Activation of Osteoclastogenesis and Rosiglitazone-Induced Bone Loss

View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Cell Metabolism Article

PGC1b Mediates PPARg Activation of Osteoclastogenesis and Rosiglitazone-Induced Bone Loss

Wei Wei,1 Xueqian Wang,1 Marie Yang,1 Leslie C. Smith,2 Paul C. Dechow,2 and Yihong Wan1,* 1Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, TX 75390, USA 2Department of Biomedical Sciences, Baylor College of Dentistry, Texas A&M University Health Sciences Center, Dallas, TX 75246, USA *Correspondence: [email protected] DOI 10.1016/j.cmet.2010.04.015

SUMMARY including osteoporosis, arthritis, and bone metastasis of cancers (Novack and Teitelbaum, 2008). Long-term usage of rosiglitazone, a synthetic PPARg Peroxisome proliferator-activated receptor g (PPARg)isa agonist, increases fracture rates among diabetic member of the nuclear receptor family of transcription factors patients. PPARg suppresses osteoblastogenesis that can be activated by lipophilic ligands (Evans et al., 2004; while activating osteoclastogenesis, suggesting that Tontonoz and Spiegelman, 2008). It regulates a diverse spec- rosiglitazone decreases bone formation while sus- trum of physiological processes, including adipogenesis (Tonto- taining or increasing bone resorption. Using mouse noz et al., 1994), lipid metabolism (Chawla et al., 2001; Tontonoz et al., 1998), insulin sensitivity (Agostini et al., 2006; Barroso models with genetically altered PPARg, PGC1b,or et al., 1999; He et al., 2003), and inﬂammation (Jiang et al., ERRa, here we show that PGC1b is required for 1998; Ricote et al., 1998; Wan et al., 2007b), as well as diseases the resorption-enhancing effects of rosiglitazone. such as diabetes, obesity, and atherosclerosis (Evans et al., PPARg activation indirectly induces PGC1b expres- 2004; Lehrke and Lazar, 2005; Tontonoz and Spiegelman, 2008). sion by downregulating b-catenin and derepressing Its importance is accentuated by the widespread use of thiazo- c-jun. PGC1b, in turn, functions as a PPARg coacti- lidinediones (TZDs), synthetic PPARg ligands, as drugs for vator to stimulate osteoclast differentiation. Com- insulin resistance and type 2 diabetes, including Avandia (rosigli- plementarily, PPARg also induces ERRa expression, tazone or BRL 49653) and Actos (pioglitazone) (Lehmann which coordinates with PGC1b to enhance mito- et al., 1995; Nolan et al., 1994; Tontonoz and Spiegelman, chondrial biogenesis and osteoclast function. ERRa 2008). Epidemiological studies suggest that skeletal fragility is knockout mice exhibit osteoclast defects, revealing increased in type 2 diabetes mellitus (Grey, 2009; Strotmeyer and Cauley, 2007). Recent clinical trials report that long-term ERRa as an important regulator of osteoclastogene- use of rosiglitazone increased fracture rates among diabetic sis. Strikingly, PGC1b deletion in osteoclasts con- patients (Grey, 2009; Home et al., 2009; Kahn et al., 2006, 2008). fers complete resistance to rosiglitazone-induced Thus, TZD administration exacerbates skeletal fragility in a bone loss. These ﬁndings identify PGC1b as an population already at increased fracture risk, and it is of para- essential mediator for the PPARg stimulation of mount importance and urgency to elucidate the cellular and osteoclastogenesis by targeting both PPARg itself molecular mechanisms by which PPARg and TZDs regulate and ERRa, thus activating two distinct transcrip- bone remodeling. Moreover, TZD treatment causes bone loss tional programs. in mice and rats, indicating that these animal models represent relevant experimental systems to dissect TZD actions in bone (Ali et al., 2005; Lazarenko et al., 2007; Li et al., 2006; Sottile INTRODUCTION et al., 2004). Emerging evidence suggests that PPARg plays important Bone is a dynamic tissue that constantly remodels by balancing roles in skeletal homeostasis. PPARg suppresses osteoblast osteoclast-mediated bone resorption and osteoblast-mediated differentiation from mesenchymal stem cells (Akune et al., bone formation. Osteoclasts are derived from hematopoietic 2004; Barak et al., 1999; Cock et al., 2004; Kubota et al., 1999; progenitors (Ash et al., 1980) in the monocyte/macrophage Rosen et al., 1999). Our recent study reveals that PPARg also lineage (Scheven et al., 1986; Tondravi et al., 1997) and differen- promotes osteoclast differentiation from hematopoietic stem tiate in response to the tumor necrosis factor family cytokine cells (Wan et al., 2007a). Loss of PPARg function in mouse receptor activator of NFkB ligand (RANKL) (Lacey et al., 1998; hematopoietic lineages causes osteoclast defects manifested Yasuda et al., 1998); in contrast, osteoblasts are of mesen- as osteopetrosis, a disease characterized by increased bone chymal lineage (Pittenger et al., 1999). Bone homeostasis is nor- mass and extramedullary hematopoiesis in the spleen. Gain of mally maintained by the tight coupling of bone resorption and PPARg function by rosiglitazone (BRL) activation enhances bone formation (Edwards and Mundy, 2008). Pathological osteoclastogenesis and bone resorption in vitro and in vivo. increases in osteoclast activity and bone resorption, thus the Thus, TZDs increase skeletal fragility by inhibiting bone forma- uncoupling of bone remodeling, can cause several diseases, tion while sustaining or increasing bone resorption, leading to

Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. 503 Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

A B PGC1b c-fos TRAP 10 10 12 -cre/Veh *** *** 8 -cre/BRL 8 9 ** 6 +cre/Veh 6 +cre/BRL*** ** 6 4 ** 4 * * * 3 2 ** 2 n.s. Relative mRNA Relative mRNA ** n.s. * n.s. 0 0 0 0h 24h 48h 72h 0h 24h 48h 72h 0h 24h 48h 72h MMP9 CTSK 8 C ** 8 WT PGC1β-/- ** 6 BRL - - + - - + 6 RANKL - + + - + + 4 ** 4 1β/actin ratio 1 2.8 6.5 0.8 0.7 0.4 ** ** ** 2 * 2 β Relative mRNA * PGC1 n.s. n.s. β-actin 0 0 0h 24h 48h 72h 0h 24h 48h 72h D F 1.5 PGC1b-A 0.080 -R/V ***** ***** +R/B t

