Introduction

The association between steroid administration and the development of osteonecrosis has been well established since 1957. However, the precise pathogenesis of osteonecrosis remains unknown [1,2]. Investigations on the mechanism underlying the pathogenesis of osteonecrosis have so far focused on either diminished blood supply to bone or marrow cell differentiation [3,4,5]

The pathobiological mechanism underlying the induction of adipogenesis and suppression of osteogenesis by steroids has not been elucidated, to our knowledge. In our initial report [6], we demonstrated that dexamethasone stimulates bone marrow mesenchymal cell differentiation into adipocytes as well as the accumulation of fat in the marrow while suppressing cell differentiation into osteoblasts; nevertheless, the mechanism is still not clear. It is likely that interconversion of stromal cells among phenotypes, as well as commitment to a particular lineage with suppression of alternative phenotypes, is governed by specific transcription factors [7,8]. Indeed, core binding factor a1/Runt-related transcription factor-2 (Cbfa1/Runx2) is a transcription factor that is required to commit mesenchymal progenitors to the osteoblast lineage [9,10,11,12]. Mice that are deficient in this factor lack osteoblasts and mineralized bone matrix [12]. Expression of Cbfa1/Runx2 in fibroblastic cells induces transcription of osteoblast specific factors [13]. By contrast, peroxisome proliferator activated receptor γ2 (PPARγ2) gene expression destines cells for adipocyte differentiation [14,15]. Transfection of fibroblastic cells with PPARγ2 and subsequent activation with an appropriate ligand causes the development of adipocytes, supporting the idea that PPARγ2 plays a crucial role in the differentiation of mesenchymal cells to adipocytes [16]. The effect of dexamethasone on cell differentiation through its influence on transcription factors needs to be elucidated.

Meanwhile, many authors have suggested that interference with the blood supply, by various means, plays a major role in the pathogenesis of osteonecrosis [2,17]. Intraosseous hypertension [18], intravascular fat emboli and coagulation [19,20] compression of vessels by progressive accumulation of marrow fat stores [21] and coagulopathies are commonly accepted theories. Such theories have focused on decreased blood supply as the underlying factor for osteonecrosis. There are few reports on the angiogenic changes that ensue upon treatment of mesenchymal cells with dexamethasone; however, normal and pathological bone physiology is inexorably tied to angiogenesis [22]. There are several indications that vasculature plays an important role for the mechanism of coupling resorption by osteoclasts and bone formation by osteoblasts [23]. There is a functional relationship between bone endothelium and osteoblasts, although the mechanism is not clear. Recent reports document the important role of angiogenic factors in skeletal development and in disorders of the skeleton. The process of bone development and repair depends on adequate formation of new capillaries from existing blood vessel [24]. Angiogenesis, the development of a microvascular network for blood supply, is critical for the development, remodeling, and healing of most tissues, including bone [24,25,26,27]. Vascular endothelial growth factor (VEGF), a dimeric heparin-binding glycoprotein, plays a central role in the development and modulation of angiogenesis and has been studied extensively; moreover, VEGF–mediated angiogenesis is important in fracture repair [28] and endochondral ossification [29]. Thus, it appeared promising to investigate the role of angiogenic factors, specifically VEGF, in osteoporosis and osteonecrosis.

Understanding the effects of dexamethasone on bone and fat cell transcription factors, and changes in VEGF during mesenchymal cell differentiation, are clearly important in order to elucidate the mechanism of steroid induced osteonecrosis. This study was designed in an attempt to answer two questions. First, do steroids induce adipogenesis and suppress osteogenesis by regulating transcription factors PPARγ2 and Cbfa1/Runx2? Second, will steroid decrease angiogenesis by decreasing VEGF expression? We examined the effect of dexamethasone on a multipotential bone marrow cell line, D1, which was previously cloned from mouse bone marrow stroma and shown to produce a mineralized matrix as well as adipogenesis in vivo [30]. The treatment of D1 cells with dexamethasone in vitro showed diminished gene expression for Cbfa1/Runx2, and an increase in PPARγ2, consistent with inhibition of osteogenesis and enhanced adipogenesis in vitro. In addition, VEGF production was suppressed upon treatment of mesenchymal cells with dexamethasone in vitro.

