Hypoxia Inhibition of Adipocytogenesis in Human Bone Marrow Stromal Cells Requires Transforming Growth Factor-β/Smad3 Signaling*

Although hypoxia and transforming growth factor-β (TGF-β) inhibit differentiation of adipocytes from preadipocytes and bone marrow-derived cells in several species, the relationship between hypoxia and TGF-β signaling in adipocytogenesis is unknown. In this study, we evaluated the mechanisms of inhibition of adipocyte differentiation by hypoxia and TGF-β in human and murine marrow stromal cells (MSCs) and the role of TGF-β/Smad signaling in the inhibition of adipocytogenesis by hypoxia. Both hypoxia-mimetic deferoxamine mesylate (DFO) and TGF-β1 inhibited adipocyte differentiation (1.0% versus the control at 15 μm DFO and 1.4% versus the control at 1 ng/ml TGF-β1) and adipocyte gene expression (peroxisome proliferator-activated receptor-γ2 and lipoprotein lipase) in human MSCs after 21 days of treatment. Hypoxia (2% O2) and DFO (but not TGF-β1) increased hypoxia-inducible factor-1α as shown by Western blotting. Macroarrays and Western and Northern blot analyses showed that hypoxia activated the TGF-β/Smad signaling pathway and that both hypoxia and TGF-β1 modulated adipocyte differentiation pathways such as the insulin-, peroxisome proliferator-activated receptor-γ-, phosphatidylinositol 3-kinase-, and MAPK-associated signaling pathways. Studies with mouse marrow stromal cell lines derived from Smad3+/+ or Smad3–/– mice revealed that the TGF-β type I receptor (ALK-5) and its intracellular signaling molecule Smad3 were necessary for the inhibition of adipocyte differentiation by both TGF-β and hypoxia-mimetic DFO. Thus, the TGF-β/Smad signaling pathway is required for hypoxia-mediated inhibition of adipocyte differentiation in MSCs.

Oxygen homeostasis represents an important organizing principle for human development and physiology. Dysregulation of oxygen homeostasis is found in inflammatory and cardiovascular diseases, cancer, cerebrovascular disease, and chronic obstructive pulmonary disease (1,2). One of the immediate sequelae of bone fracture is regional hypoxia resulting from vasculature disruption (3). In response to hypoxia or low O 2 tension, mammalian tissues show increased expression of a wide variety of genes that stimulate erythropoiesis, angiogenesis, and glycolysis (4). Most of the hypoxia-regulated genes are transcriptionally up-regulated by hypoxia-inducible factor-1␣ (HIF-1␣) 1 (5)(6)(7).
Adipocyte differentiation of murine 3T3-L1 preadipocytes is inhibited under hypoxic conditions (0.01-2% O 2 ), and its inhibition is not observed in mouse embryonic fibroblasts deficient in HIF-1␣ (8). Low levels of O 2 reduce in vitro adipocytogeneis of murine skeletal muscle satellite cells compared with 20% O 2 , a finding that has been corroborated by similar observations in the pluripotent mesenchymal cell line C3H10T1/2 and 3T3 cells (9). A recent report shows that hypoxia-dependent inhibition of adipocyte differentiation of murine 3T3-F442A preadipocytes can be mediated by mitochondrial reactive oxygen species generation (10). Transforming growth factor-␤ (TGF-␤) was reported to strongly inhibit adipocytogenesis and the amount of fat in adipocytes that develops in cultures of human bone marrow-derived stromal cells (11). It has also been observed that TGF-␤ inhibits adipocytogenesis of the murine 3T3-F442A and NIH3T3 cell lines by signaling through Smad3 (12,13). In several systems, various effects of hypoxia appear to be mediated by TGF-␤ signaling. In human peritoneal fibroblasts, hypoxia increases TGF-␤1 and TGF-␤ type I and II receptor mRNA levels, with no effect on TGF-␤2 or TGF-␤3 (14). In human umbilical vein endothelial cells, hypoxia results in phosphorylation and nuclear transportation of Smad2 and Smad3 proteins as well as stimulation of the transcriptional activities of Smad3 and HIF-1␣ and up-regulation of TGF-␤2 and TGF-␤ type II receptor gene expression (15,16). In human dermal fibroblasts, up-regulation of COL1A1 mRNA levels by hypoxia is blocked by a TGF-␤1 antisense oligonucleotide and fails to occur in fibroblasts from TGF-␤1 Ϫ/Ϫ mice (17). It is unknown whether TGF-␤ signaling has a role in the inhibition of adipocytogenesis by hypoxia.
