOXO3 – Mouse Genome Informatics) in the preosteoblast cell line C3H10T1/2 led a twofold suppression of Fgf18 mRNA levels (Fig. 5D). shRNAi-mediated knockdown of PTEN (Fig. 5E′) in C3H10T1/2 led to a twofold increase in Fgf18 (Fig. 5E), supporting the view that FGF18 expression is positively regulated by PI3K signaling, which acts to release the inhibition of FGF18 by FOXO transcription factors. This was investigated more directly using an FGF18 luciferase reporter plasmid. In previous work we showed that the Fgf18 gene expression is directly induced through a bipartite binding element that is recognized by RUNX2 and the WNT-dependent transcription factors TCF4 and LEF1 (Reinhold and Naski, 2007). In Fig. 5D′ we show that a constitutively active form of FOXO3A inhibited the activation of the FGF18 luciferase reporter, whereas in the absence of RUNX2 and beta-catenin, FOXO3A had relatively little effect on FGF18 expression. To study further the effect of activated PI3K activity on Spry2, we knocked out Pten in Pten flox/flox primary calvarial osteoblasts using AdenoCRE and studied Spry2 expression in response to basic fibroblast growth factor (bFGF). In control AdenoGFP cells, Spry2 was induced as expected; however, the induction of Spry2 was blunted in the knockdown cells (see Fig. S2 in the supplementary material). This corroborates the findings of decreased SPRY2 protein in the cko mice (Fig. 5C).

The development of osteoblasts in the perichondrium is tightly coupled to the differentiation of chondrocytes in the growth plate. During normal growth and development osteoblasts form adjacent to and synthesize bone matrix directly on hypertrophic chondrocytes. Because osteoblasts in ckos developed at sites uncoupled from the chondrocyte hypertrophy, we hypothesized that Pten knockout osteoblasts develop autonomously from chondrocytes. To test this, we used the metatarsal rudiment culture system wherein osteoblast development is stringently coupled to endochondral chondrocyte differentiation. We anticipated that if osteoblast hyperplasia and development was driven by chondrocyte hypertrophy, then the metatarsals from ckos should show similar expansion of osteoblasts and bone. Metatarsal bone rudiments were isolated from 15.5 dpc embryos of wild type and Pten cko and cultured in serum-free conditions. Fig. 6A shows deletion of PTEN using immunofluorescence. The metatarsals at 15.5 dpc lacked hypertrophic chondrocytes as well as osteoblasts. Therefore, development of perichondrial osteoblasts is strongly driven by chondrocyte differentiation. Importantly, there was no significant difference in osteoblast development in the Pten ckos compared with the wild-type littermates (Fig. 6B,C), supporting the view that expanded osteoblast development in Pten ckos is not secondary to an alteration in chondrocytes differentiation in the ckos and that expanded osteoblast differentiation is uncoupled from chondrocyte differentiation in the growth plate.

Hedgehog signaling is essential for osteoblast differentiation and the effect of hedgehog is largely determined by the GLI transcription factors. Because we observed an increase in osteoblast differentiation, we questioned whether this might be due to increased GLI2. Significantly, immunofluorescence staining showed an increase in GLI2 levels in ckos compared with the wild-type control (Fig. 7A). The increase in GLI2 prompted us to look at targets of GLI2 transcription, namely, patched. In situ hybridization for the patched transcript revealed increased signal in the Pten ckos relative to littermate controls (Fig. 7B). As our data showed that FGF signaling is augmented in the ckos, we investigated whether an activated FGF receptor (K650E) can stimulate Gli transcriptional activity. We transfected C3H10T1/2 preosteoblast cells with a Gli2 reporter plasmid, and found that the activity of GLI2 was increased when co-transfected with a plasmid expressing activated FGF receptor (Fig. 7C). Similarly, dose-dependent activation of GLI2 was observed when co-transfected with an activated form of MEK (Fig. 7C′), suggesting that activation of GLI2 by FGF signaling is through a MAP kinase cascade. We further substantiate a role for increased MAP kinase activity in Pten knockout cells. Calvarial osteoblasts were isolated from mice with homozygous floxed alleles for Pten, then cultured in vitro and infected with either AdenoCRE or AdenoGFP to produce Pten knockout or control osteoblasts, respectively. MAP kinase activity was then measured by western blotting and immunofluorescence staining for phospho-extracellular signal-related kinase (pERK). Significantly, western blotting showed that Pten-null osteoblasts have increased pERK and, in addition, knockout osteoblasts had increased nuclear pERK (Fig. 7D,E).

