Runx2 Regulates G Protein-coupled Signaling Pathways to Control Growth of Osteoblast Progenitors*

Runt-related transcription factor 2 (Runx2) controls lineage commitment, proliferation, and anabolic functions of osteoblasts as the subnuclear effector of multiple signaling axes (e.g. transforming growth factor-β/BMP-SMAD, SRC/YES-YAP, and GROUCHO/TLE). Runx2 levels oscillate during the osteoblast cell cycle with maximal levels in G1. Here we examined what functions and target genes of Runx2 control osteoblast growth. Forced expression of wild type Runx2 suppresses growth of Runx2-/- osteoprogenitors. Point mutants defective for binding to WW domain or SMAD proteins or the nuclear matrix retain this growth regulatory ability. Hence, key signaling pathways are dispensable for growth control by Runx2. However, mutants defective for DNA binding or C-terminal gene repression/activation functions do not block proliferation. Target gene analysis by Affymetrix expression profiling shows that the C terminus of Runx2 regulates genes involved in G protein-coupled receptor signaling (e.g. Rgs2, Rgs4, Rgs5, Rgs16, Gpr23, Gpr30, Gpr54, Gpr64, and Gna13). We further examined the function of two genes linked to cAMP signaling as follows: Gpr30 that is stimulated and Rgs2 that is down-regulated by Runx2. RNA interference of Gpr30 and forced expression of Rgs2 in each case inhibit osteoblast proliferation. Notwithstanding its growth-suppressive potential, our results surprisingly indicate that Runx2 may sensitize cAMP-related G protein-coupled receptor signaling by activating Gpr30 and repressing Rgs2 gene expression in osteoblasts to increase responsiveness to mitogenic signals.

Stringent regulation of mesenchymal stem cell renewal and the generation and differentiation of the appropriate number of committed osteoprogenitors are critical for development, remodeling, and repair of bone tissue. Runx2 (Runt-related transcription factor 2) is a critical regulator of bone development by driving osteoblast differentiation and formation of a bone-specific mineralized extracellular matrix (1)(2)(3)(4). Runx2 regulates osteoblast growth (5,6) in part through epigenetic mechanisms (7)(8)(9) and the ability of Runx2 to promote a nonproliferative state (e.g. senescence or quiescence) (6, 10 -12). The first evidence that Runx2 affects bone cell growth was the observation that primary calvarial cells from Runx2 null mice proliferate faster than corresponding wild type osteoblasts, and reintroduction of Runx2 into Runx2 null cells restores normal cell growth (5). Several studies indicate that Runx2 control of proliferation is cell type-specific. Runx2 inhibits proliferation of osteoprogenitors and committed osteoblasts (5,6), but it may have distinct biological roles in chondrocytes (2,6,13) and endothelial cells (14,15), as well as osteosarcoma cells (6,12), T cell lymphomas (16,17), and breast cancer (18). In non-osseous cells (e.g. T cells), Runx2 has anti-proliferative potential (19) unless c-Myc levels are elevated (17).
The growth regulatory activity of Runx2 is controlled by cell cycle-dependent modulations in its protein levels. Runx2 expression is modulated during the cell cycle at both transcriptional and post-transcriptional levels and maximal in early to mid-G 1 (6,20,21). In addition, Runx2 is phosphorylated during the cell cycle by distinct cyclin-Cdk complexes to regulate its activity or stability by a proteasome-dependent mechanism (6,15,20,22,23). Runx2 expression and activity in osteoblasts are regulated by a large number of growth regulatory factors and external signals, including fibroblast growth factor 2 (24 -26), TGF␤ 5 /BMP2 (27)(28)(29)(30)(31), and PTH (32). Importantly, retention of Runx2 protein on mitotic chromosomes provides an epigenetic memory of its developmental position in the osteogenic lineage (7,8) and poises Runx2 to control gene expression in early G 1 (21).
Runx2 is a bifunctional transcription factor that mediates the integration of multiple signaling pathways at nuclear matrixassociated subnuclear foci through interactions with a large * This work was supported, in whole or in part, by National Institutes of Health cohort of cofactors (Ͼ24 partner proteins) that support activation or repression of Runx2 target genes in osteoblasts (reviewed in Ref. 1 and see Refs. [33][34][35][36]. For example, Runx2 interacts with GROUCHO/TLE proteins that require the conserved C-terminal VWRPY pentapeptide of Runx2 for gene repression (37). Runx2 transduces TGF␤/BMP2 signaling through direct interactions with SMAD proteins that bind to a composite protein domain ("NMTS/SMID") that mediates transcriptional activation and subnuclear targeting (31,38). Furthermore, Runx2 interacts with Yes-associated protein (9), a WW domain-containing protein that transduces signals from the SRC/YES family of tyrosine kinases, as well as a growing number of other WW domain-containing proteins (e.g. TAZ, SMURF, and SCHNURRI) that all recognize a PPYP motif linked to the NMTS (39 -41). Despite the growing number of Runx2 partner proteins that modify or modulate its transcriptional activity, there is limited understanding of the signaling pathways and Runx2 cofactors that control the growth regulatory potential of Runx2 and the downstream expression programs that mediate its cellular functions.
In this study, we have examined how Runx2 regulates growth of osteoblasts using complementation assays with Runx2 null calvarial osteoprogenitors in which we introduced wild type or mutant Runx2 proteins. Our studies revealed that DNA binding and C-terminal transcriptional functions of Runx2 are essential for cell growth control. Furthermore, our data suggest that Runx2 may regulate G protein-signaling pathways to modify how osteoblasts respond to mitogenic stimuli.

EXPERIMENTAL PROCEDURES
Cell Culture-Immortalized Runx2 null cells from mouse calvaria were generated in our laboratory from the fetal calvarial region of Runx2 knock-out mice by stable integration of mTERT as described previously (30). Runx2 null cells were maintained in ␣MEM supplemented with 10% fetal bovine serum (FBS) (Atlanta Biologicals, Lawrenceville, GA), 30 mM penicillin/streptomycin, and 100 mM L-glutamine at 37°C and 5% CO 2 -humidified atmosphere. Mouse MC3T3 osteoblasts (6) and primary cultures of calvarial cells from wild type and Runx2 null mice were maintained as described previously (5,10). Primary cells were expanded in subconfluent culture for three passages before collection and RNA extraction.