0.060 1

0.040 0.5 * ChIP/10%Inpu luc/b-gal ***** ***** 0.020 *** 0 IgG c-Jun AcH3 0.000 vec g/a jun fos 0.5jun 0.5fos 0.5jun NFATc1 p65 RANKL 2.5 PGC1b-B -R/V **** 0.5vec 0.5vec 0.5fos 2 +R/B t E mPGC1b-A -1690 CCCTA TGA G TAA GTGAT -1673 1.5 rPGC1b-A -1768 CATTG TGA G TAA GTGAT -1751 hPGC1b-A -1730 CCTTA TGA G TGT AGGGT -1713 1 ***

mPGC1b-B -924 CAGAG TGA C TCA GATGG -907 ChIP/10%Inpu rPGC1b-B -971 CAGAC TGA C TCA GATCA -954 0.5 hPGC1b-B -920 TGGTA TTT C TCC CTCCT -903 0 AP-1 Consensus TGA C/G TCA IgG c-Jun AcH3

G H I ***** PGC1b 0.4 ) 6 -3 BRL - - + 0 *** RANKL - + + 0.3 4 ** β-catenin 0.2 2 β luc/b-gal -actin ****

0.1 Relative (x1 mRNA Relative mRNA (x10 mRNA Relative 0 vector cre jun 0 β vector bCA jun jun/bCA -catenin flox/flox

Figure 1. PPARg Activation Induces PGC1b Transcription during Osteoclast Differentiation (A) RNA expression of PGC1b was induced during a 72 hr time course of RANKL treatment and was further stimulated by BRL. Bone marrow cells from gf/f control mice (Àcre) or gf/fTie2cre mutant mice (+cre) were cultured with MCSF for 3 days in the presence of BRL or vehicle. On day 4 (0 hr), the macrophage precursors were differentiated with RANKL and MCSF for 3 days (72 hr) in the presence of BRL or vehicle. ‘‘Veh’’ indicates RANKL alone; ‘‘BRL’’ indicates RANKL+BRL. The p values were calculated by comparing each time point with 0 hr baseline control; n = 3. (B) The BRL induction of PGC1b preceded the BRL induction of osteoclast marker genes.

504 Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

the uncoupling of bone remodeling and a net loss of bone RESULTS (Lazarenko et al., 2007; Wahli, 2008; Wan et al., 2007a). A recent clinical study examining bone biomarkers in participants ran- PPARg Activation Induces PGC1b Expression domized to rosiglitazone in a diabetes outcome progression trial during Osteoclast Differentiation (ADOPT) demonstrated that bone resorption was significantly In our previous study, we generated the PPARgflox/flox; Tie2cre+/À increased in women taking rosiglitazone (Zinman et al., 2009). (gf/fTie2cre) mice in which PPARg was deleted in hematopoietic In this study, we aim to determine the molecular mechanisms lineages, but not in mesenchymal lineages; thus, in osteoclasts, by which PPARg stimulates osteoclastogenesis. but not in osteoblasts (Wan et al., 2007a). We isolated bone PPARg modulates transcription through ligand-mediated marrow cells from gf/fTie2cre (+cre) mutants or gf/f (Àcre) con- recruitment of coactivators. Tissue-specific differences in coac- trols and performed ex vivo osteoclast differentiation in the tivator expression can affect PPARg function, adding another presence of macrophage colony stimulating factor (MCSF) and dimension to the complexity of its gene- and cell-specific tran- RANKL, with or without BRL (rosiglitazone) treatment. The induc- scriptional regulation (Yu and Reddy, 2007). To determine the tion of PGC1b during RANKL-mediated osteoclast differentiation PPARg coactivator in osteoclasts, we measured the expression was markedly potentiated by BRL (Figure 1A). PGC1b induction of several nuclear receptor coactivators and found that PGC1b by BRL and RANKL was PPARg dependent because it was abol- (peroxisome proliferator-activated receptor-gamma coactivator ished in the PPARgÀ/À cells differentiated from the bone marrow 1b, Ppargc1b) is highly upregulated during osteoclast differenti- of gf/fTie2cre mutants (Figure 1A). In contrast, a closely related ation. Thus, we hypothesize that PGC1b may be involved in coactivator PGC1a was not induced by BRL or RANKL (data PPARg regulation of osteoclastogenesis. PGC1b is a transcrip- not shown). The BRL induction of PGC1b preceded the BRL tional coactivator that regulates energy metabolism by stimu- induction of osteoclast marker genes (Figure 1B). Furthermore, lating mitochondrial biogenesis and respiration of cells (Kamei PGC1b protein induction by BRL and RANKL was confirmed et al., 2003; Lai et al., 2008; Lelliott et al., 2006; Lin et al., 2005; by western blot analysis (Figure 1C). These results showed that Sonoda et al., 2007b; Vianna et al., 2006). Intriguingly, a recent ligand activation of PPARg strongly stimulates PGC1b expres- study reported that PGC1b deletion causes defects in both sion during osteoclast differentiation. osteoclasts and osteoblasts. PGC1b was induced during osteo- To determine how PPARg and BRL induced PGC1b transcrip- clast differentiation by reactive oxygen species. Knockdown of tion, we cloned a 1.8 Kb PGC1b promoter into a luciferase PGC1b in vitro inhibited osteoclast differentiation and mitochon- reporter. Transient transfection analyses in several cell lines indi- dria biogenesis, and global PGC1b deletion in mice resulted in cated that BRL activation of PPARg had no significant effect on increased bone mass. However, the mechanisms underlying the luciferase readout, suggesting that PPARg induced PGC1b PGC1b regulation of osteoclasts were underexplored. In addi- via indirect mechanisms. Of interest, we observed that BRL tion, defects were also observed in PGC1b-deficient osteoblasts stimulated PGC1b expression only during RANKL-induced oste- (Ishii et al., 2009); thus, it was unclear whether PGC1b deletion in oclast differentiation, but not in the macrophage precursors osteoclasts was sufficient to cause a resorption defect. Here, before RANKL treatment (Figure 1A, time point 0 hr). This sug- we report that PGC1b is highly induced by BRL during osteo- gested that there was a functional crosstalk between PPARg clast differentiation in a PPARg-dependent manner. Moreover, and RANKL signaling and that BRL induction of the PGC1b pro- PGC1b is required for the pro-osteoclastogenic effect of BRL moter required component(s) in the RANKL pathway. To identify in vivo and ex vivo. PGC1b functions as a PPARg coactivator this component, we cotransfected various RANKL-activated to stimulate osteoclast differentiation. Complementarily, PGC1b transcription factors, including c-fos,c-jun, NFAT-c1, and the also coordinates with the BRL-induced ERRa to enhance mito- p65 subunit of NFkB, to determine which one(s) could induce chondrial biogenesis and osteoclast function. Importantly, using the PGC1b promoter. RANKL treatment of the cells transfected conditional PGC1b knockout mice, we have investigated the with PGC1b-luc alone indeed activated the PGC1b promoter specific requirement for osteoclastic PGC1b in rosiglitazone- by 1.4-fold. However, this induction was largely conferred by induced bone loss. c-jun, which robustly activated the PGC1b promoter by 7-fold,

