Author contributions: C.V., R.K., and G.D.: concept design, in vivo and in vitro experiments, data collection and analysis, and preparation of the manuscript; S.B., W.L., and D.H.: in vivo and in vitro experiments, data collection and analysis, and preparation of the manuscript.
Interferon gamma (IFNγ) has been reported to induce osteoblastogenesis from mesenchymal stem cells (MSCs) both in vitro and in vivo. With ageing, adipocytes outnumber osteoblasts within the bone microenvironment leading to a decrease in bone formation. Since both osteoblasts and adipocytes are of mesenchymal origin, we hypothesized that IFNγ treatment might negatively affect adipogenesis while stimulating osteoblastogenesis in human MSC. To test this hypothesis, human MSCs were induced to differentiate into adipocytes in the presence or absence of osteogenic doses of IFNγ (1, 10, and 100 ng/ml). IFNγ-treated MSC showed a decrease in adipocyte differentiation and lipid deposition when compared with vehicle-treated controls. Additionally, adipogenic markers were significantly decreased by IFNγ treatment at the same doses that have been reported to have a strong osteogenic effect in vitro. Furthermore, DNA binding of peroxisome proliferator-activated receptor gamma was significantly lower in IFNγ-treated differentiating MSC. Subsequently, ovariectomized C57BL6 mice were treated with osteogenic doses of IFNγ three times a week for 6 weeks. In distal femur, treated mice showed significantly higher hematopoiesis concomitant with lower levels of fat volume/total volume, adipocyte number, and expression of adipogenic markers when compared with the vehicle-treated mice. Together, these findings demonstrate that, at osteogenic doses, IFNγ also acts as an inhibitor of adipogenesis in vitro and prevents marrow fat infiltration while favors hematopoiesis in ovariectomized mice. STEM CELLS2012;30:1042–1048
The maintenance of bone mass and the prevention of fractures relies on two synchronized (coupled) events, resorption by osteoclasts and bone formation by osteoblasts, a process known as bone turnover . The cellular components of the bone marrow milieu (hematopoietic, immune, bone, and adipose) play an important role in the regulation of bone turnover  either through their own activity or through cell-to-cell interactions. The particular relationship between the immune and musculoskeletal systems has been described as osteoimmunology [3, 4]. The study of this interaction has demonstrated that both, cytokines (tumor necrosis factor alpha, interleukin-6 [IL-6], etc.)  and adipokines (leptin, adiponectin, etc.) [5–7], play an important role in the regulation of bone turnover and could be responsible for the changes seen in age-related bone loss and osteoporosis.
One of those cytokines, interferon gamma (IFNγ), was recently reported to promote osteoblastogenesis and bone formation both in vitro  and in vivo . The anabolic effect of IFNγ is mediated by increasing both osteoblastogenesis and osteoclastogenesis  with a predominant stimulatory effect on the osteoblast lineage, thus increasing bone mass and rescuing oophorectomized (OVX) mice from osteoporosis . In addition, IFNγ significantly increases the expression of osteogenic markers in differentiating mesenchymal stem cell (MSC) into osteoblasts in vitro, including runt-related transcription factor 2 and osteopontin .
Overall, the effects of IFNγ on bone remain complex  with some investigators reporting contrasting findings about its effect on osteoclastogenesis  and bone resorption [3, 12–15]. High doses of IFNγ have been used as treatment in patients with osteopetrosis  to induce bone resorption and reduce bone mass. Other studies report that IFNγ directly inhibits osteoclast differentiation [9, 13] and induce osteoclast apoptosis . In addition, the effect of IFNγ on osteoclast differentiation and function could be affected, among others, by estrogen deficiency, inflammation, and bacterial toxins [12, 18].
