Glucocorticoids inhibit developmental stage-specific osteoblast cell cycle. Dissociation of cyclin A-cyclin-dependent kinase 2 from E2F4-p130 complexes.

Unique cell cycle control is instituted in confluent osteoblast cultures, driving growth to high density. The postconfluent dividing cells share features with cells that normally exit the cell cycle; p27(kip1) is increased, p21(waf1/cip1) is decreased, free E2F DNA binding activity is reduced, and E2F4 is primarily nuclear. E2F4-p130 becomes the predominant E2F-pocket complex formed on E2F sites, but, unlike the complex that typifies resting cells, cyclin A and CDK2 are also present. Administration of dexamethasone at this, but not earlier stages, results in reduction of cyclin A and CDK2 levels with a parallel decrease in the associated kinase activity, dissociation of cyclin A-CDK2 from the E2F4-p130 complexes, and inhibition of G(1)/S transition. The glucocorticoid-mediated cell cycle attenuation is also accompanied by, but not attributable to, increased p27(kip1) and decreased p21(waf1/cip1) levels. The attenuation of osteoblast growth to high density by dexamethasone is associated with severe impairment of mineralized extracellular matrix formation, unless treatment commences in cultures that have already grown to high density. Both the antimitotic and the antiphenotypic effects are reversible, and both are antagonized by RU486. Thus, glucocorticoids induce premature attenuation of the osteoblast cell cycle, possibly contributing to the osteoporosis induced by these drugs in vivo.

contributor to GC-induced bone loss (4), is poorly understood. It may involve both indirect mechanisms, such as gonadal insufficiency and impaired hematopoiesis (reviewed in Refs. 2, 3, and 5), and direct, cell-autonomous effects, such as (i) induction of apoptosis (6), (ii) inhibition of type I collagen (7)(8)(9) and alkaline phosphatase (10,11) gene expression, and (iii) decreased activity of growth factors acting in bone in an autocrine/paracrine fashion, in particular IGF-I (12)(13)(14). Inhibition of osteoblast proliferation was also suggested to contribute to GC-induced osteoporosis (15,16). Notably, the inhibitory effects of GC on osteoblasts occur at pharmacological concentrations and should not be confused with the positive effects observed at physiological concentrations (for a review, see Ref. 17).
GC are antimitogenic in several cell types. Remarkably, however, this effect is mediated via diverse mechanisms. For example, inhibition of lymphoid cell proliferation, which partly accounts for the anti-inflammatory property of GC, is mediated by a decrease in the levels of G 1 cyclin and cyclin-dependent kinases (CDKs), in particular cyclin D3 and CDK4 (18 -20), as well as c-Myc (21,22). By contrast, in both hepatoma and lung alveolar cells, GC-induced cell cycle arrest has been attributed to induction of the CDK inhibitor (CDI) p21 (23)(24)(25). GC-induced osteoblast cell cycle arrest has been recently addressed in two osteosarcoma cell lines (26). In Rb-and p53-deficient SAOS2 cells, GC-induced cell cycle arrest involves up-regulation of the CDIs p21 and p27, whereas in U2OS osteosarcoma cells this same phenotype is mediated by repression of CDK4, CDK6, cyclin D, c-Myc, and E2F-1, all of which are positive regulators of the cell cycle. Mechanisms by which GC inhibit nontransformed osteoblast cell cycle progression have not been studied in depth.
Abrogation of osteoblast cell cycle progression may contribute to impaired bone formation in two ways. First, cell proliferation is simply required to provide enough cells capable of new bone formation. Second, the cell cycle may have additional, more specific roles in phenotype development, similar to the clonal expansion necessary for adipocyte differentiation (27,28). Osteoblast differentiation in vitro follows a multistep program, microscopically typified by (i) a proliferation period, during which the cells form a monolayer; (ii) matrix maturation, morphologically characterized by cell condensation, and (iii) the formation of nodules, in which mineral is then deposited (29 -31). We hypothesized that the cell condensation, which accompanies matrix maturation, reflects a unique proliferation phase required for phenotype progression. Indeed, this report demonstrates that in confluent MC3T3-E1 osteoblastic cultures, a unique cell cycle control is instituted, which is mech-anistically different from that operative in preconfluent cultures and is sensitive to glucocorticoids. Furthermore, we provide evidence that GC-mediated inhibition of the osteoblast persistent cell cycle is tightly linked to the inhibitory effect of GC on terminal differentiation, i.e. formation of mineralized extracellular matrix.

MATERIALS AND METHODS
Cell Culture-Like other osteoblastic cell lines, MC3T3-E1 cells are phenotypically heterogeneous (32). We therefore isolated 10 MC3T3-E1 subclones and screened them for calcium deposition in the absence and presence of the synthetic GC dexamethasone (DEX) at 10 -1000 nM. Although mineralization in most of the subclones was strongly inhibited by 100 -1000 nM, we continued the study with one subclone, which reproducibly exhibited progressive extracellular calcium deposition starting on day 10 after plating. Stock plates were maintained in ␣-minimum essential medium supplemented with 10% fetal bovine serum and penicillin/streptomycin and split every 4 -7 days. To support differentiation, the medium was also supplemented with 50 g/ml ascorbic acid and 10 mM ␤-glycerophosphate as described (33).
Flow Cytometry-Cell cycle analysis was performed according to Darzynkiewicz et al. (34). Briefly, cells were lightly trypsinized, resuspended in Hanks' buffer, fixed in cold 70% ethanol, and stored at Ϫ20°C for up to 1 week. Cells were then washed with Hanks' buffer and suspended in 1 ml of Hanks' buffer containing 20 g/ml propidium iodide and 5 Kunitz units of DNase-free RNase A. The percentage of cells in G 1 , S, and G 2 was determined using an EPICS® Profile Analyzer.
