Akt promotes BMP2-mediated osteoblast differentiation and bone development

Signaling through the IGF-I receptor by locally synthesized IGF-I or IGF-II is crucial for normal skeletal development and for bone remodeling. Osteogenesis is primarily regulated by bone morphogenetic proteins (BMPs), which activate gene expression programs driven by bone-specific transcription factors. In a mesenchymal stem cell model of osteoblast commitment and differentiation controlled by BMP2, we show that an inhibitor of PI3-kinase or a dominant-negative Akt were as potent in preventing osteoblast differentiation as the IGF binding protein IGFBP5, whereas a Mek inhibitor was ineffective. Conversely, an adenovirus encoding an inducible-active Akt was able to overcome the blockade of differentiation caused by IGFBP5 or the PI3-kinase inhibitor, and could restore normal osteogenesis. Inhibition of PI3-kinase or Akt did not block BMP2-mediated signaling, because the Smad-responsive genes Sox9 and JunB were induced normally under all experimental conditions. When activated during different stages of osteoblast maturation, dominant-negative Akt prevented accumulation of bone-specific alkaline phosphatase and reduced mineralization, and more significantly inhibited the longitudinal growth of metatarsal bones in primary culture by interfering with both chondrocyte and osteoblast development and function. We conclude that an intact IGF-induced PI3-kinase–Akt signaling cascade is essential for BMP2-activated osteoblast differentiation and maturation, bone development and growth, and suggest that manipulation of this pathway could facilitate bone remodeling and fracture repair.

Introduction resulted in increased trabecular bone volume (Zhao et al., 2000). Targeting IGF-I to osteoblast precursors also gave rise to a robust bone phenotype in mice, and led to increases in femur length, cortical width and cross-sectional area (Jiang et al., 2006). Thus, regardless of the timing of IGF-I overexpression in bone of transgenic mice, net bone formation and mass were enhanced. Therefore, based on several types of evidence, IGF action via the IGF-I receptor is crucial for normal bone development and mineralization.
IGF-mediated stimulation of the IGF-I receptor triggers receptor autophosphorylation to create docking sites at phosphorylated tyrosine residues for adaptor molecules (Nakae et al., 2001). This initiates a series of protein-protein interactions that lead to activation of intracellular signal transduction pathways (Nakae et al., 2001). Although several signaling pathways mediate IGF action in bone, as well as in other tissues (Giustina et al., 2008), a growing literature supports the idea that the PI3-kinase-Akt network is critical for both osteoblast differentiation and bone growth (Fujita et al., 2004;Ghosh-Choudhury et al., 2002;Liu et al., 2007;Osyczka and Leboy, 2005;Peng et al., 2003;Raucci et al., 2008), yet the biochemical or molecular mechanisms through which the IGF-stimulated PI3kinase-Akt pathway increases osteoblast development and function have not been elucidated. Fujita and colleagues have postulated an interaction with Runx2, because the PI3-kinase inhibitor LY294002 reduced both its DNA-binding activity and its ability to stimulate target gene transcription (Fujita et al., 2004). Qiao and co-workers have reached similar conclusions (Qiao et al., 2004), whereas others have suggested collaboration at the level of nuclear translocation of BMP2-stimulated Smad proteins (Ghosh-Choudhury et al., 2002).
Here, we define a key role for the IGF-activated PI3-kinase-Akt pathway in BMP-mediated osteoblast differentiation of uncommitted mesenchymal precursor cells and their subsequent maturation. We also find that Akt-regulated signaling is crucial for longitudinal bone growth and that it exerts positive actions on both chondrocyte and osteoblast differentiation and function in developing bone. Based on these observations, we conclude that the IGF-stimulated PI3-kinase-Akt pathway is a central component in an interactive osteogenic signaling network that is necessary for both bone development and remodeling.