In contrast, the anabolic effect of low doses of IFNγ on bone is quite consistent [8, 9], thus suggesting that IFNγ could become a useful therapeutic approach to age-related bone loss in the near future. However, to understand the potential therapeutic effect of IFNγ on osteoporosis and age-related bone loss, it is essential to assess its effect on other cellular components of the bone marrow milieu and specifically on marrow fat. High levels of fat infiltration are the result of increased adipogenesis that occurs in aging bone along with a decrease in the number of osteoblasts due to a clonal switch in differentiating MSC [19–21]. It would be then expected that the anabolic effect of IFNγ on bone would be associated with a reduction in marrow adipogenesis. In fact, other molecules that have an anabolic effect on bone such as alendronate , vitamin D , and parathyroid hormone  have been reported to increase osteoblast differentiation while decreasing adipogenesis. Therefore, considering that IFNγ has been just recently reported as a bone anabolic, in this study we assessed the effect of osteogenic dose of IFNγ on adipogenesis both in vitro and in vivo.
MATERIALS AND METHODS
Cell Differentiation and Treatment
For adipogenic differentiation of MSC, human MSCs (PT-125 Lonza, Basel, Switzerland, http://www.lonza.com)  were seeded at a density of 5 × 105 cells per square centimeter in 56 cm2 Petri dishes containing basal medium (MSCGM BulletKit PT-3001, Lonza, Basel, Switzerland). These primary cells are commercially available and are obtained from bone marrow of healthy young donors (age 24–30-year old) . This model has been used in our previous studies [8, 23] and is preferred over cells from older donors due to their excellent differentiation potential. After reaching 60% confluence, cells were transferred to six-well plates and stimulated to differentiate into adipocytes alternating between adipogenic induction media (AIM), containing 0.1 μM dexamethasone, 10 μg/ml insulin, 0.2 mM indomethacin, 0.5 mM 3-isobutyl-1-methylxanthine, 10% fetal bovine serum, 0.05 U/ml penicillin, 0.05 μg/ml streptomycin, and adipogenic maintenance medium ([10 μg/ml insulin, 10% fetal bovine serum, 0.05 U/ml penicillin, and 0.05 μg/ml streptomycin]) every 3 days for 2 weeks, until obtaining an adipogenic phenotype as previously described . During differentiation, cells were treated with either vehicle or IFNγ (Sigma-Aldrich, Co., St. Louis, MO, http://www.sigmaaldrich.com) at three different doses (1, 10, and 100 ng/ml), as previously described .
Oil Red O Staining
Oil red O (ORO) staining was used to assess adipocyte differentiation as an indicator of intracellular lipid accumulation. On day 15, culture medium was removed from tissue culture well and cells were rinsed with phosphate buffered saline (PBS) once, followed by fixation using 10% formaldehyde in PBS for at least 1 hour. The fixative was then aspirated and cells were washed with 60% isopropanol before being allowed to dry completely. Cells were stained for 10 minutes at room temperature with a diluted solution of ORO (66.6%) prepared from a 0.5% (w/v) ORO dissolved in isopropanol. Cells were then washed four times with running tap water to remove excess stain. Photomicrographs were taken using an IX50 Olympus inverted microscope (Olympus, Tokyo, Japan, http://www.olympus-global.com/en/) and a Digital Sight DS-5M Nikon camera (Nikon Instruments, Inc. Melville, New York, NY, http://www.nikoninstruments.com). To quantify lipid deposition, ORO was eluted with 1 ml 100% isopropanol for 10 minutes and absorbance measured at 490 nm .
Measurement of Glycerol in Culture Media
Determination of free glycerol concentration in culture media was measured by coupled enzymatic reactions catalyzed by glycerol kinase, glycerol phosphate oxidase, and peroxidase (Sigma; cat# F6428). Ten microliters of supernatant or standard was added to 800 μl of free glycerol reagent and incubated for 5 minutes at 37°C. Absorbance was measured at 540 nm and glycerol concentration determined using a standard curve. All tests were done in duplicate.