Materials and methods

Steroid induced adipogenesis in culture

Multipotential mesenchymal cells, D1, which were cloned from Balb/c mouse bone marrow cells, were maintained in Dulbecco’s Modified Eagle’s Medium (Gibco BRL, Gaithersburg, Md., USA) containing 10% fetal bovine serum (Hyclone Laboratories, Logan, Vt., USA), 50 μg/ml sodium ascorbate, and 100 IU/ml penicillin G and 100 μg streptomycin per ml of culture media, in a humidified atmosphere of 5% carbon dioxide at 37°C. For all experiments, cells were seeded at a density of 5×103 cells per cm2 and the experiments were started when the cells reached 80% confluence.

D1 cells were treated with five different concentrations of dexamethasone (Sigma Chemical Company, St Louis, Mo., USA) between 10−9 and 10−5 mol/l. Cell morphology and the appearance of cytoplasmic lipid droplets were monitored with a phase contrast microscope. To determine the number of adipocytes, cells in culture were stained with Sudan IV, a stain for fat, counterstained with hematoxylin, and scanned using a Nikon 35 mm slide scanner (Nikon Inc., Melville, N.Y., USA).

Semiquantitative RT-PCR

Semiquantitative RT-PCR was carried out according to the Ambion (Ambion, Austin, Tex., USA) protocol based on the Competimer Technique. This method has been validated to be semiquantitative with data that has correlation coefficients of –0.97 [8,31]. By mixing primers for 18S rRNA with Ambion’s exclusive competimers (primers of the same sequence but that cannot be extended), the 18S rRNA signal can be reduced even to the level of rare messages. It provides semiquantitative data on relative changes of a given mRNA when amplified with the optimal ratio of 18S primers:competimers for a fixed number of PCR cycles within the liner range [32]. RNA (0.5 μg) was reverse transcribed in 20 μl buffer containing AMV reverse transcriptase 5×2.5 μmol/l poly dT, 1 mmol/l each of dATP, dCTP, dGTP, dTTP, 20 IU of RNAse inhibitor, and 20 IU of AMV RT. The reverse transcription reaction was incubated at room temperature for 10 min and then in a Perkin-Elmer Cetus DNA Thermal Cycler at 42°C for 15 min, 99°C for 5 min, and then at 5°C for 5 min. Aliquots of cDNA were amplified in a 100 μl PCR reaction mixture which contained 0.3 μmol/l of 5’ and 3’ oligoprimers, in high-fidelity PCR buffer containing 15 mmol/l MgCl2, 0.1 nmol/l each of dATP, dCTP, dGTP, and dTTP, 0.35 IU of high-fidelity Taq DNA polymerase. For each cDNA sample, PCR amplification was performed in triplicate. The identity of PCR products was confirmed by sequence analysis in an automated DNA sequencer (Perkin-Elmer, Norwalk, Conn., USA).

The primer sequences for PCR were: for Cbfa1/Runx2: mCbfa1/Runx2 +515.F-5’ACG ACA ACC GCA CCA TGG T-3’; mCbfa1/Runx2 +1382.R-5’CGG CTC TCA GTG AGG GAT G-3’; for PPARγ2: mPPARγ2+1111.F-5’CTG GCC TCC CTG ATG AAT AA-3’; mPPARγ2+1315.R-5’GGC GGT CTC CAC TGA GAA TA-3’.

For mCbfa1/Runx2, amplification by PCR was performed at thermal cycling parameters of 94°C for 30 s, 60°C for 30 s and 72°C for 30 s for 34 cycles followed by a final extension at 72°C for 7 min. For mPPARγ2, amplification was carried out with 30 cycles of 94°C for 15 s, 50°C for 15 s and 65°C for 30 s followed by a final extension at 72°C for 2 min. For each PCR product mixture, 5 μl was analyzed by electrophoresis in 3% (w/v) agarose gels. The amplified DNA fragments were visualized by staining with ethidium bromide. Photographs were scanned using a ScanJet II cx (Hewlett Packard, Calif., USA). Relative quantitative data was obtained using ImageQuant Software (Molecular Dynamics, Aunnyvale, Calif., USA). Quantitative differences were normalized with the QuantumRNA 18S PCR products (Ambion, Austin, Tex., USA).