This study was designed to test the hypothesis that the TGF-␤/Smad signaling pathway is required for the inhibition of adipocyte differentiation of human marrow stromal cells (hMSCs) by hypoxia. hMSCs have the potential to differentiate to lineages of mesenchymal tissues, including bone, cartilage, fat, tendon, and muscle (18). Loss in bone volume (associated with osteoporosis and age-related osteopenia) is accompanied by an increase in bone marrow adipose tissue (reviewed in Ref. 19). Recently, we demonstrated cooperation between the TGF-␤ and Wnt signaling pathways in their stimulation of chondrocytogenesis and inhibition of adipocytogenesis in hMSCs (20). The results presented here, together with those from our previous studies, demonstrate that several signaling pathways are involved in the inhibition of adipocyte differentiation of hMSCs and thus provide a basis for possible approaches to prevent and/or treat obesity and osteopenia/osteoporosis.

EXPERIMENTAL PROCEDURES
Cell Culture-Adherent hMSCs were prepared from femoral bone marrow that was obtained as discarded material from a 42-year-old woman undergoing total hip replacement for osteoarthritis (21). Low density mononuclear cells were isolated by density centrifugation with Ficoll/Histopaque 1077 (Sigma). This procedure enriches for undifferentiated cells, including a fraction of non-adherent hematopoietic cells and a stromal fraction capable of adherence and differentiation into various connective tissue cells (21). The adherent fraction was expanded in monolayer culture with phenol red-free ␣-minimal essential medium (Invitrogen), 10% heat-inactivated fetal bovine serum (FBS-HI; Atlanta Biologicals, Inc., Norcross, GA), and antibiotics (100 units/ml penicillin and 100 g/ml streptomycin; Invitrogen). Mouse Smad3 ϩ/ϩ and Smad3 Ϫ/Ϫ MSCs were derived from bone marrow cultures of wild-type and Smad3-null mice, respectively (22). The Smad3null mice were provided by Dr. K. C. Flanders (National Institutes of Health, Bethesda, MD). The murine MSCs were maintained in McCoy's 5A medium (Invitrogen) with 10% FBS until they reached confluence.
Conditions for Adipocyte Differentiation of hMSCs-After hMSCs reached confluence, the medium was changed to ␣-minimal essential medium (phenol red-free), 1% FBS-HI, and antibiotics (100 units/ml penicillin and 100 g/ml streptomycin) with different combinations of the following adipocytogenic supplements: 10 g/ml insulin (referred to as I), 1 M dexamethasone (referred to as D), 0.5 mM 3-isobutyl-1methylxanthine (referred to as M), and 100 M indomethacin (referred to as Y) (Sigma). After 6 days of treatment, cultures were inspected by light microscopy for the presence of lipid-containing cells. RNA was extracted with TRIzol (Invitrogen) and analyzed by reverse transcription (RT)-PCR for expression of the peroxisome proliferator-activated receptor-␥2 (PPAR␥2) and lipoprotein lipase (LPL) adipocyte marker genes. Lipid-containing cells were observed in four of the supplement combinations: DM, IDM, DMY, and IDMY. PPAR␥2 expression was detected in cultures treated with DM or IDM and was up-regulated in hMSC cultures treated with DY and IDY, with which adipocytes were not observed. The greatest LPL expression was detected in cells treated with IDM compared with DM, DMY, and IDMY. In cultures without visible lipid-containing cells, LPL was up-regulated only in the group treated with IDY. Neither PPAR␥2 nor LPL was detected by RT-PCR under basal conditions. These data indicate that up-regulation of PPAR␥2 and LPL was achieved with the combination of insulin, dexamethasone, and 3-isobutyl-1-methylxanthine and that addition of indomethacin to the treatment did not increase gene expression or apparent accumulation of adipocytes in hMSCs. For most experiments, IDM supplementation was used as the standard adipocytogenic condition.