MTS Viability Assay
To test whether treatment with IFNγ has any effect on cell survival, differentiating MSCs were seeded in 96-well plates. At 60% confluence, media were replaced with AIM containing either IFNγ (1, 10, and 100 ng/ml) or vehicle alone. At timed intervals (24, 48, and 72 hours), 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS)-Formazan cell viability assays (Promega, Madison, WI, http://www. promega.com) were performed and corrected for cell number. Briefly, a stock solution of MTS was dissolved in PBS at a concentration of 5 mg/ml and was added in a 1:10 ratio (MTS/Dulbecco's modified Eagle's medium) to each well, incubated at 37°C for 2 hours, and the optical density was determined at a wavelength of 490 nm on a microplate reader model 3550 (FLUOstar; BMG Labtech, Durham, NC, http://www.bmglabtech. com). The percent survival was defined as [(experimental absorbance − blank absorbance)/control absorbance − blank absorbance)] × 100, where the control absorbance is the optical density obtained for 1 × 104 cells per well (number of cells plated at the start of the experiment), and blank absorbance is the optical density determined in wells containing medium and MTS alone. This experiment was replicated three times.
Real-Time PCR and Western Blotting
Total RNA and proteins were extracted from the same cell pellet, collected at the end of the second week of differentiation using the specifications of PARIS protocol (Ambion Inc., Austin, TX, http://www.invitrogen.com/site/us/en/home/brands/ambion.html). Briefly, treated cells were kept on ice rinsed with PBS and directly disrupted by adding a cell disruption buffer. Proteins were protected by adding halt protease and phosphatase inhibitors (Thermo Fisher Scientific Inc., Rockford, IL, http://www.thermofisher.com), collected using a rubber spatula and stored at −80°C until further analysis. RNA in the lysate was treated with β-mercaptoethanol (Sigma-Aldrich, Co.) and guanidinium thiocyanate (provided) and linked to a filter cartridge according to the PARIS manufacturer conditions. RNA and protein concentrations were estimated based on UV absorbance (260 and 280 nm, respectively) readings by spectrophotometry. First strand complementary DNA (cDNA) synthesis was performed using 200 ng of total RNA, 50 ng random hexamers, and 50 units reverse transcriptase at 42°C for 1 hour, as described by manufacturer (Bioline Australia Pty, Alexandria, NSW, Australia, http://www.bioline.com; cat# BIO-65025).
Real-time polymerase chain reaction (RT-PCR) for expressed genes as markers for adipogenesis was performed in duplicate in a total reaction volume of 25 μl, 10% of which was cDNA (or water for nontemplate control), 3 mM MgCl2, and 250 nM each forward and reverse-specific primer for target genes and normalizer (Table 1). All PCRs were performed in a Corbett Rotor-Gene 3000 (QIAGEN Pty, http://www.qiagen.com) using SYBR green with no-ROX reaction mix and a standard thermal profile as described by supplier (Bioline Australia Pty; cat# QT6750-02). Quantitative RT-PCR data were defined by threshold cycle (Ct) normalized for the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Quantification of relative differences of expressed genes between different adipogenic conditions was calculated using REST software (QIAGEN Pty)  correcting for PCR reaction efficiencies (>0.90). A statistical significant change in relative expression between treated and untreated cells was taken at p < .05.
Table 1. Oligonucleotide primers used for real-time PCR
Protein samples for Western blotting were reduced with dithiothreitol 50 mM in NuPAGE lithium dodecyl sulfate Sample Buffer ×1 and were run on a NuPAGE Novex Bis-Tris mini-gel using the XCell SureLock Mini-Cell and blotted to a nitrocellulose membrane using the XCell II Blot module (all reagents and equipment from Invitrogen Corporation, Carlsbad, CA, http://www.invitrogen.com/site/us/en/home.html). After membrane blocking with 5% bovine serum albumin, they were incubated overnight at 4°C using monoclonal antibodies directed against peroxisome proliferator-activated receptor gamma (PPARγ) 1 and 2, CCAAT/enhancer binding protein alpha (CEBPα), and adiponectin (Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com). The linked antibodies were detected with the corresponding secondary antibodies conjugated with horseradish peroxidase (HRP). Blots were developed by chemiluminescence using the Quantity One 4.4.0 software (Bio-Rad Laboratories, Hercules, CA, http://www.bio-rad.com) and the Super Signal West reagents (Thermo Fisher Scientific Inc.). In all experiments, GAPDH gene and protein expression were used as internal control. These experiments were repeated three times.