Inhibition of the PI3-kinase-Akt pathway blocks BMP2mediated osteoblast differentiation
We previously demonstrated that both BMP2-stimulated osteoblast differentiation of mouse mesenchymal stem cells and growth and mineralization of mouse metatarsal bones, could be blocked by IGFBP5 . In these studies, the inhibitory effects of IGFBP5 on osteogenesis appeared to depend on its ability to bind IGF-I with high affinity, thereby sequestering IGF-I from its cell-surface receptor, and leading to impaired IGF-I receptor activity . Since an IGF-I analog with diminished affinity for IGFBPs but normal affinity for the IGF-I receptor could restore BMP2-mediated osteogenesis in the presence of otherwise inhibitory concentrations of IGFBP5 , our results indicated that sustained IGF action is required for osteoblast differentiation and bone growth. The focus of current experiments was to identify and characterize the pertinent IGF-activated signaling pathways.
To define the IGF-mediated mechanisms involved in regulation of osteoblast development and function, we first incubated confluent C3H10T1/2 mesenchymal stem cells with recombinant BMP2 in osteogenic medium. Under these conditions, BMP2 treatment was accompanied by the rapid and sustained stimulation of intracellular signaling via BMP receptors, as indicated by serine phosphorylation of Smad1, Smad5 and Smad8 in protein extracts observed by day 1 and maintained for up to 7 days (Fig. 1A), and by rapid and sustained upregulation of Dlx5 and Runx2 mRNA (Fig. 1B), two osteoblast-specific transcription factors whose genes are wellknown targets of BMP2 (Lee et al., 2003;Phimphilai et al., 2006). Subsequent events included accumulation of transcripts encoding the bone transcription factor osterix (Osx), and for the secreted osteoblast protein osteocalcin (Ocn) (Fig. 1B), followed by activity of bone-specific alkaline phosphatase, and mineralization of extracellular matrix, the latter measured by Alizarin red staining (Fig. 1C,D). None of these biological effects were observed in cells incubated in osteogenic medium without BMP2 (Fig. 1A-D).
To block potential IGF-regulated signaling cascades, we treated confluent C3H10T1/2 cells with either the Mek inhibitor UO126 or the PI3-kinase inhibitor LY294002, in the presence of BMP2 and osteogenic medium, because both the Grb-Sos-Mek-Erk and PI3-kinase-Akt pathways have been shown to be activated by the IGF-I receptor via the adaptor molecules IRS1 and IRS2 in bone cells (Akune et al., 2002;Kadowaki et al., 1996;Ogata et al., 2000). Addition of UO126 had no effect on BMP2-mediated signaling, or on the rate or extent of osteoblast gene expression or differentiation ( Fig. 2A-D), although at the concentration used (10 μM), it completely inhibited IGF-I-stimulated Erk phosphorylation in C3H10T1/2 cells (supplementary material Fig. S1). By contrast, LY294002 (20 μM), which blocked IGF-induced Akt phosphorylation (supplementary material Fig. S1), impaired expression of osteoblast differentiation genes (Fig. 2B) and completely prevented induction of alkaline phosphatase activity and mineralization (Fig. 2C,D), although like UO126, it also did not inhibit BMP2-activated Smad phosphorylation ( Fig. 2A). Immunoblots of whole-cell protein lysates for serine-phosphorylated Smad1, Smad5 and Smad8 (pSmad1,5,8), total Smads, Akt phosphorylated at Ser473 (pAkt S473 ) and total Akt. (B) Results of RT-PCR assays showing expression of osteoblast-specific genes Dlx5, Runx2, osterix (Osx) and osteocalcin (Ocn), and control gene S17 after incubation for up to 7 days in osteogenic medium with or without BMP2. (C) Representative images of qualitative alkaline phosphatase (AP) staining in cells after incubation in osteogenic medium with or without BMP2 for 7 days. (D) Measurement of mineralization assessed by Alizarin red staining 7 days after incubation in osteogenic medium with or without BMP2. We next considered whether BMP2 could activate the PI3kinase-Akt pathway, and whether LY294002 could impair BMP2mediated signaling. To address the first question, we incubated C3H10T1/2 cells with BMP2 in serum-free medium for up to 60 minutes, and measured Akt phosphorylation on Ser473 as an indicator of PI3-kinase activation. Addition of BMP2 stimulated serine phosphorylation of Smad1, Smad5 and Smad8 within 15 minutes, but had no effect on Akt. Conversely, IGF-I induced phosphorylation of Akt but not that of Smad proteins (Fig. 3A). To examine effects of PI3-kinase inhibition on the acute actions of BMP2, we measured expression of Smad target genes Sox9 and JunB (Chalaux et al., 1998;Zehentner et al., 1999) after addition of BMP2 with or without LY294002 to confluent C3H10T1/2 cells in osteogenic medium. Under these conditions, BMP2 stimulated the progressive accumulation of both mRNAs starting at 12 hours, as well as inducing transcripts for Dlx5 and Runx2 (Fig. 3B). Addition of LY294002 had no effect on the kinetics of Sox9 or JunB gene expression, or mRNA abundance, but completely prevented accumulation of Dlx5 or Runx2 mRNA. Thus, LY294002 interferes selectively with BMP2-induced Smad-regulated gene activation.
To expand these observations, we next used adenovirus-mediated gene transfer to deliver a regulated dominant-negative version of the serine-threonine protein kinase Akt (Akt DN ) to C3H10T1/2 cells, to assess whether inhibition of Akt activity also could block BMP2mediated osteoblast differentiation (see Fig. 4A). Synthesis of Akt DN in our adenoviral delivery system was prevented by the antibiotic doxycycline (Dox). In the presence of Dox, no Akt DN was produced (Fig. 4B) and osteoblast-specific genes were induced, alkaline phosphatase activity accumulated and bone matrix mineralization proceeded normally (Fig. 4C-E). However, when Dox was omitted from the culture medium, Akt DN accumulated in the cells (Fig. 4B), and as a result, the expression of osteoblast genes was impaired, alkaline phosphatase activity was eliminated and mineralization Journal of Cell Science 122 (5) prevented ( Fig. 4C-E), even though BMP2-activated Smad phosphorylation and Sox9 and JunB gene expression appeared to be normal ( Fig. 4B,C). In other mesenchymal cell derivatives, including skeletal muscle, inhibition of signaling through Akt can block differentiation by promoting cell death (Fujio et al., 2001;Lawlor Fig. 2. Inhibition of PI3-kinase activity blocks BMP2induced osteoblast differentiation. Results are shown of experiments in which C3H10T1/2 cells were incubated in osteogenic media (OM) containing BMP2 for up to 10 days without (Con, control) or with the MEK inhibitor UO126 (UO) (10 μM) or the PI3-kinase inhibitor, LY294002 (LY) (20 μM), as described in the Materials and Methods. (A) Immunoblots of whole-cell protein lysates for Akt phosphorylated at Ser473 (pAkt S473 ), total Akt, tyrosine and serine phosphorylated Erk1 and Erk2 (pErk1/2), total Erks, serine phosphorylated Smad1, Smad5 and Smad8 (pSmad1,5,8), total Smads and α-tubulin. (B) Results of RT-PCR assays showing expression of osteoblast-specific genes encoding Dlx-5, Runx2, osterix (Osx) and osteocalcin (Ocn), and the control gene S17 after incubation for up to 7 days in osteogenic medium with or without UO126 or LY294002. (C) Representative images of qualitative alkaline phosphatase (AP) staining in cells after incubation in osteogenic medium with or without UO126 or LY294002 for 5, 7 or 10 days. The graph depicts measurement of alkaline phosphatase activity in lysates of cells incubated for 5 or 10 days in osteogenic medium with or without UO126 or LY294002 (mean ± s.d., n=3; *P<0.01, **P<0.001 vs cells incubated without LY294002). (D) Measurement of osteoblast-mediated mineralization assessed by Alizarin red staining at days 5, 7 and 10 after incubation in osteogenic medium with or without UO126 or LY294002. The graph shows calculation of mineralized area at day 10 (mean ± s.d., n=5; **P<0.01 vs cells incubated without LY294002). activates Smads but not Akt. Immunoblots of whole-cell protein lysates for serine phosphorylated Smad1, Smad5 and Smad8 (pSmad1,5,8), total Smads, Akt phosphorylated at Ser473 (pAkt S473 ), and total Akt after incubation of C3H10T1/2 cells in serum-free medium with BMP2 (200 ng/ml), 10 nM IGF-I or both growth factors for 0, 15, 30 or 60 minutes. (B) Results of RT-PCR experiments for mRNA encoding Sox9, JunB, Dlx-5, Runx2 and S17 after incubation for 0, 6, 12, or 24 hours in osteogenic medium without BMP2, with BMP2 or with BMP2 plus 20 μM LY294002. and Rotwein, 2000), yet under the conditions of these experiments, cell numbers remained constant whether or not Akt signaling was impaired by expression of Akt DN (Fig. 4F). Thus, taken together, the results in Figs 2-4 show that inhibition of either PI3-kinase or Akt activity blocked all aspects of BMP2-mediated osteogenic differentiation of cultured mesenchymal stem cells, apparently without interfering with Smad function or impairing cell viability.