Regulation of Adipocyte Differentiation in Human and Murine MSCs-hMSCs were cultured in 12-well tissue culture plates; after they reached confluence, the medium was changed to ␣-minimal essential medium containing 1% FBS-HI and supplemented with IDM, which had been shown to promote PPAR␥2 and LPL gene expression and development of lipid-containing cells as described above. hMSCs were treated with 15 M hypoxia-mimetic deferoxamine mesylate (DFO; 15 mM stock solution in tissue culture-grade sterile phosphate-buffered saline (PBS); Sigma), with 1 ng/ml TGF-␤1 (R&D Systems, Minneapolis, MN), or with the vehicle control (0.1% bovine serum albumin in PBS). Three weeks later, lipid accumulation in cells was visualized by staining with 0.3% oil red O (Sigma) (20). The lipid-containing cells in three to six random areas of 1 mm 2 were enumerated in each culture well (n ϭ 9) per group. The means Ϯ S.D. are presented. The phosphatidylinositol 3-kinase (PI3K) inhibitor LY294002, the p38 MAPK inhibitor SB203580, the p42/44 MAPK inhibitor PD098059, and the TGF-␤ type I receptor ALK-5 (activin receptor-like kinase-5) inhibitor SB431542 were purchased from Sigma, and the stock solutions in Me 2 SO were kept at Ϫ20°C. For murine MSCs, after reaching confluence, the medium was changed to McCoy's 5A medium containing 1% FBS-HI and supplemented with IDM for 4 days, and then the adipocytogenic supplements were changed to 10 g/ml insulin only. Seven days after treatment with the vehicle control, 15 M DFO, or 1 ng/ml TGF-␤1 with or without 10 M SB431542, lipid accumulation in cells was visualized by staining with 0.3% oil red O. The lipid-containing cells in three to six random areas of 1 mm 2 were enumerated in each culture well (n ϭ 3) per group.
Cell Viability Assay-Confluent cultures of hMSCs (from a 42-yearold woman) were exposed to increasing concentrations of DFO (1-1,000 M). Cells were collected after 24 h by trypsinization, and, including cells floating in the medium, cell viability was assessed using the trypan blue exclusion assay. The percentage of viable cells was calculated as the number of viable cells (unstained) divided by the sum of dead (stained) and viable cells.
Western Blotting-hMSCs (from a 42-year-old woman) were cultured in 100-mm dishes. After reaching confluence, the cells were cultured in ␣-minimal essential medium containing 1% FBS-HI and supplemented with IDM under hypoxic (2% O 2 ) or standard (19% O 2 ) conditions with or without DFO (15 M) or TGF-␤1 (1 ng/ml) for 16 h. Whole cell lysates were prepared with urea denaturing buffer containing 6.7 M urea, 10 mM Tris-HCl (pH 6.8), 5 mM dithiothreitol, 1% SDS, 10% glycerol, and a mixture of protease inhibitors (Roche Diagnostics). They were homogenized with a Kontes Pellet Pestle and separated from insoluble cell materials by centrifugation at 16,000 ϫ g in an Eppendorf bench-top centrifuge at 4°C. Protein concentration was determined using the BCA system (Pierce). Western immunoblotting was performed as described (20,23). In brief, proteins were resolved electrophoretically through 4 -12% SDS-polyacrylamide gel (NuPAGE ® BisTris gel, Invitrogen) and transferred to polyvinylidene difluoride membranes (Amersham Biosciences). The membranes were blocked with 5% nonfat milk in PBS containing 0.1% Tween 20 (PBST) for 2-3 h at room temperature and then incubated overnight with primary antibodies at 4°C. The anti-HIF-1␣ primary antibody was purchased from BD Biosciences; the rabbit anti-phospho-Smad2/3 (Ser 433 /Ser 435 ) IgG antibody was purchased from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); and the anti-␤-actin antibody was purchased from Sigma. After removal of the unbound primary antibodies by three 10-min washes with PBST, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature and washed three times for 10 min with PBST. The horseradish peroxidase-conjugated anti-mouse IgG secondary antibody was purchased from Amersham Biosciences, and the horseradish peroxidase-conjugated anti-rabbit IgG antibody was purchased from Santa Cruz Biotechnology, Inc. The antibody-associated protein bands were visualized with the ECL Plus Western blotting system (Amersham Biosciences).
RNA Isolation and RT-PCR-Total RNA was isolated from MSCs with TRIzol. For RT-PCR, 2 g of total RNA was reverse-transcribed into cDNA with SuperScript II (Invitrogen) following the manufacturer's instructions. One-tenth of the cDNA was used in each 50-l PCR (30 -35 cycles of 94°C for 1 min, 55-60°C for 1 min, and 72°C for 2 min) with gene-specific primers. The human LPL forward primer is 5Ј-GAGATTTCTCTGTATGGCACC-3, and the reverse primer is 5Ј-CT-GCAAATGAGACACTTTCTC-3Ј (1261-1536 bp; GenBank TM accession number NM_000237.1). Primers for the human PPAR␥2 (24) and mouse adipsin (25) genes were used for amplification as described previously.
