SEARCH

SEARCH BY CITATION

Keywords:

  • Leukemia inhibitory factor;
  • Peroxisome proliferator-activated receptor γ;
  • Osteoblast;
  • Adipocyte;
  • Differentiation Single colonies

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Osteoblasts and adipocytes derive from a common mesenchymal precursor, and in at least some circumstances, differentiation along these two lineages is inversely related. For example, we have recently observed that concomitant with inhibition of osteoblast differentiation and bone nodule formation, leukemia inhibitory factor (LIF) induces genes regulating lipid metabolism in fetal rat calvaria (RC) cell cultures. In this study, we further investigated the adipogenic capacity of LIF-treated RC cells. Quantitative analyses revealed that LIF increased the adipocyte differentiation induced by the peroxisome proliferator-activated receptor γ agonist BRL49653 (BRL) in RC cell populations. Gene expression profiling of individual RC cell colonies in untreated cells or cells treated with LIF, BRL, or combined LIF-BRL suggested that some adipocytes arose from bipotential or other primitive precursors, including osteoprogenitors, since many colonies co-expressed osteoblast and adipocyte differentiation markers, whereas some arose from other cell pools, most likely committed preadipocytes present in the population. These analyses further suggested that LIF and BRL do not act at the same stages of the mesenchymal hierarchy, but rather that LIF modifies differentiation of precursor cells, whereas BRL acts later to favor adipocyte differentiation. Taken together, our data suggest that LIF increased adipocyte differentiation at least in part by altering the fate of osteoblastic cells and their precursors.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

A large body of experimental data indicates that multipotential mesenchymal stem cells differentiate into several lineages, including osteoblasts, adipocytes, myocytes, and chondrocytes [1, [2], [3], [4], [5], [6]7]. These terminally differentiated cell types are also thought to be related in a lineage hierarchy in which more primitive progenitor cells with multilineage capacity give rise to increasingly restricted progeny [3, 4]. Thus, for example, bipotential adipocyte-osteoblast progenitors would further differentiate into monopotential precursors, that is, committed osteoprogenitors giving rise to the postproliferative osteoblast and committed preadipocytes giving rise to adipocytes (reviewed in [8, 9]). It has also been suggested that the inverse relationship sometimes seen between expression of the osteoblast and adipocyte phenotypes in marrow stroma in osteoporosis [10, 11], in certain transgenic mice (e.g., overexpressing a dominant-negative form of N-cadherin in osteoblasts [12]), and in some culture manipulations [13, 14] may reflect the ability of single agents or combinations of agents to alter the commitment or at least the differentiation pathway these bipotential cells will transit [15].

A hallmark of adipocyte differentiation (reviewed in [16, [17]18]) is activation of the transcription factors CCAAT/enhancer-binding protein (C/EBP) β and δ [19, 20], which promote expression of the master adipocyte differentiation gene, nuclear receptor peroxisome proliferator-activated receptor γ2 (PPARγ2) [21, 22], which in turn stimulates expression of C/EBPα, with which it cooperates to promote expression of genes involved in lipid metabolism and storage [23]. Few endogenous PPARγ ligands are known, the prostaglandin (PG) derivative 15-deoxy-Δ(12–14)-PGJ2 being one of them [24]. However, several potent synthetic PPARγ agonists have been developed, including the thiazolidinediones, first known for their antidiabetic activity [25, 26], which in vitro promote the differentiation of fibroblast-like preadipocytes into mature adipocytes [25]. Osteoblast differentiation, on the other hand, involves at least two crucial transcription factors. The first, the runt domain-containing protein Runx2 (also known as core binding factor alpha 1, Cbfa1; osteoblast-specific factor-2, Osf2), is essential for osteoblast differentiation and bone formation in mice and humans [27, [28], [29]30]. The second essential transcription factor in osteoblastogenesis, Osterix, is a zinc-finger-containing protein [31] that acts downstream of Runx2 and is restricted in expression to cells of the osteoblast lineage [31, 32]. The human homologue of osterix has recently been cloned and mapped to chromosome 12q13.13; however, no skeletal abnormality has yet been associated with that locus [33, 34]. Activation of PPARγ2 has been shown to decrease osteoblastogenesis at the expense of adipogenesis, by a mechanism involving Runx2 repression [35]. However, depending on the ligand used, it is possible to uncouple the proadipogenic and antiosteoblastic activities of PPARγ2. Specifically, with BRL49653 or the natural ligands 9,10-dihydroxyoctadecenoic acid or 15-deoxy-Δ (12, 14)-PGJ2, adipogenesis was stimulated concomitantly with inhibition of osteoblastogenesis, whereas 9-hydroxyoctadecadienoic acid enhanced adipogenesis but had no effect on osteoblastogenesis [36]. Thus, osteoblast and adipocyte lineages can be independently regulated by factors modulating or cross-talking with PPARγ2 to influence these fate decisions.