Transfection and Luciferase Assays-For transient transfection assays, MC3T3-E1 cells were plated in six-well plates (80,000 cells/well), and the calcium phosphate co-precipitation method of Chen and Okayama (35) was employed with 5 g of Qiagen-purified plasmid DNA. The cells were harvested on the indicated days, always 24 h after medium change, and lysed in "reporter lysis buffer" (Promega). Luciferase activity was determined using a microtiter plate luminometer (MLX Dynex Technologies) and protein concentration determined using the Micro BCA protein assay reagent kit (Pierce). Reporter constructs contained the luciferase gene driven by the mouse mammary tumor virus (MMTV) promoter (a kind gift from Dr. Ron Evans, La Jolla, CA), the Id2 promoter (36), the p21 promoter (37), and an artificial promoter, G13, containing 13 binding sites for p53 (37).
Nuclear, Cytoplasmic, and Whole Cell Extracts-Whole cell lysates were prepared using 0.5% (v/v) Nonidet P-40 buffer containing 50 mM Tris-HCl (pH 7.4), 250 mM NaCl 20 g/ml tosylphenylalanyl chloromethyl ketone, 10 g/ml aprotinin, 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 10 mM NaF, and 0.1 mM Na 3 VO 4 . Lysates were passed through a 27-gauge needle, centrifuged at 13,000 ϫ g for 20 min, and the supernatant was stored at Ϫ80°C. Nuclear and cytoplasmic extracts were prepared essentially according to Verona et al. (38). Cells were briefly trypsinized and washed with phosphate-buffered saline. Cell pellets were resuspended, by flicking the tube, in two packed cell volumes of hypotonic buffer containing 10 mM HEPES (pH 7.5), 10 mM KCl, 3 mM MgCl 2 , 0.05% Nonidet P-40, 1 mM EDTA (pH 8), 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol, 10 g/ml aprotinin, 10 g/ml leupeptin, 10 mM NaF, and 0.1 mM Na 3 VO 4 . Cells were left to swell on ice for 10 min and then vortexed for 10 s and spun at 500 ϫ g for 5 min. The supernatant, containing the cytoplasmic lysate, was supplemented with one-third the volume of 80% glycerol and clarified by centrifugation at 20,000 ϫ g for 30 min. The nuclear pellet was washed twice in hypotonic buffer and lysed in two pellet volumes of lysis buffer containing 100 mM HEPES (pH 7.4), 0.5 M KCl, 5 mM MgCl 2 , 28% glycerol, and protease and phosphatase inhibitors as above. The nuclear extracts were centrifuged at 20,000 ϫ g for 1 h to remove cell debris. Protein concentration was determined using the Micro BCA protein assay reagent kit (Pierce).
Western Analysis-Between 60 and 100 g of protein was subjected to SDS-polyacrylamide gel electrophoresis and transferred to a 0.2-m nitrocellulose membrane using Mini Trans-Blot Transfer Cell (Bio-Rad), and immunodetection was performed using ECL (Amersham Pharmacia Biotech) according to the manufacturer's recommendations, followed by exposure of the membrane to x-ray film. The fluorographic signals were quantified by densitometry of the film using the AlphaImager 2000 system (Alpha Innotech Corp.) Results from one representative experiment out of three independent experiments ate shown for each analyte.
Retrovirus-mediated p27 Overexpression-The murine p27 cDNA (45) was amplified by polymerase chain reaction using Pfu polymerase (Stratagene). The forward primer, 5Ј-GAGCGTTCGAATGTCTAACG-GGAGCCCGA, contained an SfuI recognition site (underlined), and the reverse primer, 5Ј-ATATAGCGGCCGCTTACGTCTGGCGTCGAAGG, contained a NotI recognition site (underlined). The 602-base pair amplicon was digested with SfuI and NotI and cloned into the retroviral replication-competent vector pZAPd, downstream of the encephalomyocarditis virus internal ribosomal entry site (IRES), 2 to yield pZAPd-p27. In preliminary experiments with pZAPd-emd, which contains the emerald mutant (Packard Bioscience) of the green fluorescent protein (GFP) under the control of the encephalomyocarditis virus IRES, 2 calcium phosphate transfection (35) of MC3T3-E1 cells resulted in nearly complete transduction of the entire cell population (see Fig. 4B), which was as fast as transduction with virus prepared in 293 cells. 2 Therefore, calcium phosphate transfection was subsequently employed to introduce the pZAPd vectors into MC3T3-E1 cells, evidently resulting in efficient spread of the virus throughout the entire cell populations.
E2F Electromobility Shift Assays-Nuclear and cytoplasmic extracts (5 g) were subjected to electrophoretic mobility shift assay as described (38) with 1.5 ng of 32 P-end-labeled double-stranded E2F oligonucleotide probe (5Ј-ATTTAAGTTTCGCGCCCTTTCTCAA). The binding reaction (20 -30 l) contained 50 mM KCl, 20 mM HEPES (pH 7.4), 1 mM MgCl 2 , 1 mM EDTA, and 14% glycerol. Prior to the addition of the probe, the protein extracts were preincubated with 0.5-1 g of salmon sperm DNA and either (i) 150 ng of unlabeled E2F oligonucleotide, (ii) 150 ng of unlabeled E2F mutant oligonucleotide (5Ј-ATTTAAGTTTC-GatCCCTTTCTCAA), or (iii) for supershift analysis, antibodies to either E2F-4, p107, p130, or CDK2. Preincubation with the cold DNAs and the antibodies was for 10 min on ice, and incubation with the probe was for an additional 10 min on ice and 15 min at room temperature. For cyclin A supershift analysis, the antibody was added after the probe as described (40). Protein-DNA complexes were resolved by electrophoresis at 4°C in 0.25% TBE, 4% native polyacrylamide gels containing 5% glycerol.
