FGF2 Stimulation of the Pyrophosphate-Generating Enzyme, PC-1, in Pre-Osteoblast Cells Is Mediated by RUNX2

Abstract

Pyrophosphate is an established inhibitor of hydroxyapatite deposition and crystal growth, yet when hydrolyzed into phosphate, it becomes a substrate for hydroxyapatite deposition. Pyrophosphate-generating enzyme (PC-1), Ank, and tissue nonspecific alkaline phosphatase (Tnap) are three factors that regulate extracellular pyrophosphate levels through its generation, transport, and hydrolysis. We previously showed that fibroblast growth factor 2 (FGF2) induces PC-1 and Ank while inhibiting Tnap expression and mineralization in MC3T3E1(C4) calvarial pre-osteoblast cells. In this study, we showed similar FGF2 regulation of these genes in primary pre-osteoblast cultures. In contrast to Ank and Tnap that are regulated by FGF2 in multiple cell types, we found regulation of PC-1 to be selective to pre-osteoblastic cells and to require the osteoblast-related transcription factor, Runx2. Specifically, FGF2 was unable to induce PC-1 expression in Runx2-negative nonbone cells or in calvarial cells from Runx2-deficient mice. Transfection of these cells with a Runx2 expression vector restored FGF2 responsiveness. FGF2 was also shown to stimulate recruitment of Runx2 to the endogenous PC-1 promoter in MC3T3E1(C4) cells, as measured by chromatin immunoprecipitation. Taken together, our results establish that FGF2 is a specific inducer of PC-1 in pre-osteoblast cells and that FGF2 induces PC-1 expression through a mechanism involving Runx2.

INTRODUCTION

Fibroblast growth factor (FGF) signaling plays a critical role in skeletal development, yet the mechanism of FGF action in bone is not well understood. In fact, a close review of the literature yields a complex and potentially paradoxical story. FGF2 knockout mice exhibit significantly diminished bone mass, bone formation rate, and trabecular bone volume, suggesting that FGF2 is a positive regulator of bone formation, yet FGF2 overexpression in mice also results in significantly diminished BMD and trabecular bone volume.(1,2) Similarly, bone marrow stromal cells (BMSCs) from both FGF2 knockout and FGF2 overexpressor mice show a profoundly diminished ability to form mineralized nodules in culture. Furthermore, whereas systemic FGF1 and FGF2 are known to have anabolic effects on bone in vivo, this effect does not occur until after cessation of FGF treatment. Significantly, close inspection of FGF-treated mice shows that mineralization is completely inhibited during the course of FGF treatment.(3,4) Taken together, these results indicate that FGF has conflicting direct and indirect effects on bone mineralization and suggest that FGFs stimulate expression of factors that prevent mineralization in the short term and enhance mineralization in the long term. Pyrophosphate is an ideal example of a factor having both negative and positive effects on mineralization; pyrophosphate itself inhibits mineralization but also serves as an essential source of phosphate to enhance mineralization when it is hydrolyzed to inorganic phosphate.(5,6) Altered expression of pyrophosphate elaborating factors downstream of FGF signaling could provide a potential mechanism for the apparently contradictory immediate versus long-term effects of FGF2.

PC-1 (ectonucleotide pyrophosphate/phosphodiesterase-1 [Enpp1]), tissue nonspecific alkaline phosphatase (Tnap), and Ank are three factors that regulate pyrophosphate levels through its generation, transport, and hydrolysis into phosphate. PC-1 is a nucleoside triphosphate pyrophosphohydrolase that generates pyrophosphate by hydrolysis of ATP.(7) PC-1 activity increases extracellular levels of pyrophosphate, which inhibits hydroxyapapatite crystal propagation.(8–10) However, this pyrophosphate also serves as a source of phosphate for hydroxyapatite crystal generation when it is hydrolyzed by the osteoblast enzyme, Tnap.(6,11) The superficially paradoxical role of PC-1 and pyrophosphate elaboration in tissue calcification is evidenced by the fact that Enpp1 (PC-1)-null mice exhibit diminished bone mineralization but increased calcification of normally nonmineralized tissues.(12) This finding, plus the fact that PC-1 is upregulated during osteoblast differentiation, supports the idea that, while pyrophosphate is itself an inhibitor of tissue mineralization, it is also acts as a critical source of phosphate for bone mineralization.(10) It is important to note here that, although PC-1 enzyme activity continuously increases with osteoblast differentiation, levels of pyrophosphate only rise transiently for the first few days of culture. Pyrophosphate levels then drop to predifferentiation levels, presumably as a result of hydrolysis by Tnap, which is then expressed. Significantly, a cross of the Tnap^−/− mouse with the Enpp1 ^−/− mouse generates a double knockout mouse with apparently normal long bone mineralization.(13,14) This result supports the idea that PC-1 and Tnap work together to produce normally mineralized bone matrix through the generation and hydrolysis of pyrophosphate.