An inducible Akt promotes osteoblast differentiation in the presence of IGFBP5 or the PI3-kinase inhibitor LY294002
To test the hypothesis that the PI3-kinase-Akt pathway has an essential role in IGF-regulated osteogenic differentiation, we next asked whether an inducible-activated Akt (iAkt) could reverse the inhibition of differentiation seen with either IGFBP5 or LY294002 (see Fig. 5A and Fig. 6A). Synthesis of iAkt by the adenoviral gene delivery vehicle was prevented by Dox, and stimulated in its absence ( Fig. 5B and Fig. 6B), and full Akt enzymatic activity of the fusion protein was induced by the selective estrogen receptor modulator 4hydoxytamoxifen (4-HT), thus bypassing normal regulatory mechanisms (Tureckova et al., 2001). As shown in Fig. 5C, BMP2mediated osteoblast-specific gene expression was blocked by IGFBP5 but was restored by iAkt, as was accumulation of alkaline phosphatase and mineralization (  E, lanes '+LY294002, -Dox'). These latter data additionally show that the dose of LY294002 used was not toxic, because its effects could be reversed. Based on the results depicted in Figs 2, 4, 5 and 6, we conclude that the IGF-stimulated PI3-kinase-Akt pathway is required for BMP2-mediated osteogenic differentiation of cultured mesenchymal stem cells. Since both Smad phosphorylation and activation of Smad-dependent genes appeared to be normal when IGF-stimulated PI3-kinase or Akt were blocked, it seems likely that the point of interaction between the two growth-factor-initiated signaling cascades is downstream of Smad action and upstream of induction of osteoblast-specific gene expression.