Immunoprecipitation and Kinase Assays-Whole cell extracts (see above) were subjected to immunoprecipitation essentially as described previously (41). For cyclin A, cyclin E, and CDK2 immunoprecipitation, 100 g of cell lysate was incubated with 1 g of primary antibody, followed by 20 l of A/G-agarose bead suspension. For cyclin D1, 500 g of cell extract was immunoprecipitated with 3 g of primary antibody. The immunoprecipitates were washed twice with 0.5% (v/v) Nonidet P-40 buffer as above and once with kinase buffer (41) containing protease and phosphatase inhibitors as above. Immunocomplexes were subjected to kinase assay in the presence of [␥-32 P]ATP and either histone H1 (for cyclin A, cyclin E, and CDK2) or pRb (residues 769 -921) (for cyclin D1) as substrate, followed by SDS-polyacrylamide gel electrophoresis (10% gel containing 5% glycerol) and autoradiography. For kinase inhibition assays (41,42), immunocomplexes were resuspended in 0.5% (v/v) Nonidet P-40 buffer as above boiled for 5 min, and centrifuged, and the supernatant was added to active extracts prior to immunoprecipitation and kinase assay.
Extracellular Matrix Mineralization-Calcium deposition was demonstrated histochemically by Alizarin red staining of 70% ethanol-fixed cultures as described previously (33). For quantitation of calcium accumulation, cell layers were initially scraped in saline solution containing 10 mM Tris-HCl (pH 7.8) and 0.2% Triton X-100 and centrifuged, and an aliquot was removed for protein determination using the BCA protein assay reagent kit (Pierce). HCl was then added to 0.5 M, and the dissolved calcium was quantitated based on the light absorbance of complexes formed between calcium ions and o-cresolphthalein, using Sigma Procedure no. 587. Results are hence expressed as calcium per protein.

GC Inhibit Cell Cycle Progression in Postconfluent MC-3T3-E1 Preosteoblasts-MC3T3-E1
is a preosteoblastic cell line that undergoes a staged developmental process, leading to deposition and then mineralization of the extracellular matrix within ϳ2-4 weeks after plating (43,44). Chronic exposure of MC3T3-E1 cells to DEX results in severe attenuation of phenotype development, but only when treatment commences during the first week of culture (33). This suggests that GC alleviate a commitment stage during early differentiation. In pursuit of early effects of GC in MC3T3-E1 cell cultures, we noticed that, while the nontreated cells exhibited the postconfluent condensation typical of most osteoblast culture systems, this process was attenuated by DEX ( Fig. 1). No morphological difference was observed between nontreated and DEX-treated cultures until confluency (see Fig. 1, day 4). We postulated that a unique cell cycle control was instituted during the condensation period, and that this cell cycle control is susceptible to the antimitogenic effect of GC. Flow cytometric analysis was performed to examine the cell cycle profile before and after confluency, in the absence and presence of DEX. The results show that the cultures maintain an active cell cycle for several days after confluency (Fig. 2). In multiple experiments, the percentage of postconfluent cells in S, G 2 , and M was 20 -30%, and was equal to or only slightly lower than that in the preconfluent cultures. DEX consistently inhibited cell cycle progression in the postconfluent cultures, reflected by a ϳ2-fold reduction in the percentage of cells in S, G 2 , and M ( Fig. 2). Remarkably, DEX did not inhibit the cell cycle in preconfluent cultures (Fig.  2, day 3). The GC-induced inhibition of cell cycle progression, occurring specifically in the postconfluent cultures, did not result from prolonged exposure to the steroid, because, as shown in Fig. 2, B and C, an acute (20-h) exposure of the postconfluent cells to DEX was sufficient for inhibition of cell cycle progression to levels observed in the chronically treated cultures ( Fig. 2A).
The Specific Inhibition of Cell Cycle Progression in Postconfluent Osteoblasts Does Not Reflect GR Potentiation-Lack of inhibition of cell cycle progression by GC in preconfluent cultures (Fig. 2, A-C) could result from unavailability of functional GC receptor (GR). To address this possibility, we compared the effects of DEX in preconfluent versus postconfluent osteoblasts on several promoter-reporter constructs. As shown in Fig. 3A, DEX strongly activated the MMTV promoter in preconfluent cultures, even more than in postconfluent cultures (130-versus 3.5-fold, respectively). The promoter of Id2, which has been implicated in cell cycle control (45), was responsive to DEX in preconfluent but not postconfluent MC3T3-E1 cells (Fig. 3B). Thus, preconfluent MC3T3-E1 cultures evidently have a functional receptor. Interestingly, additional transient transfection experiments revealed one promoter, that of the CDI p21, to be specifically affected by DEX in postconfluent MC3T3-E1 cells. However, unlike the effects observed in hepatoma (24) and osteosarcoma (26) cells, p21 promoter activity (as well as protein levels; see Fig. 4) is decreased, not increased, in postconfluent MC3T3-E1 cells treated with DEX ( Fig. 3C), probably in a p53-dependent manner ( Fig. 3D; see Ref. 46).

