Human Trabecular Bone Cells Are Able to Express Both Osteoblastic and Adipocytic Phenotype: Implications for Osteopenic Disorders


  • Presented in part at the 17th Annual Meeting of American Society for Bone and Mineral Research in Baltimore, MD, U.S.A., 1995.


The decrease in bone volume associated with osteoporosis and age-related osteopenia is accompanied by increased marrow adipose tissue formation. Reversal of this process may provide a novel therapeutic approach for osteopenic disorders. We have shown that cells cultured from human trabecular bone are not only osteogenic, but are able also to undergo adipocyte differentiation under defined culture conditions. Osteoblast differentiation was induced by 1,25-dihydroxyvitamin D3 (1,25(OH)2D3) and adipocyte differentiation by dexamethasone (dex) plus 3-isobutyl-1-methylxanthine (IBMX) treatment. Adipogenesis was characterized by lineage-specific enzyme and gene activities, α-glycerophosphate-3-dehydrogenase activity, fatty acid binding protein, aP2 and lipoprotein lipase expression. Osteoblastogenesis was assessed by osteoblast characteristic 1,25(OH)2D3 induction of alkaline phosphatase activity and osteoblast-specific 1,25(OH)2D3-induced osteocalcin synthesis and release. We provide evidence for a common pluripotent mesenchymal stem cell that is able either to undergo adipogenesis or osteoblastogenesis, using clonal cell lines derived from human trabecular bone cell cultures. Adipogenesis can be induced also by long chain fatty acids and the thiazolidinedione troglitazone. Dex plus IBMX-induced adipogenesis can be inhibited by interleukin-1β, tumor necrosis factor-α, and transforming growth factor-β. Interestingly, and in contrast to extramedullary adipocyte differentiation as shown by mouse 3T3L-1 and a human liposarcoma SW872 cell line, trabecular bone adipogenesis was unaffected by insulin. Also, the formation of fully differentiated adipocytes from trabecular bone cells after troglitazone treatment and long chain fatty acids was dependent on increased expression of the nuclear hormone receptor peroxisome proliferator-activated receptor γ2 caused by dex plus IBMX. Specific inhibition of marrow adipogenesis and promotion of osteoblastogenesis of a common precursor cell may provide a novel therapeutic approach to the treatment of osteopenic disorders.


Bone marrow stromal cells are morphologically heterogeneous and include fibroblasts, osteogenic cells, and adipocytes. There is evidence to suggest that these cells are derived from a multipotential stem cell in the marrow that gives rise to progenitors of each mesenchymal cell type.(1–7)

The decrease in bone volume associated with osteoporosis and age-related osteopenia is accompanied by an increase in marrow adipose tissue.(8,9) Indeed an increase in marrow adipocytes is observed in all conditions that lead to bone loss, such as ovariectomy(10,11) immobilization,(12) or treatment with glucocorticoids.(13)

One mechanism that could account for this decrease in bone volume, and hence mechanical strength, is an imbalance in the differentiation of progenitors of the osteoblast and adipocyte lineages at the expense of the osteoblast. Indeed early histomorphometric observations(8) suggested that a change in bone cell dynamics causing osteoporosis is the consequence of the adipose replacement of the marrow functional cell population. With age, the number and size of marrow adipocytes increases in a linear manner.(14) In a mouse model of accelerated aging, the SAMP 6 strain, osteoblastogenesis is decreased and there is a concordant increase in the number of marrow adipocytes.(15) Previous studies have provided evidence that cultures of bone marrow stromal cells are capable of both osteogenic and adipocytic differentiation.(16–20) While these reports have all utilized marrow-derived stromal cells, we report here the first direct evidence of bipotentiality of human adult trabecular-derived bone cells.