(C) Western blot analysis showed that PGC1b protein level was induced by RANKL and further elevated by BRL in the WT, but not PGC1bÀ/À bone marrow differentiation culture after 72 hr. (D) Transient transfection assays showed that the PGC1b promoter was induced by c-jun. HEK293 cells were cotransfected with a PGC1b-luc reporter and each indicated transcription factor or were alternatively treated with RANKL overnight, 24 hr after transfection (RANKL, right). The same amount of total DNA was transfected; thus, half c-jun was transfected in ‘‘0.5 jun 0.5 fos’’ compared to ‘‘jun.’’ The p values were calculated by comparing each condition to the vector transfected and untreated control (left); n = 3. g, PPARg; a, RXRa; vec, vector. (E) Alignment of mouse, rat, and human PGC1b-A and PGC1b-B AP-1-binding regions, together with the AP-1 consensus. (F) ChIP analysis of c-jun binding to the endogenous mouse PGC1b-A and PGC1b-B promoter regions in bone marrow differentiation cultures with or without BRL and RANKL treatment; n = 3. (G) Both basal expression and c-jun induction of PGC1b promoter were significantly inhibited by b-catenin. An expression vector encoding a constitutive active b-catenin mutant (bCA) was cotransfected; n = 3. (H) Western blot analysis of bone marrow differentiation culture showed that the b-catenin protein level was downregulated by RANKL and further diminished by BRL. (I) b-catenin deletion or c-jun overexpression induced PGC1b expression. Macrophages were differentiated from the bone marrow of b-cateninflox/flox mice retro- virally transduced with cre, c-jun, or vector control. PGC1b mRNA expression was measured by RT-QPCR. Error bars in (A), (B), (D), (F), (G), and (I) represent means ± SD.

Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. 505 Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

A erc-b1 erc+b1 RANKL+Veh RANKL+BRL RANKL+Veh RANKL+BRL

B c-fos NFATc1 TRAP CAR2 Calcr +++++****** ****+++ 9 ++++*** *** 5 -R/V 7 3 +++ 8 ++++*** -R/B 6 4 +R/V 5 6 6 2 3 +R/B 4 4 3 2 3 1 2 ** 2 1

Relative mRNA 1 * * 0 * 0 0 0 0 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre

C PGC1b Ndg2 Aco2 IDH3a ATP5b *** 6 ** 5 ++ 4 4 3 ++ -R/V 5 **+ *** ***++ -R/B 4 3 3 ++ 4 2 +R/V 3 3 +R/B 2 2 2 2 *** 1 *** 1 * ** 1 1 *** Relative mRNA 1 *** **** 0 * 0 0 0 0 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre

ERRa MCAD VLCAD SCAD 3 2 3 2 *** ***++ ***++ ***++ + 2 2 * 1 * ** 1 1 1 Relative mRNA 0 0 0 0 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre

D E 0.9 -R/V **** mERRa (-1874) TCAA AGGTCA T AGGGGA CACT +R/B rERRa (-2070) TCAA AGACCA T AGGGGA CACT t 0.6 hERRa (-5255) GGAG AGTTCA A ACGACA GCCC ***

aP2 CTCT GGGTGA A ATGTGC ATTT PEPCK AACT GTGGTA A AGGTCT TGTT 0.3 ChIP/10%Inpu Acyl-CoA synthase AGGGCA T CAGTCA DR-1 consensus AGGTCA A AGGTCA 0 IgG PPARg PGC1b

Figure 2. PGC1b Is Required For Rosiglitazone Stimulation of Osteoclast Differentiation Bone marrow cells were isolated from PGC1bflox/flox; Tie2cre+/À (1b+cre) mutants or PGC1bflox/flox (1b-cre) controls and were differentiated ex vivo with RANKL and MCSF in the presence or absence of BRL treatment. (A) Representative images of the TRAP-stained osteoclast differentiation culture. Mature osteoclasts were identified as multinucleated TRAP+ cells. Scale bar, 25 mm. (B) BRL induction of osteoclast marker genes was impaired during osteoclast differentiation from PGC1bÀ/À bone marrow isolated from 1b+cre mutants. R, RANKL; V, vehicle; B, BRL. (C) BRL induction of ERRa and PGC1b, as well as their downstream targets of mitochondrial genes, was abolished during osteoclast differentiation from the PGC1bÀ/À bone marrow. Truncated PGC1b transcripts were detected using primers specific for sequences in exon 8 (Sonoda et al., 2007b). The p values