DEX-induced Growth Inhibition of Postconfluent MC3T3-E1 Cells Is Not Mediated by p21 or p27 Cyclin-dependent Kinase
Inhibitors-To address the mechanism underlying the GC-mediated attenuation of cell cycle progression in postconfluent MC3T3-E1 cells, we initially determined the effect of DEX on expression of the CDIs p21 and p27, which play a role in the antimitogenic activity of GC in hepatoma, lung alveolar, and osteosarcoma cells (23)(24)(25)(26). As shown in Fig. 4A, the level of p21 is decreased, not increased, in DEX-treated MC3T3-E1 cells, thus ruling out this CDI as the mediator of the GC effect.
In contrast to p21, Fig. 4A demonstrates a Ͼ2-fold DEXinduced increase in p27 levels (2.3 Ϯ 0.2-fold in three independent experiments), which could contribute to the inhibition of G 1 /S transition in DEX-treated postconfluent MC3T3-E1 cells. This possibility was further addressed by overexpression of p27. We employed the replication-competent retroviral vector ZAPd, which comprises the Moloney murine leukemia virus genome into which was inserted the IRES from the encephalomyocarditis virus (Fig. 4B). 2 Efficiency of this vector in MC3T3-E1 cells was evaluated by infecting cells with pZAPdemd, in which the GFP gene was inserted downstream of the encephalomyocarditis virus IRES (Fig. 4B). As revealed by flow cytometry analysis for GFP expression, 9% of the cells in the infected culture expressed GFP by day 8 postinfection, and this increased to 99.1% within the following 6 additional days after passing (Fig. 4B). Next, p27 cDNA was cloned in place of GFP, and the expression of p27 in infected MC3T3-E1 cultures was evaluated by Western analysis. As shown in Fig. 4C (top), the pZAPd-p27-infected cells expressed p27 at levels severalfold higher than those observed in control pZAPd-emd-infected cells. These levels are also higher than those of DEX-treated ( Fig. 4A) or serum-starved cells, used as control (Fig. 4C, top). If p27 plays a pivotal role in the DEX-induced growth inhibition, then this may be mimicked by postconfluent p27-overexpressing cells in the absence of the steroid. However, as demonstrated by propidium iodide staining and flow cytometry analysis, the cell cycle profile of postconfluent p27-enriched cells was only marginally affected, whereas each of the DEXtreated or serum-deprived cells, serving as positive controls, exhibited the expected decrease in the percentage of cells at the S/G 2 /M phases (Fig. 4C). These results suggest that the upregulation of p27 in DEX-treated postconfluent MC3T3-E1 cultures does not mediate, at least not alone, the attenuation of cell cycle progression.
Dissociation of Cyclin A and CDK2 from E2F4-p130 Complexes in GC-treated Osteoblasts Growing to High Density-E2F proteins play a pivotal role in the transcriptional regulation of numerous cell cycle regulatory genes (reviewed in Ref. 47). To address whether the osteoblast developmental stagespecific DEX-mediated G 1 /S attenuation resulted from alterations in E2F complexes, electrophoretic mobility shift assay was performed with nuclear extracts of nontreated and GCtreated MC3T3-E1 cells, either growing to confluency (day 3) or to high density (day 6). Similar to a variety of other cell types (see, for example, Refs. 48 -50) the E2F binding activity in MC3T3-E1 cells consists of a number of fast migrating free E2F complexes and a cluster of slow migrating E2F-pocket complexes (Fig. 5A). As compared with the preconfluent cells (day 3), a prominent reduction in the free E2F complexes is observed in the postconfluent cultures. There are also significant alterations in the composition of the E2F-pocket cluster ( Fig. 5A; see below). DEX did not affect the E2F binding activities in the preconfluent stage ( Fig. 5A; see Fig. 5, B and C). However, at the postconfluent stage (day 6), DEX decreased the free E2F-DNA binding activity, and the E2F-pocket cluster was significantly altered, exhibiting a new, faster migrating complex. As opposed to the nuclear fraction, cytoplasmic E2F binding activity in the postconfluent cultures was not affected by DEX ( Fig. 5A, Cyt). The cytoplasmic activity consisted of one major free E2F complex and a light E2F-pocket, the latter co-migrating with that seen in the DEX-treated postconfluent nuclear extracts (Fig. 5A).
Competition analyses were performed to characterize the protein-DNA complexes, initially in preconfluent cultures (Fig.  5, B and C). With the exception of the fastest migrating band, all of the complexes exhibited DNA sequence-specific binding (compare lanes wt and mt). Supershift analysis with antibodies to E2F4, p107, and p130 suggest that in the preconfluent stage, the E2F complexes are similar between nontreated (Fig. 5B) and DEX-treated (C) cultures. In both cases, the slowest migrating free E2F complex was identified as E2F4, since anti-E2F4 antibodies attenuated its migration to the position marked by the closed circle (Fig. 5, B and C; see Fig. 6). E2F4 is also present in one of two E2F-pocket complexes (arrow), which is supershifted to the position marked by an open circle. The pocket component of this complex (arrow) is mainly p130, because anti-p130 antibodies block its formation. The composition of the other E2F-pocket complex (arrowhead) has not been determined. The light supershifted band observed with anti-p107 antibodies indicates that this protein is also represented in the preconfluent E2F-pocket complexes from both nontreated and DEX-treated cells (Fig. 5, B and C).