It is well established that cultured explants of human trabecular bone provide a simple means of obtaining large numbers of human cells which express reproducibly an osteoblast phenotype.(21–25) These cells have been shown to produce type I collagen and to express high levels of osteoblast-specific osteocalcin and the bone/kidney/placenta form of alkaline phosphatase, which are further elevated in response to 1,25-dihydroxyvitamin D3 (1,25(OH)2D3).(26–29) In addition, these cells have the capacity to mineralize and produce all the major noncollagenous proteins of the extracellular bone matrix.(22,30,31) Diffusion chambers, which prevent mixing of cells containing human osteoblast-like cell (hOB) preparations, form bone when implanted in vivo.(32)

Adipocyte differentiation is accompanied not only by gross changes in cellular morphology, but also by the transcriptional activation of many genes. Several of these gene products are known to be involved in triglyceride synthesis and include the early marker of adipocyte conversion lipoprotein lipase (LPL) as well as markers associated with the late and very late stages of adipocyte differentiation, such as glycerol-3-phosphate dehydrogenase (G3PDH)(33) and the fatty acid binding protein aP2.(34,35) The differentiation of adipocytes in vitro is initiated by a number of factors including insulin, growth hormone, and glucocorticoids(36–38) and has been shown to be inhibited by interleukin-1β (IL-1β), retinoic acid, transforming growth factor β (TGF-β), and tumor necrosis factor-α (TNF-α).(39,40) In recent years, many studies have focused on the regulation of the expression of these adipocyte-specific genes, and several candidate transcription factors have been identified. Recently, a member of the nuclear receptor superfamily of ligand-activated transcription factors identified as peroxisome proliferator-activated receptor γ2 (PPARγ2)(41) has been shown to be expressed early in the adipocyte differentiation program.(42) It acts synergistically with CCAATT enhancer-binding protein α(43) to coordinate the adipocyte differentiation cascade.(44,45) Importantly, the peroxisome proliferator response element has been identified in the promotor regions of several enzymes associated with triglyceride metabolism.(41,46) PPARγ2, in common with other PPARs, is activated by long chain fatty acids,(47,41) peroxisome proliferators,(48,41) and the thiazolidinedione class of antidiabetic agents.(49)

In this study, we report that osteogenic cells derived from explants of adult human trabecular bone are capable of differentiation to adipocytes. We also describe the regulation of this differentiation by a range of physiological and pharmacological agents and correlate this with induction of osteoblast and adipocyte activities. This, therefore, clearly demonstrates the reciprocal relationship between osteoblast and adipocyte differentiation and has important implications in understanding osteopenic disorders.


Cell culture and materials

All chemicals were purchased from Sigma (St. Louis, MO, U.S.A.) unless otherwise stated. Explants were isolated and cultured from trabecular bone fragments from knee joints as described previously.(22) Briefly, specimens of trabecular bone taken at surgery were dissected into 0.3–0.5 cm diameter pieces and washed extensively in phosphate-buffered saline (PBS), pH 7.4, supplemented with penicillin (10 U/ml), streptomycin (100 mg/ml; Gibco BRL, Life Technologies, Grand Island, NY, U.S.A.), and amphotericin (250 mg/ml). The fragments were seeded as explants into 80 cm2 culture dishes and cultured at 37°C in a humidified atmosphere of 95% air–5% CO2. Unless otherwise stated, culture medium was Eagle's modified minimal essential medium containing 10% heat inactivated fetal calf serum (FCS; Gibco BRL) supplemented with penicillin, streptomycin, and amphotericin as above and 2 mM glutamine. The medium was replaced 14 days later and thereafter at 7-day intervals. At confluence (4–6 weeks), cells were subcultured after exposure to trypsin-EDTA. Mouse embryo 3T3-L1 and human liposarcoma SW872 cell lines were obtained from American Type Culture Collection, (Rockville, MD, U.S.A.) and grown routinely under the same conditions as the hOBs described above.