Ank is a 12-membrane spanning protein associated with two separate human diseases of mineralization (chondrocalcinosis and craniometaphyseal dysplasia).(15–17) Ank decreases intracellular pyrophosphate levels and increases extracellular pyrophosphate levels.(18) This function of Ank is blocked by treatment with probenicid (an anion channel inhibitor), which suggests that Ank is a transmembrane channel allowing for the transport of pyrophosphate from the intra- to the extracellular space.(18) The critical role of Ank in the calcification of tissues is evidenced by the fact that inactivating mutations in Ank result in a phenotype that is similar to the PC-1 knockout mice in terms of pathologic calcification.(12)

We previously showed that FGF2 specifically upregulates expression of PC-1 and Ank and inhibits expression of Tnap in the MC3T3E1(C4) calvarial pre-osteoblast cell line.(19) As noted above, pyrophosphate has the potential to both inhibit and enhance (when hydrolyzed to phosphate) mineralization. Therefore, the regulation of these three factors by FGF2 may explain the apparently contradictory short-term and long-term effects of FGF2 on bone mineralization. Because pyrophosphate generation and hydrolysis downstream of FGF signaling may be a primary mechanism for controlling bone mineralization, in this study, we examine FGF2 regulated expression of PC-1, Tnap, and Ank in primary osteoblastic cells and in non-osteoblastic cell types. Because PC-1 is the enzymatic generator of pyrophosphate and because we previously showed that the regulation of PC-1 gene expression is osteoblast differentiation stage dependent, we also study the underlying mechanism of FGF2-induced PC-1 expression.

Here we show that FGF2 increases PC-1 and Ank expression while inhibiting Tnap expression in primary pre-osteoblast cells. Additionally, we show that the induction of PC-1 by FGF2 is cell type specific and mediated by the transcription factor, Runx2.

MATERIALS AND METHODS

Cell culture

The mTERT immortalized calvarial cell line from Runx2-null mice(20) was a generous gift from Dr Jane Lian (University of Massachusetts School of Medicine). MC3T3E1(C4), C3H10T1/2, and Runx2-null cells were cultured in custom formulated αMEM containing no ascorbate, supplemented with 10% FBS and 10,000 μg/ml penicillin/streptomycin (P/S). COS7 cells were cultured in DMEM supplemented with 10% FBS and P/S. Primary calvarial and BMSCs were isolated from 6-wk-old C57BL mice as previously described.(21)

Cytokine treatments were conducted in media containing 0.5% FBS. Recombinant FGF2 (Peprotech) was added to a final concentration of 50 ng/ml along with 1 μg/ml heparin, unless otherwise indicated. Recombinant BMP2 (R&D Systems) was added to a final concentration of 100 ng/ml. For transcription and translation inhibitor treatments, cells were preincubated with actinomycin D (Sigma) or cycloheximide (Sigma) for 1 h before cytokine treatment.

Transfection

Superfect transfection reagent (Qiagen) was used for all transfection experiments, following the manufacturer's recommended protocol. Briefly, cells were incubated with DNA/Superfect complexes for 4 h in media containing 10% FBS and penicillin/streptomycin. Cells were washed with PBS, and media were replaced. Cells were incubated for an additional 24 h before cytokine treatment or cell lysis.

Viral transduction

Transductions were performed using attenuated adenoviral vectors expressing either β-galactosidase (LacZ; pAd-LacZ) or Runx2 (pAd-Runx2).(22) Briefly, cells were transduced in media containing no serum or antibiotics and ∼100 pfu/cell of virus for 4 h at 37°C. An equivalent volume of media containing 4% FBS was added, and cells were incubated for an additional 24 h before cytokine treatment.

DNA constructs

A 2864-bp region of the proximal PC-1 gene promoter was cloned by PCR (Table 1). This PCR product was digested with the restriction enzymes Kpn1 and Xho1, purified, and ligated into PGL3basic vector (Promega) to create PGL3/PC1. Renilla luciferase constructs, PGL4/hRluc/TK or PGL4/hRluc (Promega), were used as internal controls for transfection efficiency. pCMV5/LacZ and pCMV5/Runx2 expression vectors were previously described.(21)

Table 1.

Oligonucleotide Primers Used in This Study

Mutagenesis

PCR-based in vitro mutagenesis was performed to mutate putative Runx2 and msx2 binding sites located within the PC-1 gene promoter (Table 1). Four potential PC-1 promoter Runx2 binding sites were individually mutated: site 1 found at −2554 to −2549, site 2 found at −2500 to −2495, site 3 found at −810 to −805, and site 4 found at −312 to −307.

mRNA quantification

RNA was isolated using Trizol reagent (Invitrogen) following manufacturer's protocols. mRNA levels were assayed by reverse transcription and semiquantitative or real-time PCR. Semiquantitative RT-PCR was conducted using the Titanium One-Step RT PCR Kit (Clontech) (Table 1). Semiquantitative PCR parameters were established by initial experimentation varying PCR cycle number (data not shown). Five-microliter aliquots of PCR reactions were visualized by agarose gel electrophoresis and ethidium bromide staining. Densitometry was performed using NIH image software. Real-time PCR was performed using the murine PC-1 primer/probe set Mm01193752_m1, the murine Tnap primer/probe set Mm00475834_m1, the murine Ank primer/probe set Mm00445050_m1, the murine GAPDH primer/probe set Mm9999915_g1, the murine Runx2 primer/probe set Mm00501578_m1 (does not distinguish endogenous from exogenous Runx2 mRNA), and Taqman Universal PCR Master Mix (Applied Biosystems). Real-time PCR was performed on a GeneAmp 7700 thermocyler (Applied Biosystems) and quantified by comparison with a standard curve. PC-1 and Runx2 mRNA levels are reported after normalization to GAPDH mRNA levels.