Transient reporter gene assays

D1 cells at a density of 3×105 per well in six-well culture plates were maintained in culture media until they reached 60–80% confluence. Cells were co-transfected with two DNA constructs: (1) human osteocalcin promoter luciferase-fusion plasmid (phOC-luc; 3 μg), and (2) pSV β-gal plasmid (0.5 μg), using a DOSPER Liposomal Transfection protocol (Roche, Boehringer Mannheim). The cells were transfected with 60 μl DOSPER/DNA complexes for 6 h in 1 ml fresh culture medium without serum and maintained in culture for an additional 24–48 h in complete growth media. Luciferase activity was measured using a luciferase assay kit (Promega, Madison, Wisc., USA) and luminescence was detected with an Optocom I luminometer (MGM Instruments, Cambridge, Mass., USA). β-Galactosidase activity was measured with a commercial kit using a colorimetric assay (Promega). Luciferase activity was normalized against β-galactosidase activity.

VEGF assay

Cell culture supernatants were collected at 12, 24 and 48 h for the time-course studies with 10−7 mol/l dexamethasone, or after 24 h for the dose-response studies with 10−9 mol/l, 10−7 mol/l, 1−5 mol/l dexamethasone. VEGF was measured by ELISA (R&D Systems, Minneapolis, Minn., USA), which recognizes mouse VEGF 164, the predominant isoform (cross-reactivity >95%). The detection limit of the assay was 3 pg/ml VEGF. Intra-assay coefficient of variation was 4.7%, and the inter-assay coefficient of variation was 6.4%.

Statistical methods

All experiments were performed in triplicate, and evaluated by ANOVA for statistical significance. The data are presented as the mean±SD at a significance level of P<0.05.

Results

Cells treated with increasing concentrations of dexamethasone, from 10−9 to 10−5 mol/l, accumulated lipid vesicles that were small initially and increased in size with time. The lipid vesicles were clearly distinguishable from the surrounding cells by phase contrast microscopy (Fig. 1) and with Sudan IV, a stain that is characteristic of neutral lipid in fat cells. Adipogenic changes were not found in D1 cells that were not treated with dexamethasone.

Fig. 1
figure 1

Phase contrast micrographs of D1 cells in culture showing adipogenesis when the cells are treated with dexamethasone. A D1 cells; B after treatment with 10−7 mol/l dexamethasone showing accumulation of lipid vesicles; C and D D1 cells treated with 10−9 mol/l (C) and 10−7 mol/l (D) dexamethasone for 7 days and stained with Sudan IV (original magnification ×200)

Dexamethasone increases PPARγ2 mRNA

To determine the effect of different concentrations of dexamethasone on PPARγ2 mRNA, D1 cells were incubated for 48 h with dexamethasone concentrations from 10−9 to 10−5 mol/l. Gene expression was quantified by RT-PCR analysis and is presented graphically in Fig. 2A. Within the range of dexamethasone concentrations used in our study, PPARγ2 mRNA levels increased to between 121% and 235% of controls (Fig. 2A). The increase was significant from 10−8 mol/l to 10−5 mol/l dexamethasone treatment (P<0.05), while there was no significant PPARγ2 mRNA expression increase with 10−9 mol/l dexamethasone treatment (P>0.05). To examine the time course effect, RNA was prepared from D1 cells after culture in the presence of 10−7 mol/l dexamethasone for 12 h and 1, 2, and 4 days, and compared with non-treated control cells at the same time points. Analysis by RT-PCR (Fig. 2B) showed a significant increase in PPARγ2 mRNA at 12 h (P<0.05), and further increases through the final time point of 4 days. At this time, PPARγ2 mRNA had increased to approximately 190% of control levels (P<0.05).

Fig. 2
figure 2

PPARγ2 up-regulation by dexamethasone is concentration and time dependent. Quantitation of the relative PPARγ2 gene expression of the D1 cells treated with different concentrations of dexamethasone (A) for 48 h and 10−7 mol/l dexamethasone for different time points (B). Graphs depict quantitative data from densitometric scanning of the gel stained with ethidium bromide. Values are based on band density relative to internal control 18S ribosomal RNA band density and plotted as a percentage of the control (zero dexamethasone). Error bars represent the standard deviation of triplicate experiments. Significance is denoted by an asterisk (*) compared with control