A class of regulators active on both adipocyte and osteoblast differentiation is the interleukin-6 (IL-6)/leukemia inhibitory factor (LIF) family of cytokines. In addition to IL-6 and LIF, this family includes oncostatin M (OSM), IL-11, ciliary neurotrophic factor, cardiotrophin-1, cardiotrophin-like cytokine/novel neurotrophin 1/B cell-stimulating factor-3, and neuropoietin, all of which use glycoprotein 130 as the signaling component of their multimeric receptors [37, [38], [39], [40]41]. Use of a common signaling receptor subunit may explain the redundancy observed in at least some biological responses elicited by members of this cytokine family. For example, we have previously shown that IL-6, LIF, mouse OSM, and IL-11 all inhibit osteoblast differentiation under particular conditions in vitro [42].

Many reports suggest that IL-6/LIF cytokines affect adipocyte differentiation and function, although the reported effects are diverse. For example, IL-6/LIF family members inhibit lipid-metabolizing enzymes such as lipoprotein lipase (LPL) and glycerol-phosphate dehydrogenase (GPDH) in adipocyte cell lines [43, [44], [45], [46]47]. On the other hand, LIF stimulates GPDH activity in preadipocyte cell lines when they are pulse-treated during the first week postconfluence, and LIF acts synergistically with the PPARγ agonist BRL49653 to stimulate adipocyte differentiation of mouse embryonic fibroblasts [48]. Mouse ESCs deficient in the LIF receptor (lifr) gene and normal mouse ESCs treated with a LIFR antagonist have a reduced capacity to undergo adipocyte differentiation, likely reflecting a requirement for cytokines using this receptor component in the early steps of adipocyte progenitor differentiation [48].

The effects of LIF/IL-6 family members on osteoblasts and bone are also complex, with increased expression of family members after menopause [49] and parathyroid hormone treatment [50], consistent with a possible role in bone turnover and osteoporosis. Notably, inactivation of the LIFR gene in mice results in abnormalities characterized in part by bone bowing, decreased bone density, reduced bone spicules, and few well-formed trabeculae [51], phenotypic features typical of those in human Stüve-Wiedemann/Schwartz-Jampel type 2 syndrome [52], which has also recently been shown to be due to null mutations in LIFR. On the other hand, engraftment of LIF-overexpressing hematopoietic cells in mice increases bone thickness, leading to occlusion of the marrow cavity by osteosclerotic tissue [53, 54]. Mice overexpressing OSM display bone hypertrophy with immature osteoblasts, disrupted and disorganized growth plates, the formation of incomplete Haversian systems, and no sign of remodeling [55], whereas mice engrafted with OSM-overexpressing hematopoietic cells develop a myeloproliferative disease [56]. Mice overexpressing IL-11 have augmented cortical thickness and strength of long bones, attributed to an increase in osteoprogenitors, detected in a colony forming unit-osteoblast assay in cultures of bone marrow stromal cells [57].

Many in vitro studies further demonstrated that IL-6/LIF cytokines stimulate alkaline phosphatase activity, an early marker of osteoblast differentiation [14, 57, [58], [59], [60]61], but decrease expression of the mature osteoblast marker osteocalcin (OCN) [14]. Using two primary culture models of osteoblast differentiation, the fetal rat calvaria (RC) and the adult rat bone marrow (RBM) stromal cell systems, we have also shown previously that LIF has a biphasic effect on osteoblast development, depending on the stage of precursor cell differentiation and duration of exposure, with stimulation of cell proliferation and a resultant increase in bone formation with short pulse treatments early in stromal cell cultures [42, 62] but inhibition of bone formation with longer treatments of RBM cells and in RC cell cultures. In the latter case, precursor cells were blocked at the late osteoprogenitor/early osteoblast stages [42, 63, 64].