Gene Macroarray-GEArray TM human TGF-␤/bone morphogenetic protein signaling pathway and insulin pathway macroarrays (Super-Array Bioscience Corp., Bethesda, MD) were performed as described previously (20,26). In brief, 3 g of total RNA was reverse-transcribed with [␣-32 P]dCTP (PerkinElmer Life Sciences) and the reagents provided by the manufacturer. The labeled cDNAs were hybridized overnight to the macroarrays; washed; and exposed to x-ray film for 5, 16 -24, 48, and 72 h. Digital images were obtained by scanning autoradiographs with an Epson transparency adapter. Data were extracted from the images with ScanAlyze software as described in detail elsewhere (26). Data were organized by functional category of the genes, and a qualitative measure of relative expression under control conditions is indicated. Genes whose expression levels under control conditions were measured during short film exposures (5-24 h) were classified as high expression (ϩϩϩ); 48-h film exposures were classified as low expression (ϩϩ); and 72-h film exposures were classified as very low expression (ϩ). Undetectable or absent calls (Ϫ) were made for pixel intensity equal to the background level. The expression value for each gene on the array was calculated with GEArrayAnalyzer TM software provided by the manufacturer and was verified by visual comparison with the autoradiographs. The usable linear range of pixel intensity was established as 9,000 -65,000 as described previously (26). The quantitative magnitude of change evoked by treatment is expressed as fold change upon treatment relative to the control. A qualitative increase or decrease or no change evoked by treatment relative to the control is indicated for genes that were expressed at levels too low for quantification or where a neighboring strong spot bled into the feature of interest.
Northern Blot Analysis-Northern blot analysis was performed with modification as described previously (26). In brief, 10 g of total RNA was electrophoretically resolved on formaldehyde-containing 1% agarose gel and blotted onto a positively charged nylon membrane (Roche Diagnostics) by downward capillary transfer in 20ϫ SSC. The nylon membrane was prehybridized in formamide prehybridization/hybridization buffer (5ϫ SSC, 5ϫ Denhardt's solution, 1% SDS, 50% formamide, and 0.1 mg/ml denatured salmon sperm DNA) for 3 h at 42°C. Hybridization was performed overnight in formamide prehybridization/ hybridization buffer with random-labeled probes at 42°C. The probes for human TGFBI, plasminogen activator inhibitor-1 (PAI-1), PKM2, and INSR (insulin receptor) were labeled with a random-primed DNA labeling kit (Roche Diagnostics) with 25 ng of RT-PCR products of each gene and purified using the QIAquick nucleotide removal kit (Qiagen Inc.). The primers for human TGFBI (26), PAI-1 (27), PKM2 (28), and INSR (29) were described previously. The internal control 18 S oligonucleotide probe (Ambion Inc.) was labeled with a DNA 5Ј-end labeling kit (Roche Diagnostics). For removal of nonspecifically bound probes, membranes were washed twice with 2ϫ SSC with 0.1% SDS (5 min each) and twice with 0.2ϫ SSC and 0.1% SDS (5 min each) at room temperature and twice with 0.2ϫ SSC and 0.1% SDS (15 min each) at 42°C. Hybridization was visualized by autoradiography at Ϫ80°C with Eastman Kodak X-Omat Blue XB-1 film.
Statistical Analysis-All experiments were performed two to five times. Data are presented as the means Ϯ S.D. Quantitative data were analyzed by one-way analysis of variance. A p value Ͻ0.05 was considered significant.