The abundance of E2F4-p130 complexes in preconfluent actively proliferating MC3T3-E1 cells was unexpected, because E2F4-p130 complexes typically form in cells that undergo growth arrest, replacing E2F4-p107 complexes (51). To ascertain the expression and nuclear localization of E2F4 and p130 in proliferating MC3T3-E1 cells, the nuclear and cytoplasmic fractions were subjected to Western analysis. As shown in Fig.  5D, both E2F4 and p130 are present abundantly in MC3T3-E1 cell nuclei, regardless of developmental stage or DEX treatment. Interestingly, E2F4 cellular localization is developmentally regulated, exhibiting an equal contribution to nuclear and cytoplasmic proteins until confluency but a clear nuclear preference thereafter (Fig. 5D).
Analysis of the E2F complexes in postconfluent untreated MC3T3-E1 cells is presented in Fig. 6A. The predominant E2F-pocket complex (arrow) consists of E2F4 and p130 as antibodies against each of these proteins block or supershift the majority of the E2F-pocket cluster. This complex thus appears to be identical to that of preconfluent cultures (compare with Fig. 5, B and C, arrow), based on co-migration (Fig. 5A) and supershift analyses. In contrast, the postconfluent cultures seem to have lost the other E2F-pocket complex present in the preconfluent cells (compare with Fig. 5, B and C, arrowhead).
As shown in Fig. 5A, postconfluent DEX-treated cultures exhibit one major E2F-pocket complex but with faster migration compared with that of nontreated cells. Fig. 6B demonstrates that, similar to the major E2F-pocket complex of the postconfluent nontreated cells (Fig. 6A, arrow), the faster migrating complex of the treated cells (white arrow) also consists of E2F4 and p130. Therefore, the migration difference between the E2F4-p130 complexes in postconfluent treated versus nontreated cells is probably accounted for by the loss of some additional component(s). These could be cyclin-CDK2 complexes, which typically bind to the inhibitory E2F-pocket complexes in proliferating cells, resulting in release of the pocket protein to yield free E2F (52).
To address the hypothesis that DEX inhibits postconfluent MC3T3-E1 cell cycle progression by dissociating cyclin-CDK2 from E2F4-p130 complexes, supershift analysis was performed with antibodies to CDK2, cyclin A, and cyclin E. Indeed, CDK2 antibodies strongly affected the E2F4-p130 complex (arrow) in nontreated postconfluent cultures (Fig. 6C), while no effect was observed on the E2F4-p130 complex of postconfluent DEXtreated cells (Fig. 6D, white arrow). Supershift analysis of the nontreated cells with CDK2 in combination with either E2F4 or p130 antibodies demonstrated a further shift compared with that observed with CDK2 antibodies alone, indicating co-existence of the three proteins in the same complex (Fig. 6C). The E2F4-p130 complex of the preconfluent cells (Fig. 5, B and C, arrow) also contains cyclin-CDK2 (data not shown).
We next addressed the identity of the cyclin component, responsible for tethering (and probably activating) CDK2 to the E2F4-p130 complex of postconfluent nontreated cells. To this end, we added cyclin A antibodies to the supershift reaction that already contained anti E2F4 antibodies. Fig. 6C (right panel) demonstrates complete and specific disappearance of the E2F4-p130-anti-E2F4 complex with the addition of the cyclin A antibodies. Some of the anti-cyclin A-containing complexes were retained in the well, which is characteristic of supershift with this specific antibody (40) and is not attributable to interaction with the free probe (Fig. 6C, rightmost lane). Not surprisingly, the cyclin A antibody did not affect the CDK2-deficient E2F4-p130-anti-E2F4 complexes of the DEX-treated postconfluent cells (Fig. 6D, right panel). As expected from the complete shift by the cyclin A antibody (Fig. 6C), cyclin E antibody did not affect any of the E2F complexes (data not shown). In summary, E2F4-p130-cyclin A-CDK2 becomes the predominant E2F-pocket complex in postconfluent MC3T3-E1 proliferating osteoblasts (Fig. 6C). Treatment of the postconfluent cells with DEX leads to elimination of cyclin A-CDK2 from this complex, resulting in a predominant stable E2F4-p130 complex (Fig. 6D), thus leading to attenuation of G 1 /S transition (Fig. 2).