Cells were plated for infections in 6-well plates (12.5 ϫ 10 4 cells/well). After 24 h, cells were infected with 100 multiplicities of infection of each virus in 600 l of ␣MEM complemented with 1% FBS for 4 h. Upon addition of 400 l of media contain-ing 1% FBS, cells were incubated for an additional 10 h. Cell numbers were determined at daily intervals during the next 4 days after infection to determine growth curves and to perform expression analyses. Infection efficiencies were assessed by expression of a green fluorescent protein (GFP) under the control of an IRES signal, and images of GFP-expressing cells were taken using a Carl Zeiss fluorescence microscope with a SPOT camera (Diagnostic Instruments, Inc., Sterling Heights, MI).
Quantitative Real Time Reverse Transcriptase-PCR (qRT-PCR) Analysis-Total RNA for qRT-PCR assays was isolated using TRIzol reagent (Invitrogen), subjected to DNase I digestion, and purified using an RNA purification kit (Zymogen, Orange, CA). Aliquots of RNA (1 g) were used for reverse transcription (first strand cDNA synthesis kit, Invitrogen) with random hexamer primers. Quantitative PCR was performed with commercially available reagents (Power SYBR Green PCR Master Mix, Applied Biosystems, Foster City, CA) using an automated system (Applied Biosystems 7300 Real Time PCR denaturation occurred at 95°C for 10 min followed by 40 cycles of two-step PCR (95°C for 15 s denaturation and 60°C for 1 min synthesis).
Affymetrix Analysis-Total RNA for Affymetrix analysis (and subsequent qRT-PCR validation) was isolated with TRIzol reagent and purified using the RNeasy mini kit (Qiagen). Approximately 1-5 g of total RNA was used for two strand cDNA synthesis using the Affymetrix cDNA synthesis kit using oligo(dT) primers (Affymetrix, Santa Clara, CA). Following cRNA synthesis, samples were labeled with biotin using the BioArray HighYield RNA transcript labeling kit (Enzo Lifesciences, Farmingdale, NY) and purified with Affymetrix Cleanup Module for cRNA (Affymetrix). Aliquots of the full-length cRNA products (ϳ20 g) were fragmented using metal-induced hydrolysis at a concentration of 1 g/l. The quality of cRNA fragments (35-200 bases) was evaluated in 1% formaldehyde/ MOPS gel. Aliquots of fragmented cRNA (12 g) were used for hybridization with Affymetrix cDNA microarrays ("chips") (Mouse Genome 430 2.0 Array) in hybridization mixture for 16 h at 42°C in a hybridization oven. After hybridization, chips were washed on an Affymetrix Fluidics Station and stained with solutions containing streptavidin/phycoerythrin solution and goat IgG antibody to amplify the signals of the transcripts. The resulting signals were scanned using an Affymetrix scanner, and numerical files (CEL) were generated from the resulting images. Data processing and sample comparisons were performed using an open source library for statistical analysis (Bio-Conductor library for R environment). In brief, robust multiarray average expression measure as part of the Affy software package was applied to calculate the average value of signals on the arrays (43). Following robust multiarray average background correction, array values were subjected to quantile normalization assuming identical signal distributions in each of the arrays. Statistically significant differences between probe sets were evaluated using Student's t test (p Ͻ 0.05). Functional annotation of Affymetrix probe sets and gene ontology relationships between groups of co-regulated genes were assessed using the data base for Annotation, Visualization and Integrated Discovery (DAVID 2.0) (44).
Western Blot Analysis-Cell lysates were prepared from cell pellets that were boiled in 100 l of Direct Lysis buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 12% urea, 25 M MG132, 100 mM dithiothreitol, and 1ϫ Complete protease inhibitors) (Roche Applied Science). Aliquots of each lysate (5 l) were separated in 10% SDS-PAGE. Proteins were transferred using semi-wet blotting to nitrocellulose membranes (Millipore, Billerica, MA). Phosphate-buffered saline (PBS: 20 mM phosphate, 150 M NaCl, pH 7.4) with 5% nonfat milk was used for 1 h at room temperature to block nonspecific protein binding. Primary and secondary antibodies were used at 1:2,000 dilutions for 1 h room temperature in PBS, 0.1% Tween (PBST) with 1% milk. Signal was detected with ECL (Western Lighting Chemiluminescence Reagent Plus, PerkinElmer Life Sciences). Runx2-specific mouse monoclonal antibodies were a generous gift of Dr. Yoshiaki Ito (Institute for Molecular and Cellular Biology, Singapore). CDK2 rabbit polyclonal antibody (SC-163) was purchased from a commercial vendor (Santa Cruz Biotechnology, Santa Cruz, CA).
In Situ Immunofluorescence-Immunofluorescence microscopy was performed according to standard procedures (9,37). For detection of Gpr30 and Rgs2 protein in adenovirus-infected Runx2 null cells, we applied the following procedure. Runx2 null cells were plated on gelatin-coated coverslips at 0.6 ϫ 10 6 cells per 6-well plate. Cells were infected after 24 h with adenovirus vectors expressing wild type Runx2, the ⌬361 mutant, or GFP alone. Cells were fixed with 3.8% formaldehyde in PBS for 10 min and permeabilized with 0.25% Triton X-100 in PBS for 20 min. Nonspecific binding was blocked for 20 min with PBSA solution (PBS containing 0.5% bovine serum albumin). Cells were incubated with the following primary antibodies at a 1:250 dilution in PBSA at 4°C overnight: rabbit polyclonal specific for Gpr30 (Abcam, Ab-12564) and goat polyclonal specific for Rgs2 (Santa Cruz Biotechnology, C18/sc7678).
For siRNA transfections, cells were plated at a density of ϳ0.6 ϫ 10 5 cells/well in 6-well plates and transfected in duplicate after 24 h with 50 nM of each siRNA in Opti-MEM using Oligofectamine reagent according to the recommendations of the supplier (Invitrogen). Concentrated complete ␣-MEM (2.5ϫ) was added 4 h after transfection to obtain final concentrations of 30 mM penicillin/streptomycin, 100 mM L-glutamine, and 10% FBS. Cells were re-plated 24 h after siRNA transfection to prevent contact inhibition. Samples were collected following trypsinization and either counted or fixed for flow cytometry 36 h after re-plating. An expression vector for Rgs2 (pcDNA3.1-Rgs2) was purchased from University of Missouri Rolla cDNA Resource Center, Rolla, MO. A kanamycin-resistant vector expressing enhanced green fluorescent protein (EGFP) we have used previously (45) was transfected in parallel to evaluate the percentage of transfected cells. For transient DNA transfections, cells were plated in 6-well plates (ϳ0.6 ϫ 10 5 cells/well; same as above for siRNA transfections) and transfected with Lipofectamine (Invitrogen) for 16 h with 0.5 g of Rgs2 or EGFP-expressing control vector according to protocols from the supplier.