In the course of further analysis of the RC cell model, we recently found among the genes upregulated by LIF specifically during the differentiation-sensitive time window two genes involved in lipid metabolism: diphosphate dimethylallyl diphosphate isomerase 2 (Idi2) and hormone-sensitive lipase (D. Falconi and J.E. Aubin, manuscript in preparation). In this study, therefore, we investigated whether LIF acts in part by altering the fate choice of RC progenitor cells (i.e., whether the LIF-induced decrease in osteoblast differentiation was compensated for by a reciprocal increase in adipocyte differentiation).

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Cell Culture

Osteogenic cells were isolated from 21-day-old fetal RC by successive collagenase digestions, as described elsewhere [65]. The cells were plated at a density of 5,000 cells per square centimeter and cultured for 15 days in α-minimal essential medium (University of Toronto Tissue Culture Media Preparation, Toronto, ON, Canada), supplemented with 10% fetal bovine serum (PAA Laboratories, Etobicoke, ON, Canada, http://www.paa.com), 50 μg/ml ascorbic acid (Fisher Scientific Ltd., Nepean, ON, http://www.fisherscientific.com), and 10 mM β-glycerophosphate (Sigma-Aldrich Canada Ltd., Oakville, ON, Canada, http://www.sigmaaldrich.com) without dexamethasone (Decadron; Merck & Co., Whitehouse Station, NY, http://www.merck.com) (Dex), conditions known to support osteoblast differentiation [65]. To generate single-cell-derived colonies, RC cells were plated at a density of ∼6.3 cells per square centimeter (500 cells per 100-mm dish) and cultured for 31 days in the same medium as above, supplemented with 10−8 M Dex to increase osteoprogenitor cells [66]. Recombinant mouse LIF (high cell density: 5 ng/ml, 500 U/ml, 0.25 × 10−9 M; low cell density: 42 ng/ml, 4,200 U/ml, 2.08 × 10−9 M; Chemicon, Temecula, CA, http://www.chemicon.com) was used to block osteoblast differentiation, as previously described [42, 63]. BRL49653 (rosiglitazone) was synthesized at Sankyo Pharmaceuticals, Tokyo (http://www.sankyo.co.jp). Supplied lyophilized, it was dissolved in sterile dimethyl sulfoxide (EMD Chemicals Inc., Gibbstown, NJ, http://www.emdchemicals.com) and used at a final concentration of 10−6 M. At the end of the culture period, cells were fixed in 10% neutral formalin buffer and stained with silver nitrate/sodium carbonate (von Kossa) to reveal mineralization and/or oil red O (Sigma-Aldrich) to reveal lipid foci. Bone nodules and lipid foci were counted manually with a grid and a dissecting microscope.

Lipid Quantification

In some cases, lipid accumulation was quantified [67]. Briefly, cells grown in 24-well plates and stained with oil red O after fixation as described above were extracted with 200 μl per well of a solution of 4% Nonidet P-40 (BioShop Canada Inc., Burlington, ON, Canada, http://www.bioshopcanada.com) in isopropanol. Four wells were used for each treatment analyzed. The volume of each sample was adjusted to 1 ml with the 4% Nonidet P-40/isopropanol solution, and quantification was performed by spectroscopy at 520 nm with a Beckman DU640 spectrophotometer (Beckman Coulter Canada Inc., Mississauga, ON, Canada, http://www.beckmancoulter.com). For each sample, optical density measurements were done in quadruplicate. The average values were then normalized for the number of lipid foci, measured as described above.

Polymerase Chain Reaction

For cells grown at high density, total RNA was extracted with Trizol reagent (Invitrogen Canada Inc., Burlington, ON, Canada, http://www.invitrogen.com) according to manufacturer's instructions. Total RNA was subjected to RNase-free DNase I (Fermentas Canada Inc., Burlington, ON, Canada, http://www.fermentas.com) treatment and purification with Qiagen RNA Cleanup Kit (Mississauga, ON, Canada, http://www1.qiagen.com) before reverse transcription. Two μg of DNA-free purified RNA was reverse-transcribed with SuperScript II RNase H Reverse Transcriptase (Invitrogen Canada) and oligo(dT)12–18 primers (Amersham Biosciences, Baie d'Urfé, PQ, Canada, http://www.amersham.com). The primers used to amplify the genes presented in this chapter are summarized in Table 1.