Dexamethasone down-regulates Cbfa1 mRNA

To determine Cbfa1/Runx2 mRNA in response to dexamethasone, D1 cells were incubated for 48 h with concentrations of 10−9 to 10−5 mol/l dexamethasone and analyzed by RT-PCR (Fig. 3A). With 10−9 mol/l dexamethasone, the level of Cbfa1/Runx2 mRNA was essentially the same as in non-treated control cells (P>0.05), while treatment with 10−8 mol/l dexamethasone decreased Cbfa1/Runx2 mRNA significantly to around 85% of control (P<0.05). At 10−7 to 10−5 mol/l dexamethasone treatment, expression of Cbfa1/Runx2 mRNA was approximately 50% of the untreated control cells (P<0.05) (Fig. 3A). To examine the time course of down-regulation of Cbfa1/Runx2 mRNA, RNA was prepared from D1 cells after culture in the presence of 10−7 mol/l dexamethasone for 12 h and 1, 2, and 4 days, and from non-treated control cells at the same time points. Analysis by RT-PCR showed that a small decrease in Cbfa1/Runx2 mRNA (to 75% of the control cells) at 12 h (P<0.05), and a further decrease through the final time point of 4 days to 47% of control levels (Fig. 3B) (P<0.05).

Fig. 3
figure 3

Concentration and time response of Cbfa1/Runx2 mRNA down-regulation by dexamethasone. Quantitation of the relative Cbfa1/Runx2 gene expression of the D1 cells treated with different concentrations of dexamethasone (A) for 48 h and 10−7 mol/l dexamethasone for different time points (B). Graph depicting quantitative data from densitometric scanning of the gel stained with ethidium bromide. Values are based on band density relative to internal control 18S ribosomal RNA band density and plotted as a percentage of the control (zero dexamethasone). Error bars represent the standard deviation of triplicate experiments. Significance is denoted by an asterisk (*) compared with control

Dexamethasone decreases activity of the osteocalcin promoter

To characterize the inhibitory effect on osteoblastic transactivation potential of dexamethasone, we studied the effect of dexamethasone on osteocalcin promoter activity. Treatment of cells with 10−7 mol/l dexamethasone resulted in a significant decrease in osteocalcin promoter activity (P<0.05) at 1–3 days (Table 1).

Table 1 Effect of dexamethasone on osteocalcin promoter activity. Values are expressed as the ratio of luciferase activity to β-Gal activity (mean±SD) and represent results from two separate experiments. Each sample was analyzed in triplicate

Dexamethasone decreases VEGF

To investigate the effect of dexamethasone on VEGF gene expression, D1 cells were treated with dexamethasone concentrations from 10−9 to 10—5 mol/l. Cell culture supernatants were collected at 12, 24, and 48 h for the time-course study and at 24 h for the comparison of the effect of different concentrations of dexamethasone. A concentration-dependent inhibition effect was detected after 24 h of treatment with dexamethasone (Fig. 4A). VEGF decreased by 30% within 24 h of exposing the cells to dexamethasone and by 55% after 48 h (Fig. 4B).

Fig. 4
figure 4

Inhibition of VEGF production of D1 cells by dexamethasone is dose and time dependent. VEGF expression in supernatant medium of the D1 cells treated with different concentrations of dexamethasone (A) for 24 h and 10−7 mol/l dexamethasone for different time points (B) was assessed by ELISA. Absorbance at 405 nm was measured. Each column shows mean±SD of data from three experiments

Discussion

The increased use of steroids for immunosuppression after organ transplantation, as treatment for rheumatoid diseases and severe adult respiratory syndrome (SARS), and for chemotherapy, has resulted in an increased risk of osteonecrosis [33]. Glucocorticoids have been the focus of studies on the pathogenesis of osteonecrosis. Although statistical data show that steroids may be implicated in one-third of all cases of osteonecrosis [34], the precise mechanism of action of the steroids on cells has not been determined [19,20] to our knowledge.