Flow Cytometry-Cells in which the levels of proteins of interest were modulated were examined by fluorescence-activated cell sorting (FACS) to assess the distribution of cells in different cell cycle phases. We performed FACS analysis by trypsinizing cells that were washed with PBS and fixed in 85% ethanol at 4°C for 24 h. Cells were stained for DNA content with 50 g/ml propidium iodide in PBS plus 1.8 mM MgCl 2 and 50 g/ml RNase A solution by incubation at 37°C for 20 min in the absence of light. After the sorting FACS data were analyzed with the Synchronization Wizard module of the ModFit LT 3.0 program (Verity Software House, Topsham, ME).

Runx2 Growth Regulation of
Osteoblast Progenitors Requires DNA Binding, as Well as C-terminal Transactivation and Repression Domains-We have previously shown that Runx2 deficiency increases the proliferative potential of osteoprogenitors and that re-introduction of wild type Runx2 into primary calvaria cells restores stringent cell growth control (5). To characterize the mechanisms by which Runx2 controls cell growth, we used deletion and point mutants of Runx2 to assess which functions of Runx2 are required for growth regulation. Wild type and mutant Runx2 proteins (Fig. 1A) were expressed from adenoviral vectors that were used to infect immortalized calvarial osteoprogenitors from mice with a homozygous Runx2 null mutation. Cells were maintained under standard culture conditions without osteogenesis-promoting agents. We achieved Ͼ95% infection efficiency for all adenoviral vectors based on fluorescence detection of co-expressed GFP proteins (data not shown), and our set of Runx2 proteins is expressed at comparable levels as assessed by Western blot analysis (Fig. 1B) and real time qRT-PCR (Fig. 1C). Furthermore, as expected, exogenous expression of wild type Runx2, but not a subset of Runx2 mutants, increases bone phenotypic markers (e.g. osteocalcin and osteopontin) (Fig. 1C). Re-introduction of wild type Runx2 into Runx2 null cells inhibits osteoblast proliferation as reflected by a prominent decrease in the slope of the growth curve when compared with control cells expressing GFP alone (Fig. 2, left), consistent with our previous observations (5,6). We note that the inhibitory effects of Runx2 are reversible and that osteoblasts resume proliferation after a delay when transiently expressed Runx2 levels decrease (6).
We tested three different point mutations of Runx2 affecting specific known signaling functions of the protein. We have shown that the C terminus of Runx2 is multifunctional and, for example, integrates BMP/TGF␤ signaling, supports nuclear matrix targeting, as well as mediates signaling by WW domain FIGURE 1. Runx2 reconstitution in osteoprogenitors from Runx2 null mouse calvaria. Cells were infected with adenoviral vectors expressing an IRES-driven GFP marker and either Runx2 wild type (WT) or different mutants defective for interactions with distinct regulatory cofactors. Virus expressing GFP alone was used as a control (GFP-EV). A, schematic representation of specific Runx2 mutations used in the context of functional protein domains of Runx2. Numbering is for the P1 isoform of Runx2 that starts with the amino acids MASNS and ends with the conserved residues VWRPY. Also indicated are a poly(Q/A) stretch, the runt homology domain (RHD) that controls DNA binding, C-terminal activation, and repression domains, as well as a nuclear matrix targeting signal. Micrographs were taken of transfected cells expressing an IRES-driven GFP at day 2 after infection to establish comparable infection efficiencies of each adenoviral vector (data not shown). B, Western blot analysis of exogenous Runx2 protein levels and different mutants shows comparable expression of Runx2 at day 1 after infection. The figure is a representative collage of comparable exposures of four distinct experiments with different sets of adenoviral vectors. Spliced lanes are indicated with dashed lines. C, expression of Runx2 mRNA levels in Runx2 null cells and induction of bone marker genes were monitored by qRT-PCR analysis. The mRNA levels for exogenous Runx2 were plotted relative to the average endogenous mRNA level in an equivalent amount of RNA from MC3T3 cells that was examined in parallel as external standard. RNA quantity was normalized relative to GAPDH as internal control. The mRNA levels of osteocalcin and osteopontin in control cells (no virus) are minimal, and values of 1 and below are close to background levels.
proteins (e.g. c-SRC/YES signaling and WOXX signaling) and GROUCHO/TLE-dependent transcriptional repression (1). Because we considered it plausible that each of these functions could be directly linked to control of bone cell proliferation, we focused on mutants with defects in the C terminus of Runx2 (Fig. 1A). One Runx2 mutant ("HTY") has a triple amino acid substitution (H426A/T427A/Y428A) that abrogates subnuclear targeting and blocks the interaction of BMP/TGF␤responsive SMAD proteins with Runx2 (31,38). The HTY protein blocks cell proliferation (Fig. 2, middle) indicating that the BMP/TGF␤-SMAD signaling pathway is dispensable for Runx2 growth regulation.
The second mutant ("Y433A") has a single tyrosine mutation that blocks binding of WW domain proteins, including the YES-associated protein (YAP), which transduces signaling by YES/C-SRC tyrosine kinase pathways (9). Expression of the mutant Y433A Runx2 protein also rescued the hyper-proliferative phenotype of Runx2 null cells (Fig. 2, middle) establishing that YES/C-SRC signaling is not required for growth inhibition. In contrast, the third point mutant ("R197Q") (7) abrogates the DNA binding activity of Runx2 and thus recognition of target gene promoters. Expression of the mutant R197Q protein did not inhibit cell proliferation, and the growth curve for this mutant was indistinguishable from those obtained for control cells infected with the empty vector (Fig. 2, middle). Taken together, these data show that DNA binding is essential for the growth regulatory function of Runx2.
To define a specific region of Runx2 that reduces cell growth potential, we tested several C-terminal deletion mutants. One mutant ("⌬361") lacks amino acids 362-528 that encompass distinct transcriptional activation and repression domains. Another deletion mutant lacks amino acids 433-528 ("⌬432") that retains trans-activation potential but does not have a C-terminal repression module. The results show that expres-sion of the ⌬361 mutant does not impede proliferation of Runx2 null cells, demonstrating that the removal of key transcriptional functions renders Runx2 ineffective (Fig.  2, right). Expression of the ⌬432 deletion mutant appears to suppress cell proliferation but only to a limited degree (Fig. 2, right). This intermediate growth inhibitory phenotype suggests that gene repression and trans-activation by Runx2 may each contribute to control of cell proliferation.