Through several lines of evidence we have shown that FGF signaling is increased in the ckos and that this leads to increased MAP kinase activity. We hypothesized that the increased osteoprogenitor cell proliferation and osteoblast differentiation are a consequence of increased signaling through FGF receptors. We then reasoned that removal of an FGF receptor should reverse or ameliorate the phenotype. Therefore, we crossed a null allele for fibroblast growth factor receptor 2 (FGFR2) into the mating scheme. Deletion of one Fgfr2 gene has no deleterious effects and is silent phenotypically (Yu et al., 2003). We choose Fgfr2 because it is the major FGFR expressed in bone. Fig. 8A shows a partial rescue of the phenotype with a decrease of the perichondral bone phenotype. As seen previously, histological sections showed an increased perichondrial bone present in the Pten cko; however, there was a dramatic decrease in the amount of perichondrial bone in the Pten cko lacking an allele of Fgfr2 (Fig. 8A). In situ hybridization for the molecular marker Col1a1 (Fig. 8A′) showed an increase in expression in the Pten cko and this was attenuated in the cko lacking an allele for FGFR2. The rescue of the phenotype was probed further by examining GLI2 protein in the perichondrium. Significantly, protein levels of GLI2 were diminished in the perichondrium of the Pten cko lacking an allele of Fgfr2 (Fig. 8B).

Our data of metatarsal rudiment cultures suggest cell autonomous differentiation of the Pten-null osteoprogenitor cells. To address this more directly we prepared Pten-null calvarial osteoblasts and analyzed the differentiation of these cells in culture. Pten flox/flox calvarial osteoblasts were isolated from newborn pups and recombinant adenoviruses transducing cre-recombinase or GFP were used to make knockout and control cells, respectively. Western blotting showed efficient deletion of PTEN and a corresponding increase in pAKT. Using real-time PCR we assayed for indicators of osteoblast differentiation. Significantly, the relative expression of osteocalcin, osteopontin and osterix genes was increased in the PTEN cko cells. In addition, FGF18 expression was increased and Spry2 was decreased, supporting our data indicating that FGF signaling is increased. We also characterized the proliferation of calvarial cells lacking PTEN and found increased proliferation (BrdU labeling) in PTEN-null cells. (Fig. 9A-C). Von Kossa staining on calvarial cells lacking PTEN showed increased bone nodule formation compared with control cells (Fig. 9D).

PTEN is a lipid phosphatase that directly antagonizes PI3K (Stambolic et al., 1998). Genetic disruption of Pten therefore leads to unchecked PI3K signaling (Sun et al., 1999). Recent data demonstrate that PTEN is essential for normal growth and development of skeletal cells. Conditional deletion of Pten in chondrocytes using collagen2a1 cre deleter mice leads to activation of stress response pathways and disruption of chondrocyte differentiation with development of a chondrodysplasia along with increased lipoma formation (Ford-Hutchinson et al., 2007; Yang et al., 2008; Hsieh et al., 2009). Pten deletion in end-stage, mature osteoblasts using osteocalcin cre leads to an increase in bone mass with a parallel increase in cell proliferation (Liu et al., 2007). Other studies knocking-out the downstream effectors AKT1 and AKT2 showed a delay in ossification (Peng et al., 2003). Deletion of Pten in osteoblasts certainly results in increased PI3K signaling by activating downstream kinase AKTs, but how this affects skeletal cell proliferation or skeletal cell differentiation is not clear. This study looks at these crucial issues by deleting Pten in osteoprogenitor cells and identifies increased FGF signaling as the major mediator of increased bone formation in the absence of PTEN.