hOBs were then cultured in medium as described above with the exception that for all differentiation studies, all the cell lines were grown in medium supplemented with 10% heat-inactivated, charcoal-stripped, FCS (Hyclone, Logan, UT, U.S.A.). hOBs were treated with either 10 nM 1,25(OH)2D3 (Biomol Research Laboratories, Plymouth Meeting, PA, U.S.A.) for 48 h, or 1 μM dexamethasone plus 100 μg/ml 3-isobutyl-1-methylxanthine (IBMX) for 2–3 weeks until >30% of the cells had acquired an adipocyte morphology.(50) Medium containing 1 μM dexamethasone (dex) and 100 μg/ml IBMX is termed adipocyte differentiation medium. In all experiments, cells were cultured continuously in the presence of adipogenic agonists added freshly every 3 days. α-bromopalmitate (Fluka Chemical Corp., Ronkonkoma, NY, U.S.A.) was dissolved in ethanol and complexed to fatty acid–free bovine serum albumin (BSA) in a 6:1 molar ratio of fatty acid to BSA before being diluted in culture medium and added to the cells at a final concentration of 25 μM. The thiazolidenedione, troglitazone (CS-045) ((+/−5-[4-(6-hydroxy-2,5,7,8-tetramethylchroman-2-ylmethoxy)benzyl]-2,4-thiazolidinedione; Product Development Laboratories of Sankyo Co. Ltd., Tokyo, Japan) was added at a final concentration of 5 μM. Cytokines TGF-β, recombinant human (rh) IL-1β, and rhTNF-α (Genzyme, Boston, MA, U.S.A.) were added directly to the differentiation medium at a final concentration of 100 ng/ml.

Determination of bipotentiality of single-cell clones

hOBs were plated at 0.33 cells/well in a 96-well plate and grown for up to 3 weeks or until colonies appeared in a subset of wells. The cloning efficiency of the hOBs was typically 20% (20 out of 96 wells). These cells were then replated at the same seeding density and grown for up to 3 weeks or until colonies formed. The colonies were treated with 1,25(OH)2D3 (10 nM) for 48 h, and media was removed for osteocalcin measurement by enzyme-linked immunosorbent assay (ELISA) as described below. The cells were then treated with 1 μM dex plus 100 μg/ml IBMX for up to 2 weeks or until microscopically apparent triglyceride accumulation was detected and the colonies were stained with oil Red O.


Cell cultures were washed briefly in PBS and fixed in 60% isopropanol for 1 minute. Cells were stained with oil Red O for 30 minutes, rinsed briefly in 60% isopropanol, and counterstained with hematoxylin.(51)

Alkaline phosphatase assay

Cell lysates were obtained by solubilization of the monolayer of cells with 0.1% (v/v) Triton X-100. ALP activity was determined by colorimetry using p-nitrophenylphosphate as the substrate.(27) Protein was determined using the bicinchoninic acid assay (Pierce, Rockford, IL, U.S.A.). Results are expressed as units per microgram of protein/10 minutes. One unit is defined as the amount of enzyme that will produce 1 nmol of p-nitrophenol/minute.

Determination of osteocalcin production

hOBs were grown to confluence in culture medium and then switched to culture medium containing 10% charcoal stripped FCS and 10−8 M 1,25(OH)2D3 for 48 h. The medium was then collected and stored at −20°C until assay for osteocalcin. Osteocalcin released into the culture medium was measured using an intact human osteocalcin ELISA(52) (Biomedical Technologies, Inc., Stoughton, MA, U.S.A.).

α-glycerophosphate-3-dehydrogenase activity

G3PDH activity was determined using the method of Wise and Green,(33) with modifications. Briefly, cells were washed twice with ice-cold PBS and then scraped into lysis solution (50 mM Tris, pH 7.5, 1 mM EDTA, 5 mM DTT). Cell extracts were prepared by sonication on ice. G3PDH activity was measured in the cell extract by continuous spectrophotometry at 340 nm. Results are expressed as units per milligram of protein. One unit of enzyme activity corresponds to the oxidation of 1 nmol of nicotinamide adenine dinucleotide (NADH)/minute.

Isolation and analysis of RNA

Total cellular RNA was prepared from confluent cell cultures using TRI-Reagent.(53) Total RNA (5–10 μg/sample) was denatured in formamide and formaldehyde and separated by electrophoresis through formaldehyde-containing agarose gels according to the method of Ausubel.(54) Following electrophoresis, the RNA was transferred to Duralon-ultraviolet membrane (Stratagene, La Jolla, CA, U.S.A.) and cross-linked to the membrane by ultraviolet exposure. Northern blots were hybridized and washed according to the membrane manufacturer's instructions under high stringency conditions.

cDNA probes

cDNA probes were labeled with [α32-P]dCTP by the random primer method(55) using a commercially available kit (Pharmacia, Piscataway, NJ, U.S.A.). A plasmid containing human aP2 was provided courtesy of Dr. D.A. Bernlohr (University of Minnesota). The plasmids containing human glyceraldehyde phosphate dehydrogenase (GAPDH) and β-actin were obtained from ATCC. PPARγ2 cDNA was from +103 to +534 of the coding region of an expression sequence tag (EST) from a human placenta library. The probe for lipoprotein lipase (LPL) was obtained by reverse transcribed polymerase chain reaction of human adipose RNA using commercially available primers amplified from a human placenta library (Perkin Elmer, Branchburg, NJ, U.S.A.).