Smad3 Is One of the Mediators of Inhibition of Adipocyte Differentiation in Murine MSCs by Hypoxia or TGF-␤-Murine
MSCs were used to assess the role of Smad3 in mediating the inhibition of adipocytogenesis. Smad3 Ϫ/Ϫ MSCs were derived from bone marrow cultures of neonatal Smad3-null mice; Smad3 ϩ/ϩ MSCs were from wild-type mice. There were two notable differences in adipocyte development in bone marrow cultures from Smad3 Ϫ/Ϫ and Smad3 ϩ/ϩ mice. First, there was more extensive (37-fold) adipocyte differentiation in bone marrow cultures from Smad3 Ϫ/Ϫ mice than from Smad3 ϩ/ϩ mice cultured under adipocytogenic conditions (Fig. 5, A and  B); this finding suggests that Smad3 is a critical inhibitor of adipocyte differentiation. Second, there was a 7-fold attenuation of TGF-␤ inhibition and a 2-fold attenuation of DFO inhibition of adipocytogenesis in cells lacking Smad3 (Fig. 5B). Hypoxia-mimetic DFO (15 M) decreased adipocyte number in Smad3 ϩ/ϩ MSCs (32.9% versus the control; p Ͻ 0.05) compared with Smad3 Ϫ/Ϫ MSCs (67.7% versus the control; p Ͻ 0.05). TGF-␤1 (1 ng/ml) decreased adipocyte number in Smad3 ϩ/ϩ cells (5.5% versus the control; p Ͻ 0.05) compared with Smad-3 Ϫ/Ϫ cells (39.2% versus the control; p Ͻ 0.05). These data support the conclusion that Smad3 is required for the inhibition of adipocyte differentiation by either TGF-␤ or hypoxia. The effect of Smad3 on lipid accumulation was reflected in the expression pattern of the adipocyte marker gene adipsin in Smad3 ϩ/ϩ and Smad3 Ϫ/Ϫ MSCs. After 7 days of treatment, 15 M DFO or 1 ng/ml TGF-␤1 down-regulated adipsin gene expression in Smad3 ϩ/ϩ MSCs, but not in Smad3 Ϫ/Ϫ MSCs (Fig.  5C). SB431542 (10 M), a specific inhibitor of the TGF-␤ type I receptor (ALK-5), antagonized the inhibitory effects of DFO and TGF-␤1 on adipsin gene expression in Smad ϩ/ϩ MSCs (Fig.  5C). This is confirmation that TGF-␤/Smad is required for the inhibition of adipocyte differentiation by either TGF-␤ or hypoxia.

DISCUSSION
Although hypoxia and TGF-␤ both inhibit adipocyte differentiation, the relationship between their signaling pathways in adipocytogenesis was unknown. In this study, we investigated the mechanisms of inhibition of adipocyte differentiation of MSCs by hypoxia and TGF-␤ as well as the role of the TGF-␤/ Smad signaling pathway in this process.
Hypoxia-induced gene expression is mediated by HIF-1␣. Hypoxia stabilizes the HIF-1␣ protein, which is otherwise (under normoxic conditions) degraded by a ubiquitin-dependent proteasome (5-7). Our results show that hypoxia and its mimetic DFO stabilized the HIF-1␣ protein in hMSCs and that continued treatment with DFO inhibited adipocyte differentiation and PPAR␥2 and LPL adipocyte gene expression in hM-SCs. It has been reported that many growth factors, cytokines, and circulatory factors, including insulin, insulin-like growth factor-1 and -2, platelet-derived growth factor, epidermal growth factor, epidermal growth factor-2, TGF-␤1, hepatocyte growth factor, tumor necrosis factor-1, interleukin-1␤, etc., stimulate HIF-1␣ through pathways distinct from that employed by the classical hypoxic pathway (30). Our results show that 1 ng/ml TGF-␤1 blocked adipocyte differentiation and down-regulated PPAR␥2 and LPL adipocyte gene expression, but did not change HIF-1␣ protein levels in hMSCs. These data indicate that hypoxia may inhibit adipocytogenesis of hMSCs by stabilizing HIF-1␣, but that, under these conditions, TGF-␤ does not require the stabilization of HIF-1␣ for its inhibition of adipocytogenesis of hMSCs.
To compare the effects of hypoxia and TGF-␤ on adipocyteassociated signaling, we used a targeted macroarray that included several insulin receptor-associated genes and insulin signaling target genes, PI3K pathway components and target genes, MAPK pathway components and target genes, insulinactivated transcription factor SREBP1 target genes, and PPAR␥ target genes. The macroarray data were confirmed by Northern blot analysis of several highly expressed genes.
b Arrows indicate the qualitative change by hypoxia in genes for which quantification was not possible: 1, increase; 2, decrease; 7, no change.