Akt activity is required during all phases of osteoblast differentiation and function
Osteogenic differentiation is a multi-step process, which begins with expression of osteoblast-specific transcription factors (Lian et al., Journal of Cell Science 122 (5) 2006), and proceeds with production of bone-specific proteins, deposition of extracellular matrix and its subsequent mineralization (Balcerzak et al., 2003;Hoshi et al., 2000;Lian et al., 2006). As depicted in Figs 2, 4, 5 and 6, IGF-mediated Akt activity appears to be a necessary collaborator with BMP2-stimulated signaling pathways for initiating osteoblast differentiation. To begin to address whether Akt actions also are needed for later events in bone cell maturation and function, we devised a way to activate Akt DN by removal of Dox from the medium at different times during the course of BMP2-mediated osteogenesis of mesenchymal stem cells (see Fig. 7A). As shown in Fig. 7B, in the presence of Dox, no Akt DN was synthesized, whereas its sustained production was seen beginning 1 day after Dox removal from culture medium. Analysis of osteoblast gene expression revealed complete inhibition when Akt DN was present at the onset of differentiation (Fig. 7C, Dox removal on day 0; also see Fig. 4C). By contrast, Dlx5 and Runx2   7. Continual Akt activity is necessary for osteoblast differentiation, maturation and function. C3H10T1/2 cells were infected with Ad-Akt DN and Ad-tTA and incubated in osteogenic medium with BMP2 and Dox. Dox was removed sequentially at days 0, 2, 4 or 6 to induce expression of Akt DN . (A) Experimental scheme. (B) Immunoblots of wholecell protein lysates for Akt and Akt DN . (C) Results of RT-PCR assays for mRNA encoding Dlx-5, Runx2, Osx, Ocn and S17 at day 3, 5, 7 and 10. (D) Results of alkaline phosphatase activity measured at day 3, 5, 7 and 10 by staining and by in vitro enzymatic assay (graph) after removal of Dox on different days (mean ± s.d., n=3 experiments; *P<0.001, **P<0.01 vs +Dox). (E) Mineralized area assessed by Alizarin red staining at day 10 after removal of Dox on different days (mean ± s.d., n=5 experiments; *P<0.001, **P<0.05 vs +Dox at day 10). Representative images are depicted above the graph. mRNA was fully induced even when Akt DN was present from day 3 onwards (Fig. 7C, Dox removal on day 2), although expression of genes encoding Osx and Ocn was diminished by ~50-60%, and only reached maximal values when normal differentiation conditions were sustained for a longer interval (Fig. 7C, Dox removal on day 4). Similarly, alkaline phosphatase enzymatic activity, a measure of differentiated osteoblast function, was inhibited by >85% if Akt DN was present by day 3, but was fully induced in cells if the inhibitor did not appear until day 5 (Fig. 7D, compare Dox removal on day 2 and day 4). Taken together, the results in Fig. 7C,D indicate a requirement for continuous IGF-stimulated Akt activity for at least the first few days of BMP2-directed osteoblast differentiation in order for sufficient bone-specific mRNAs and proteins to be produced to sustain differentiated functions. However, matrix mineralization, a later event in the process of osteoblast maturation, appears to have a qualitatively different set of requirements, because the presence of Akt DN from day 5 onward prevented 50% of full mineralization by day 10, and expression of Akt DN from day 7 onwards still blocked ~25% of the normal accumulation of mineralized matrix at day 10 ( Fig. 7E, compare Dox removal on day 4 and day 6 with +Dox), even though at these time points Akt DN had no inhibitory effects on expression of bone genes or on alkaline phosphatase activity. Our provisional interpretation of the results in Fig. 7 is that there are several temporally distinct Akt targets in differentiating osteoblasts that govern different aspects of bone cell development and function.
Signaling through the PI3-kinase-Akt pathway is required for growth of isolated mouse metatarsal bones Short-term culture of neonatal mouse metatarsal bones has been used previously to study bone growth and endochondral ossification (Krishnan et al., 2003;Mukherjee et al., 2005). As we have shown recently, these bones increased in length by >35% in serum-free medium over a 10-day culture period, but growth was blocked by IGFBP5 . As also depicted in Fig.  8, a single addition of 20 μM LY294002 at 1 day after plating reduced growth to <5% over the same time course. We observed similarly dramatic inhibitory effects on longitudinal growth in metatarsals infected with Ad-Akt DN at the beginning of ex vivo culture, whereas, by contrast, infection with Ad-EGFP or Ad-Akt DN in the presence of Dox, was completely ineffective (Fig. 9). Thus, in this model system, the PI3-kinase-Akt pathway appears to be needed for normal bone growth.