Expression of osteoblast markers

hOBs exhibited a high basal level of ALP activity that was induced further, following culture in the presence of 1,25(OH)2D3 (Table 1). Osteocalcin was undetectable (<0.78 ng/ml) under basal conditions but was induced following culture in the presence of 1,25(OH)2D3. Stimulation of ALP and osteocalcin production in human bone cells in vitro by 1,25(OH)2D3 has been described previously, and these proteins serve as useful markers in the identification of osteoblasts.(56)

Table Table 1. Osteoblast Characteristics of Human Trabecular Bone (hOB) Cells
original image

Expression of adipocyte markers: Response to adipogenic agonists

hOBs grown in culture medium displayed a fibroblastic morphology typical of these cells (Fig. 1A). However, when cultured in the presence of dex plus IBMX, the cells became rounded. In some cells, small perinuclear granules appeared, which over a 2–3 week period fused to form highly refractile vacuoles that stained positive for neutral lipids by oil Red O, indicating an adipogenic phenotype (Fig. 1B). Adipogenesis occurred in a maximum of 30–40% of cells in discrete clusters separated by fibroblast-like cells. Under higher magnification, prior to staining, microscopic accumulation of fat droplets could be observed (data not shown). In the presence of dex alone the cells became more rounded and only very occasionally did they acquire any morphological characteristics of adipocytes over the 2–3 week period (data not shown).

Figure FIG. 1.

hOBs were grown as described in the Materials and Methods in normal medium (A) or in medium supplemented with 1 μM dex plus 100 μg/ml 3-isobutyl-1-methylxanthine (B) for 21 days and stained with oil Red O.(51) Original magnification ×20.

Expression of several key stage-specific markers of adipocyte differentiation were examined by enzymatic assays and Northern blotting. In other cell systems, we have found that culturing cells in the presence of increased lipid content, i.e., rabbit serum, results in positive oil Red O staining due to passive lipid uptake, and this is not accompanied by adipogenic enzyme and gene activities (unpublished observations). G3PDH activity was increased in cells treated with dex in the presence or absence of IBMX, although the increase in activity was significantly greater in those cells treated with IBMX (Fig. 2). Following adipocyte differentiation, the expression of mRNA for the adipocyte associated genes, aP2 and LPL were detected only in those cells treated with dex plus IBMX (Fig. 3A). This was consistent for multiple hOB preparations from different patient samples (unpublished observations). LPL expression was induced in all (n = 5) and ap2 in four out of five patient hOB samples by cotreatment with IBMX plus dex only.

Figure FIG. 2.

hOBs were grown as in Fig. 1, in the presence or absence of 1 μM dex or 1 μM dex plus 100 μg/ml IBMX, for 21 days, as described in the Materials and Methods. α-glycerophosphate dehydrogenase activity was determined as described in the Materials and Methods and the results expressed as the mean ± SD (n = 3). Significance was evaluated in an unpaired Student's t-test (*p < 0.0005). One unit of enzyme activity corresponds to the oxidation of 1 nmol of NADH per minute.

Figure FIG. 3.

hOBs were grown for 21 days either in normal FBS-containing medium or medium supplemented with dex (1 μM) or dex (1 μM) plus IBMX (100 μg/ml). (A) RNA was extracted, and Northern blots were run as described in Materials and Methods with RNA from normal FBS-containing medium (lane 1), dex plus IBMX (lane 2), and dex alone (lane 3). Probes for LPL, aP2, and GAPDH were labeled, and hybridization was as described in the Materials and Methods. GAPDH was used to correct for loading efficiencies. (B) RNA was prepared as above, and Northern Blots were probed for β-actin. Ribosomal RNA were stained with bromophenol blue to correct for loading efficiencies.