Table Table 1.. Primers used
Thumbnail image of

For cells grown at low density, total RNA was obtained from individual colonies with the aid of a cloning ring. All purification steps and reverse transcription were carried as described above. However, before proceeding with the polymerase chain reaction (PCR), the cDNA was first precipitated according to the protocol developed by Liss [68] to remove inhibitory components from the reverse transcription reaction. Semiquantitative real-time PCR experiments were done on 2.5 ng of cDNA per reaction using the SYBR Green PCR Master Mix in an ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA, http://www.appliedbiosystems.com). Expression levels of each analyzed gene, measured as a threshold value (Ct), were reported as a difference over the levels of the ribosomal protein L32 in the same sample (ΔCt) and compared with the relative expression levels of the same gene in vehicle-treated samples collected at the same time point (ΔΔCt); the results therefore represent fold changes in levels of expression relative to control cells (calculated as 2−ΔΔCt, according to the manufacturer's instructions). The primers used, designed with the Primer Express 2.0 software (Applied Biosystems), are summarized in Table 1. For each gene, analyses were conducted in triplicate on samples from two separate cell isolations.

Statistical Analyses

For colony counts at high cell density, analyses of variance were performed by the Statistical Consulting Services at the University of Toronto using SAS software (SAS Institute, Cary, NC, http://www.sas.com). Whenever a statistically significant interaction was found, a Tukey-Kramer multiple comparison test was done. For cells grown as single colonies, statistical differences were assessed by Fisher's exact test. In all cases, a p value <.05 was considered significant.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

In RC cell cultures grown under standard conditions supporting osteoblast differentiation (serum-containing medium supplemented with ascorbic acid and β-glycerophosphate [65]), numerous bone nodules form but few, if any, adipocyte foci (Figs. 1, 2). As we found previously [42, 63], the effect of LIF was time/differentiation stage-dependent: bone nodules were present in cultures treated with LIF early [day 1 (d1)–d4]), but longer treatments inhibited nodule formation (Fig. 2A; vehicle vs. LIF, p < .0001) and treatments from day 1 to day 8 had the same effect as chronic exposure (days 1–15; Fig. 2A; LIF d1–d4 vs. LIF d1–d8, p < .0001). Continuous exposure to the PPARγ agonist BRL49653 (BRL) alone slightly but nonsignificantly decreased bone nodule formation (Figs. 1, 2A; vehicle vs. BRL, p = .3315) but potently stimulated the formation of adipocyte foci (Figs. 1, 2B), an effect enhanced in the presence of LIF (Fig. 2B; BRL vs. BRL+LIF, p < .0001). Although cells treated with LIF for longer periods (d1–d8 and beyond) tended to form more lipid foci than those treated for shorter times (Fig. 2B), no significant effect was detected for the incubation time variable. Interestingly, there was a trend that did not reach significance for cells treated with BRL+LIF to form more bone nodules than those treated with LIF alone (Fig. 2A; LIF vs. BRL+LIF, p = .1398). No lipid foci formed in cells treated with vehicle or LIF alone (Fig. 2B). These results indicate that BRL promotes adipogenesis in RC cultures, an effect enhanced by LIF, and slightly but nonsignificantly abrogated the LIF inhibitory effect on osteoblast differentiation.

thumbnail image

Figure Figure 1.. Osteoblast and adipocyte development in rat calvaria (RC) cell cultures treated with LIF and/or BRL. Fetal RC cells were grown in 24-well plates and treated continuously without (vehicle; dimethyl sulfoxide/phosphate-buffered saline) and/or with BRL or LIF. Cells were fixed at day 15, stained with silver nitrate/sodium carbonate (von Kossa) for bone nodules and oil red O for lipid foci, and visualized with a bright-field microscope. Magnification, ×14. Scale bar = 5 mm. Abbreviations: BRL, BRL49653; LIF, leukemia inhibitory factor.