Identification of Runx2-responsive Genes That Contribute to Cell Growth Regulation-We used gene expression profiling to assess what genes respond to the re-introduction of Runx2 into Runx2 null calvarial progenitors during growth inhibition. As above, reconstitution experiments were performed by introducing wild type Runx2 or the ⌬361 deletion mutant that does not inhibit cell growth into Runx2 null cells using adenoviral vectors. Control cells were infected with a vector expressing GFP.
The advantage of our experimental design is that it permits delineation of the effects of Runx2 mutations in the absence of endogenous wild type Runx2 expression. Furthermore, it enables us to detect induction of gene expression by Runx2 in cells that have not yet biologically activated Runx2-responsive genes as a consequence of developing into maturing osteoblasts; the latter is not possible with alternative strategies involving RNA interference with Runx2 siRNA or forced expression of the ⌬361 deletion mutant as a competitive inhibitor of wild type Runx2 function.
RNA samples were prepared in triplicate after induction of exogenous Runx2 gene expression at a time when growth suppression is initiated (day 1) or overtly evident (day 2) (see Fig. 2, left). Samples from two independent experiments were subjected to gene expression profiling using Affymetrix mouse cDNA micro-arrays (Fig. 3). Validation of expression values was performed using qRT-PCR with the same samples and new replicates (see Fig. 4). Consistent with expression results for Runx2 and bone marker genes (see Fig. 1C), we observed that Affymetrix expression values for exogenous Runx2 mRNA are robust in cells producing the wild type and the ⌬361 deletion mutant but not in GFP-expressing cells (data not shown).
Because growth suppression requires the C terminus of Runx2, we focused our analysis on genes that are preferentially modulated (i.e. activated or repressed) by the wild type but not the mutant ⌬361 Runx2 protein. We calculated expression ratios for genes modulated by the wild type or ⌬361 mutant proteins. Genes were rank ordered based on a statistically significant (p Ͻ 0.05) fold increase or decrease in their expression. By comparing expression patterns observed for the wild type and ⌬361 Runx2 proteins, our data were filtered to eliminate  OCTOBER 10, 2008 • VOLUME 283 • NUMBER 41 genes that are not directly linked to cell growth control. A small set of 12 genes that is significantly modulated by Runx2 ⌬361 relative to cells expressing GFP from the empty vector control was excluded from the analysis.

Runx2 Control of G protein Signaling
Our analysis revealed strong up-regulation of several well known Runx2 target genes in osteoblasts including Bone Gla Protein 2/Osteocalcin (BGP2/OC), matrix metalloproteinase 13 (MMP13), and secreted phosphoprotein/osteopontin (SPP/ OP) (Fig. 3A). Comparison of genes modulated between wild type Runx2 and the ⌬361 mutant or wild type Runx2 and GFP (control vector) revealed very similar changes in the expression of specific genes. Interspersed among established target genes for Runx2, we observed several other genes that similarly exhibited pronounced changes in expression with the wild type but not the mutant ⌬361 protein. These genes include tissue inhibitor of matrix metalloproteinase-3 (TIMP-3), procollagen, type VI, ␣2 (Col6a2) and pan-hematopoietic expression (Phemx)/tetraspannin 32 (Tspan32). Interestingly, Phemx/Tspan32 is a known target gene for RUNX1/ AML1 in hematopoietic cells, whereas distinct members of the TIMP and collagen multigene families are Runx2 target genes in other cells and tissues (46 -49). Taken together, our genome-wide expression profiling experiments reveal that the C terminus is required for changes in the expression of known target genes.
To understand which cell functions and regulatory pathways are controlled by the C terminus of Runx2 during cell growth inhibition, we applied functional clustering of genes by gene ontology analysis (e.g. David 2.0). We selected 313 Affymetrix probe sets that are expressed at measurable basal levels (arbitrary signal strength above 25), are statistically significant (p Ͻ 0.05), and exhibit a 2-fold or greater change in expression between the wild type and the mutant ⌬361 protein. Of these, 194 probe sets genes are repressed by Runx2, and 119 probe sets exhibited activation. Functional annotation clustering revealed that these 313 probe sets represent at least 254 independent genes that can be grouped into multiple clusters with different biological functions (Fig. 3B). We detected modulations in the expression of several osteoblastrelated genes (e.g. osteocalcin, osteopontin, bone sialoprotein, alkaline phosphatase, collagens, matrix metalloproteinases, and proteoglycans), mRNAs for proteins involved in cell adhesion and cell to cell communication (e.g. cadherins and miscellaneous cell surface receptors), as well as a number of genes broadly related to growth factor/cytokine-mediated cell signal-FIGURE 3. Runx2 controls osteoblast proliferation by modulating the expression of several groups of related genes. Genes dependent on the growth regulatory C-terminal region of Runx2 were obtained by comparing Affymetrix cDNA expression microarrays. A, representative Runx2 target genes are shown that exhibit a consistent modulation on both day 1 and day 2 in cells expressing wild type Runx2 versus Runx2 ⌬361 mutant or GFP (i.e. control vector). The genes selected here have a greater than 2.4-fold increase or decrease in expression (p Ͻ 0.05) on either day 1 or 2 (except Gpr54, which was selected as a member of the Gpr family of genes). Comparison of Runx2 wild type (WT) versus Runx2 ⌬361 or wild type versus GFP yielded very similar changes in the expression of specific genes. Among the top genes identified in our list are classical bonespecific targets of Runx2 (e.g. osteopontin (Spp1), osteocalcin (Bglap2), and matrix metalloproteinase 13 (MMP13)). B, functional annotation clustering of 254 genes (Ͼ2-fold modulation with wild type Runx2 relative to Runx2 ⌬361 at Day 1; p Ͼ 0.05) was performed using David 2.0 (data base for annotation, visualization, and integrated discovery, available on line). The pie chart shows the classification of annotated genes into different biological and cellular categories. C, Runx2-dependent genes related to G protein-coupled receptors encode proteins that operate at different steps within a signaling cascade.
ing. Strikingly, of the 254 genes we identified, 33 genes (ϳ13%) represent ligands, receptors, G proteins, regulators, second messengers, or nuclear effectors of GPCR-signaling pathways (Fig. 3C). Because many G protein-signaling pathways are functionally linked to control of cell proliferation, it is plausible that Runx2 may suppress cell proliferation at least in part by modulating the sensitivity of selected G protein-coupled receptor signaling pathways.