506 Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

whereas the other factors tested had no significant effect (Fig- expression of RANKL-induced transcription factors (c-fos and ure 1D). Moreover, c-jun induced PGC1b promoter in a dose- NFATc1) and osteoclast function genes (TRAP, CAR2, and Calcr) dependent manner because the luciferase output was reduced (Figure 2B). This indicated that PGC1b was required for BRL by 56% in cells transfected with half the amount of c-jun. In addi- stimulation of osteoclast differentiation, possibly as a coactivator tion, c-fos exerted neither activity nor interference because c-fos for either PPARg or other transcription factors in the specific had no effect on either the basal- or the c-jun-induced PGC1b context of osteoclastogenesis. promoter expression, suggesting that c-jun functioned as homo- dimers (Figure 1D). Two conserved AP-1 sites were identified in PGC1b Coordinates with ERRa to Activate the PGC1b promoter (Figure 1E). Chromatin immunoprecipita- Mitochondrial Function in Osteoclasts tion (ChIP) assay confirmed c-jun binding to these sites at the Previous studies have shown that PGC1b functions as a endogenous mouse PGC1b promoter in bone marrow-derived ligand-independent coactivator (or protein ligand) for estrogen osteoclast precursors upon BRL and RANKL treatment, which receptor-related receptor a (ERRa) to induce the expression was associated with increased histone H3 acetylation, indicating of medium-chain acyl-CoA dehydrogenase (MCAD), a pivotal activation of PGC1b transcription (Figure 1F). enzyme in mitochondrial fatty acid b-oxidation (FAO) (Kamei Surprisingly, cotransfection of a constitutively active b-catenin et al., 2003). PGC1b activation of ERRa also induces other mito- mutant repressed both the basal expression and the c-jun induc- chondrial target genes involved in the tricarboxylic acid (TCA) tion of the PGC1b promoter by 56% and 72%, respectively cycle and oxidative phosphorylation (OXPHOS), such as Ndg2 (Figure 1G). b-catenin is the downstream effector in the canon- (Nur77 downstream gene 2), Aco2 (aconitase 2), IDH3a (isoci- ical Wnt-signaling pathway and is regulated by protein degrada- trate dehydrogenase 3), and ATP5b (ATP synthase 5b) (Sonoda tion. Western blot analysis revealed that the b-catenin protein et al., 2007a). In light of the role of PGC1b in mitochondrial level was reduced upon RANKL treatment, which was further biogenesis during osteoclast activation (Ishii et al., 2009), we diminished by BRL in bone marrow-derived osteoclast precur- examined whether BRL could induce these ERRa/PGC1b target sors (Figure 1H). Furthermore, b-catenin deletion or c-jun genes. As shown in Figure 2C, BRL in conjunction with RANKL, overexpression is sufficient to induce PGC1b expression in but not BRL or RANKL alone, increased the expression of ERRa, bone marrow-derived macrophages (Figure 1I). Together, these thus resulting in the significant induction of its target genes, results demonstrated that BRL activation of PPARg indirectly including Ndg2, Aco2, IDH3a, ATP5b, MCAD, VLCAD, and induced PGC1b expression by downregulating b-catenin protein SCAD. Strikingly, BRL induction of these genes was completely level, thus derepressing c-jun, which directly activates the abolished in PGC1bÀ/À differentiation culture, demonstrating PGC1b promoter. that it was PGC1b dependent. The fact that most of these genes were only induced when ERRa was increased suggested that PGC1b Is Required for Rosiglitazone Stimulation they were ERRa targets. Furthermore, we identified a conserved of Osteoclast Differentiation Ex Vivo PPAR response element (PPRE) in the ERRa promoter (Fig- To determine the functional significance of the PPARg induction ure 2D). ChIP assay demonstrated that both PPARg and PGC1b of PGC1b, we examined whether PGC1b deletion affected were associated with this PPRE during BRL- and RANKL- PPARg stimulation of osteoclastogenesis (Wan et al., 2007a). induced osteoclast differentiation (Figure 2E), suggesting that We generated conditional PGC1b KO mice by crossing PGC1b ERRa is a direct PPARg target gene in the specific context of flox mice (Sonoda et al., 2007b) (1bf/f, see Figure S1 available osteoclastogenesis. Together, these results showed that BRL online) with Tie2cre transgenic mice (Constien et al., 2001; activation of PPARg induced the expression of both PGC1b Wan et al., 2007a). Our previous studies showed that Tie2cre and ERRa, which coordinately upregulated mitochondrial genes deletes flox alleles in all hematopoietic lineages, but not in mes- during osteoclastogenesis. enchymal lineages; thus, in osteoclasts, but not in osteoblasts in bone (Wan et al., 2007a; Wan et al., 2007b). Bone marrow ERRa Deletion Causes Osteoclast Defects cells were isolated from 1bf/fTie2cre mutants (1b+cre) or 1bf/f and Decreased Bone Resorption littermate controls (1b-cre) and were differentiated ex vivo in To determine the role of ERRa in osteoclast differentiation the presence of MCSF and RANKL, with or without BRL treat- ex vivo, we compared the osteoclastogenic potential of the bone ment. The result demonstrated that PGC1b deletion severely marrow cells isolated from ERRa KO mice or ERRa heterozy- impaired the pro-osteoclastogenic effect of BRL. In the control gous (Het) control mice (Luo et al., 2003). ERRa deletion severely differentiation culture, BRL robustly stimulated the formation compromised RANKL induction of several key osteoclast genes; of multinucleated TRAP+ (tartrate-resistant acid phosphatase) more strikingly, it blunted their stimulation by BRL (Figure 3A). mature osteoclasts, whereas this effect was abolished in the In addition, ERRa deletion also completely abolished the BRL 1b+cre mutant culture (Figure 2A). Furthermore, PGC1b deletion induction of the aforementioned mitochondrial genes, further resulted in a markedly reduced ability of BRL to potentiate the demonstrating that they were ERRa/PGC1b targets during designated as * were calculated by comparing 1bÀcre controls and 1b+cre mutants under the same treatment conditions; the p values designated as + were calculated by comparing +R/B and +R/V treatment conditions in the 1b-cre cells; n = 3. (D) Alignment of mouse, rat, and human ERRa promoter PPRE region, together with known PPREs for ap2, PEPCK, ACS, and DR-1 consensus. (E) ChIP analysis of PPARg and PGC1b binding to the endogenous mouse ERRa promoter PPRE region in bone marrow differentiation cultures with or without BRL and RANKL treatment; n = 3. Error bars in (B), (C), and (E) represent means ± SD.

Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. 507 Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

A c-fos TRAP B Ndg2 Aco2 IDH3a -R/V 9 7 15 20 ** 18 ** **** **** -R/B 6 **** +R/V 15 5 +R/B 12 6 10 4 10 3

* 6 * 3 **** 2 5 ****

5 ** 1 Relative mRNA Relative mRNA 0 0 0 0 0 ERRaHet ERRaKO ERRaHet ERRaKO ERRaHet ERRaKO ERRaHet ERRaKO ERRaHet ERRaKO

CTSK CAR2 ATP5b MCAD VLCAD 18 2 7 3 * 9 ** * * 6 ** 12 6 5 2 4 1 3 6 * * 1 3 2 * Relative mRNA Relative mRNA 1 0 0 0 0 0 ERRaHet ERRaKO ERRaHet ERRaKO ERRaHet ERRaKO ERRaHet ERRaKO ERRaHet ERRaKO C D ERRaHet 70 250 ERRaKO teHaRRE OKaRRE 60 * RANKL+Veh RANKL+BRL RANKL+Veh RANKL+BRL 200 50 40 *** 150 30 100 20 BM Cell (x106) 50

10 Spleen Cell (x106) 0 0

E F 0.12 4 0.9 G 80 ERRaHet ERRaKO ** ***

3 60 0.08 0.6 *** 2 40

BV/TV *** 0.04 0.3 Tb.Sp (mm) Tb.N (1/mm) Tb.N 1 20 Urinary CTX-1 (ng/mg Cr) 0 0 0 0 14 180 3 H * *** 35 n.s. 12 30 10 * 120 2 25 8 20

6 SMI 15

BS (mm2) 60 1 4 10 Conn.D. (1/mm3) Conn.D.