Download figure to PowerPoint

thumbnail image

Figure Figure 2.. Effect of LIF on osteoprogenitor and adipoprogenitor differentiation. Rat calvaria cell cultures were treated chronically with BRL (or vehicle dimethyl sulfoxide) and pulse-treated with LIF (or vehicle phosphate-buffered saline) for the indicated periods. Counting was done on cells dually stained with silver nitrate/sodium carbonate (von Kossa) for BNs (A) and oil red O for lipid foci (B). The sum of BNs and LF in each well is reported as total number of colonies (C). Each point represents mean values from five to eight wells ± SEM. On BN formation (A), analyses of variance indicated a significant two-way interaction between time × LIF (p < .0001) and BRL × LIF (p = .0074). On LF (B), LIF had a significant effect over vehicle (p < .0001). On the total number of colonies (C), a significant three-way interaction was found between time × BRL × LIF (p = .0333), and the relevant significant differences are indicated (∗∗, p < .01 and ∗∗∗, p < .001 vs. vehicle; ##, p < .01 and ###, p < .001 vs. BRL; ††, p < .01 and †††, p < .001 vs. BRL+LIF). Abbreviations: BN, bone nodule; BRL, BRL49653; d, day(s); LIF, leukemia inhibitory factor; LF, lipid foci.

Download figure to PowerPoint

It is also worth noting that BRL alone had no effect on the total number of colonies (bone nodules + lipid foci) formed. On the other hand, LIF significantly decreased the total number of colonies formed when present longer than 4 days (LIF alone) or 6 days (BRL+LIF) (Fig. 2C). Because of the effect of LIF on BRL-induced adipogenesis, the decrease in total number of colonies was not as steep as that observed with LIF alone. In fact, BRL+LIF-treated cells formed a significantly higher number of colonies compared with cultures treated with LIF alone between d1–d6 and d1–d11 (Fig. 2C).

In spite of the increase in number of adipocyte foci documented above, qualitative morphological assessment indicated that adipocytes in LIF-treated cultures contained fewer lipid droplets, based on the fainter oil red O staining detected microscopically (Fig. 1) and quantification of oil red O extracted from stained cells (Fig. 3A; BRL+LIF, 0.01674 ± 0.00035 vs. BRL, 0.01930 ± 0.00035; p69]; the secreted enzyme LPL, whose expression is upregulated at the preadipocyte stage [70]; and the adipocyte-specific serine protease adipsin, which characterizes mature adipocytes [70]. The increased expression of both C/EBPδ and adipsin in BRL+LIF compared with BRL alone suggests that LIF increases the recruitment of cells toward the adipocyte lineage and adipocyte maturation (Fig. 3B). In keeping with our previous results [42, 63, 64] and the inhibition of bone nodule formation observed here (Fig. 2A), both bone sialoprotein (BSP) and OCN were inhibited by LIF (Fig. 3B). Together, these results suggest that the reduced lipid accumulation in LIF-treated cultures results not from reduced adipocyte maturity but from reduced lipid storage and/or enhanced lipolysis.

thumbnail image

Figure Figure 3.. Effect of LIF on adipocyte and osteoblast differentiation. (A): Rat calvaria (RC) cells were treated chronically with BRL (or vehicle dimethyl sulfoxide) and pulse-treated with LIF (or vehicle PBS) for the time periods indicated. After cell fixation and staining, bound oil red O was extracted and quantified; results from each well were normalized to the number of lipid foci. Each point represents the mean value from four wells ± SEM (p < .0001, LIF+BRL vs. BRL-treated samples; analysis of variance). (B): Polymerase chain reaction analyses were conducted to evaluate the relative expression of BSP, OCN, C/EBPδ, LPL, and adipsin in BRL+LIF-treated samples, as compared with samples with BRL alone. Similar results were found in an independent RC cell culture. Abbreviations: BRL, BRL49653; BSP, bone sialoprotein; C/EBP, CAAT/enhancer-binding protein; d, day(s); LIF, leukemia inhibitory factor; LPL, lipoprotein lipase; OCN, osteocalcin; O.D., optical density; PBS, phosphate-buffered saline.