Runx2 Regulates Components of G Protein-coupled Signaling in Osteoblast Progenitors-Among the targets of the full-length Runx2 protein, we identified four genes that encode regulators of G protein signaling (RGS) proteins (i.e. RGS-2, -4, -5, and -16), as well as five genes encoding G proteins (i.e. Gna13) or G protein-coupled receptors (i.e. Gpr23, Gpr30, Gpr54, and Gpr64). The four RGS genes represent members of the R4-subfamily of proteins, whereas the other five genes encode a more diverse set of proteins that are linked to signaling events involving cyclic AMP, steroid hormone biology, and/or phospholipids. This set of nine genes that we collectively refer to as the "GPCR gene panel" is of particular interest, because of extensive conceptual links between GPCR signaling and processes controlling cell proliferation (50). Furthermore, several genes in this panel were consistently and reproducibly detected (i.e. two or more times within the same microarray, in each of the Affymetrix microarray replicates and/or on two different days) and exhibited a statistically robust modulation (p Ͻ 0.05) in expression upon introduction of Runx2.
We used qRT-PCR analysis to validate changes in the expression levels of the GPCR panel of genes as detected by Affymetrix analysis (Fig.  4, A and B). The results show that qualitative changes in gene expression observed by Affymetrix analysis are also observed by qRT-PCR analysis. However, qRT-PCR analysis consistently equaled or exceeded values measured by Affymetrix analysis. Strikingly, the qPCR data show that Gpr30 and Gpr54 are strongly up-regulated at day 1 (up to 6 -10-fold), whereas Gna13 is only modestly up-regulated (ϳ1.5-fold) (Fig. 4, A and B). For comparison, the four Rgs genes (i.e. Rgs-2, -4, -5, and -16), as well as Gpr23 and Gpr64 are each down-regulated in the presence of Runx2 by 2-6-fold. Expression of this subset of six genes is modulated by Runx2 at both days we analyzed (day 1 and day 2). Thus, our data show selective changes in the expression of several G protein signaling-related genes that are either stimulated or inhibited by Runx2.
We validated changes in expression of two of these genes, Gpr30 and Rgs2, at the protein level by in situ immunofluorescence in Runx2 null cells. Gpr30 protein expression is below the level of detection in the absence of Runx2 or in the presence of Gene expression values measured by qRT-PCR were normalized using GAPDH as internal control. Statistical significance of the data was determined by Student's t test, and values with p Ͻ 0.05 are indicated by an asterisk, and values with p Ͻ 0.01 have two asterisks. C, Runx2 null cells were fixed and stained for Gpr30 and Rgs2 immunofluorescence upon infection with adenovirus expressing wild type Runx2, ⌬361 mutant, or GFP (control vector). Wild type Runx2 modestly reduces Rgs2 protein level in the nucleus, albeit that Rgs2 detection is evident at bright aggregates in the cytoplasm. DAPI, 4Ј,6-diamidino-2-phenylindole. D, forced expression of Runx2 upon adenoviral infection in MC3T3 cells modulates the endogenous mRNA levels of the indicated genes as determined by qRT-PCR at day 1 after infection. The mRNA levels of GPCR-related genes in cells expressing wild type Runx2 were plotted as fold change over GFP and no virus controls and normalized using GAPDH as internal control. Statistical significance of the data was determined by Student's t test, and values with p Ͻ 0.05 are indicated by one asterisk, and values with p Ͻ 0.01 have two asterisks. the ⌬361 mutant protein, but it is clearly detectable upon induction by wild type Runx2 (Fig. 4C), consistent with RNA expression data (Fig. 4B). Similarly, Rgs2 protein expression is detectable in the absence of Runx2, but decreased upon expression of wild type Runx2 (Fig. 4C).
To address whether these Runx2dependent genes are also modulated by Runx2 in committed preosteoblasts that endogenously express Runx2 at low levels, we examined the expression of our GPCR gene panel in MC3T3 cells infected with adenoviral vectors expressing Runx2 (and GFP) or the corresponding empty vector that expresses only GFP. Corroborating the data we obtained in Runx2 null cells (Fig. 4, A and B), expression of Gpr30 and Gpr54 is enhanced, and expression of Gpr23, Rgs2 and Rgs4 is down-regulated (Fig. 4D). Expression of Gpr64, Rgs5, and Rgs16 is below the level of detection in MC3T3 osteoblasts, indicating differences in the expression of these GPCR-related genes in osteogenic progenitors and committed osteoblasts (see below). The only gene that exhibited opposite changes in expression upon elevation of Runx2 is the gene for the G␣13 protein (Gna13); Gna13 expression is enhanced in Runx2 null cells but down-regulated in MC3T3 osteoblasts (Fig. 4D). Hence, although many Runx2-responsive genes are consistently up-or down-regulated, individual genes may be differently regulated by Runx2 depending on the cell type or stage of lineage commitment.

Selective Expression of Runx2-dependent Genes Encoding Factors Involved in G Protein Signaling during Different Stages of Osteoblast
Phenotype Maturation-Studies from our group and others suggest that Runx2 mediates lineage commitment of osteoprogenitors, as well as controls the proliferative expansion of pre-osteoblasts and subsequent maturation of quiescent pre-osteoblastic cells into mature osteoblasts. To understand the physiological role that genes involved in G protein signaling play in control of osteoblast growth by Runx2, we examined expression of the Runx2-dependent GPCR gene panel in two different cell culture models that reflect key stages of osteoblast lineage progression (Fig. 5).
We first compared expression profiles of our selected gene set in primary calvarial osteoprogenitors expressing normal FIGURE 5. Expression of G protein-coupled receptor signaling components at distinct biological stages of osteogenic differentiation. A, expression of Runx2-dependent genes related to G protein-coupled receptor signaling and other markers was examined by qRT-PCR in mesenchymal progenitor cells obtained from the calvarial regions of Runx2 wild type and Runx2 null (KO) mice at embryonic day 17.5 and maintained in subconfluent culture for three passages. Several housekeeping genes were used as internal controls to validate differences in mRNA levels (Gapdh, Mcox, rRNA, and Hprt). mRNA levels of all different genes were plotted as fold change relative to each other (Gpr64 expression in Runx2 KO cells was arbitrary set as 1). Error bars represent standard error between four different mRNA preparations, each derived from multiple mouse embryos. Statistical significance of the data was determined by Student's t test, and values with p Ͻ 0.05 are indicated by one asterisk, and values with p Ͻ 0.01 have two asterisks. B, expression of representative GPCRrelated genes, as well as Runx2 and bone phenotypic genes, was monitored by qRT-PCR analysis during growth of MC3T3 osteoblasts. MC3T3 cells were grown in proliferating media for 4 days until confluency. The levels for each mRNA were plotted relative to day 1 and normalized by 18 S rRNA as internal control. Error bars represent mean Ϯ S.E. between three independent plates.