GC Inhibit Cyclin A and CDK2 Expression Levels in Postconfluent MC3T3-E1 Cultures-DEX-induced elimination of cyclin
A and CDK2 from E2F4-p130 complexes of postconfluent MC3T3-E1 cells (Fig. 6) could reflect either a decrease in the expression of these proteins or failure to associate with the E2F4-p130 moiety. To address this issue, Western analysis of nontreated and DEX-treated cells was performed. As demonstrated in Fig. 7A, DEX induced a ϳ5-fold decrease in the level of cyclin A (5.2 Ϯ 2.6-fold in three independent experiments), and this was specific to the postconfluent stage (day 6). A   FIG. 4. p21 and p27 do not mediate the antimitogenic effect of DEX. A, Western blot analysis. MC3T3-E1 cells growing to confluency (day 3) or to high density (day 6) were treated with DEX (1 M) or vehicle. The steroid was administered with the last medium change, 20 h prior to harvest. Protein extracts were prepared, and 80 g was subjected to Western analysis as described under "Materials and Methods," using either p21 or p27 antibodies. B, high efficiency retrovirusmediated delivery of GFP to MC3T3-E1 cells. Top, schematic illustration of the replication-competent pZAPd-emd retrovirus, encoding viral enzymes (pol), glycoprotein antigen (gag), and envelope (env) proteins, as well as the GFP, translatable via the encephalomyocarditis virus IRES. Bottom, MC3T3-E1 cells were transfected with p1ZAPd-emd using calcium phosphate co-precipitation (35). Nontransfected cells and cells aliquoted during passaging on days 8 and 14 post-transfection were analyzed by flow cytometry. The progressive right shift indicates that the successfully transfected cells become virus producers and infect neighboring cells. C, forced p27 expression does not mimic DEX-mediated cell cycle attenuation. MC3T3-E1 cells were infected with pZAPd-p27 (top) using calcium phosphate co-precipitation (35) and passaged on days 8, 14, and 20. Six days following the last passage, cells were subjected to either p27 Western analysis (middle) or to flow cytometry (bottom) as in Fig. 2. pZAPd-emd-infected cells and 24 h serum-starved cells were analyzed as controls. The effect of DEX (0.1 M, 20 h) on the cell cycle profile was also determined as in Fig. 2. A representative cell cycle profile is shown for each experimental condition, with depiction of the percentage of cells in the G 1 , S, and G 2 /M phases of the cell cycle. smaller or no reduction was observed in the abundance of Dand E-type cyclins (Fig. 7A). DEX also induced a significant decrease in the expression levels of CDK2 in the postconfluent cultures (2.6 Ϯ 0.3-fold, n ϭ 3) and in the level of CDK4 in both the pre-and postconfluent stages. Thus, the stage-specific effects of GC on MC3T3-E1 cell cycle progression (Fig. 2) and E2F-pocket complex composition (Fig. 6) best correlate with inhibition of cyclin A and CDK2 expression levels. These effects were also observed at the mRNA level (data not shown).

GC Decrease Cyclin A-, CDK2-, and Cyclin E-associated Kinase Activity in Postconfluent MC3T3-E1
Cells-Persistence of inhibitory E2F-p130 complexes requires that CDK2 kinase activity be reduced (51). Immunoprecipitation and kinase assays were therefore performed with extracts of pre-and postconfluent MC3T3-E1 cells, either treated with DEX or untreated. As shown in Fig. 8A, the kinase activities associated with cyclin A and CDK2 were strongly reduced specifically in DEX-treated postconfluent cultures. In contrast, DEX only minimally decreased the kinase activity associated with cyclin D1 (Fig. 8A). Surprisingly, the kinase activity associated with postconfluent cyclin E was dramatically reduced by DEX (Fig. 8A) in the face of conserved protein levels (Fig. 7A). This led us to address the possibility that the reduced cyclin E-, as well as cyclin A and CDK2-associated kinase activity could partly result from an increased association with CDK2 inhibitors. To this end, we exploited the heat resistance property of some CDK2 inhibitors, including p21 and p27, which allows estimation of their activity using a boiling/kinase inhibition assay (41,42). Cyclin E immune complexes from nontreated or DEX-treated postcon-  (A and B), cyclin-CDK2 antibodies were also employed here (C and D). Note that the E2F4-p130 complex of the nontreated cells (arrow), but not that of the DEX-treated cells (white arrow) contains cyclin A and CDK2. Ab in C indicates incubation of the probe with the cyclin A antibody in the absence of nuclear extracts.

FIG. 7. DEX decreases cyclin A and CDK2 levels in postconfluent MC3T3-E1 cultures.
MC3T3-E1 cells growing to confluency (day 3) or to high density (day 6) were treated with either DEX (1 M) or vehicle. The steroid was administered with the last medium change, 20 h prior to harvest. Protein extracts were prepared, and 60 -80 g was subjected to Western analysis as described under "Materials and Methods," using anti-cyclin (A) or anti-CDK (B) antibodies. fluent cultures were boiled, and the supernatant was assayed for CDI activity upon active cyclin E complexes of nontreated cells. As shown in Fig. 8B, the active complexes were only slightly inhibited by the boilate of the DEX-treated as compared with that of the nontreated cells (column 2 versus column 1). This inhibition did not approach the diminution of cyclin E-associated kinase activity observed in postconfluent cells treated with DEX (Fig. 8A). Similarly, active cyclin A and CDK2 immune complexes of nontreated cells were not inhibited by the respective boiled complexes of DEX-treated compared with nontreated postconfluent cells (Fig. 8B). Thus, the decrease in cyclin E-associated kinase activity in postconfluent DEX-treated cells is not attributable to either a decrease in cyclin E (Fig. 7A) or an increase in associated heat-resistant CDI activity (Fig. 8B). Instead, the DEX-induced inhibition of cyclin E-associated kinase activity in postconfluent MC3T3-E1 cells may result from the reduction in CDK2 protein levels (Fig.  7B). Finally, Western analysis of pocket proteins, physiological CDK2 substrates, revealed that the DEX-mediated reduction of CDK2 kinase activity in postconfluent MC3T3-E1 cells is reflected in hypophosphorylation of both pRB and p130 (Fig. 8C). Taken together, our data are consistent with a scenario in which DEX treatment of postconfluent MC3T3-E1 cultures results in inhibition of cyclin A and CDK2 expression, leading to inhibition of both cyclin A-and cyclin E-associated CDK2 kinase activity and thus stabilization of inhibitory E2F-pocket complexes on promoters of cell cycle regulatory genes.