Download figure to PowerPoint

Given that BRL alone induces adipocyte foci, an effect augmented by LIF, we next asked whether the effects of LIF and BRL are maintained at the single-colony level, to gain insight into the cellular mechanism(s) underlying this fate change. We therefore cultured fetal RC cells at low density, either treated or not with LIF, BRL, or LIF+BRL, and analyzed individual selected colonies for expression of osteoblast and adipocyte differentiation markers under the different treatment regimens. Colonies selected had either cells with a cuboidal morphology, typical of osteoblastic cells, or cells with accumulated lipid in cytoplasmic vesicles, distinctive of adipocytes. Colonies were assigned to one of four phenotypic groups based on marker expression: BSP/Fabp4 (designated precursors), BSP+/Fabp4 (osteoblasts), BSP+/Fabp4+ (colonies expressing markers of both lineages), and BSP/Fabp4+ (adipocytes) (Fig. 4). LIF alone tended to increase and decrease the frequency of colonies expressing the BSP/Fabp4 (p = .1021 vs. vehicle) and BSP+/Fabp4 (p = .1152 vs. vehicle) phenotypes, respectively (Fig. 4); although this is not statistically significant, it suggests that LIF blocks the acquisition of a mature osteoblast phenotype, as observed at high cell density. On the other hand, BRL alone significantly reduced the number of BSP+/Fabp4 (p < .001 vs. vehicle) colonies and significantly increased the number of BSP+/Fabp4+ (p < .05 vs. vehicle) colonies (Fig. 4). This may indicate that colonies that would otherwise acquire an osteoblast (BSP+/Fabp4) phenotype were forced to co-express Fabp4. BRL also tended to decrease the number of BSP/Fabp4 (p = .4146 vs. vehicle) colonies and increase the number of BSP/Fabp4+ (p = .0596 vs. vehicle) colonies (Fig. 4). When LIF and BRL were applied simultaneously, the number of BSP+/Fabp4 colonies was significantly reduced compared with vehicle-treated cultures (p < .05 vs. vehicle; Fig. 4), suggesting that BRL further enhanced the effect of LIF. However, the number of BSP/Fabp4 colonies was the same in treatments with vehicle versus combined LIF and BRL (p = .7833 vs. vehicle; Fig. 4), suggesting that BRL abrogated the positive effect of LIF on that population. In addition, LIF/BRL cotreatment increased the number of Fabp4-positive colonies versus vehicle treatment (BSP+/Fabp4+, p = .1895, and BSP/Fabp4+, p = .2167, both vs. vehicle; Fig. 4), suggesting, as seen with cells cultured at high density, that LIF-arrested progenitors are more apt to undergo the adipocyte differentiation program. Taken together, the data suggest that both LIF and BRL affect mesenchymal cell fate and that different developmental mechanisms underlie the LIF- versus BRL-induced changes in fate.

thumbnail image

Figure Figure 4.. Effect of LIF and BRL, alone and together, on gene expression in single-cell-derived rat calvaria (RC) cell colonies. Fetal RC cells were plated at low density; cultured in the presence of LIF, BRL, or vehicle; and assayed for expression of the osteoblast marker BSP and the adipocyte marker Fabp4. Top, schematic representation of the four phenotypes observed. Bottom, number of colonies observed (n and %) with each of the marker phenotypes. ∗, p < .05 and ∗∗, p < .01 versus vehicle. Abbreviations: BRL, BRL49653; BSP, bone sialoprotein; Fabp4, fatty acid-binding protein 4; LIF, leukemia inhibitory factor.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

Although many studies have examined the effects of LIF and other IL-6-related cytokines on osteoblast and adipocyte differentiation, none has addressed their simultaneous effects in a model capable of differentiation along the two lineages. Fetal RC cells represent such a model, comprising a heterogeneous mixture of progenitor cell types, including bi- and multipotential progenitors with capacity to differentiate into multiple mesenchymal lineages, including osteoblasts, adipocytes, myocytes, and chondrocytes [3, 4, 71, 72]. As we showed previously [42, 62, [63]64] and confirm here, LIF inhibits bone nodule formation in RC cell cultures. We have also shown recently that genes associated with lipid metabolism are increased in LIF-treated RC cells concomitant with inhibition of osteogenesis (D. Falconi and J.E. Aubin, manuscript in preparation). We therefore tested here the hypothesis that LIF-treated osteoprogenitors are reprogrammed to the adipocyte lineage but require a proadipogenic stimulus to complete the differentiation program.