To investigate the pathways downstream of PTEN that influence bone, we created conditional null mice using the Dermo1cre deleter mice and mice having floxed Pten alleles. The majority of cko mice died in the perinatal period. Although the cause of death is not entirely clear it is probably due to respiratory failure, which might be secondary to inelasticity of the ribs. The limbs from ckos exhibited increased bone mass and asynchronous development of perichondrial osteoblasts. Our evidence indicates that this is due to increased proliferation of perichondrial cells and subsequent cell autonomous differentiation of these perichondrial cells into osteoblasts. Normally, the differentiation of the osteoblasts in the perichondrium is strictly coupled to the maturation of chondrocytes in the growth plate. Hypertrophic chondrocytes in the growth plate produce indian hedgehog and this in turn signals to perichondrial cells to induces osteoblast differentiation (Vortkamp et al., 1996). Significantly, we have found that osteoblast differentiation in the perichondrium of Pten ckos is uncoupled from the growth plate. Interestingly, the differentiation of Pten-null perichondrial cells still appears to require hedgehog-dependent signals; however, our data suggest that differentiation might not require a hedgehog ligand. Activation of hedgehog signaling might occur downstream of receptor-ligand dependent events. Activation of MAP kinase cascades through FGFR signaling might stabilize the transcription factor GLI2, which in turn might regulate differentiation of the cells into osteoblasts.

Pten cko mice demonstrated increased cell proliferation in the perichondrium (Fig. 3). Because of this, we hypothesized that there is increased growth factor signaling in Pten cko mice. To investigate this possibility we examined FGF18. We chose to study FGF18 because (1) it is expressed in the perichondrium at the site of increased proliferation and (2) it is a principle regulator of bone cell growth (Liu et al., 2002; Ohbayashi et al., 2002). Significantly, we showed that the expression of FGF18 is increased in the perichondium of ckos, consistent with the view that expansion of the perichondrial osteoprogenitors is FGF dependent. The increase in FGF18 expression appears to be, in part, a reflection of increased transcription of FGF18. We provide evidence that the forkhead transcription factor FOXO3 represses FGF18 expression. The repressor activity is therefore relieved when FOXO3 is exported from the nucleus in the face of increased PI3K signaling. Other evidence also points to increased FGF signaling. Specifically, we showed a relative decrease in the levels of the FGF signaling antagonist SPRY2. This implies that FGF signaling is amplified by two pathways: increased ligand and decreased post-receptor antagonism. These results suggest that a reduction of FGF receptor signaling could have a major effect on the consequences of deleting Pten in osteoprogenitor cells. To test this, we dampened FGF signaling by introducing a null allele for FGF receptor 2 (Fgfr2) into the mating scheme. Fgfr2 is the major FGFR expressed in the perichondrium (Yu et al., 2003). Significantly, removal of a single allele of Fgfr2 resulted in a partial reversion of the phenotype. There was no apparent rescue of the phenotype seen in the growth plate and the primary ossification center as Fgfr1 and Fgfr3 are the major receptors expressed in the growth plate whereas osteoblasts express all three receptors, suggesting that the loss of one allele of Fgfr2 is not enough to rescue the Pten cko phenotype completely. Even then, these data provide strong evidence that increased FGF signaling contributes to the phenotype of increased bone following disruption of Pten in osteoprogenitor cells.