Inhibition of Akt activity impairs cartilage growth and osteoblast development and function in isolated mouse metatarsal bones
We analyzed histological sections of mouse metatarsal bones incubated ex vivo for different intervals to address potential mechanisms by which Akt signaling was required for bone growth. At the start of the 10-day culture period, proliferating cartilage made up nearly half of the bone length and, together with hypertrophic cartilage, comprised 85% of the total, with the mid-diaphyseal mineralizing zone comprising the remaining 15% (Fig. 10, top  panel). By day 10, this central mineralizing zone had increased to almost 30% of the now longer bone, with proliferating cartilage remaining proportionately constant (54%), and the hypertrophic zone decreasing from 39% to 18% (Fig. 10, second panel). We observed nearly identical results on day 10 of culture with bones infected with Ad-Akt DN and incubated with Dox (Fig. 10, third  panel), thus illustrating the lack of effect of adenoviral infection on the proportion of different cell types in the developing and growing bone. By contrast, expression of Akt DN not only prevented longitudinal metatarsal growth, but also completely inhibited lengthening of the zone of proliferating cartilage and expansion of the mineralized zone (Fig. 10, bottom panel). As a consequence, the histological profile after 10 days of Akt DN expression resembled  that of control metatarsals at the onset of ex vivo incubation (Fig.  10, compare top and bottom panels). Thus, Akt signaling appears to be required for both cartilage and bone growth in isolated metatarsal bones.
We further examined the effects of inhibiting Akt signaling on osteoblasts within the mid-diaphyseal mineralizing zone. Cellcounting experiments demonstrated that the number of morphologically recognizable osteoblasts per ϫ400 microscopic field nearly doubled (from 16±3 to 28±7) during 10 days of organ culture, whereas total cell density did not change (Fig. 11A). As the mineralizing zone also expanded twofold in absolute length (Fig.  10), it appears that the total number of osteoblasts increased by a factor of four during ex vivo bone development. We recorded nearly identical results in bones infected with Ad-Akt DN and incubated with Dox. By contrast, when Akt DN was expressed in the absence of Dox, both osteoblast numbers and total cell density were unchanged at day 10 compared with control metatarsals at day 0 (Fig. 11A). Our interpretation of these results is that Akt appears to be required for osteoblast recruitment and/or differentiation within bone during the early postnatal period in mice, although alternatively, Akt signaling might be necessary for osteoblast viability.
We next assessed osteoblast maturation and function in metatarsal cultures by examining bone-cell-specific gene expression and mineralization. Akt DN reduced induction of mRNA encoding Ocn by 70%, but had no effect on transcripts encoding Runx2 or Osx (Fig. 11B). Since Runx2 and Osx are produced by both chondrocytes and osteoblasts, whereas Ocn is synthesized exclusively by differentiated osteoblasts (Karsenty, 1998), we interpret these data to indicate that Akt DN interfered with osteoblast development Journal of Cell Science 122 (5) within bone. Similarly, as measured by both calcein labeling of living bones (Fig. 11C) and Alizarin red staining of histological sections (Fig. 11D), Akt DN prevented the normal accumulation of mineralized matrix. Thus, based on these results, we conclude that continuous Akt signaling is necessary for full osteoblast maturation and function in vivo, as well as in vitro.

Discussion
BMPs have a central role in bone development and osteoblast differentiation (Li and Cao, 2006), but require interactions with other growth-factor-activated signals. Here, we demonstrate essential crosstalk between BMP2 and the IGF-activated PI3-kinase-Akt pathway to initiate osteogenic differentiation in uncommitted mesenchymal precursor cells, and to promote maturation of committed osteoblasts. We also show that IGF-stimulated and Aktmediated signaling is crucial for longitudinal bone growth by exerting facilitating effects on both chondrocyte and osteoblast development and function. Based on these results, we conclude that the PI3-kinase-Akt pathway is a crucial component of an interactive osteogenic signaling network.
An IGF-activated PI3-kinase-Akt pathway regulates BMP2mediated osteoblast differentiation We previously found that IGFBP5 could block BMP2-regulated osteogenic differentiation of mesenchymal stem cells by sequestering IGF-I and IGF-II from the IGF-I receptor , and postulated that through this mechanism it also prevented osteoblast maturation and blocked longitudinal growth of mouse metatarsal bones. We now show that both a chemical PI3-kinase inhibitor and a dominant-negative version of Akt (Akt DN ), also can inhibit BMP2-initiated osteogenesis (Figs 2  and 4). By contrast, blocking the Mek-Erk signaling pathway had no effect on the onset or progression of BMP2-activated osteoblast differentiation (Fig. 2). As the inhibitory actions of IGFBP5 or LY294002 on BMP2-stimulated osteogenesis could be reversed by a recombinant adenovirus encoding an activated version of Akt (Figs 5 and 6), our results document that Akt-regulated signaling is the key pathway of IGF action in promoting osteoblast differentiation in collaboration with BMPs.
Several previous studies have supported facilitating roles for IGFmediated signaling in osteogenesis in cell culture models, but with disparate results regarding the intracellular pathways implicated (Merriman et al., 1990;Niu and Rosen, 2005;Strong et al., 1991;Strong et al., 1994). Fujita and colleagues found that either PI3kinase or Mek inhibitors could reduce the amount of bone-specific alkaline phosphatase produced by cells overexpressing Runx2 (Fujita et al., 2004). Raucci and co-workers also showed that these chemical inhibitors could decrease alkaline phosphatase activity in two additional osteogenic cell lines, and attributed the negative effects of the PI3-kinase inhibitor to enhanced cell death (Raucci et al., 2008). These authors also found that a constitutively active Akt led to increased accumulation of several bone-specific mRNAs, including those encoding Runx2 and Osx (Raucci et al., 2008). Several other groups also showed that chemical inhibition of PI3kinase or Mek could reduce markers of osteoblast differentiation (Ghosh-Choudhury et al., 2002;Ghosh-Choudhury et al., 2007;Hanai et al., 2006;Noda et al., 2005;Osyczka and Leboy, 2005). Our observations thus appear to be one of the few studies to clearly discriminate between the PI3-kinase-Akt and Mek-Erk pathways in terms of osteogenic outcomes.
What mechanisms might mediate Akt-regulated osteoblast differentiation? A dominant-negative Akt decreased the transcriptional actions of Runx2 in a model system in which Runx2 was overexpressed (Fujita et al., 2004). These authors additionally found that overexpression of Runx2 led to an increase in abundance of Akt and of both regulatory and catalytic PI3-kinase subunits, and postulated the existence of a positive-feedback loop in which Runx2 upregulated components of the PI3-kinase-Akt pathway, which then enhanced the functions of Runx2 (Fujita et al., 2004). Others have found that a dominant-negative Akt could reduce the activity of a BMP2-dependent promoter-reporter gene, possibly by inhibiting the nuclear accumulation of activated Smad1 and Smad5 (Ghosh-Choudhury et al., 2002), but these results have not been replicated, and we see no inhibitory effects of dominant-negative Akt on Smad-mediated Sox9 and JunB gene expression (Fig. 4B). In another mesenchymal derivative, skeletal muscle, Akt has been shown to collaborate with myogenic transcription factors to enhance the abundance of transcriptional co-activators on muscle gene promoters (Wilson and Rotwein, 2007), and to reduce co-repressors (Serra et al., 2007). It thus might be reasonable to postulate an analogous role for Akt signaling in osteoblast differentiation, although to date there is little experimental evidence for or against this idea.