Runx2 Control of G protein Signaling
wild type Runx2 that can be induced to differentiate into mature osteoblasts and primary Runx2 null cells that are phenotype-arrested and exhibit a more immature mesenchymal progenitor phenotype. To investigate which genes of the GPCR panel are expressed in either of these two stages of osteogenic differentiation, we determined their mRNA levels by qRT-PCR (Fig. 5A). We found that 7 of 9 genes analyzed are expressed at significantly lower levels in primary calvarial osteoblasts from wild type mice than in pre-osteoblastic progenitors that lack Runx2. For comparison, several osteoblast related genes (e.g. alkaline phosphatase (Alk Phos), Wnt7B, Serpinb2, and the Cdk inhibitor p21) are more prominently expressed in calvarial osteoblasts, whereas a series of housekeeping genes does not show appreciable changes in expression (e.g. glyceraldehyde-3phosphate dehydrogenase (Gapdh), 18 S rRNA, mitochondrial cytochrome oxidase (Mcox), and hypoxanthine-phosphoribosyltransferase (Hprt)) (Fig. 5A). Because Runx2 null calvarial cells are less mature than calvarial osteoblast, the results suggest that the GPCR genes are preferentially expressed at elevated levels at the early stages of osteogenic lineage commitment.
The expression of two other Runx2-responsive genes (Gpr30 and Gpr54) we identified in complementation assays with Runx2 null cells is not significantly different between primary calvarial osteoblasts from Runx2 wild type or Runx2 null mice. One plausible explanation of these data is that short term forced expression of Runx2 in Runx2 null cells will yield an acute response, whereas calvarial cells with a Runx2 null phenotype may have undergone long -term adaptation (during fetal development) perhaps by invoking compensatory mechanisms that bypass a dependence on regulation by Runx2.
We subsequently assessed how G protein signaling-related genes are controlled during the switch from active proliferation to density inhibition of osteoblast growth. To reduce contributions of differentiation-related processes, these experiments were performed in subconfluent cultures in nondifferentiating medium lacking exogenous agents that expedite osteoblast maturation (e.g. absence of ascorbic acid and ␤-glycerol phosphate). Based on our experience, MC3T3-E1 cells achieve maximum proliferation (monitored by FACS analysis) between days 2 and 3 after plating, with contact inhibition of cell growth occurring around day 4 (5, 6). Accordingly, there is a slight decline in histone H4 mRNA levels and a concomitant increase in expression of the Cdk inhibitors p21 and p27 at day 4 after plating (Fig. 5B).
Runx2 mRNA expression increases throughout the 4-day culturing period (Fig. 5B), consistent with previous findings (6). Under our experimental conditions, osteocalcin mRNA is 4-fold enhanced above a minimal level that is evident at the three earlier days. The results show that Rgs2 and Rgs4 are reciprocally regulated with Rgs4 diminishing and Rgs2 increasing as the proliferative potential of cells decreases by day 4 (Fig.  5B). Similar results were obtained during osteoblast differentiation using osteogenic conditions (data not shown). In contrast, expression of Gpr23, Gpr30, Gpr54, and Gna13 does not change appreciably but tends to increase by day 4 (Fig. 5B). Our data indicate that cessation of proliferation correlates with loss of Rgs4 and stimulation of Rgs2, whereas expression of the other GPCR-related genes we detected is sustained throughout the culturing period.
One reason for the absence of significant changes in the expression of Runx2-responsive genes during osteoblast growth inhibition is that these genes are most likely not solely regulated by Runx2. For example, although Runx2 represses Rgs2 in Runx2 null cells, it may attenuate the expression of this gene to counteract excessive activation by another factor during osteoblast growth and differentiation. Similarly, modest elevation of Gpr30 mRNA levels in our time course could be the result of an attenuating factor that is up-regulated in parallel with Runx2-dependent activation of Gpr30.
Runx2-activated Gene Gpr30 Is Required for Proliferation of MC3T3 Osteoblasts-To understand the biological consequences on osteoblast proliferation of Runx2 regulating specific classes of genes, we focused initially on Gpr30 that stimulates cyclic AMP signaling and is prominently expressed in proliferating osteoblasts. We used a loss-of-function approach and depleted Gpr30 by siRNA transfection in MC3T3 osteoblasts (Fig. 6). Treatment of cells with siRNA against Gpr30 but not with control nonsilencing RNA reduces Gpr30 expression and FIGURE 6. GPR30 regulates proliferation of MC3T3 osteoblasts. MC3T3 cells were transfected with siRNAs (50 nM each) for Gpr30, Gpr54, Rgs4, or nonsilencing (NS) siRNA control in semi-confluent cultures and replated in a 1:3 ratio 24 h after transfection. Following a 36-h culture period, the subconfluent cell population was trypsinized, and cells were counted and harvested for RNA analysis. A, efficacy of each siRNA to knock down target gene expression was estimated by qPCR analysis. Levels of mRNA were plotted relative to RNA samples from untransfected control cells. Error bars represent standard error between two different RNA samples. Statistical significance of the data was determined by Student's t test, and values with p Ͻ 0.05 are indicated by one asterisk, and values with p Ͻ 0.01 have two asterisks. The efficiency of siRNA transfection was monitored using rhodamine-labeled control siRNA signal. B, biological effects of distinct siRNAs on proliferation of MC3T3 osteoblasts were determined by cell counting at 60 h after initiating siRNA treatment. Error bars represent the range of cell counts from triplicate samples. Statistical significance of the data was determined by Student's t test, and values with p Ͻ 0.01 are indicated by an asterisk. C, growth curves were obtained by cell counting at daily intervals until 4 days after treatment with nonsilencing (NS) RNA or siRNAs against Rgs4 or Gpr30 (#A, Ambion; #B, Dharmacon). Day 0 is the day when siRNA-transfected cells were replated. Error bars reflect variation observed in at least three independent measurements. OCTOBER 10, 2008 • VOLUME 283 • NUMBER 41 inhibits cell proliferation (Fig. 6). Parallel experiments using specific siRNAs against Gpr54 and Rgs4, which are both coexpressed with Gpr30 in proliferating osteoblasts, show the expected reductions in gene expression but did not reveal siRNA-mediated changes in cell number (Fig. 6). Furthermore, combined treatment of Gpr30 and Gpr54 siRNA or any other combination of siRNAs for GPCR-related proteins does not magnify cell growth inhibition caused by reducing Gpr30 alone (data not shown). Taken together, our data indicate that although the majority of GPCR-related proteins are not rate-limiting under standard cell culture conditions, the Runx2-activated gene Gpr30 stimulates proliferation of MC3T3 osteoblasts.