PC-1 enzyme activity assay

PC-1 enzyme activity was assayed using the colorimetric substrate, p-nitrophenol thymidine 5′ monophosphate (Sigma). Cells were lysed in 1% triton and equivalent amounts of protein were incubated with 1 mg/ml p-nitrophenol thymidine 5′ monophosphate for 1 h; 100 mM NaOH was added to stop the reaction and absorbance was measured at 405 nm.

Luciferase assays

PC-1 promoter activity was assayed using the Dual Glo Luciferase Assay System (Promega) following the manufacturer's protocol. Cells were co-transfected with experimental firefly and control renilla luciferase constructs. Firefly luciferase activity was normalized to renilla luciferase activity to account for potential differences in transfection and/or cell lysis efficiency.

Nuclear extraction

Nuclei were isolated and nuclear extract was purified as previously described.(22) Cells were lysed in hypotonic buffer and homogenized with a type B pestle. Cell lysis was monitored by trypan blue staining. Lysed cells were centrifuged and resuspended in a low salt buffer. An equivalent volume of high salt buffer was added, and cells were incubated with gentle rotation at 4°C. Nuclear pellets were centrifuged and dialyzed, before storage at −80°C.

EMSAs

Double-stranded DNA oligonucleotides were designed to include 20 bp of the PC-1 promoter sequence surrounding each of the four potential Runx2 binding sites. Each oligonucleotide was designed to include a two guanine overhang for [³²P]dCTP labeling (Table 1). Mutant PC-1 promoter Runx2 binding site oligonucleotides were also designed for this assay. These mutant oligonucleotides include the same mutations that were incorporated into the PGL3/PC1 luciferase constructs to eliminate the potential for Runx2 binding (Table 1).

EMSAs were performed as previously described.(22) Duplexed oligonucleotide was labeled with [³²P]dCTP, purified on G-50 columns (Amersham), and verified for radioactivity by scintillation counting. Radiolabeled oligonucleotides were incubated with nuclear extract and 0.1 μg/ml poly dI-dC and separated on 10% Tris-Borate-EDTA buffered acrylamide gels at 4°C. For supershift assays, Runx2 (D130-3; MBL International), or control IgG (sc-2027; Santa Cruz) antibody was added to the binding reaction before gel electrophoresis.

Chromatin immunoprecipitation assays

Chromatin immunoprecipitation (ChIP) assays were conducted as previously described.(22) Chromatin (10 μg; input DNA) and 2 μg of Runx2 or control IgG antibody were used for each reaction. PCR was performed using primers generated to detect DNA segments located near the putative Runx2 binding sites within the distal promoter region, the putative Runx2 binding sites within the proximal promoter region, or within exon 18 of the PC-1 gene (Table 1).

Immunoblot assay

Cell lysate was solubilized in RIPA buffer (50 mM Tris-Cl, pH 7.4, 100 mM NaCl₂, 1% sodium deoxycholate, 1% Triton-X 100, 0.1% SDS) containing protease inhibitor cocktail (Sigma). Cell lysate (50 μg) or 10 μg nuclear extract was separated by SDS-PAGE and transferred onto Immobilon (Millipore). Immunoreactive protein bands were visualized by incubation with mouse anti-Runx2 (D130-3; MBL) or mouse anti-β-actin (A3853; Sigma) and horseradish peroxidase–conjugated goat anti-mouse antibody (sc-2005; Santa Cruz), followed by enhanced chemiluminescence (Pierce).

RESULTS

FGF2 induction of PC-1 expression in primary pre-osteoblastic calvarial cells (BMSCs)

We previously showed that FGF2 induces PC-1 expression in undifferentiated MC3T3E1(C4) calvarial osteo-progenitor cells.(19) To confirm the physiological relevance of this phenomenon, primary calvarial cells and BMSCs were isolated, treated with FGF2, and assessed for changes in PC-1 expression. Results confirm that FGF2 significantly induces PC-1 mRNA expression (Figs. 1A and B) and PC-1 enzymatic activity in these cells (Figs. 1C and 1D). The ability of FGF2 to induce PC-1 expression in the non-osteoblastic cell lines NIH3T3, C2C12, and 3T3L1 was also assessed and compared with MC3T3E1(C4) cells. FGF2 strongly induced PC-1 expression in the osteoblastic cell line, but not in the non–bone-related cell lines, indicating that the FGF2 response may be selective to bone-related cells (Fig. 1E). It is worthy to note here that each of these non-osteoblastic cell types express FGF receptors and are responsive to FGF2.(23–25)

FIG. 1.