Akt signaling is required in all phases of osteoblast differentiation and maturation
Osteoblast differentiation can be divided into several phases, including lineage commitment, characterized in part by expression of Runx2 and Osx (Komori, 2008), early differentiation, in which other bone-cell-specific mRNAs and proteins are produced (Deng et al., 2008), and maturation, marked by accumulation of bonespecific alkaline phosphatase, extracellular matrix deposition and mineralization (Balcerzak et al., 2003;Hoshi et al., 2000;Zaidi, 2007). We now find that Akt activity appears to be required for each of these stages of differentiation, because Akt DN can block progression from one step to the next (Fig. 7). Most remarkable in this regard is the inhibitory action of Akt DN on mineralization, which was seen even when Akt DN was added relatively late in the differentiation process, at a time when alkaline phosphatase activity was already maximal (compare Fig. 7D with 7E). These latter results, which complement our previous data using IGFBP5 to block osteogenesis , are supported by the inhibitory effects of Akt DN on osteoblast maturation and function in cultured metatarsal bones (Fig. 11), and also are consistent with Fig. 11. Akt DN inhibits osteoblast development and function in cultured neonatal mouse metatarsal bones. Neonatal mouse metatarsals were uninfected (Con) or were infected with Ad-Akt DN and Ad-tTa, and incubated in DMEM plus 0.5% BSA with or without Dox for 10 days. (A) Cell counts of osteoblasts per field (h.p.f.) in histological sections of the mineralized zone at ϫ400 magnification [mean ± s.d., n=4; *P<0.001, vs Con (day 0)]. (B) Results of RT-PCR experiments at day 10 for mRNA encoding Runx2, Osx, Ocn, and S17. (C) Representative fluorescent images of calcein-labeled mineralizing zone after incubation for 10 days; graph shows the relative difference in length of the calceinlabeled mineralizing zone at day 10 in metatarsals incubated with or without Dox (mean ± s.d., n=4; *P<0.001, vs +Dox). (D) Representative images showing Alizarin-red-stained histological sections after a incubation for 10 days; graph shows the difference in the Alizarin-redstained area at day 10 in metatarsals incubated with or without Dox (mean ± s.d., n=4; *P<0.01).
the defective mineralization phenotype seen in mice lacking the IGF-I receptor in mature osteoblasts (Zhang et al., 2002). Mineralization represents the outcome of a complex series of steps that include active transport of calcium and inorganic phosphate into osteoblasts, release of matrix vesicles into the extracellular space, and nucleation and deposition of hydroxyapatite granules in the osteoid (Balcerzak et al., 2003;Hoshi et al., 2000). Among factors that control mineralization is the sodium-dependent phosphate transporter Pit-1, which appears to be regulated by IGF-I in osteoblasts (Kavanaugh and Kabat, 1996;Palmer et al., 1997;Selz et al., 1989). As mineralization is an important step in fracture healing (Schindeler et al., 2008), an understanding of the regulatory mechanisms has the potential to lead to better treatment options.