The increase in adipocyte characteristic genes was accompanied also by a decrease in the expression of β-actin mRNA (Fig. 3B). This is consistent with adipogenesis studies reported in the 3T3-L1 preadipocyte cell line.(57,58) There was a decrease in β-actin expression only in the cells that had been cotreated with dex plus IBMX.

IBMX and dex treatment after both 3 and 10 days resulted in a stimulation of expression of a member of the steroid hormone receptor superfamily, the PPARγ2 (Fig. 4). This is consistent with models of extramedullary fat differentiation.(41) The up-regulation of PPARγ2 preceded the positive triglyceride staining, which under these conditions takes 14–21 days, suggesting that this is indeed an early event in the differentiation process.(41)

Figure FIG. 4.

hOBs were grown for 3 and 10 days, and RNA was isolated and Northern blots were prepared as described in Materials and Methods. Lanes 1 and 3 represent hOBs grown for 3 and 10 days, respectively. Lanes 2 and 4 represent hOBs grown in normal medium supplemented with 1 μM dex plus 100 μg/ml IBMX. Blots were probed for PPARγ2 expression. GAPDH was probed to correct for loading efficiencies.

Bipotentiality of single-cell clones

Since cells from trabecular bone are a heterogeneous mixture, it was important to determine whether different cells in the cultures were responding to osteogenic and adipogenic signals, or whether the same cell could give rise to both phenotypes. A protocol was designed to generate colonies derived from single cells by limiting dilution and then sequential treatment with 1,25(OH)2D3, followed by exposure to adipogenic agonists (see Methods). Osteocalcin could not be measured in the media under basal conditions, but after 1,25(OH)2D3 treatment for 72 h, all of the 20 colonies isolated secreted osteocalcin (Table 2). The values ranged from 4.3–20.2 ng/ml (mean 10.7 ± 4.0 ng/ml, n = 20). All 20 clones reported were from the same donor, but preparations of cells from other donors behaved similarly. All the colonies that released osteocalcin also accumulated triglyceride and stained positively for oil Red O when treated with dex plus IBMX (Figs. 5A and 5B). In all clones, the proportion of cells that stained positively for oil Red O was similar (>90%). Pretreatment was not a prerequisite for the clones to subsequently undergo adipogenesis (data not shown). After dex plus IBMX treatment, the medium contained very low levels (≤1.0 ng/ml) of osteocalcin, indicating not only that there was an induction of adipogenesis but also an inhibition of osteoblastogenesis. The inhibition of osteoblastogenesis was inferred from the very low level of osteocalcin secretion. This could be a direct effect of cAMP and glucocorticoids on osteocalcin secretion. This indicates at least bipotentiality of a subset of these cells to release osteoblast-specific osteocalcin after 1,25(OH)2D3 and accumulate triglyceride after dex plus IBMX treatment.

Figure FIG. 5.

Isolated colonies were grown up as described in the Materials and Methods, and these were then considered individual cell lines. (A) Each cell line was treated for 72 h with 1,25(OH)2D3, and osteocalcin was measured in the media. All the cell lines exhibited a fibroblastic appearance. (B) Colonies were subsequently treated with the adipogenesis-inducing conditions of dex (1 μM) plus IBMX (100 μg/ml) and stained for oil Red O as described by Chayen and Bitensky.(51) (A and B) show one of the cell lines (No. 8; see Table 2). Clone 8 was selected for presentation because it is a representative clone. Adipogenic conditions caused a change from the normal fibroblastic morphology to rounding of the cells in every cell line tested (B).

Table Table 2. Osteocalcin Secretion by Clonal Cell Lines of Human Trabecular Bone (hOB) Cells
original image

Regulation of adipogenesis

Previous studies have shown that some cytokines regulate key enzymes in adipocytes, resulting in a decrease in fat storage. Consistent with these studies, adipogenesis was inhibited in bone cells when cultured with dex plus IBMX in the presence of TGF-β (Figs. 6A and 6B). Adipogenesis was also inhibited when hOBs were grown in the presence of dex plus IBMX either with IL-1β or TNF-α (results not shown). TGF-β in the presence of dex plus IBMX did not induce osteocalcin secretion, suggesting simply a block of adipogenesis and not a promotion of osteoblastogenesis.