2 Osteocalcin (ng/ml) 5 0 0 0 0 I ERRaHet ERRaKO J 9 14 12 10 6 ****** 8 ****** 6 3 4 Oc.S/BS (%) Oc.S/BS Oc.N/B.Ar (%) Oc.N/B.Ar 2 0 0 ERRaHet ERRaKO ERRaHet ERRaKO

Figure 3. ERRa Deletion Results in Osteoclast Defects and Decreased Bone Resorption (A and B) Bone marrow cells were isolated from ERRaKO mice or ERRaHet controls and differentiated ex vivo with RANKL and MCSF in the presence or absence of BRL treatment. RANKL-mediated and BRL-stimulated induction of osteoclast markers (A), as well as BRL induction of mitochondrial genes (B), were severely impaired in ERRaKO differentiation culture compared to ERRaHet controls; n = 4. R, RANKL; V, vehicle; B, BRL.

508 Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

osteoclastogenesis (Figure 3B). Consequently, the RANKL- osteoclastogenesis and have provided in vivo evidence for ERRa mediated and BRL-stimulated formation of mature osteoclasts as a key PGC1b target to promote bone resorption. was impaired in the ERRa KO differentiation culture (Figure 3C). These data suggested that ERRa deletion caused osteoclast PGC1b Functions as a PPARg Coactivator to Promote defects. Osteoclast Differentiation Consistently, the ERRa KO mice exhibited osteopetrosis Next, we examined whether PGC1b also functioned as a tran- and extramedullary hematopoiesis in the spleen similar to the scriptional coactivator for PPARg. First, we tested the effect of gf/fTie2cre mice (Wan et al., 2007a). Bone marrow cell number PGC1b on PPARg activation of a consensus PPAR response was decreased by 41% in ERRaKOs compared to ERRaHet element (PPRE)-driven luciferase reporter in a transient transfec- controls (Figure 3D, left), which was compensated by a 46% tion assay of the RAW 264.7 macrophage cell line. Cotransfec- increase in spleen cell numbers (Figure 3D, right). MicroCT anal- tion of PGC1b with PPARg and RXRa potentiated the BRL acti- ysis of the trabecular bone in the proximal tibia (Figure 3E) vation of PPRE, resulting in a 12.5-fold induction, compared to revealed that the ERRa KO mice displayed higher bone volume. a 5.2-fold induction for the receptors alone and a 2.6-fold induc- Quantification of bone structure and architecture demonstrated tion for PGC1b alone (Figure 4A). Second, we tested whether that the trabecular bone volume/tissue volume ratio (BV/TV) was PGC1b functioned through the PPARg ligand-binding domain increased by 102% in ERRaKOs compared to ERRaHet con- (LBD) by examining its ability to stimulate the BRL activation trols, accompanied by 79% greater bone surface (BS), 106% of a Gal4DBD-PPARgLBD fusion protein (Forman et al., 1995). greater trabecular number (Tb.N), 55% less trabecular separa- To determine whether PGC1b could increase the potency of tion (Tb.Sp), 3.3-fold greater connectivity density (Conn.D), and BRL to activate PPARg LBD, we conducted a dose curve of 28% less structure model index (SMI) (Figure 3F). This observa- BRL treatment (0.01, 0.1, and 1 mM). To investigate whether tion was confirmed by the statistically significant increases in the coactivation was PGC1b dose dependent, PGC1b expres- both the trabecular apparent density and the BV/TV of the entire sion plasmid was transfected at a low (20 ng) or high (100 ng) tibia, although the BV/TV of the cortical bone was not signifi- amount. The results demonstrated that PGC1b indeed acted cantly altered (Figure S2). through PPARg LBD and sensitized PPARg LBD to ligand activa- To determine whether the bone defects in ERRa KO mice tion at a lower BRL concentration in a PGC1b dose-dependent resulted from decreased bone resorption and/or increased bone manner (Figure 4B). Third, PGC1b enhanced the PPRE activation formation, we measured urinary CTX-1 (C-terminal telopeptides by a constitutively active VP16-PPARg fusion protein (Saez et al., of type-1 collagen) and serum osteocalcin levels, respectively. 2004), but not an AF2 domain-deleted PPARg mutant (Figure 4C), CTX-1 was significantly reduced (À39%, Figure 3G), whereas further confirming that PPARg LBD was required for PGC1b osteocalcin was elevated (+18%) though statistically nonsignifi- coactivation. cant (Figure 3H). Consistently, histomorphometric analysis of In our previous study, we identified c-fos, a key regulator of femoral metaphyses showed that ERRa KO mice exhibited osteoclast differentiation (Grigoriadis et al., 1994), as a direct significantly less osteoclast surface (Oc.S/BS, À39%) and oste- PPARg target (Wan et al., 2007a). Thus, we tested whether oclast number (Oc.N./B.Ar, À49%) (Figures 3I and 3J); in con- PGC1b could enhance the ability for PPARg to induce the trast, they had greater osteoblast surface and number (Fig- c-fos promoter in RAW 264.7 cells upon BRL stimulation. The ure S3). Moreover, BRL-induced bone resorption and bone results showed that PGC1b potentiated the BRL induction of loss were severely diminished in ERRa KO mice (Figure S4). c-fos by 2.1-fold (Figure 4D). Furthermore, ChIP assays demon- Together, these results demonstrated that the osteopetrosis- strated that PGC1b was indeed recruited to the two previously like phenotype in the ERRa KOs was mainly caused by identified PPREs (Wan et al., 2007a) in the endogenous c-fos decreased bone resorption but also contributed to by increased promoter upon BRL stimulation of the bone marrow-derived bone formation, leading to the uncoupling of bone remodeling osteoclast precursors, resulting in increased histone H3 acetyla- and a net gain of bone. Importantly, these results have identified tion and thus c-fos transcriptional activation (Figure 4E). a previously unrecognized role for ERRa as a critical regulator of Together, these data provided strong evidence that PGC1b