togenesis is the insulin-activated transcription factor SREBP1c (sterol regulatory element-binding protein-1c)/ADD1 (32,33). It is a basic helix-loop-helix protein expressed abundantly in adipose tissue (34) and has been shown to increase fatty acid and fat synthesis; this has been attributed in part to its proposed influence on PPAR␥ activity (35). Expression of SREBP1/ ADD1 is also increased during osteoblast differentiation; however, this indicates that increased expression of this protein is not specific to adipocyte differentiation (36). Our results show that hypoxia and TGF-␤ had no effect on SREBP1 gene expression in hMSCs. Insulin, the major anabolic hormone, promotes in vivo accumulation of adipose tissue. Insulin is a potent inducer of adipocytogenesis, and differentiation of adipocytes requires many components of the insulin signaling pathway, including the insulin receptor, insulin receptor substrate-1, and PI3K (37). There is growing interest in the effects of age on redistribution of body fat deposits, on insulin actions on site-specific adipocytes and their progenitors, and on dysdifferentiation of mesenchymal precursors into mesenchymal adipocyte-like default cells (38). Our macroarray results show strikingly similar effects of hypoxia and TGF-␤ in decreasing expression of the insulin receptor-associated genes INSR, NCK2, CRK, and PPP1CA and the insulin target gene v-jun. These findings suggest that both treatments block insulin signaling as part of their inhibition of adipocyte differentiation in MSCs.
PI3K is important in a wide variety of cellular processes, including intracellular trafficking; organization of the cytoskeleton; cell growth, differentiation, and transformation; and pre-vention of apoptosis (39,40). Pharmaceutical inhibition of PI3K or overexpression of dominant-negative mutant subunits abrogates differentiation of adipocytes in vitro (41,42); this suggests that the insulin/PI3K/AKT signaling pathway is required for adipocytogenesis. Our RT-PCR results show that the PI3K inhibitor LY294002 down-regulated PPAR␥2 and LPL adipocyte gene expression; this demonstrates that the PI3K pathway promotes adipocyte differentiation in hMSCs. Our macroarray results show that hypoxia down-regulated PI3K pathway-associated genes and increased expression of PI3K target genes such as PAI-1 and VEGF. TGF-␤1 had similar effects on PI3K signaling gene expression compared with hypoxia. Although PAI-1 and VEGF were named as PI3K pathway target genes in this commercial macroarray, the induction of the PAI-1 (43,44) and VEGF (45) genes by hypoxia involves HIF-1, hypoxia response elements in the PAI-1 and VEGF gene promoters, and PI3K/AKT (protein kinase B). Therefore, we cannot conclude whether hypoxia activates the PI3K pathway through up-regulation of the PAI-1 and VEGF genes by hypoxia. The effects of hypoxia and TGF-␤ on the PI3K pathway in adipocyte differentiation in hMSCs need to be further evaluated. Down-regulation of PI3K pathway components, including PIK3R1, PIK3R2, GSK3A, and GSK3B, etc., may contribute to the inhibition of adipocytogenesis in hMSCs by hypoxia and TGF-␤ because the PI3K inhibitor down-regulated PPAR␥2 and LPL gene expression.
Both hypoxia and TGF-␤ inhibited the p42/44 MAKP signaling pathway during adipocyte differentiation of hMSCs. Reports in the literature concerning the role of the p42/44 MAPK pathways in adipocyte differentiation have been contradictory (46). p42/44 MAPK is required (47,48) or has no effect or inhibits (49) adipocyte differentiation of preadipocytes under different conditions. Under our experimental conditions (1% FBS-HI, IDM), the p42/44 MAPK inhibitor PD98059 up-regulated the PPAR␥2 gene and had no effect on LPL gene expression in hMSCs; this indicates that p42/44 MAPK may inhibit adipocyte differentiation in hMSCs. Thus, the inhibition of the p42/44 MAPK signaling pathway by both hypoxia and TGF-␤ is unlikely to mediate their inhibition of adipocyte differentiation in hMSCs. It has been reported that JNK suppresses the process of adipocyte differentiation (50), and p38 MAPK has been shown to promote adipocyte differentiation (51). It was unknown whether p38 MAPK and JNK play roles in the effects of hypoxia or TGF-␤ on adipocyte differentiation. It has been reported that hypoxia activates JNK and p38 MAPK in carcinoma cells (52). Like hypoxia, TGF-␤ has also been reported to activate the JNK pathway in skeletal muscle cells (53) and the p38 MAPK signaling pathway in murine mesangial cells (54). The effects of hypoxia and TGF-␤1 on JNK and p38 MAPK need to be evaluated further in our system. Nevertheless, our RT-PCR data show that the p38 MAPK inhibitor blocked PPAR␥2 and LPL gene expression in hMSCs and imply that p38 MAPK is an activator of adipocytogenesis in hMSCs. Thus, the inhibition of adipocyte differentiation is not through activation of p38 MAPK by hypoxia or TGF-␤ in hMSCs.