FGF2 induction of PC-1 expression in primary pre-osteoblastic cells. Primary mouse calvarial cells and BMSCs were treated with FGF2 as indicated, before mRNA isolation or cell lysis. (A and B) FGF2 increases PC-1 mRNA expression in (A) primary calvarial cells and (B) BMSCs. RNA was isolated and used for semiquantitative RT-PCR. PC-1 mRNA levels are presented as normalized to GAPDH. (C and D) FGF2 increases PC-1 enzyme activity in (C) primary calvarial cells and (D) BMSCs. Equivalent amounts of cellular extract were incubated with colorimetric substrate and absorbance was read at 405 nm. *p < 0.05 vs. no treatment. (E) The induction of PC-1 mRNA by FGF2 is cell type specific. RNA was isolated and used for semiquantitative RT-PCR. PC-1 mRNA was normalized to GAPDH and is reported as fold induction by FGF2.

FGF2 regulation of Ank and Tnap expression in primary pre-osteoblast cells and non-osteoblastic cells

We previously showed that FGF2 induces Ank and inhibits Tnap expression in undifferentiated and osteoblast differentiated MC3T3E1(C4) cells.(19) To assess the physiological relevance of this phenomenon, Ank and Tnap mRNA expression levels were evaluated in FGF2-treated primary calvarial cells and BMSCs. Results confirm that FGF2 significantly induces Ank (Figs. 2A and 2B) and inhibits Tnap (Figs. 2C and 2D) expression in these cells. The ability of FGF2 to regulate Ank and Tnap mRNA expression in non-osteoblastic cell types was also assessed. FGF2 significantly induced Ank expression in all studied cell types, although the most dramatic induction was seen in pre-osteoblast cells (Fig. 2E). FGF2 also inhibited Tnap expression in all studied cell types. Of note, inhibition of Tnap mRNA expression by FGF2 was strongest in pre-myoblastic and pre-adipocytic cells types (Fig. 2F).

FIG. 2.

FGF2 regulation of Ank and Tnap expression in primary pre-osteoblastic cells. Primary mouse calvarial cells and BMSCs were treated with FGF2 as indicated, before mRNA isolation. (A–D) FGF2 increases Ank and inhibits Tnap mRNA expression in (A and C) primary calvarial cells and (B and D) BMSCs. RNA was isolated and used for quantitative real-time PCR. mRNA levels are presented as normalized to GAPDH. *p < 0.05 vs. no treatment. (E) The regulation of Ank and Tnap mRNA expression by FGF2 is not cell type specific. RNA was isolated and used for quantitative real-time PCR. mRNA was normalized to GAPDH and is reported as fold induction (Ank) or inhibition (Tnap) by FGF2.

Time- and dose-dependent induction of PC-1 gene transcription by FGF2

Because PC-1 gene expression seems to be tightly regulated in pre-osteoblast and osteoblast cells, we became interested in understanding the mechanism of this regulation. As an initial step toward understanding the mechanism by which FGF2 induces PC-1 expression in pre-osteoblasts, we examined the time course and dose responsiveness of this induction. FGF2 dose dependently increased PC-1 mRNA expression, with maximal induction seen between 25 and 50 ng/ml (Fig. 3A). Induction of PC-1 mRNA expression was rapid (seen within 3 h) and peaked 12 h after the addition of FGF2 (Fig. 3B).

FIG. 3.

Rapid and direct induction of PC-1 mRNA by FGF2. MC3T3E1(C4) cells were treated with FGF2 as indicated. RNA was isolated and used for semiquantitative RT-PCR. PC-1 mRNA levels are presented as normalized to GAPDH. (A) Dose-dependent induction of PC-1 mRNA by FGF2. Cells were incubated with increasing doses of FGF2 for 12 h before RNA isolation. (B) Time-dependent induction of PC-1 mRNA by FGF2. Cells were incubated with FGF2 for increasing lengths of time, as indicated, before RNA isolation. (C) FGF2 induces transcription of PC-1. Cells were pretreated with 1 μM actinomycin D for 1 h before treatment with FGF2. (D) FGF2 induction of PC-1 does not require a protein intermediate. Cells were pretreated with 10 μM cycloheximide for 1 h before treatment with FGF2. *p < 0.05 vs. no treatment.

Translational and transcriptional inhibitor studies indicate that FGF2-dependent increases in PC-1 are the result of a direct change in PC-1 gene expression. Pretreatment of cells with actinomycin D completely abolished the increase in PC-1 mRNA by FGF2, indicating that increased levels of PC-1 result from increased transcription (Fig. 3C). In contrast, pretreatment of cells with cycloheximide did not inhibit PC-1 mRNA induction, indicating that FGF2 induces expression of PC-1 in the absence of a protein intermediate (Fig. 3D).