We have demonstrated previously that the multipotential bone marrow cell line D1 responds to treatment with dexamethasone by differentiating into fat cells and by expressing an adipose-specific gene, 422(aP2), that increases with higher doses and prolonged treatment. There are also reports that glucocorticoids affect osteoblasts by down-regulating type-I collagen and osteocalcin gene expression [35] and inducing osteoblast and osteocyte apopotosis [36,37]. Dexamethasone has also been reported to promote the expression of markers of the osteoblast phenotype in cultures of rat marrow stromal fibroblasts [38]. Glucocorticoids can have various effects on bone cells, depending on the state of cellular differentiation, animal species, and the dose and duration of treatment [39]. The effect on human marrow cells and mouse cells is to inhibit matrix synthesis. Consistent with other reports [16,40,41], our results demonstrate that dexamethasone down-regulates the expression of Cbfa1/Runx2 and osteocalcin promoter activity while it increases the expression of PPARγ2. The previously unexplained effects of glucocorticoids on bone loss may result from down-regulation of osteoblast transcription factor expression with concomitant up-regulation of the adipocyte transcription factor, leading to the differentiation of bone marrow stromal cells along the adipocytic lineage. Since adipocytes and osteoblasts share a common pool of progenitor cells, when exogenous stimulators shift the differentiation of marrow stem cells into the adipocyte lineage, the osteoprogenitor cell pool may not be sufficient to provide enough osteoblasts in order to meet the need for bone remodeling, fracture healing, or repair of necrotic bone. In addition, marrow stem cells that have already differentiated into osteoblasts would develop apoptotic changes [36,42] on treatment with glucocorticoids, accounting for the decline in bone formation and trabecular width. Thus, decreased production of osteoblasts and accumulation of apoptotic osteocytes may contribute to osteonecrosis eventually. Furthermore, the increase in volume of fatty marrow with a concomitant increase in intraosseous pressure could decrease blood flow in a semi-intact osseous compartment, eventually leading to ischemia and osteonecrosis [18].

Blood vessels are present in each bone multicellular unit (BMU). The vessels are located in the center of the BMU between the osteoclasts advancing at the end of the cavity and the osteoblasts lining the walls farther behind. Since the capillaries in bone grow at the same rate as the BMU advances, angiogenesis, the formation of new vessels from pre-existing capillaries, might be an integral process during bone remodeling and the sequential expression of peptides during angiogenesis might have a strong impact on the mechanism of coupling of bone resorption and bone formation.

VEGF is a key regulator of vasculogenesis in the embryo and angiogenesis in adult tissues. Impaired VEGF expression has been reported in some diseases that are related to angiogenic disorders such as diabetic retinopathy, fibrovascular disease and tumor growth. An important physiological role for VEGF in bone was reported by Gerber et al. [29], as blocking VEGF-mediated capillary invasion inhibited both bone formation and resorption in the growth plate. More recent evidence indicates that expression of VEGF is important for normal endochondral bone development, not only to mediate bone vascularization but also to allow normal differentiation of hypertrophic chondrocytes, osteoblasts and osteoclasts [24,37,43]. VEGF is expressed in the normal rat tibia. Additionally, VEGF expression in osteoblasts and osteoblast-like cells is increased by several cytokines and growth factors, including prostaglandin E1 (PGE1) and PGE2, insulin-like growth factor (IGF), platelet-derived growth factor, and 1,2,3-dihydroxyvitamin D3 [44,45,46]. The VEGF gene has been reported to have 5’ AP-1 binding sites. AP-1, a dimeric transcription factor composed of the Fos and Jun proteins, has been shown to be potently inhibited by dexamethasone [47]. Down-regulation of VEGF expression by dexamethasone has previously been demonstrated in rat glioma cells [48], human vascular smooth muscle cells [49], mouse and rat pituitary folliculostellate cells [50] and porcine brain endothelial cells [51]. Dexamethasone has been shown to block prostaglandin stimulation of VEGF production in osteoblastic cells [44]. In vitro, dexamethasone produced a dose-dependent inhibition of TGF-β1-induced VEGF protein production and this observation may provide an additional explanation for the well-documented phenomenon of both impaired fracture vascularization and healing in glucocorticoid-treated patients.

In this study, we have demonstrated that the decrease in VEGF upon treatment with dexamethasone is both dose and time dependent. Down-regulation of VEGF by glucocorticoids is directly responsible for disturbed angiogenesis, resulting in the observed defects in capillary architecture, which eventually lead to osteonecrosis. It is suggested that the mechanism of steroid-induced necrosis is through the inhibition of angiogenesis by suppressing VEGF.

In conclusion, there may be at least two mechanisms that are involved in the pathogenesis of osteonecrosis. First, glucocorticoids down-regulate osteoblast transcription factor gene expression and, by up-regulating adipocyte transcription factor gene expression, could lead to the differentiation of bone marrow stromal cells along the adipocytic lineage. Second, glucocorticoids decrease angiogenesis by suppressing VEGF. Consequently, the action of glucocorticoids may impair mesenchymal cell differentiation and thereby decrease blood supply leading to bone cell death.