Runx2 Control of G protein Signaling
Runx2-repressed Gene Rgs2 Attenuates Proliferation of Runx2 Null Osteoprogenitors and Lineagecommitted MC3T3 Osteoblasts-Unlike the Runx2-activated gene Gpr30, which is an agonist for cAMP signaling, Rgs2 attenuates this pathway. The function of Rgs2, which is repressed by Runx2, was examined by elevating its low levels in proliferating MC3T3 osteoblasts by forced expression (Fig. 7). The effect of up-regulating Rgs2 expression levels in MC3T3 osteoblasts on cell cycle progression was measured using immunofluorescence microscopy with Ki67 antibodies (Fig. 7). Ki67 is a microscopic marker that permits assessment of the cell cycle stage of individual cells, and the subnuclear localization of Ki67 is related to nucleolar reorganization. For example, the Ki67 pattern in osteoblasts changes from a micropunctate pattern that is evident in G 1 phase cells to a focal organization that becomes successively more prominent as cells progress through S and G 2 phases (Fig. 7A). In parallel, we transfected cells with expression vectors for RGS2 or GFP. Fluorescence visualization of RGS2 or GFP in cells reveals which cells have high levels of either protein (Fig. 7B). In transfected cells that exogenously express Rgs2, there are more G 1 cells and less S phase based on the cell cycle stage-specific staining patterns of Ki67 (Fig. 7) compared with cells that express GFP. Thus, these data indicate at the single-cell level that RGS2 inhibits cell growth of MC3T3 osteoblasts under basal (i.e. un-induced) conditions.
To corroborate the conclusion that Rgs2 is a key attenuator of osteoblast proliferation, we performed RNA interference experiments. Rgs2 expression is normally suppressed by Runx2 (see Fig. 4), and Rgs2 expression is increased in Runx2 null osteoprogenitors compared with wild type calvarial cells (see Fig. 5A). Treatment of Runx2 null cells with siRNA for Rgs2 counteracts this elevation in Rgs2 levels (Fig. 8). Rgs2 siRNA but not nonsilencing siRNA reduces mRNA levels by at least 2-fold (Fig. 8A). Cells were also co-treated with forskolin or phorbol myristate acetate (PMA), each of which stimulates cell cycle progression. Treatment with either agent promotes the inhibitory activity of Rgs2 (51), and thus the effect of Rgs2 siRNA is  expected to be more prominent. Flow cytometry results show that Rgs2 siRNA increases the percentage of S phase cells in all treatment groups, and more overtly in cells treated forskolin or PMA (Fig. 8, B and C). We conclude that the Runx2-dependent gene Rgs2 reduces the mitogenic response of Runx2 null osteoprogenitor cells. In addition, because Runx2 regulates Gpr30 and Rgs2, and because the activities of Gpr30, Rgs2 and Runx2 proteins are all linked to cAMP levels in osteoblasts (52)(53)(54), we propose that Runx2 may regulate cAMP-related signaling pathways to influence osteoblast proliferation (Fig. 9).

DISCUSSION
In this study, we have characterized mechanisms that account for the ability of Runx2 to control osteoblast proliferation using an integrated strategy in which we investigated, first, what functions of Runx2 are required for cell growth control and, second, what downstream targets mediate its growth regulatory function. Using complementation assays where wild type and mutant Runx2 proteins are reintroduced into Runx2 null calvarial cells, we established that neither TGF␤/BMP2-SMAD-Runx2 nor WW domain signaling (e.g. WWOX, SRC/YES-YAP) was required for growth regulatory effects in osteogenic cells. However, DNA binding and C-terminal transcriptional functions of Runx2 are essential for cell growth inhibition, emphasizing the importance of gene expression programs that respond to modulations in Runx2 levels.
We have advanced the idea that Runx2 may control commitment for cell proliferation or osteogenic lineage progression by regulating the integrated cellular response to hormonal and mitogenic stimuli through modulation of the levels of key regulatory proteins that participate in diverse signaling pathways. Because Runx2 attenuates osteoblast proliferation, this factor may directly activate the expression of cell cycle inhibitory proteins or repress mitogenic proteins. Affymetrix gene expression profiling upon expression of either wild type or mutant Runx2 proteins revealed more than 250 genes that are either repressed or activated by wild type Runx2 during cell growth suppression. One striking finding of these studies is that Runx2 coordinately controls multiple components of G protein-coupled receptor (GPCR) signaling in osteoblasts.
GPCRs are prominently expressed in bone and mediate responses to a broad range of physiological stimuli, including growth factors, neurotransmitters, cytokines/chemokines, and peptide hormones (e.g. PTH and PTH-related peptide), as well as steroid hormones (e.g. estrogen and glucocorticoids) and calcium ion signaling (32,51,(55)(56)(57)(58)(59). For example, our data show that Runx2 controls expression of several G protein-coupled receptors (e.g. Gpr30, Gpr54, Gpr23, and Gpr64), regulators of G protein signaling (e.g. Rgs2, Rgs4, Rgs5, and Rgs16), and G-␣ proteins (e.g. Gna13). GPCR signaling is functionally linked to control of cell proliferation, because many growth factors signal through small G proteins and many GPCR pathways signal through kinases (e.g. PKA, PKC, I3P, and mitogen-activated protein kinases) that actively regulate cell cycle progression. Our study shows that at least two of the nine Runx2 responsive genes we characterized (i.e. Gpr30 and Rgs2), which have previously been linked to PKA signaling, have functional roles in control of osteoblast proliferation.