Analysis of the proximal PC-1 gene promoter

PC-1 expression increases with osteoblast differentiation,(10) and our results suggest that the induction of PC-1 by FGF2 is limited to osteoblast-related cells. Because Runx2 is an essential regulator of osteoblast differentiation,(26) we hypothesized that Runx2 regulates expression of this gene. To examine this hypothesis, we cloned a 2866-bp region of the proximal 5′-untranscribed region of the murine PC-1 gene and analyzed it for potential Runx2 transcription factor binding sites (Fig. 4A). Sequence analysis identified four consensus Runx2 sites: two within a distal region of the promoter (−2554 to −2549; −2500 to −2495) and two within a more proximal region (−810 to −805; −312 to −307). Of note, three of these four potential binding sites are conserved between mouse and human promoter sequences.

FIG. 4.

FGF2 specifically induces activity of a PC-1 gene promoter/firefly luciferase reporter construct in pre-osteoblastic cells. (A) PC-1 gene promoter/firefly luciferase construct. A 2.8-kb 5′-untranslated region of the PC-1 gene was cloned by PCR and ligated into the PGL3 firefly luciferase reporter vector. Four potential Runx2 binding sites are present within the promoter sequence: two within a distal region of the promoter and two within a more proximal region. Regions of high homology between mouse and human gene sequences are highlighted in gray. Also indicated are primer sets designed to amplify Runx2 binding site regions for ChIP analysis. (B) FGF2 specifically drives the PC-1 promoter in pre-osteoblastic cells. MC3T3(E1(C4) cells were transfected with PGL3/PC1 firefly luciferase and a control renilla luciferase construct, treated with FGF2 or BMP2, and analyzed for luciferase activity. *p < 0.05 vs. no treatment.

FGF2 regulation of PC-1 promoter activity

To more directly study the mechanism by which FGF2 induces expression of PC-1, the 2.8-kb PC-1 gene promoter fragment was subcloned into the PGL3 luciferase expression vector to produce PGL3/PC1. This construct was transfected into MC3T3E1(C4) cells and assayed for FGF2 responsiveness. As shown in Fig. 4B, FGF2 dramatically stimulated promoter activity. In contrast, this promoter was not affected by treatment with BMP2.

Runx2 mediation of FGF2-induced PC-1 expression

The presence of four potential Runx2 binding sites within the PC-1 gene promoter suggests that Runx2 may mediate transcription of this gene. To test this hypothesis, we examined PC-1 promoter activity in osteoblastic and non-osteoblastic cell lines in the presence or absence of added Runx2. Luciferase assay results indicate that Runx2 can induce PC-1 promoter activity in both osteoblastic and non-osteoblastic cells. Whereas FGF2 alone increased PC-1 promoter activity in Runx2-containing MC3T3E1(C4) cells, expression of exogenous Runx2 lead to a further increase in the FGF2 response (Fig. 5A). In contrast, FGF2 was unable to stimulate PC-1 promoter activity in COS7 cells, which do not express Runx2. Strikingly, introduction of Runx2 into these cells rendered them responsive to FGF2 (Fig. 5B).

FIG. 5.

Runx2 is required for the induction of PC-1 expression by FGF2. (A and B) Runx2 increases FGF2-induced PC-1 promoter activity in MC3T3E1(C4) (A) and COS7 (B) cells. Cells were transfected with PGL3/PC1 firefly luciferase, control renilla luciferase constructs, and expression vectors for LacZ or Runx2. Cells were treated with FGF2 and analyzed for luciferase activity. *p < 0.05 vs. no treatment or between indicated groups. (C and E) Runx2 enhances PC-1 mRNA expression in response to FGF2. C3H10T1/2 cells (C) or Runx2^−/− calvarial osteoblastic cells (E). Cells were transduced with adenoviral vectors expressing LacZ or Runx2 and treated with FGF2 as indicated. PC-1 and GAPDH mRNA levels were measured by real-time PCR, and PC-1 mRNA levels are presented as normalized to GAPDH. *p < 0.05 vs. no treatment or between indicated groups. (D and F) FGF2 treatment does not enhance Runx2 expression. C3H10T1/2 cells (D) or Runx2^−/− calvarial osteoblastic cells (F) were transduced with adenoviral vectors expressing LacZ or Runx2, and treated with FGF2. Runx2 and GAPDH mRNA levels were measured by real-time PCR, and Runx2 mRNA levels are presented as normalized to GAPDH. *p < 0.05 vs. LacZ. (G) FGF2 treatment does not increase Runx2 nuclear expression in calvarial pre-osteoblastic cells. MC3T3E1(C4) nuclear extracts were analyzed for Runx2 protein and control β-actin protein expression levels by immunoblot analysis. (H) Runx2 binding site mutations in the PC-1 promoter significantly diminish promoter responsiveness to FGF2. MC3T3E1(C4) cells were transfected with wildtype PGL3/PC1 or Runx2 binding site mutant PGL3/PC1 firefly luciferase, and a control renilla luciferase construct. Cells were treated with FGF2 and analyzed for luciferase activity. *p < 0.05 vs. wildtype PGL3/PC1.