We showed increased expression of FGF18 combined with reduced SPRY2, which implies increased FGFR signaling. Additional evidence of increased FGF signaling followed from studies of mitogen-activated protein kinase (MAP kinase) activity. Activated MAP kinases are a hallmark of receptor tyrosine signaling. Therefore, we examined whether deletion of PTEN increased activation of the MAP kinases ERK1 (MAPK3 – Mouse Genome Informatics) and ERK2 (MAPK1 – Mouse Genome Informatics). Using cultured osteoblasts lacking PTEN we found an increase in the phosphorylation of ERK1 and ERK2. This coincided with increased nuclear localization of ERK. This discovery is germane both to the proliferation and differentiation of osteoprogenitor cells. Activation of ERK is a well-recognized signal for the stimulation of activation of cell proliferation (Pagès et al., 1993; Seger and Krebs, 1995). Thus, osteoprogenitor cell expansion in the absence of PTEN is likely to require increased MAP kinase signaling. The effects on osteoprogenitor cell differentiation are likely to be coupled to the increase in GLI2. Studies show that GLI2 is stabilized in the presence of activated MAP kinases, possibly due to direct phosphorylation of the transcription factor (Riobó et al., 2006). Our unpublished data demonstrates that active MAPK signaling stabilizes GLI2 protein by increasing its half life (Z. Liu and M.C.N. unpublished). Further evidence for stabilization of GLI2 has been shown previously by Brewester et al. (Brewester et al., 2000) who showed that GLI2 regulation is cycloheximide dependent. Therefore, the activation of differentiation might rely on activation and/or stabilization of the transcription factor GLI2.

Our data demonstrate increased growth and differentiation of osteoprogenitor cells following conditional deletion of Pten in mice. These data suggested that these were direct, cell-autonomous effects. To test this possibility, we examined the differentiation of osteoprogenitor cells derived from the calvarium of newborn mice. Pten-null osteoprogenitor cells were prepared from the calvarium of mice with floxed Pten alleles. Subsequently, the cells were infected with adenoviruses transducing cre-recombinase or GFP. Similarly, cells were prepared from the calvarium of cko mice. Using real-time PCR we showed an increase in osterix, osteocalcin and Fgf18 in the Pten-deleted calvarial osteoblasts. Similar results were found in primary osteoblasts isolated from Pten ckos (see Fig. S3 in the supplementary material). These data strongly support a cell-autonomous pathway of osteoblast differentiation in the perichondrium of Pten cko mice.

Additional evidence of cell-autonomous differentiation was derived from organ culture experiments. We prepared metatarsal rudiments from 15.5 dpc mouse embryos (wild type and cko) and cultured the samples in serum-free medium. In this assay, perichondrial osteoblast development is strictly linked to chondrocyte hypertrophy. We reasoned that if amplified osteoblastogenesis is secondary to accelerated or magnified chondrocyte maturation in Pten ckos, then this will be recapitulated in the organ culture assay. We observed, however, no increase in bone formation in the Pten cko metatarsals relative to the wild-type controls. Moreover, the onset of hypertrophy was similar in wild-type and cko rudiments. This leads us to conclude that the dramatic increase in perichondrial bone formation is not due to augmented chondrocyte hypertrophy in the ckos. In view of the results discussed, we propose the following model which discusses the mechanism through which bone formation is regulated in the absence of Pten (Fig. 10). Osteoprogenitor cells in the perichondrium lacking Pten regulate osteoblast proliferation and differentiation by activating FGF signaling. Activation of FGF signaling due to increased ligand (FGF18) and decreased SPRY2 leads to stabilization of GLI2 protein levels, which in turn promotes osteoblast differentiation. These data also show that FOXO3 has a role in regulating FGF signaling. The importance of the FOXO group of transcriptional factors in bone development has been shown recently. Ambrogini et al. (Ambrogini et al., 2010) showed that FOXOs regulate osteoblasts via oxidative stress pathways and Rached et al. (Rached et al., 2010) demonstrated regulation of osteocalcin gene expression by FOXO1. We provide additional evidence that FOXOs regulate bone development through antagonism of Fgf18 expression.

This study was submitted as part of a PhD thesis to the Department of Biochemistry, UTHSCSA by A.R.G. We thank Dr Charles Keller and Suresh Prajapati (GCCRI, San Antonio) for help with microCT analysis. This study was funded by NIH grant R01AR053100 to MCN. Deposited in PMC for release after 12 months.