Akt signaling in bone growth and endochondral ossification
We showed previously that incubation with IGFBP5 prevented both longitudinal growth and mineralization of cultured neonatal mouse metatarsal bones, and found that the inhibitory effects of IGFBP5 depended on its ability to bind IGFs with high affinity . We now show that a chemical PI3-kinase inhibitor and adenoviral-delivered Akt DN also block metatarsal growth (Figs 8 and 9). In these experiments, the metatarsals were incubated in serum-free medium, and because IGFBP5, LY294002 and Akt DN all exerted similar inhibitory effects, we conclude that locally produced IGFs are responsible for activating the PI3kinase-Akt signaling pathway that is essential for longitudinal bone growth. The defects seen in our metatarsal model resemble the bone phenotype in mice lacking the IGF-I receptor (Zhang et al., 2002) or both Akt1 and Akt2 (Peng et al., 2003), in which ossification was delayed and osteopenia resulted. Conversely, the opposite phenotype was observed in transgenic mice lacking Pten in osteoblasts, in which progressive increases in bone volume and density were seen throughout life (Liu et al., 2007). As Pten dephosphorylates and inactivates phosphatidylinositol (3,4,5)trisphosphate (PIP3), the product of PI3-kinase, and because PIP3 is essential for membrane targeting and activation of Akt (Franke, 2008;Hanada et al., 2004), these results predict enhanced Akt activity in Pten-deficient osteoblasts. In fact, cultured calvarial osteoblasts engineered to lack Pten did show increased Akt phosphorylation and phosphorylation of several Akt substrates, and the cells differentiated more extensively than controls (Liu et al., 2007). No results have been reported yet on mice in which a constitutively active Akt has been targeted to osteoblasts, but a similarly high bone density phenotype might be anticipated.

Akt is essential for optimal chondrogenesis and osteogenesis
Histological analysis of growing mouse metatarsal bones revealed a growth-associated proportional increase in the zone of proliferating chondrocytes, a decline in the extent of terminally differentiated hypertrophic chondrocytes and a more-than-proportional rise in the length of the central mineralized zone, which contained osteoblasts as well as other cell types (Fig. 10). These growth-related changes in the profile of cell types within the cultured metatarsals were completely inhibited by Akt DN (Fig. 10), thus demonstrating negative effects on both chondrocyte and osteoblast development. In their analysis of isolated tibias from E15.5 mouse embryos, Ulici and colleagues also found that LY294002 could impair chondrocyte differentiation and inhibit longitudinal bone growth (Ulici et al., 2008). In addition, Akt DN blocked the normal twofold increase in osteoblast number seen in the central mineralizing zone, and also impaired both osteoblast maturation and function, as measured by diminished osteocalcin gene expression and reduced mineralization (Fig. 11). Thus, sustained Akt activity appears to be required for the normal cartilage and bone cell development that leads to longitudinal bone growth during the early postnatal period, at least in metatarsals. These observations are additionally supported by previous studies, which have suggested that Akt signaling is required for proteoglycan and collagen production in chondrocytes mediated by Runx2 (Fujita et al., 2004), and that Akt1 is important for normal rates of bone formation and for preventing osteoblast apoptosis (Kawamura et al., 2007).
In summary, we have shown that the IGF-activated PI3kinase-Akt signaling pathway is a potent facilitator of osteoblast differentiation, bone growth, and mineralization. Our results point to a key role for IGF-mediated signaling in all phases of osteogenesis, and provide an impetus to define the mechanisms of interaction with BMPs and other regulators of cartilage and bone development and function.

Reagents
Fetal calf serum, horse serum, Dulbecco's modified Eagle's medium (DMEM), and phosphate-buffered saline (PBS) were purchased from Mediatech-Cellgrow (Herndon, VA). Okadaic acid was from Alexis Biochemicals (San Diego, CA), and NBT/BCIP tablets and protease inhibitor tablets were from Roche Applied Sciences (Indianapolis, IN). Calcein, sodium orthovanadate, alizarin red, ascorbic acid and β-glycerol phosphate were purchased from Sigma (St Louis, MO). Trypsin-EDTA solution and Superscript III first-strand synthesis kit were from Invitrogen (Carlsbad, CA). The BCA protein assay kit was from Pierce Biotechnologies (Rockford, IL) and Immobilon-FL was from Millipore Corporation (Billerico, MA). AquaBlock EIA/WIB solution was from East Coast Biologicals (North Berwick, ME). IGF-I (Gropep) was stored at -80°C at a concentration of 10 mM in 0.01 M HCl until use. Doxycycline (Dox, Clontech, Palo Alto, CA) was used at a final concentration of 1 mg/ml. LY294002 (Biomol Research Laboratories, Plymouth Meeting, PA) was stored in dimethyl sulfoxide at -20°C at a concentration of 20 mM until use; 4hydroxytamoxifen (HT) was from Sigma, and was stored in ethanol at -20°C at a concentration of 50 mM until use. UO126 (Promega, Madison, WI) was stored in ethanol at -80°C at a concentration of 20 mM until use. Other chemicals were purchased from commercial vendors.

Production of BMP2 and IGFBP5
C3H10T1/2 mouse embryonic fibroblasts (ATCC: CCL226) were incubated at 37°C in humidified air with 5% CO 2 in DMEM with 10% fetal calf serum. Cells were infected at ~50% of confluent density with Ad-BMP2 [multiplicity of infection (MOI) of 500]. The following day, medium was replaced with DMEM plus 2% horse serum; 2 days later, conditioned medium was collected, clarified, and stored in aliquots at -80°C until use. The concentration of BMP2 was determined by immunoblotting with purified standards purchased from R&D systems (Minneapolis, MN) (see supplementary material Fig. S2). Mouse IGFBP-5 was produced in C3H10T1/2 cells following infection with Ad-IGFBP-5, and was purified by heparin affinity chromatography .