Promoter activity assays do not always correlate well with corresponding changes in endogenous gene expression. For this reason, we examined the requirement for Runx2 in FGF2 induction of endogenous PC-1 mRNA. Studies used both MC3T3E1(C4) cells and C3H10T1/2 cells, a mesenchymal cell line that expresses low but detectable levels of Runx2. Although C310T1/2 cells transduced with control vector showed modest FGF2 induction of PC-1 mRNA, the FGF2 response was dramatically increased on transduction with Runx2 (Fig. 5C). Similar results were observed using the MC3T3E1(C4) cell line (data not shown). Of note, transduction with Runx2 did not enhance the ability of FGF2 to regulate Ank or Tnap expression in C310T1/2 cells (data not shown).

To further assess if Runx2 is necessary for induction of PC-1 gene expression by FGF2, we also examined PC-1 mRNA levels in a Runx2 null, immortalized calvarial osteoblastic cell line.(20) FGF2 did not induce PC-1 mRNA expression in these cells. Significantly, strong FGF2 induction of PC-1 was seen after transduction of these cells with Runx2 adenovirus (Fig. 5E).

FGF2 effect on Runx2 expression or nuclear localization

It has been previously suggested that FGF receptor signaling can increase Runx2 expression in pre-osteoblastic cells.(27) To determine whether FGF2 stimulates PC-1 gene expression by increasing Runx2, we next examined Runx2 mRNA in C310T1/2 cells and Runx2^−/− immortalized calvarial osteoblastic cells that had been treated with FGF2. Whereas transduction with Runx2 adenovirus dramatically upregulated Runx2 expression in both of these cell types, FGF2 treatment had minimal effect (Figs. 5D and 5F).

It is also possible that FGF2 regulates PC-1 expression by affecting Runx2 nuclear levels. To determine whether FGF2 induces PC-1 by changing Runx2 nuclear localization, protein levels in FGF2-treated MC3T3E1(C4) cell nuclear fractions were examined. Western blot results show that FGF2 does not increase nuclear Runx2 (Fig. 5G).

Mutation of Runx2 binding sites and effect on PC-1 promoter activity

To confirm the involvement of Runx2 in the induction of PC-1 gene expression by FGF2, each of the four putative Runx2 binding sites within the PC-1 gene promoter were mutated to eliminate the potential for Runx2/PC-1 promoter binding. Results from luciferase assays showed that mutation of any one of the four potential Runx2 binding sites lead to a significant drop in FGF2-dependent induction of PC-1 promoter activity (Fig. 5H). This indicates that each of the four putative Runx2 binding sites contributes to maximal promoter activity downstream of FGF2. However, mutation of combinations of two, three, or four Runx2 binding sites did not lead to further attenuation of FGF2 responsiveness (Fig. 5H). This may indicate that the proximal PC-1 gene promoter contains additional Runx2 binding sites or that other factors in addition to Runx2 mediate the FGF2 response.

Runx2 association with PC-1 gene promoter

Both EMSAs and ChIP assays were performed to determine whether Runx2 associates with the putative binding sites found within the PC-1 gene promoter. EMSAs were undertaken to establish that Runx2 present in nuclear extracts of pre-osteoblastic cells can bind to each of the four potential Runx2 binding sites. Incubation of each of the double-stranded DNA oligos (containing one of the four putative Runx2 binding sites) with nuclear extract purified from MC3T3E1(C4) cells resulted in a significant shift in migration of the oligonucleotide. Competition with excessive amounts of nonradioactive oligonucleotide successfully blocked the shift in oligonucleotide migration. In contrast, oligonucleotides containing point mutations in core Runx2 binding sites were unable to compete with bound species (Fig. 6A). Supershift assays using specific anti-Runx2 antibody confirmed that the shifted complex contained Runx2 (Fig. 6B).

FIG. 6.

Runx2 associates with the PC-1 gene promoter. (A) Nuclear extract specifically binds to Runx2 binding sites within PC-1 promoter oligonucleotides. MC3T3E1(C4) cell nuclear extract was incubated with radiolabeled Runx2 binding site oligonucleotides or mutant Runx2 binding site oligonucleotides (Table 1), and increasing amounts of nonradiolabeled competitor wildtype or competitor mutant oligonucleotide. Samples were analyzed by PAGE. (B) Runx2 from nuclear extract binds to Runx2 binding sites within PC-1 promoter oligonucleotides. Radiolabeled oligonucleotides were incubated with MC3T3E1(C4) nuclear extract and Runx2 or control IgG antibody. Samples were analyzed by polyacrylamide gel electrophoresis. (C) FGF2 promotes Runx2 recruitment to the endogenous PC-1 gene promoter. Cross-linked chromatin isolated from untreated or FGF2 treated MC3T3E1(C4) cells was immunoprecipitated with Runx2 or control IgG antibody. PCR was performed using primers generated to detect Runx2 binding site regions of the distal or proximal PC-1 gene promoter (Figure 3; Table 1). Arrow indicates PC-1 promoter-specific amplification product. (D) ChIP PCR amplification is specific to the PC-1 gene promoter. PCR of cross-linked chromatin was performed using primers specific to exon 18 of the PC-1 gene.