(C) Representative images of the TRAP-stained osteoclast differentiation culture. Mature osteoclasts were identified as multinucleated TRAP+ (purple) cells. Scale bar, 25 mm. (D) ERRa KO mice exhibited extramedullary hematopoiesis in the spleen, evidenced by reduced bone marrow cell numbers and increased splenocyte numbers compared to littermate ERRaHet controls; n = 4. (E and F) ERRa KO mice displayed increased bone mass. Tibiae from ERRaKOs or littermate ERRaHet controls (10- to 12-month-old male; n = 4) were scanned and analyzed by mCT35. (E) Representative images of the trabecular bone of the tibial metaphysis (top; scale bar, 10 mm) and the entire proximal tibia (bottom; scale bar, 1 mm). (F) Quantification of trabecular bone volume and architecture. BV/TV, bone volume/tissue volume ratio; Tb.N, trabecular number; Tb.Sp, trabecular separation; BS, bone surface; Conn.D., connectivity density; SMI, structure model index. (G) Urinary concentration of a bone resorption marker CTX-1 (normalized to urinary creatinine concentration) was significantly decreased in ERRa KO mice; n = 4. (H) Serum concentration of a bone formation marker osteocalcin was increased in ERRa KO mice, but the difference was statistically nonsignificant; n = 4. (I and J) Histomorphometric analysis showed decreased osteoclasts in the ERRa KO mice; n = 4. (I) Representative images of TRAP-stained femoral sections from ERRaKO or littermate ERRaHet controls. Osteoclasts were identified as multinucleated TRAP+ (purple) cells. Scale bar, 100 mm. (J) Quantification of osteoclast surface (Oc.S/BS) and osteoclast number (Oc.N/B.Ar); B.Ar, bone area. Error bars in (A), (B), (D), (F), (G), (H), and (J) represent means ± SD.

Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. 509 Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

A B 14 PPREx3-TK-luc 2.5 UASGx4-TK-luc veh 12 ** 2 0.01uM BRL 10 0.1uM BRL 8 1.5 1uM BRL * 6 1

luc/b-gal ** 4 Fold (BRL/Veh) * 0.5 2 *

0 0 vec g/a 1b g/a/1b gLBD 1b gLBD/lo 1b gLBD/hi 1b 20ng 100ng C 4 PPREx3-TK-luc

Veh 3 BRL * 2

luc/b-gal * 1

0 vec vec/1b g g/1b vp16g vp16g/1b g-dAF2 g-dAF2/1b

D E c-fos1.1kb-luc c-fos-PPRE-A c-fos-PPRE-B 300 *** 5 1 15 Veh ** -R/V **** **** ***** 4 +R/B t

BRL t

200 n.s. 10 3 ** 0.5 2 luc/b-gal 100 n.s. 5 ChIP/10%Inpu ChIP/10%Inpu 1 ****

0 0 0 0 vec vec/1b gg/1b IgG PGC1b AcH3 IgGPGC1bAcH3

Figure 4. PGC1b Functions as a PPARg Transcriptional Coactivator (A) PGC1b potentiated the BRL activation of the consensus PPAR response element (PPRE). Expression plasmids for PPARg (g), RXRa (a), and PGC1b (1b) or vector control (vec) were transfected into the RAW 264.7 macrophage cell line, along with the reporter PPREx3-TK-luc. (B) PGC1b potentiated the BRL activation of a Gal4DBD-PPARgLBD fusion protein and sensitized PPARg LBD to ligand activation at lower BRL concentration in a PGC1b dose-dependent manner. Expression plasmids for Gal4DBD-PPARgLBD (gLBD) and/or PGC1b (1b) were transfected as indicated, along with the Gal4 reporter UASGx4-TK-luc. PGC1b were transfected at 20 ng (low 1b) or 100 ng (hi 1b). BRL treatment was at the indicated concentration. (C) PGC1b increased the PPRE activation by a constitutively active VP16-PPARg fusion protein (VP16 g), but not an AF2 domain deleted PPARg mutant (g-dAF2). (D) PGC1b enhanced the ability for PPARg to induce the c-fos promoter upon BRL stimulation. (E) ChIP analysis of PGC1b recruitment to the endogenous mouse cfos-A and cfos-B PPRE regions in bone marrow differentiation cultures with or without BRL and RANKL treatment. The p values were calculated by comparing BRL and vehicle treatment; n = 3. Error bars in (A)–(E) represent means ± SD. also functions as a PPARg coactivator to effectively mediate the and 1bf/f littermate controls (1b-cre) with BRL (10 mg/kg/day) BRL stimulation of osteoclast differentiation. or vehicle daily by oral gavage for 8 weeks. MicroCT imaging of the proximal tibiae revealed that the BRL-mediated reduction Osteoclastic PGC1b Is Required in trabecular bone in control mice was completely abolished in for Rosiglitazone-Induced Bone Loss In Vivo the 1b+cre mutants (Figures 5A and 5B). Quantification of To further delineate the specific requirement for osteoclastic bone parameters showed that BRL significantly decreased the PGC1b in rosiglitazone-induced bone resorption and bone loss BV/TV in the controls (À31%), whereas a statistically nonsignifi- in vivo, we treated 8-month-old 1bf/fTie2cre mutants (1b+cre) cant increase (+2%) was found in the mutants. Consistently, in

510 Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

A Veh BRL B 1b-cre 1b+cre

Veh BRL Veh BRL 1b-cre

1b+cre

C **** Veh 0.16 **** 25 5 * 0.4 ** 3 BRL 20 ** 4 0.12 0.3 ** *** 2 15 3 0.08 0.2 SMI

BV/TV 10 2 1 BS (mm2) Tb.Sp (mm) 0.04 (1/mm) Tb.N 0.1 5 1

0 0 0 0 0 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre

D erc-b1 erc+b1 heV LRB heV LRB

E F G 20 *** 30 60 ****** Veh ***** **** ***** ***** ****** 50 BRL 160 15 *** ****** n.s. 20 40 n.s. 120 10 30 n.s. n.s. 80 10 20 * Oc.S/BS (%) Oc.S/BS 5 (%) Oc.N/B.Ar 40 10 Osteocalcin (ng/ml) 0 0 0 0 Urinary CTX-1 (ng/mg Cr) 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre 1b-cre 1b+cre

Figure 5. Osteoclastic PGC1b Is Required for Rosiglitazone-Induced Bone Loss In Vivo PGC1bf/fTie2cre mutant mice (1b+cre) and PGC1bf/f littermate control mice (1b-cre) (8-month-old male) were treated with BRL at 10 mg/kg/day or vehicle daily by oral gavage for 8 weeks; n = 4 or 5 in each group. (A–C) MicroCT imaging and analysis of the tibiae. (A) Representative images of the trabecular bone of the tibial metaphysis. Scale bar, 10 mm. (B) Representative images of the entire proximal tibia. Scale bar, 1 mm. (C) Quantification of trabecular bone volume and architecture. BV/TV, bone volume/tissue volume ratio; BS, bone surface; BS/BV, bone surface/bone volume ratio; Tb.N, trabecular number; Tb.Sp, trabecular separation; SMI, structure model index. (D and E) Histomorphometric analysis showed that BRL increased osteoclast surface and number in control mice (1b-cre), but not in mutants (1b+cre). (D) Representative images of TRAP-stained femoral sections. Osteoclasts were identified as multinucleated TRAP+ (purple) cells. Scale bar, 100 mm. (E) Quantification of osteoclast surface (Oc.S/BS) and osteoclast number (Oc.N/B.Ar). B.Ar, bone area. (F) Serum concentration of a bone formation marker osteocalcin. (G) Urinary concentration of a bone resorption marker CTX-1 (normalized to urinary creatinine concentration). Bars in (C), (E), (F), and (G) represent means ± SD.

Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. 511 Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

PGC1β

c-fos … PPARγ/RXRα/BRL

Osteoclast PPARγ Differentiation RXRα Osteoclast β-catenin c-jun [ PGC1 β ] Function Mitochondrial BRL Biogenesis/ Activation ERRα ↑

PGC1β ERRα FAO/OXPHOS…

Figure 6. A Simplified Model for How PGC1b Mediates PPARg Activation of Osteoclast Differentiation and Bone Resorption Rosiglitazone-activated PPARg, in concert with RANKL signaling, indirectly induces PGC1b expression by downregulating b-catenin protein, thus stimulating both basal- and c-jun-induced PGC1b transcription. PGC1b, in turn, forms a positive feedback loop by functioning as a PPARg coactivator to induce PPARg target genes such as c-fos, thereby stimulating osteoclast differentiation. Complementarily, rosiglitazone-activated PPARg also induces ERRa expression during osteoclast differentiation. PGC1b acts as an ERRa coactivator (or protein ligand) to induce mitochondrial genes involved in fatty acid b-oxidation (FAO) and oxidative phosphorylation (OXPHOS), thereby promoting mitochondrial biogenesis and activation. By coordinating two distinct transcriptional programs enhancing osteoclast differentiation and mitochondrial activation, PGC1b mediates PPARg stimulation of osteoclastogenesis and rosiglitazone-induced bone loss. the control mice, BRL resulted in a 27% less bone surface (BS) CTX-1 (À45%) and osteocalcin (À11%) in the vehicle-treated along with an 18% greater bone surface/bone volume ratio 1b+cre mutants compared to the 1b-cre controls (Figure 5F–5G) (BS/BV, data not shown), as well as a 19% less trabecular despite the unaltered BV/TV (Figure 5C). This indicated that oste- number (Tb.N) along with a 28% greater trabecular separation oclastic PGC1b deletion indeed suppressed basal bone resorp- (Tb.Sp), all indicating a lesser trabecular bone apparent density. tion, yet this defect was largely compensated by a simultaneous Moreover, BRL also led to skeletal fragility in the control mice, reduction in bone formation, presumably through the coupling evidenced by a 23% greater SMI, a parameter that quantifies mechanism, thus preserving skeletal homeostasis. Collectively, the characteristic form of a three-dimensional structure in terms these results provided compelling evidence that PGC1b deletion of the relative amount of plates (SMI = 0, strong bone) and rods in the osteoclast lineage confers a complete resistance to BRL- (SMI = 3, fragile bone) independent of the physical dimensions induced bone resorption and bone loss; therefore, PGC1b is an (Hildebrand and Ru¨ egsegger, 1997). Strikingly, all of these essential mediator of PPARg activation of osteoclastogenesis parameters were unaltered by BRL in the 1b+cre mutants (Fig- in vivo. ure 5C). Of interest, no statistically significant differences were found in any structural measurements between the mutants DISCUSSION and controls under vehicle-treated conditions (Figure 5C). Histomorphometric analysis of femoral metaphyses revealed This study has elucidated the molecular mechanisms for that BRL significantly increased both osteoclast surface (Oc.S/ how PPARg and rosiglitazone stimulate osteoclastogenesis by BS, +76%) and osteoclast number (Oc.N/B.Ar, +69%) in the orchestrating the downstream targets PGC1b and ERRa. Using controls, but this induction was abolished in the 1b+cre mutants several mouse models with genetically altered PPARg, PGC1b, (Figures 5D and 5E). In contrast, osteoblast surface and number or ERRa, we have provided in vitro, ex vivo, and in vivo evidence were not significantly altered (Figure S5). Consistently, although that PGC1b is required for the pro-osteoclastogenic and serum osteocalcin was nonsignificantly decreased in both con- bone resorption-enhancing effects of PPARg and rosiglitazone. trols (À25%) and mutants (À13%) by BRL (Figure 5F), urinary PPARg and PGC1b form a positive feedback loop. On one CTX-1 was significantly increased in the controls (+45%) but hand, PPARg activation indirectly induces PGC1b expression decreased in the mutants (À22%) by BRL (Figure 5G). This dem- by downregulating b-catenin protein level, thus derepressing onstrated that, in the control mice, bone resorption and bone c-jun, which directly activates the PGC1b promoter; on the other formation were uncoupled by BRL, thus leading to a net loss hand, PGC1b functions as a PPARg coactivator to stimulate the of bone. In contrast, in the 1b+cre mutants, loss of PGC1b in transcription of its target genes such as c-fos, thus promoting hematopoietic progenitors rendered them refractory to the osteoclast differentiation (Figure 6). Moreover, PGC1b also coor- osteoclast-stimulating effect of BRL; therefore, the coupling of dinates with ERRa to induce genes required for mitochondrial bone resorption and bone formation was maintained, and bone biogenesis and fatty acid oxidation, thereby activating osteo- loss was prevented. Intriguingly, there was a reduction in both clast function (Figure 6). Strikingly, targeted deletion of PGC1b