Based on these new data, we propose that hypoxia inhibition of adipocytogenesis in hMSCs requires TGF-␤/Smad signaling. Hypoxia up-regulated TGF-␤1, its intracellular signaling molecule Smad3, and many of its target genes such as TGFBI, PAI-1, TIMP1, COL1A2, and p21 WAF1/CIP1 in hMSCs. The experiments with mouse Smad3 cell lines showed that Smad3 was required for inhibition of adipocytogenesis by either DFO or TGF-␤1. In addition, SB431542 abrogated the inhibitory effects of DFO on adipocyte gene expression. SB431542 has been identified as an inhibitor of ALK-5 (TGF-␤ type I receptor) (55). These data demonstrate that hypoxia inhibition of adipocyte differentiation in MSCs requires activation of the TGF-␤/ Smad signaling pathway. There is no information about bone marrow or body fat in Smad3 Ϫ/Ϫ mice or about the mechanism of their substantial perinatal mortality. Recently, however, we reported that long-term bone marrow cultures established from neonatal Smad3 Ϫ/Ϫ mice show 16.5-fold more adipocytes in the adherent layer and prolonged hematopoiesis (Ͼ20 weeks) compared with Smad3 ϩ/ϩ mice (56). Those data confirm the negative regulation by endogenous TGF-␤ of adipocytogenesis and long-term hematopoiesis in vitro.
Hypoxia activation through TGF-␤/Smad signaling is known in other systems. It has been reported that hypoxia stimulates collagen synthesis and COL1A1 transcription through the action of TGF-␤1 in human dermal fibroblast (17). Cooperation between the hypoxia and TGF-␤ pathways is required for regulation of several genes, including endoglin (57), erythropoietin (58), VEGF (59), etc. Hypoxia results in a significant increase in TGF-␤1 and TGF-␤ type I and II receptor mRNA levels in human peritoneal fibroblasts (14). Exposure of human umbilical vein endothelial cells to hypoxia results in phosphorylation and nuclear transportation of Smad2 and Smad3 proteins as well as stimulation of the transcriptional activities of Smad3 and HIF-1␣ and culminates in up-regulation of TGF-␤2 gene expression (16). Thus, Smad proteins may play an important role in vascular responses to hypoxia and ischemia (15). It has been reported that reactive oxygen species mediate TGF-␤induced expression of PAI-1 in rat mesangial cells (60) and TIMP3 in human and bovine primary articular chondrocytes (61) and apoptosis in rat fetal hepatocytes (62). It is unknown whether reactive oxygen species have a role in the TGF-␤/ Smad-mediated inhibition of adipocytogenesis by hypoxia in hMSCs.
Understanding the balance between positive and negative regulators of adipogenesis has important health-related implications for anti-obesity medical therapy and lipodystrophy (63). TGF-␤ has been shown to be increased in adipose tissue in obese mice (64). However, overexpression of TGF-␤1 in adipose tissue in transgenic mice results in a dramatic reduction in total body fat (65). Adipose tissue growth is angiogenesis-dependent (66). Hypoxia has been reported to increase the expression of pro-angiogenic factors in adipocytes and fat tissues (67,68). Our results show that hypoxia and TGF-␤ modulated expression of pro-angiogenic genes such as VEGF and angiogenin and inhibited adipocyte differentiation in hMSCs. The relationship between the effects of hypoxia and TGF-␤ on adipocyte differentiation of MSCs and adipose tissue in obesity merits further investigation. Because of the relationship between bone marrow fat and skeletal aging, these results may also have implications for the regulation of MSC differentiation into adipocytes and osteoblasts.
In summary, hypoxia and TGF-␤ inhibited adipocyte differentiation and modulated adipocyte-associated signaling pathways in MSCs. Hypoxia activated the TGF-␤/Smad signaling pathway, and the TGF-␤ intracellular signaling molecule Smad3 was necessary for the inhibition of adipocytogenesis by both TGF-␤ and hypoxia. These findings indicate that TGF-␤/ Smad signaling is required for the inhibition of adipocyte differentiation by hypoxia.