Figure FIG. 6.

hOBs were grown, as described previously, for 21 days and stained for triglycerides by oil Red O. (A) Medium was supplemented with adipogenesis differentiation medium (dex plus IBMX). (B) shows cells grown in the adipogenesis medium supplemented with TGF-β (100 ng/ml). (C) shows hOB cells grown in normal medium, and (D) shows cells grown in normal medium supplemented with insulin (1 μM). (E) shows 3T3-L1 cells grown for 8 days in normal medium, and (F) in medium supplemented with insulin (1 μM) and the cells were stained with oil Red O. (G) shows SW872 cells grown in normal medium for 8 days, and (H) shows cells supplemented with insulin (1 μM) and stained with oil Red O.

Extramedullary adipocyte differentiation is induced by insulin; however, in the hOBs culture medium it did not appear to cause adipogenesis (Figs. 6C and 6D). The presence of insulin did not potentiate the adipogenesis induced by dex plus IBMX alone, as indicated by the same number of oil Red O–stained cells and the same extent of triglyceride staining per cell (data not shown). This is in contrast to examples of extramedullary adipocyte differentiation, the mouse embryo 3T3-L1 cells, and the human liposarcoma SW872 cells that when grown in medium supplemented with insulin alone underwent extensive adipocyte differentiation (Figs. 6E, 6F, 6G, 6H).

Effect of fatty acids on adipogenesis

Previous studies have shown that long chain fatty acids promote adipogenesis of preadipocyte cell lines by induction of the expression of aP2 and acyl-CoA synthetase.(59–61) Because this enhancement of adipogenesis may be due to an increased concentration of fatty acids or fatty acid metabolites in the medium, the effect of the nonmetabolized fatty acid α-bromopalmitate was investigated.

Confluent cultures of hOBs were treated with α-bromopalmitate (25 μM)(62) in the presence or absence of dex plus IBMX for 8 days. This time period is not long enough for the adipocyte differentiation medium (dex plus IBMX) alone to cause the cells to accumulate triglyceride (Fig. 7A). However, inclusion of α-bromopalmitate in adipocyte differentiation medium resulted in the appearance of lipid vacuoles in a proportion of the cells (Fig. 7B). α-bromopalmitate alone was unable to induce adipogenesis (data not shown).

Figure FIG. 7.

hOBs were grown for 8 days in (A) adipocyte differentiation medium (dex plus IBMX) or (B) adipocyte differentiation medium supplemented with α-bromopalmitate (25 μM). (C) shows hOBs grown for 8 days in adipocyte differentiation medium and (D) the same medium supplemented with CS-045 (5 μM). Original magnification ×20.

Effects of troglitazone on adipogenesis

The new class of antidiabetic agents, thiazolidinediones, that improve insulin sensitivity in models of non–insulin-dependent diabetes mellitus have been shown to have dramatic effects on in vitro adipocyte differentiation.(63–65) To investigate the effect of troglitazone on the differentiation of hOBs, cells were treated for 8 days either in the presence of differentiation medium alone (Fig. 7C) or supplemented with troglitazone; CS-045 (5 μM; Fig. 7D). Troglitazone alone was unable to cause adipocyte differentiation of hOBs (data not shown), but in combination with dex and IBMX potentiated adipogenesis when compared with cells treated with dex plus IBMX in the absence of compound (Figs. 7C and 7D).

Effect of α-bromopalmitate and troglitazone on gene expression in hOBs

RNA from hOBs treated with α-bromopalmitate and troglitazone for 8 days in the presence or absence of dex plus IBMX was analyzed by Northern blot. Cells treated with troglitazone alone showed no increase in expression of aP2, LPL or PPARγ2 (Figs. 8A and 8B). α-bromopalmitate treatment resulted in induction of LPL expression (Fig. 8A). Treatment of hOBs with dex plus IBMX caused an increased expression of LPL and PPARγ2. In cells treated with troglitazone, in addition to dex plus IBMX, there was also an increase in the expression of aP2 (Fig. 8A). No increased expression of these adipocyte markers was observed when α-bromopalmitate was included with dexamethasone and IBMX (Figs. 8A and 8B).