Osteoblasts That Have Grown to High Density Are Resistant to the Inhibitory Effect of GC on Differentiation-To begin addressing the significance of the developmental stage-specific antimitogenic effect of GC to bone formation, we initiated exposure of MC3T3-E1 cultures to DEX on day 3, 4, 5, or 6 (i.e. just before confluency, at confluency, and 1 or 2 days after confluency). The cultures were treated until day 21, and calcium deposition was evaluated histochemically and biochemically. The results demonstrate that DEX treatment commencing before confluency impedes the differentiation process, as reflected by Ͼ90% inhibition of calcium deposition (Fig. 9). However, 1 day of DEX-free postconfluent condensation (i.e. commencement of DEX treatment on day 4) is sufficient to render the cultures partially resistant to GC exposure for the remaining of the culture period. When DEX treatment was initiated on day 6, after 3 days of DEX-free postconfluent condensation, calcium accumulation amounted to 64% of that measured in nontreated cultures (Fig. 9). Because calcium deposition in our MC3T3-E1 cultures starts around day 10 -12, these data suggest that GC do not inhibit the mineralization process per se. Rather, a commitment step(s) associated with the postconfluent cell cycle is abrogated, leading to impaired differentiation.
Alleviation of GC Effects on Osteoblast Postconfluent Proliferation and Differentiation by the Antagonist/Partial Agonist RU486 -The pleotropic effects of GC in cells may be explained, in part, by a multitude of mechanisms of action of the activated GR. These mechanisms include transcriptional activation and repression; autonomous action and interactions with other transcription factors (e.g. AP1, NF-B); and DNA binding-dependent and -independent modes of action (reviewed in Ref. 53). Synthetic analogues, such as the partial agonist/antagonist RU486 have been employed to selectively mimic or block specific GC effects (54 -56). The partial agonist/antagonist activity of RU486 in MC3T3-E1 cells is demonstrated in Fig. 10A using the MMTV promoter in transient transfection assays; some activation is observed with 0.1 M RU486 alone (ϳ6% of that observed with 0.1 M DEX), and ϳ20-fold repression is observed when RU486 is added together with DEX. We rationalized that, if GC inhibit osteoblast proliferation and differentiation via independent mechanisms, the two effects might be separable using RU486. However, RU486 alone did not mimic the effect of DEX on either osteoblast differentiation (Fig. 10B) or cell cycle progression (Fig. 10C). When administered together with DEX, RU486 blocked the inhibition exerted by DEX on both mineral deposition (Fig. 10B) and cell cycle progression (Fig. 10C). Thus, neither the partial agonist nor the antagonist activity of RU486 could dissociate the effects of FIG. 8. Dex inhibits cyclin A-, CDK2-, and cyclin E-associated kinase activity in postconfluent MC3T3-E1 cultures. A, kinase assays. MC3T3-E1 cells were plated in 100-mm plates and harvested either before (day 3) or after (day 6) confluency, 20 h after the last feeding, which was without (Cont) or with (DEX) 1 M DEX. Immunoprecipitates obtained with the indicated antibodies or IgG control antibodies were subjected to kinase assays as described under "Materials and Methods," using either Rb (for cyclin D1) or histone H1 as substrate. The arrow in the cyclin D1 panel (cycD1) indicates a specific phosphorylation product, as opposed to the faster migrating product (ns), which is also phosphorylated by immunoprecipitates obtained with nonspecific IgG antibodies. B, kinase inhibition assays. Immunoprecipitates of cyclin E, cyclin A, or CDK2 were boiled, and the released proteins were assayed for their ability to inhibit active complexes in kinase assays as in A. All of the lysates were from postconfluent (day 6) cells. The active complexes were obtained, using the indicated antibodies, always from nontreated cultures. Lane 1, the boiled complexes were obtained with the indicated antibodies from nontreated cells; lane 2, the boiled complexes were obtained with the indicated antibodies from DEX-treated cells; lane 3, the boiled complexes were obtained with nonspecific IgG antibodies from DEX-treated cells. C, pocket protein Western analysis. Extracts were prepared and subjected to Western analysis as described in the legend to Fig. 4A, using antibodies against the indicated pocket proteins. Faster migrating bands observed with anti-pRB and anti-p130 antibodies represent hypophosphorylated forms of the respective proteins. GC on MC3T3-E1 cell proliferation and differentiation.
GC-mediated Inhibition of Osteoblast Postconfluent Proliferation and Differentiation Is Reversible-The antimitogenic activity of GC can be either reversible, such as in murine Tlymphoma (18), fibrosarcoma (57), SAOS-2 osteosarcoma (56), and mast (58) cells, or it can be irreversibly cytotoxic, leading to apoptosis, such as in murine thymoma (59), osteoclasts (60), granulocytes (61), MEL (62), and U2OS osteosarcoma (56) cells. As shown in Fig. 11A by flow cytometry analysis, the antimitogenic effect of DEX in postconfluent MC3T3-E1 cells is reversible, since withdrawal of the steroid following 10 days of treatment results in a rapid restoration of a cell cycle profile similar to that seen in control cultures (compare Fig. 11A with Fig. 2).
The restoration of the cell cycle profile following DEX withdrawal prompted us to examine whether the inhibitory effect of DEX on development of the osteoblast phenotype would also be reversible. Cells were treated for 10 days as in Fig. 11A, and calcium accumulation was evaluated 5 days following DEX withdrawal. As shown in Fig. 11B, withdrawal of DEX resulted in robust extracellular matrix mineralization, restoring within 5 days more than half of the calcium deposition observed in cultures that had not been treated at all. In summary, the requirement for the presence of DEX during the time period of cell condensation (Fig. 9) along with the parallel effects of RU486 (Figs. 10) and of DEX withdrawal (Figs. 11) on postconfluent cell cycle progression and calcium accumulation together suggest a linkage between the antimitogenic and antiphenotypic properties of glucocorticoids in osteoblasts. DISCUSSION Cell proliferation and differentiation are partially overlapping processes. During the differentiation course of MC3T3-E1 osteoblastic cells, a developmental switch occurs, which is responsible for persistent proliferation after confluency. Most notably, the cell cycle that drives growth to high density is inhibited by pharmacological doses of glucocorticoids, while the cell cycle driving growth toward confluency is not.