PPARγ activity as a transcription factor, binding to the peroxisome proliferator response element, was determined using the enzyme-linked immunosorbent assay-based PPARγ activation TransAM kit (Active Motif, Rixensart, Belgium, http://www.activemotif.com) as previously described . In brief, each one of the 96 wells in the Trans-AM PPARγ-Kit contains the immobilized oligonucleotide containing a PPARγ consensus-binding site (5′-AACTAGGNCAAAGGTCA-3′). PPARγ, if present, specifically binds to this oligonucleotide. The primary antibody used in the probe recognizes an accessible epitope on PPARγ protein upon DNA binding. The secondary HRP-conjugated antibody provides a sensitive colorimetric readout easily quantified by spectrophotometry (450 nm). To quantify PPARγ activity in the cell lysate, 10 μg of proteins was added to the wells, incubated 1 hour at room temperature, and revealed by posterior addition of primary and secondary antibodies. After washing steps, developing and stop solutions were added, as recommended by manufacturer.
Animal Tissue Preparation
Eight-week-old virgin female C57BL/6 mice (Charles River, Quebec, Canada, http://www.criver.com) were OVX under general anesthesia. Two weeks after surgery, mice received intraperitoneal injections of either 2,000, 5,000, or 10,000 international units (IU) IFNγ (R&D Systems Inc., Minneapolis, MN, http://www.rndsystems.com) or vehicle (PBS) three times a week for a total of 6 weeks. Mice were housed in cages in a limited access room. Animal husbandry adhered to Canadian Council on Animal Care Standards and all protocols were approved by the McGill University Health Center Animal Care Utilization Committee. One side femur from each animal in each group was removed at the time of euthanization, fixed in 70% ethanol, dehydrated, and embedded undecalcified in methyl methacrylate (J-T Baker, Phillipsburg, NJ, http://www.jtbaker. com). At 50-mm intervals, longitudinal sections 5- and 8-mm thick were cut using a Polycut-E microtome (Reichert-Jung Leica, Heerbrugg, Switzerland, http://www.leica.com), placed on gelatin-coated glass slides, deplasticized, and stained with von Kossa. For fat quantification, after von Kossa staining, fat fraction (fat volume vs. total volume), adipocyte number, adipocyte diameter, and adipocytic/hematopoietic (A/H) ratio were quantified using von Kossa stained sections, as previously described .
The details of these methods were described previously . Briefly, for immunohistochemistry, sections were incubated overnight at 4°C with a goat polyclonal antibody IgG against either PPARγ2 or CEBPα (Santacruz Biotechnology Inc.). Primary antibody was detected by incubation with an anti-goat IgG secondary antibody conjugated with horseradish peroxidase (1:300 in BSA 1%, Sigma-Aldrich). Antibody complexes were visualized with DAB, a 3,3-diaminobenzine solution containing hydrogen peroxide (Zymed Laboratories Inc., San Francisco, CA, http://www.invitrogen.com/zymed.html) and then counterstained in 1% hematoxylin. Photographs were taken under an Olympus fluorescence microscope controlled by an IPLab system. Brightness and contrast adjustments were performed in Photoshop (Adobe). Levels of PPARγ2 and CEBPα expression were quantified as percentage of bone marrow surface using the Bioquant Image Analysis Software.
All results were expressed as the mean ± SE. Differences of the structural and static parameters of bone histomorphometry between different groups of mice were determined using Levene's test for homogeneity of variances and the unpaired t test for equality of means. In all experiments, a value of p < .05 was considered significant.
Lipid Accumulation In Vitro Is Impaired by IFNγ Treatment
As shown in Figure 1A, the proportion of cells having easily identifiable fat globules (adipocytes) was inversely related to the IFNγ concentration in the media. To assess lipid deposition, ORO was eluted and quantified by measuring the absorbance at 490 nm by spectrophotometry (Fig. 1B). Whereas cells treated with the lowest dose of IFNγ showed no difference when compared with vehicle-treated cells, higher concentrations of IFNγ in the media induced a significant decrease in lipid deposition (p < .01). In addition, IFNγ treatment induced a significant reduction in glycerol secretion by differentiating cells in a dose-dependent manner (Fig. 1C). Finally, treatment with IFNγ had no effect on cell survival (Fig. 1D).