Figure FIG. 8.

hOBs were grown for 21 days in normal medium supplemented with 10% heat-inactivated, charcoal-stripped FBS or medium supplemented with adipogenesis inducing medium (dex plus IBMX) either in the presence or absence of α-bromopalmitate or CS-045. RNA was extracted and Northern blots were performed as described in the Materials and Methods. (A) RNA was run from cells grown in normal medium (lane 1), medium supplemented with either 5 μM CS-045 (lane 2), or 25 μM α-bromopalmitate (lane 3), adipogenesis medium (dex + IBMX) (lane 4) supplemented with either 5 μM CS-045 (lane 5) or 25 μM α-bromopalmitate (lane 6). The blots were probed for aP2, LPL, or GAPDH as described previously. GAPDH was used as a probe to correct for loading efficiencies (A). The Northern blot was then probed for PPARγ2 expression (B).


The results demonstrate that human bone-derived cells, when cultured in the presence of dex plus IBMX, differentiate into adipocytes. Adipogenesis was confirmed by a variety of functional, morphological, and molecular markers. In the time frame used in this study, dex treatment alone was unable to promote significant adipocyte differentiation. Previous studies have shown the ability of glucocorticoids to promote adipogenesis in some preadipocytes(4,18,37,66,67) and a dramatic acceleration of the adipocyte differentiation process is seen following exposure to both glucocorticoid and IBMX.(4) Glucocorticoids and IBMX are thought to exert their effects on terminal differentiation of adipocytes by increasing intracellular concentrations of cAMP and Ca2+.(68)

We have provided further evidence of a trans-differentiation process, whereby a differentiated osteoblast, as indicated by osteocalcin release, will undergo adipogenic differentiation. This suggests that, even though osteocalcin is considered to be a late marker of a developing osteoblast in vitro,(69) the cells are able to revert to an adipogenic phenotype after treatment with dex plus IBMX and, therefore, full commitment to the osteoblast lineage is at a later check point than osteocalcin release. We favor the concept that plasticity exists among cells of the stromal lineage and this inter-relationship may have important consequences in terms of formation of fully functional differentiated cells, which may be critical in the progression of skeletal diseases. The excess adipogenesis in postmenopausal women may occur at the expense of osteogenesis and, therefore, be an important factor in the fragility of adult bone. It is not known at this time whether this, at least bipotential, cell exhibits greater multipotentiality and could be induced to differentiate to other distinct lineages, such as chondrocytes, reticular cells, or fibroblasts.

Although marrow adipocytes have not been studied to the same extent as extramedullary adipocytes, there appear to be distinct differences in the control of adipogenesis at the two sites.(70–72) Most important is the lack of dependency on the adipogenic agent, insulin,(4,20,66,73–75) which in contrast appears to be essential for differentiation of extramedullary adipocyte cell lines. Consistent with these studies, adipocyte differentiation of bone-derived cells was unaffected by insulin at concentrations (1 μM) shown to induce adipocyte differentiation in 3T3-L1 and SW872 cells, which are models of extramedullary adipogenesis. However, the observation that troglitazone induced aP2 expression and potentiated adipogenesis in the presence of dex plus IBMX suggests that there is a component of this differentiation pathway that is not dependent, but is enhanced, by the presence of insulin sensitizers.

Cytokines have been shown to play a key role in fat metabolism, leading to an increase in serum-free fatty acid levels.(76) Studies on the effect of cytokines on adipogenesis of 3T3-L1 cells and the bone marrow adipocyte cell line BSM2 have shown that they are both subjected to the same negative regulators.(72,77) Likewise, adipogenesis in human bone-derived cells was inhibited by TGF-β, IL-1β, and TNF-α. These are also in agreement with other studies in marrow-derived systems that have shown inhibition of adipocyte formation by IL-1β(78) and the bone morphogenetic proteins.(79) The inhibition of adipogenesis by these cytokines has important implications because a number of these growth factors, in particular the TGF superfamily, which include the bone morphogenetic proteins, has been shown to have effects on osteogenesis.(80) They could play an important role in the balance of this differentiation process if differences in their local concentrations occur as a result of menopause.