Gel mobility shift assays are useful for showing that a nuclear factor can bind a putative DNA binding site in vitro but are unable to assess the physiological significance of this interaction. Consequently, ChIP assays were carried out to confirm that Runx2 binds to the endogenous PC-1 gene promoter in intact cells. Two separate primer sets were designed to amplify regions of the PC-1 gene promoter containing putative Runx2 binding sites. PCR amplification of the proximal promoter region will detect Runx2 binding at sites 1 (−312) or 2 (−810). PCR amplification of the distal promoter region will detect Runx2 binding at sites 3 (−2500) or 4 (−2554) (Fig. 4A). Using ChIP with anti-Runx2 antibody, positive signals were obtained using both PCR primer pairs, confirming that Runx2 associates with both of these chromatin regions (Fig. 6C). Significantly, ChIP with chromatin from FGF2-treated cells resulted in stronger positive signals, indicating that FGF2 promotes the recruitment of Runx2 to the PC-1 promoter in pre-osteoblastic cells. Anti-Runx2 antibody ChIP did not yield a positive signal using PCR primer pairs for exon 18 of the PC-1 gene, indicating specificity of amplification in the previous reactions (Fig. 6D).

DISCUSSION

We previously showed that FGF2 upregulates expression of the pyrophosphate generating enzyme, PC-1; the pyrophosphate membrane channel, Ank; and the pyrophosphate hydrolyzing enzyme, Tnap in the MC3T3E1(C4) calvarial pre-osteoblast cell line.(19) Here we showed that FGF2 also stimulates expression of PC-1 and Ank while inhibiting expression of Tnap in primary calvarial pre-osteoblasts and BMSCs. These results support the idea that pyrophosphate generation, transport, and hydrolysis downstream of FGF signaling is physiologically relevant and likely to play a significant role in mediating the effects of FGFs on bone. Of note, it seems that, unlike MC3T3E1(C4) cells, primary osteoblastic cells express basal levels of PC-1. This may be explained by the fact that primary cultures contain a mixed pool of cells and are, therefore, expected to contain both undifferentiated pre-osteoblast cells and a number of more mature osteoblastic cells. PC-1 expression is known to increase with osteoblast differentiation in culture, and basal PC-1 expression has been previously noted in primary osteoblastic cells isolated from both calvarial and long bones.(14)

Whereas other cell types (fibroblastic, pre-myeloblastic and pre-adipocytic cells) express basal levels of PC-1, FGF2 only induces PC-1 expression in osteoblast-related cells. We previously found that FGF2-induced PC-1 expression in undifferentiated but not differentiated MC3T3E1(C4) cells.(19) Together, these results indicate that the induction of PC-1 expression by FGF2 is specific to pre-osteoblasts. Alternatively, we find that expression of Ank and Tnap are regulated by FGF2 in multiple cell types. These latter results are in accordance with our previous findings that FGF2 regulated Ank and Tnap expression in both undifferentiated and differentiated MC3T3E1(C4) cells.(19)

Because FGF2 seems to induce PC-1 gene expression in both a cell type– and differentiation stage–dependent manner, we became interested in understanding the mechanism of this regulation. To begin exploring how PC-1 is regulated by FGF2, a series of experiments were conducted to determine whether the FGF2 response is the result of a direct increase in transcription of the PC-1 gene. Our results indicate that effects of FGF2 are immediate and direct. FGF2 actions on PC-1 are both dose and time dependent and do not require de novo protein synthesis.

To more directly study the regulation of PC-1 in osteoblast cells, we cloned a 2864-bp region of the proximal PC-1 gene promoter and constructed a PC-1 promoter/firefly luciferase reporter construct for promoter activity assays. Luciferase reporter assays using this construct established that FGF2 specifically induces PC-1 promoter activity in calvarial pre-osteoblast cells. Treatment of cells with FGF2 significantly upregulated PC-1 promoter activity, whereas BMP2 was without effect. This result is consistent with our previous demonstration that FGF2, but not BMP2 or BMP4, upregulated expression of PC-1 mRNA in MC3T3E1(C4) cells.(19) The consistency of these results indicates that the activity of this promoter construct reflects changes in endogenous PC-1 gene expression.

Initial sequence analysis of the 2.8-kb proximal PC-1 promoter identified four consensus Runx2 binding sites: two located within a distal promoter region and two located within a more proximal region. Of note, three of these four putative Runx2 binding sites are conserved between mouse and human PC-1 gene sequences. The transcription factor, Runx2, is a master regulator of osteoblastic differentiation.(26) Runx2 expression in mice is required for normal skeletal development and ossification.(28–30) Runx2 has also been shown to drive expression of many osteoblast marker genes including osteocalcin, bone sialoprotein, and alkaline phosphatase.(21,22,31) Discovery of multiple putative Runx2 binding sites within the PC-1 gene promoter, combined with the knowledge that PC-1 expression increases with osteoblast differentiation and that the induction of PC-1 by FGF2 is limited to osteoblast-related cells, led us to hypothesize that Runx2 mediates transcription of this gene. Luciferase assays using the PC-1 gene promoter/firefly luciferase reporter construct indicate that Runx2 is needed for FGF2 responsiveness. FGF2-induced PC-1 promoter activity in Runx2-positive cells like MC3T3E1(C4) and, to a lesser extent, C3H10T1/2. In contrast, cells lacking Runx2 such as COS7 or calvarial cells from Runx2-null mice only exhibited FGF2 induction of PC-1 after addition of exogenous Runx2. These results are all consistent with the concept that Runx2 mediates PC-1 expression downstream of FGF2.