Our data from both high- and low-density cultures suggest that both LIF and BRL modify mesenchymal cell development. In high-cell-density cultures treated with BRL, both bone nodules and adipocyte foci develop, but the total number of colonies (combined bone and adipocyte colonies) formed is not significantly different from that in vehicle, suggesting that bone nodules and lipid foci formed from a common pool of progenitors. On the other hand, more colonies developed in LIF+BRL versus LIF alone, and these new colonies were adipocytic. In addition, LIF synergized with BRL to increase the number of lipid foci formed above the values obtained with BRL alone, a finding in agreement with that of Aubert et al., who found that LIF synergized with BRL to enhance the adipocyte differentiation of two mouse preadipocyte cell lines (Ob1771 and 3T3-F442A) and pluripotent mouse embryonic fibroblasts [48]. Thus, it seems plausible that the adipocytes formed in LIF+BRL-treated RC populations reflect reprogramming of LIF-arrested osteoprogenitors toward the adipocyte lineage.

Analysis of gene expression in single colonies formed in low-density RC cultures showed not only that LIF increased the proportion of colonies negative for both osteoblast and adipocyte markers (i.e., blocked maturation of progenitors) but also apparently redirected osteoprogenitors toward an adipogenic fate that is enhanced by BRL. There is other support for the hypothesis that LIF acts early to render cells competent to respond to PPARγ activation by BRL. Adipocytic differentiation in cultures of mouse embryoid bodies is characterized by a commitment phase not requiring PPARγ activity followed by a terminal differentiation phase requiring PPARγ activity [73]. LIFR-deficient mouse ESCs also have a reduced capacity to undergo adipogenesis [48]. Thus, LIF may act by stimulating the expression of proteins involved in PPARγ expression or activity. Among the probable mediators are C/EBPβ and C/EBPδ, two transcription factors whose expression is induced by LIF in preadipose cells [48, 74] (Fig. 3) and that stimulate PPARγ expression [17, 18, 75].

BRL increased the proportion of colonies expressing only the adipocyte marker (possibly reflecting that some adipocyte foci arose directly from primitive adipoprogenitors), increased the number of colonies positive for both osteoblast and adipocyte lineages markers, and in low- but not high-density cultures decreased the number of colonies positive for the osteoblast marker only (BSP+Fabp4 colonies and alkaline phosphatase-positive colonies [data not shown]). Although relatively few colonies (∼10 per treatment group) from low-density cultures were analyzed, reducing the robustness of statistical testing of the LIF-BRL effects, the data suggest that in addition to synergizing with LIF, BRL by itself can act on later precursors and possibly mature osteoblasts to promote acquisition of an adipogenic fate. Our data agree with results from others, who found co-expression of osteoblast- and adipocyte-“specific” genes in differentiating cells from neonatal mouse calvaria [76], fetal RC [72], and human bone marrow stromal cells [77, 78], and are consistent with data suggesting that at least under certain conditions the osteoblast and adipocyte lineages are reciprocally related [11, 13, 72].

It is, however, also worth noting that fewer colonies formed with LIF+BRL than with BRL alone, consistent with the observation that LIF tended to decrease the number of more mature adipo- and osteoprogenitors present and suggesting that the conversion between osteoblast and adipocyte is not strictly reciprocal. We did not evaluate whether osteoblast and adipocyte lineage cells generated in RC cell cultures have the same capacity to respond to LIF. However, we have previously shown that LIFR expression is constitutive during osteoblastic differentiation of RC cells [79] and that LIFR signaling is active throughout osteoblast development (D. Falconi and J.E. Aubin, manuscript in preparation). Also, LIFR is abundantly expressed in cultured preadipocytes (3T3-F442A) [48]. Although we cannot exclude the possibility that different treatments might change the relative abundance of LIFR, our results suggest that LIFR expression in RC cells is a nonlimiting factor under the culture conditions used. Recently, Lecka-Czernik et al. demonstrated with selective agonists that the proadipogenic and antiosteoblastogenic activities of PPARγ2 are mediated by distinct regulatory pathways [36]. Therefore, differences may result from an effect of LIF on modulators of PPARγ2 activity, such as co-activators and corepressors. Also, since RC cells contain heterogeneous progenitor pools with the capacity to differentiate into other mesenchymal lineages (muscle, fat, cartilage, and bone), it will be interesting to see whether LIF enhances the capacity of the progenitor cells to differentiate toward chondrogenic and/or myogenic lineages and account for the missing colonies in a manner similar to recent results with 1,25(OH)2D3 [72]. However, unambiguous assessment of the fate and potential reprogramming of LIF- and BRL-treated RC cells will require detailed studies of the developmental potential of appropriately marked single progenitors and their progeny, since we cannot exclude the possibility that LIF and/or BRL have regulatory effects on the genes analyzed independent of an ability to alter lineage fate choices.