Effect of IFNγ on Transcription and Protein Expression of Adipogenic Factors
To assess the effect of IFNγ administration on differentiating adipocytes, we tested two major adipogenic genes by quantitative real-time PCR. As shown in Figure 2, gene relative expression of the adipogenic markers, adiponectin and CEBPα, was significantly reduced by IFNγ treatment in a dose-dependent manner (p < .05). In addition, Western blot analysis showed that levels of adiponectin and CEBPα were significantly reduced upon treatment with IFNγ (p < .05) (Fig. 2).
Furthermore, we performed a comprehensive assessment of the effect of IFNγ administration on PPARγ, a transcription factor essential for marrow fat differentiation. Treatment with IFNγ induced a significant decrease in PPARγ gene (Fig. 3A) and reduced PPARγ2 protein expression without affecting protein expression of PPARγ1 (Fig. 3B). In addition, cells treated with IFNγ showed lower levels of binding of PPARγ to DNA (and hence ability to initiate transcription) in a dose-dependent manner (p < .05) (Fig. 3C). Finally, and concomitantly with the lower levels of DNA binding of PPARγ, we also observed a reduced expression of the PPARγ target genes adipocyte fatty acid binding protein (aP2) (Fig. 3D) and lipoprotein lipase (Fig. 3E) in the IFNγ-treated cells.
IFNγ Treatment Decreases Marrow Fat and Increases Hematopoiesis in OVX Mice
We treated OVX and SHAM mice with IFNγ, as described before, and marrow adipogenesis was determined by means of quantification of fat fraction (fat volume/total volume), adipocyte number, and adipocyte diameter. As shown in Figure 4A–4E, treatment with IFNγ significantly reduced the amount of marrow fat (fat fraction) and adipocytes number in a dose-dependent manner when compared with vehicle-treated OVX mice (p < .001). In contrast, adipocyte diameter was not affected by IFNγ treatment. Finally, A/H ratio was significantly decreased in the IFNγ-treated group in a dose-dependent manner (p < .001).
To further assess a potential mechanism for the reduction in adipocyte parameters induced by treatment with IFNγ, we assessed the changes in PPARγ2 and CEBPα expression within the bone marrow of OVX mice treated with either IFNγ or vehicle. As shown in Figure 4F–4H, after 6 weeks of treatment, IFNγ significantly reduced the levels of PPARγ2 and CEBPα expression within the bone marrow when compared with vehicle-treated animals (p < .01).
The pro-osteogenic effect of IFNγ has been previously reported both in vitro and in vivo [8, 9]. In this study, we tested the effect of osteogenic doses of IFNγ on adipocyte differentiation from MSC in vitro and on marrow fat infiltration in vivo. Our results support the hypothesis that, at osteogenic doses, IFNγ concurrently inhibits adipogenesis and marrow fat infiltration.
IFNγ is a cytokine produced by innate and adaptive cells in the immune system  as well as by MSC . IFNγ interacts with its receptors expressed constitutively to promote cell signaling and transcription through Jak1/STAT1 pathway . Its production is regulated by cytokines (positively, IL-12 and IL-18; negatively, IL-4, IL-10, transforming growth factor-β, and glucocorticoids) to exert antiviral activity (principally) and both directly and indirectly affecting a number of other biological pathways . In terms of osteoimmunology, IFNγ plays an important role in the regulation of both bone formation and bone resorption in a dose-dependent manner [3, 11]. Depending of the dose and the experimental model, IFNγ could regulate osteoclastogenesis and osteoclastic activity either positively [15, 18, 31] or negatively . In addition, IFNγ is required in early phases of osteoblastogenesis  while treatment with low doses of exogenous IFNγ induces osteoblastogenesis both in vitro  and in vivo .
Considering that the principle of a bone anabolic is the stimulation of bone formation through either the activation of osteoblastogenesis or prolonging osteoblast survival, in the case of IFNγ, it was initially proposed that its anabolic effect is explained by induction of osteoblast differentiation and bone formation . However, since osteoblasts and adipocytes share the same precursors, it was tempting to hypothesize that, as in the case of other bone anabolics [22–24], IFNγ could also have an inhibitory effect on adipogenesis, thus increasing the number of MSC suitable to differentiate into osteoblasts.