Runx2 is known to regulate gene transcription by binding to promoter sequences in target genes. To establish that Runx2 mediates PC-1 gene expression by binding to the PC-1 promoter, we mutated each of the four putative Runx2 binding sites found within the proximal 2.8-kb fragment of this promoter. Mutation of each site resulted in a significantly drop in FGF2-dependent PC-1 promoter activity. This indicates that each site contributes to maximal FGF2 responsiveness. However, mutation of all four binding sites did not result in further attenuation of the FGF2 response. This may indicate that other Runx2 binding sites exist within this promoter or that factors in addition to Runx2 are involved in the regulation of PC-1 expression by FGF2. Significantly, it has been noted that the proximal bone sialoprotein (Bsp) gene promoter contains two cryptic, nonconsensus Runx2 binding sites, which, in combination with an adjacent homeodomain binding site, account for most of the osteoblast-specific activity of this gene.(22) Re-examination of the PC-1 gene promoter shows a minimum of five additional cryptic Runx2 sites. Future work will be required to determine the significance of these sites in PC-1 transcription.

It has been previously reported that signaling through FGF receptors can increase Runx2 expression in pre-osteoblastic cells.(27) We find that FGF2 induces PC-1 promoter activity and gene expression in pre-osteoblastic cells and that this effect is mediated in part by Runx2. However, neither Runx2 mRNA levels nor nuclear protein localization were increased by FGF2 in our experiments. This suggests that FGF2 induces PC-1 expression by altering the activation state of Runx2. It has been proposed that protein kinase C (PKC) and MAPK signaling can increase Runx2 transcriptional activity through serine phosphorylation.(32–35) We have previously shown that specific inhibitors of MAPK and PKC, but not Phospholipase C gamma (PLCγ), block the ability of FGF2 to induce PC-1 mRNA expression in pre-osteoblastic cells.(19) Future studies will be needed to determine the role of FGF2-mediated Runx2 phosphorylation in mediating transcription of this gene.

Using EMSAs, we subsequently confirmed that Runx2 specifically binds to each of the four potential binding sites found within the PC-1 promoter in vitro. ChIP assays confirmed that Runx2 binds to the endogenous PC-1 gene in untreated cells and that signaling through FGF receptors leads to increased recruitment of Runx2 to these promoter sites. PCR amplification using primers spanning either upstream or downstream Runx2 binding sites yielded a positive ChIP signal, and this signal was stronger when ChIP was performed with chromatin isolated from FGF2-treated cells. These results confirm that Runx2 binds to the endogenous PC-1 promoter in intact pre-osteoblastic cells and that Runx2 is recruited to the PC-1 gene promoter on FGF2 treatment. It is worthy to note here that we show Runx2 binding to the PC-1 promoter in cells that do not express PC-1 [untreated MC3T3E1(C4) cells]. This result is consistent with the proposal by Young et al.(36) that Runx2 remains associated with target gene promoters during periods of gene silence and that Runx2 likely acts as a scaffolding protein to recruit co-factors to promoter sites for control of gene expression.

Significantly, the finding that FGF2 induces PC-1 expression in pre-osteoblast cells but not in differentiated osteoblasts suggests that the primary function of PC-1 in pre-osteoblasts is distinct from that in differentiated cells. Of note, PC-1 is highly expressed in precursor cells of developing molars several days before the onset of tooth mineralization in a Runx2-dependent manner.(37) This result also supports the idea that PC-1 function in precursor cells is not the same as that in mineralizing cells. As previously discussed, the primary function of PC-1 in differentiated osteoblasts is believed to be the generation of extracellular pyrophosphate for hydrolysis into phosphate by alkaline phosphatase, which is also expressed in differentiated osteoblasts. The primary function of pre-osteoblastic PC-1 could be the generation of extracellular pyrophosphate to temporarily inhibit bone mineralization, perhaps allowing for continued proliferation or the early differentiation of precursor cells before matrix mineralization. However, PC-1 is known to have pyrophosphatase and phosphodiesterase activities, giving it the potential to influence purinergic receptor signaling. Purinergic signaling can modulate gene expression and the osteoblastic cell phenotype.(38) The influence of PC-1 on purinergic signaling in osteoblastic precursor cells has yet to be studied but may explain some or all of the function of PC-1 in this cell type. Future studies will be needed to establish the significance of pre-osteoblastic PC-1 expression in the overall effects of FGF signaling on bone mineralization.