The cell cycle machinery of postconfluent MC3T3-E1 cultures is distinct from that operative in preconfluent cells. Some of the alterations in the cell cycle machinery, which occur during MC3T3-E1 growth to high density, typify cells undergoing growth arrest. Among these alterations are (i) an increase in p27 levels with a reciprocal decrease in p21 levels, as previously reported for rapamycin-arrested T lymphocytes (42) and contact-inhibited human fibroblasts (63); (ii) reduced free E2F DNA binding activity, as seen, for example, in growthinhibited BALB/c 3T3 whole cell extracts (50) and in quiescent REF52 nuclear extracts (64); and (iii) preferential nuclear localization of E2F4, as in serum-starved NIH3T3 cells (65). The occurrence of these modifications in postconfluent MC3T3-E1 cultures suggests that the cell cycle persists in these cells despite some growth-inhibitory signal(s). Cooperativity be- tween these signals and those elicited by GC may induce postconfluent cell cycle inhibition.
MC3T3-E1 cells display a unique behavior of the RB-related protein p130. First, as opposed to many cell types in which p130 is a marker of growth arrest (51), cycling MC3T3-E1 cells constitutively express high p130 levels. Further, the protein is localized to the nucleus and found in E2F4-DNA complexes. In several previous studies demonstrating the unusual occurrence of E2F4-p130 complexes in cycling cells, these complexes have been shown to reside in the cytoplasm (38), or cellular localization has not been addressed (50,66,67). Significantly, the E2F4-p130 complexes of both pre-and postconfluent cycling MC3T3-E1 cells contain cyclin-CDK2; specifically at the postconfluent stage, this complex is the main E2F-pocket complex. It has been suggested that similar to E2F4-p107-cyclin-CDK2 complexes (52), the E2F4-p130-cyclin-CDK2 complexes represent an intermediate stage leading to release of free E2F4 (66). Free E2F4 binding activity would then promote G 1 /S transition (39).
Administration of DEX to postconfluent MC3T3-E1 cells results in attenuation of the G 1 /S transition. This antimitogenic effect is attributable to elimination of cyclin A and CDK2 from E2F4-p130 complexes, promoting the stability of the inhibitory E2F4-p130 complexes. Dissociation of CDK2 from E2F-p107 (52) and E2F-p130 (66) complexes was suggested to be regulated by the CDI p21. However, the DEX-induced elimination of cyclin A-CDK2 from E2F-p130 complexes in MC3T3-E1 cells is accompanied by a decrease, not an increase, in p21 levels. The specific loss of cyclin A-CDK2 from E2F4-p130 complexes at the postconfluent stage reflects down-regulation of their overall expression level, which also results in attenuation of the associated kinase activity. Although cyclin-CDK complexes are generally thought to be regulated by the cyclin, not the CDK levels, the importance of the CDK2 down-regulation is suggested by the decreased cyclin E-associated kinase activity, despite retention of normal cyclin E protein levels.
Whether the down-regulation of cyclin A and CDK2 alone is sufficient to mediate the GC attenuation of the osteoblast persistent cell cycle remains to be addressed by forced expression of these proteins. However, it is already clear that the antimitogenic mechanism of GC in MC3T3-E1 cells differs from that observed in lymphoid (18 -20), hepatoma (23,24), lung alveolar (25), SAOS2, or U2OS osteosarcoma cells (26). Further, we have noticed a kinetic difference, in that the decline in cyclin A and CDK2 in MC3T3-E1 cells is not detectable until several hours after initiation of treatment (data not shown). A plausible interpretation is that the stage-specific antimitogenic effect of GC in postconfluent MC3T3-E1 cells is related to the well established suppression of bone extracellular matrix production by these drugs.
Chronic exposure to GC in vivo results in bone loss. Our study with MC3T3-E1 osteoblastic cells is consistent with the notion that inhibition of osteoblast proliferation contributes to GC-induced osteoporosis (2,3). In MC3T3-E1 cells, GC inhibit developmental stage-specific cell cycle machinery, operative in postconfluent cultures prior to terminal differentiation. It is possible that this cell cycle is an integral component of osteoblast differentiation, similar to the postconfluent proliferative phase (clonal expansion) observed in 3T3-L1 preadipocytes induced to differentiate (27,28). The commonly observed cell condensation, almost invariably accompanying full development of the skeletal phenotype in vitro, may reflect the requirement for a postconfluent proliferative phase during osteoblast differentiation. A possible linkage between the postconfluent cell cycle and osteoblast differentiation is further supported by the following correlations observed in MC3T3-E1 cells: (i) DEX treatment induces a concurrent abrogation of both postconfluent cell cycle progression and terminal differentiation; (ii) the effects of DEX on both the cell cycle and terminal differentiation are antagonized by RU486; and (iii) the effects of DEX on both the cell cycle and terminal differentiation are reversible. It would be interesting to test whether terminal differentiation would occur normally, despite the presence of GC, in osteoblasts whose postconfluent growth is temporally restored by manipulation of DEX-responsive cell cycle regulatory genes. It would also be important to identify an osteoblast developmental stage in vivo that resembles the glucocorticoid-sensitive osteoblast persistent cell cycle, demonstrated here with our MC3T3-E1 subclone.