In fact, the effect on IFNγ on adipogenesis has been previously tested in differentiating 3T3-L1 murine preadipocytes  and in mice embryonic-differentiating fibroblasts . In both studies, IFNγ showed an inhibitory effect on adipogenesis; however, the significance of this effect on balancing osteoblastogenesis and hematopoiesis remained untested.
To obtain this lacking evidence on the effect of IFNγ on the bone marrow milieu, in this study, we looked at the effect of IFNγ on both a model of adipocyte-differentiating human MSC in vitro and on marrow fat of OVX mice, a mouse model that was selected since OVX is a potent inducer of marrow fat infiltration . In both models, we used the same osteogenic doses as previously reported [8, 9]. Our results showed that the higher (and osteogenic) doses (10 and 100 ng/ml) of IFNγ significantly reduced adipocyte differentiation and lipid deposition in vitro. Furthermore, expression of adipocyte-specific proteins was negatively affected by IFNγ treatment at the transcriptional and protein levels in a dose-dependent manner. In addition, when testing the effect of IFNγ on PPARγ we found that, as in the case of other compounds that simultaneously inhibit adipogenesis while favor osteoblastogenesis [22–24], the effect of IFNγ was mostly targeted to PPARγ2 expression. This effect is relevant since the activation of the PPARγ2 (with or without activation of PPARγ1) directs MSC differentiation toward the adipocyte lineage at the expense of osteoblast formation , an effect that was inhibited by IFNγ administration. Finally, we also observed that the inhibitory effect of IFNγ on PPARγ is not only limited to its transcription and translation but also limited to its DNA-binding, thus decreasing the expression of PPARγ target genes.
When compared with previous evidence on the osteogenic effect of IFNγ on differentiating MSC , this evidence suggests that the osteogenic doses of IFNγ enhance the shift of human MSC from adipogenesis into osteoblastogenesis, thus increasing bone formation. In addition, from a mechanistic correlation, the previously reported stimulatory effect of IFNγ on several transcription factors for osteoblastogenesis [8, 9] correspond with the inhibitory effect on the critical transcription factors for adipogenesis evaluated in this study. This is a dual and dose-dependent action of IFNγ on MSC that could be used for therapeutic purposes in the near future.
Interestingly, when we treated OVX animals with osteogenic doses of IFNγ, we found a dose-dependent effect on marrow fat in the treated mice, which also correlated with low levels of expression of adipogenic factors within the bone marrow. This finding is relevant since it seems to be independent of the anabolic effect on bone mass as suggested by the high amount of hematopoietic bone marrow and by the increase in A/H ratio noticed in the mice treated with the higher osteogenic doses of IFNγ. This increase in hematopoiesis in IFNγ-treated mice is unexpected because IFNγ has been reported to inhibit hematopoiesis through the activation of lymphocytes . Considering that IFNγ increases the T lymphocyte repository within the bone marrow in OVX mice , it would be expected that treatment with IFNγ would also decrease hematopoiesis. In this case, higher levels of hematopoiesis observed in IFNγ-treated OVX mice could be mostly associated with the significant reduction in marrow adipocytes, which are known to be strong inhibitors of hematopoiesis , thus confirming previous theories that shifting the balance between adipogenesis and osteoblastogenesis substantially influences hematopoiesis .
In conclusion, we are reporting that osteogenic doses of IFNγ negatively affect adipogenesis. Decreasing adipocyte differentiation and hence adipocyte numbers not only favors osteoblastogenesis and hematopoiesis but also improves the overall bone microenvironment. This dual and dose-dependent effect of IFNγ on bone marrow cellularity is an important finding for the understanding of the effect of IFNγ on bone cells and for the potential use of IFNγ in the treatment of osteoporosis in older persons in the near future.
This study was supported by the Australian National Health and Medical Research Council (NHMRC 632767), the Nepean Medical Research Foundation to Dr. Duque, and by the Canadian Institutes for Health Research (CIHR MOP 10839) to Dr. Kremer.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.