It should be recognized that the lipid content of LIF-induced adipocyte foci was reduced in comparison with that in non-LIF-treated cultures. Since gene expression analyses indicated that adipocyte maturation was not affected, it seems likely that LIF modified lipid metabolism, a possibility supported by the observation that LIF decreases LPL activity in differentiated adipocyte cell lines [44, 45]. Even though we did not detect a change in LPL mRNA abundance, Berg et al. [44] reported that LPL activity is not correlated with mRNA or protein abundance in differentiated 3T3-L1 adipocytes. In addition, in 3T3-L1 adipocytes, LIF was recently shown to reduce triacylglyceride accumulation during differentiation and to decrease fatty acid synthase protein levels in mature cells [74]. Thus, LIF reduces the capacity of adipocytes to store lipids, both during adipogenesis and in mature cells.

Taken together, our data support a multilineage, multistage model for LIF and BRL regulation of osteoblast and adipocyte differentiation in RC cell populations: BRL acts alone on committed adipocyte precursors to stimulate adipogenesis, BRL acts alone to reduce osteoblast maturation and/or activity by inducing adipocyte genes in committed osteoblastic cells, and LIF acts alone to block osteoblast maturation at late differentiation stages and redirect these cells toward an adipogenic pathway responsive to BRL (Fig. 5). Whether LIF is unique among members of this cytokine family or whether other family members that inhibit osteoblast differentiation (Introduction) act similarly to LIF to reprogram mesenchymal cell fate choice remains to be determined.

thumbnail image

Figure Figure 5.. Schematic of LIF-BRL action on rat calvaria (RC) cell differentiation. The RC cell population is heterogeneous and contains cells capable of differentiation along the adipocyte (see also Ailhaud et al. [70]) and osteoblast (adapted from Aubin [8]) lineages. Some differentiation markers of each lineage are indicated in parentheses. Under osteogenic culture conditions, osteoblast differentiation is observed, a program inhibited by LIF at the mature osteoprogenitor-preosteoblast stage. This effect of LIF is consistent with the increased number of BSPFabp4 colonies detected at low cell density. Addition of BRL stimulates adipocyte differentiation from committed adipoprogenitors and may trigger the expression of adipocyte differentiation markers in cells of the osteoblast lineage. The simultaneous presence of LIF and BRL inhibits osteoblast differentiation and enhances adipocyte development. BSP+/Fabp4+ cells, detected at low cell density, are considered less mature than cells committed to expression of either lineage alone. Abbreviations: ALP, alkaline phosphatase; BRL, BRL49653; BSP, bone sialoprotein; C/EBP, CAAT/enhancer-binding protein; COL1A1, type I collagen α1; Fabp4, fatty acid-binding protein 4; LIF, leukemia inhibitory factor; LPL, lipoprotein lipase; OCN, osteocalcin.

Download figure to PowerPoint

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References

We gratefully acknowledge the help of Julia Bodnar and Dr. Debbie Kipp with data collection, Mark Kane for statistical analyses, and Dr. Norman N. Iscove for discussions regarding the studies with single colonies. D.F. was supported by scholarships from the Ontario Graduate Scholarship in Science and Technology/Edward Dunlop Foundation and a Doctoral Research Award from the Canadian Institutes of Health Research. This work was supported by Canadian Institutes of Health Research Operating Grant MOP-49419 to J.E.A.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Disclosures
  8. Acknowledgements
  9. References