It is well established that fatty acids and thiazolidinediones act in vitro as potent inducers of adipocyte differentiation.(42,64,65,81) Their exact mode of action is unclear but is thought to involve members of the steroid hormone receptor superfamily, namely the PPARs. In this study, cells cultured in the presence of the nonmetabolized fatty acid α-bromopalmitate or the thiazolidinedione, troglitazone in combination with dex and IBMX displayed lipid vacuoles after 8 days. This was in contrast to cells treated with dex and IBMX, α-bromopalmitate, or troglitazone alone in which adipocytes were absent at this time point. Previous studies, however, have shown that α-bromopalmitate and thiazolidinediones alone are capable of causing differentiation of the preadipocyte cell line Ob1771 to adipocytes by increasing expression of the terminal markers of adipocyte differentiation such as aP2 and G3PDH. In contrast to hOBs, however, Ob1771 preadipocyte cells already express reasonably high levels of the early markers of adipocytes.(61) Indeed, troglitazone and α-bromopalmitate were able to potentiate adipogenesis only in combination with dex plus IBMX, which was shown to have caused an increase in the expression of the early adipocyte markers LPL and PPARγ2 during the 8-day treatment period.

We found consistently that troglitazone was a more potent inducer of hOBs adipogenesis than α-bromopalmitate, and this may be due to an increased induction of gene expression as demonstrated by the increase in aP2 mRNA in response to this compound. Interestingly, even though dex plus IBMX treatment stimulated PPARγ2 mRNA expression, neither α-bromopalmitate nor troglitazone affected expression levels, and this had previously been proposed to be the molecular target of thiazolidinediones.(49) This suggests that the effect of troglitazone and α-bromopalmitate in hOBs may be mediated by some other mechanism, possibly involving other members of the steroid hormone superfamily. This could represent a further difference between extramedullary and bone marrow adipocytes or could be due to species differences, as shown by previous studies that have involved cell lines of nonhuman origin. Indeed Murkherjee(82) has shown different responsiveness of rat and human PPARs to PPAR activators. Moreover, preliminary experiments in our laboratory have shown a difference in the response of rat and human bone marrow cells to thiazolidinediones (unpublished observations). The PPAR family has been shown previously to have many splice forms.(83) Human peripheral monocytes have been reported to express a message for PPARγ2 of 0.65 kb.(84) The transcript in adipocytes derived from trabecular bone had a transcript for PPARγ2 of approximately 2.1 kb, which is in agreement with the size of the murine adipocyte PPARγ2.

The precise role of adipocytes in bone marrow is unknown. It has been suggested that they act in a purely passive role filling marrow cavities not required for active hematopoiesis.(4,71) They may play a role in lymphohematopoiesis,(72,85) they may serve an active role in the energy metabolism of the resorbing osteoclast where fatty acid oxidation appears to be the major source of acetyl-CoA to support a predominantly oxidative metabolism,(86) or they may participate in the animal's overall metabolism by clearing and storing circulating triglycerides.(87) These roles may not be mutually exclusive and it is possible that the role of marrow adipocytes changes in response to age, menopause, or emergency situations affecting hematopoiesis or osteogenesis. Whatever the role of adipocytes in bone marrow, there is little doubt that adipogenesis increases as bone volume decreases, and this, together with the possibility that osteoblast and adipocyte differentiation pathways are coordinately regulated, suggests that marrow adipogenesis has important implications in osteogenic disorders.

The observation that adipocyte gene expression, i.e., aP2 and LPL was up-regulated by dex plus IBMX in hOB cultures from multiple patient samples indicates consistency of this differentiation phenomenon. This leads also to the possibility that this differentiated process may reflect patient type/severity from which the primary cells were prepared. There were no clear indications in this limited study as to whether extent or type of disease was linked to adipogenic potential of the osteoblast-like cultures, and this is currently being investigated.

We have shown that bone cells derived from explants of human trabecular bone represent a multipotential system, which may facilitate greatly studies of the changes in cellular composition of bone marrow associated with osteoporosis. Indeed the inhibition of marrow adipocyte differentiation and a concommitant enhancement of osteoblastogenesis may provide a novel therapeutic approach to the treatment of postmenopausal osteoporosis.


We thank Dr. Rothman of the Rothman Institute, Pennsylvania Hospital, for providing the bone samples.