Osteogenic differentiation
Confluent C3H10T1/2 cells were incubated in osteogenic medium (DMEM, 10% fetal calf serum, 50 μg/ml ascorbic acid, 10 mM β-glycerol phosphate and 200 ng/ml BMP2) in the absence or presence of LY294002 (20 μM), UO126 (10 μM), or IGFBP-5 (150 nM). osteogenic medium was replaced every 48 hours for up to 10 days. Cell counting was performed as described (Wilson and Rotwein, 2007). Alternatively, C3H10T1/2 cells were infected at ~50% of confluent density with Ad-tTA at an MOI of 125, and either Ad-EGFP, Ad-iAkt, or Ad-Akt DN at MOIs of 500. One day later, cells were washed and osteogenic medium was added, along with other chemicals as described in individual figure legends.

Mouse metatarsal bone culture
Metatarsal bones were isolated from newborn C57BL6 mice (days 0-3 after birth), as described (Krishnan et al., 2001; and were incubated in DMEM containing 0.5% bovine serum albumin, 50 μg/ml ascorbic acid, 1 mM β-glycerol phosphate and 100 μg/ml penicillin-streptomycin solution at 37°C in humidified air with 5% CO 2 for up to 10 days. In some experiments LY294002 (20 μM) was added the next day. In others, bones were infected the next day with Ad-tTA (1ϫ10 7 PFU/ml) plus either Ad-EGFP or Ad-Akt DN (6ϫ10 7 PFU/ml). Images were captured at days 1, 4, 7 and 10 with a Nikon DXL1200 camera attached to a Lieca MZ FLIII microscope. Mineralization was assessed by addition of medium containing calcein (500 ng/ml) for 2 hours. After rinsing three times with PBS, fluorescent images were captured with a Roper Scientific Cool Snap FX CCD camera attached to a Nikon Eclipse T300 microscope using an Apple PowerPC computer running IP Labs Scientific Image Processing software v3.9.4r2 (Scanalytics, Rockville, MD).

Bone histology
Metatarsals were fixed in 4% paraformaldehyde for 18 hours at 4°C and stored in 70% ethanol. Bones were embedded in paraffin blocks and sectioned. Staining with hematoxylin and eosin or with alizarin red was performed after hydrating slides in graded concentrations of ethanol and water, as described (Mukherjee et al., 2005;Serra et al., 1999). After staining, sections were dehydrated, coverslips added and images were captured using a MicroPublisher cooled CCD camera (QImaging, Surrey, British Columbia) attached to a Nikon Eclipse E800 compound microscope. For cell counting, osteoblasts were identified in the central mineralized zone by the cuboidal morphology of their nuclei, and were counted at ϫ400 magnification in histological sections stained with hematoxylin and eosin.

Alkaline phosphatase staining
Cells were washed with PBS, fixed with 70% ethanol for 10 minutes, and incubated with 500 ml NBT/BCIP solution (1 tablet in 10 ml distilled water) for 20 minutes at 20°C . After three washes with distilled water, images were captured and analyzed with the LiCoR Odyssey Infrared Imaging System, using software version 1.2 (LiCoR, Lincoln, NE). Alkaline phosphatase activity was determined spectrophotometrically at 405 nM after incubating cell lysates (10 μg) in a 96-well format for 20 minutes at 20°C in 50 μl of a 1 mg/ml solution of pnitrophenyl phosphate .

Alizarin red staining
Cells were fixed in 70% ethanol for 10 minutes, and stained with 2% alizarin red solution (pH 4.1-4.5) for 1 minute at 20°C . Images were obtained by scanning plates on a Canon flat-bed scanner or with the LiCoR Odyssey and results were quantified as described .

Protein extraction and immunoblotting
Whole-cell protein lysates and conditioned cultured medium were prepared from C3H10T1/2 cells as described , and aliquots were stored at -80°C until use. Metatarsals were homogenized in cell lysis buffer with protease inhibitors using a hand-held Teflon homogenizer. After centrifugation at 14,000 r.p.m. for 10 minutes at 4°C in a microcentrifuge, supernatants were collected and stored at -80°C until use. Protein samples (30 μg/lane) or medium (25 μl/lane) were resolved by SDS-PAGE and transferred to Immobilon-FL membranes. After blocking with 25% AquaBlock solution for 1 hour at 20°C, membranes were incubated sequentially with primary and secondary antibodies . The following primary antibodies were used at a dilution of 1:1000: anti-Akt, anti-phospho-Akt (Ser 473 ), anti-IGFBP-5, anti-phospho-Smad, anti-Smad1, anti-HA, anti-Erk1/2 and anti-phospho-Erk1/2. Anti-T7 was used at 1:5000 and anti-α-tubulin at 1:15,000. Secondary antibodies were used at 1:5000.
Results were visualized and images captured using the LiCoR Odyssey and version 1.2 analysis software.