Previous studies have shown that Gpr30, which is the nongenomic estrogen receptor (60), promotes or inhibits cell proliferation depending on the presence and type of ligand, the cell type, and the hormonal context (61)(62)(63)(64)(65). Expression of Gpr30 is evident in the growth plate in both hypotrophic and quiescent chondrocytes but not in the proliferating region (66), suggesting that its function in chondrocytes is metabolic rather than developmental. We show that Gpr30 is expressed in proliferating osteoblasts, although its level increases by at most 2-fold in post-proliferative osteoblasts. Also, Gpr30 regulates epidermal growth factor release and can both induce and inhibit activity of mitogen-activated protein kinases (62)(63)(64)(65)(66). Our findings establish that Gpr30 is a Runx2-responsive gene and is required for osteoblast proliferation.
The G protein regulator Rgs2 modulates signaling from the PTH/PTH-related peptide receptor in osteoblasts and chondrocytes, as well as permits switching between G␣ s and G␣ q signaling pathways that stimulate PKC and PKA activity. Previous results have indicated that the Rgs2 gene promoter is activated by Runx2 (67), but our data clearly indicate that expression of Rgs2, as well as other Rgs proteins, is coordinately down-regulated by Runx2 under our experimental conditions. Forced expression of Rgs2 impedes progression through the cell cycle in G 1 in uninduced osteoblast progenitors, whereas Rgs2 depletion using siRNA increases the proliferative response upon stimulation of PKC (with PMA) or PKA (with forskolin); both PMA and forskolin promote the inhibitory function of Rgs2 (11). Our results indicate that Rgs2, like GPR30, is also a Runx2dependent regulator of osteoblast growth.
The functions of Gpr30 and Rgs2 have been linked to cAMP stimulation of PKA. The resulting activation of cAMP-responsive element-binding protein (CREB) and subsequent CREB-dependent expression of A-and D-type cyclins promotes cell cycle progression through induction of Cdk activity (68 -70). We note that at least three CREB proteins (CREB1, CREB3, and CREB5) are also Runx2-responsive (see Fig. 3C), consistent with the concept that Runx2 may control a putative growth stimulatory Gpr30-Rgs2-cAMP-PKA-CREB-cyclin-Cdk circuit in osteoblasts.
Our results show that Runx2 stimulates the pro-mitogenic function of Gpr30 in osteoblasts by enhancing expression of the FIGURE 9. Model for the reciprocal regulatory relationship between osteoblast proliferation and Runx2 control of gene expression. Runx2 levels are modulated during the cell cycle and are maximal in G 1 . We propose that Runx2 controls responsiveness of osteoblasts to mitogenic cues by altering the levels of components for multiple G protein-coupled receptor signaling pathways. In one of the pathways that is depicted here, Runx2 activates expression of Gpr30 and represses Rgs2. The net effect of these modulations is sensitization of cAMP/PKA signaling.
Gpr30 gene. Runx2 inhibits the anti-proliferative activity of Rgs2 by repressing Rgs2 mRNA expression. The observation that Runx2 activates Gpr30 and suppresses Rgs2 to promote cell proliferation is somewhat surprising considering that these genes were identified in our screen designed to define Runx2dependent growth-suppressive pathways. Apparently, Runx2 is both pro-and anti-proliferative within the same cell type. Several other findings on Runx2 may illuminate this paradox. First, Runx2 has been characterized as a c-Myc cooperating oncogene in T cell lymphoma (16,71), and Runx2 expression is frequently elevated in osteosarcoma (6) 6 clearly suggesting that Runx2 may have pro-mitogenic functions apart from growth suppressive activities. Furthermore, our other studies suggest that Runx2 is up-regulated by mitogens in osteoblasts during early G 1 where it may have a positive role in cell proliferation (21). For example, osteoblasts express the Cdk inhibitor p21 (72); Runx2 represses p21 transcription (34), and decreased levels of p21 would promote cell cycle progression by increasing Cdk activity. Runx2 is only down-regulated in late G 1 when cells have committed to a new round of cell division (6,20). These previous findings indicate that the growth-suppressive function of Runx2 operates during and beyond late G 1 , and there may be an obligatory requirement to reduce its protein levels prior to the onset of S phase. Therefore, it is plausible that control of osteoblast growth by Runx2 is cell cycle stage-specific; Runx2 may be pro-mitogenic in early G 1 and anti-mitogenic in late G 1 .
This proposed bi-functional role of Runx2 may be analogous to that of c-Myc; c-Myc is pro-mitogenic as an early response factor to growth factor stimulation, but oncogenic activation of c-Myc can trigger the growth-suppressive p14/p19 ARF pathway. Similarly, the pro-mitogenic up-regulation of Gpr30 and down-regulation of Rgs2 by Runx2 in mesenchymal cells may be temporally distinct from the ability of Runx2 to control the p14/p19 ARF pathway to inhibit cell growth (10,11). Further studies on the biological bi-functionality of Runx2 during the cell cycle are warranted.
Upon our initial demonstration that Runx2 null osteoprogenitors proliferate robustly compared with wild type calvarial osteoblasts (5), and the subsequent demonstration that Runx2 levels oscillate during the cell cycle (6,20), a major focus of our research group has been the identification of cellular programs that are controlled by Runx2 in proliferating cells. Our recent studies have shown that Runx2 controls rRNA transcription thereby modulating protein synthesis (7). Furthermore, loss of Runx2 results in a bypass of senescence and is pre-tumorigenic by increasing the accumulation of DNA damage and formation of phospho-H2A.X foci (10,11), consistent with the idea that elevation of Runx2 is functionally linked to acquisition of a nonproliferative state in osteoblasts. The finding that Runx2 controls at least two genes (Gpr30 and Rgs2) that are linked to cAMP signaling, as well as a large group of other proteins related to G protein-coupled receptor signaling, indicates that Runx2 modulates the integration of biological information from multiple G protein signaling networks.
The results from this study suggest that Runx2 is a principal regulator of the biological response of osteoblasts to external cues in the bone micro-environment and that Runx2 both sensitizes and desensitizes mitogenic pathways that determine competency for cell cycle progression and osteoblastic lineage commitment. Our studies have several interesting clinical ramifications. First, because Runx2 expression and activity are regulated by estrogen (2), it is possible that the bone-anabolic effects of estrogen and related compounds may occur via a positive feed-forward pathway (estrogen-Runx2-Gpr30-cAMP) that promotes osteoblast proliferation. Furthermore, our results suggest that the beneficial bone-anabolic effects of intermittent PTH administration (51,59,73,74), which has been associated with modulations in Runx2 levels (75), may be biologically linked to Runx2 and cAMP-dependent control of osteoblast proliferation.