512 Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

in the osteoclast lineage results in complete resistance to rosigli- should have been an increase in osteoblast number and bone tazone-induced bone loss. Together, these findings demon- formation as well, yet we observed a reduction in the osteocalcin strate that PGC1b mediates the pro-osteoclastogenic function bone formation marker (Figure 5F) and unaltered osteoblast of PPARg by targeting both PPARg itself and ERRa, thus acti- number (Figure S5). Therefore, relatively speaking, rosiglitazone vating two distinct transcriptional programs (Figure 6). indeed suppressed osteoblast number and bone formation in As acid- and proteinase-secreting polykaryons, osteoclasts our study. This is consistent with previous findings that ligand are in a state of high-energy demand and possess abundant activation of PPARg inhibits osteoblastogenesis from the mes- mitochondria (Brown and Breton, 1996; Ishii et al., 2009). enchymal stem cells by favoring adipogenesis (Akune et al., A recent study revealed that PGC1b coordinates with iron uptake 2004; Barak et al., 1999; Cock et al., 2004; Kubota et al., 1999; to orchestrate mitochondrial biogenesis during osteoclast devel- Rosen et al., 1999); on the other hand, repression of PPARg by opment (Ishii et al., 2009). However, the molecular mechanism canonical or noncanonical Wnt signaling enhances osteoblasto- by which PGC1b exerts this function was unknown, and how genesis by reducing adipogenesis (Kang et al., 2007; Takada PGC1b interacts with the osteoclast differentiation program et al., 2007). Our study identifies osteoclastic PGC1b as an provoked by RANKL and PPARg remained underexplored. Our essential mediator of the bone resorption-enhancing effect present study demonstrates that PGC1b functions as a transcrip- by rosiglitazone, which acts in concert with the bone forma- tional coactivator for both PPARg and ERRa to induce the tion-suppressing effect by rosiglitazone to induce uncoupling expression of c-fos and mitochondrial genes, thus linking osteo- and bone loss. clast differentiation with osteoclast activation. In summary, this study demonstrates that rosiglitazone stimu- Osteoclast differentiation and mitochondrial activation cross- lates osteoclastogenesis and bone resorption via a transcrip- talk with each other. For example, reactive oxygen species gen- tional network comprised of PPARg, PGC1b, and ERRa. Pro- erated by the mitochondria can stimulate osteoclast differentia- vocatively, rosiglitazone-mediated activation of adipogenesis tion by inducing Ca2+ oscillations and NFATc1 activation (Kim and suppression of osteoblastogenesis has been shown to be et al., 2010); conversely, transcription factors activated during partially attributed to the coactivator SRC-2 (Mo¨ dder et al., osteoclast differentiation induce target gene expression to pro- 2009). Therefore, PPARg recruits distinct transcriptional coacti- mote osteoclast function and mitochondrial biogenesis (Ishii vators in hematopoietic and mesenchymal lineages to confer et al., 2009; Novack and Teitelbaum, 2008). Our proposed model differential regulation of osteoclast and adipocyte development. (Figure 6) illustrates that the direct downstream targets of ERRa Importantly, this mechanistic understanding of cell type-specific are mitochondrial genes, but ERRa also indirectly regulates gene regulation by PPARg will facilitate the design of improved osteoclast differentiation. Moreover, we have previously shown diabetic drugs, such as selective PPARg modulators (SPPARMs) that PPARg deletion impairs osteoclast differentiation, demon- that retain the insulin-sensitizing benefits but dampen the detri- strating that basal PPARg activity, potentially induced by endog- mental bone loss effects. enous PPARg ligands, is required for efficient osteoclastogenesis (Wan et al., 2007a). Therefore, both PGC1b and ERRa EXPERIMENTAL PROCEDURES deletions compromise the basal PPARg activity, resulting in fewer osteoclasts induced by RANKL ex vivo and lower bone Mice gflox/flox +/À bflox/flox resorption in vivo. PPAR ; Tie2cre mice (Wan et al., 2007a), PGC1 mice (Sonoda et al., 2007b), ERRa KO mice (Luo et al., 2003), and b-cateninflox/flox mice A recent report suggested that ERRa inhibits osteoblastogen- (Brault et al., 2001) have been described. To specifically delete PGC1b in esis and enhances adipogenesis (Delhon et al., 2009). Our hematopoietic lineages and endothelial cells, we bred PGC1bflox/flox (1bf/f) present study reveals a previously unrecognized role for ERRa mice (backcrossed to C57BL/6J for at least six generations) with Tie2cre trans- in promoting osteoclastogenesis by inducing the expression genic mice (Kisanuki et al., 2001) to generate 1bf/fTie2cre+/À mice and 1bf/f of mitochondrial genes via a PGC1b-dependent mechanism. littermate controls. All protocols for mouse experiments were approved by Furthermore, a polymorphic autoregulatory hormone response the Institutional Animal Care and Use Committee of University of Texas South- western Medical Center. element on the human ERRa promoter (Laganie` re et al., 2004) has been found to be associated with bone mineral density Bone Analyses (Laflamme et al., 2005). Together, these findings not only identify To evaluate bone volume and architecture by micro-computed tomography ERRa as a critical regulator of skeletal and mineral homeostasis, (microCT), mouse tibiae were fixed in 70% ethanol and scanned using but also highlight a functional link between PPARg and ERRa a Scanco mCT-35 instrument (SCANCO Medical) at several resolutions for pathways, converging at the transcriptional coactivator PGC1b. both overall tibial assessment (14 micron resolution) and the structural analysis Rosiglitazone induces bone loss by stimulating bone resorp- of trabecular and cortical bone (7 micron resolution). Trabecular bone parameters were calculated using the Scanco software to analyze the bone scans tion and inhibiting bone formation, both of which are required from the trabecular region directly distal to the proximal tibial growth plate. for the uncoupling of bone remodeling. However, the rela- Histomorphometric analyses were conducted using the BIOQUANT Image tive effect of rosiglitazone on osteoclast and osteoblast is age Analysis software (Bioquant). TRAP staining of osteoclasts was performed dependent. In old mice, rosiglitazone increases bone resorption using the Leukocyte Acid Phosphatase staining kit (Sigma). ALP staining of while sustaining bone formation; in contrast, in young mice, osteoblasts was performed using the Alkaline Phosphatase staining kit rosiglitazone decreases bone formation while sustaining bone (Sigma). As a bone resorption marker, urinary C-terminal telopeptide frag- ments of the type I collagen (CTX-1) were measured with the RatLaps EIA kit resorption (Lazarenko et al., 2007). In our study, 8 week rosiglita- (Immunodiagnostic Systems) and normalized by urinary creatinine measured zone treatment significantly increased osteoclast number by the Infinity Creatinine Reagent (Thermo Scientific). As a bone formation (Figure 5E) and bone resorption (Figure 5G) in 8- to 10-month- marker, serum osteocalcin was measured with the mouse osteocalcin EIA old mice. If the coupling of bone remodeling was intact, there kit (Biomedical Technologies, Inc.).

Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. 513 Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

Ex Vivo Osteoclast Differentiation Received: December 10, 2009 Osteoclasts were differentiated from mouse bone marrow cells as described Revised: February 25, 2010 (Kawano et al., 2003; Wan et al., 2007a). In brief, cells were differentiated Accepted: April 16, 2010 with 40 ng/ml of M-CSF (R&D Systems) in a-MEM containing 10% FBS for Published: June 8, 2010 3 days and then with 40 ng/ml of MCSF and 100 ng/ml of RANKL (R&D Systems) for 3 days, in the presence or absence of BRL (1 mM, unless other- REFERENCES wise stated). Retroviral gene transduction was performed as previously described (Wan et al., 2007a). Mature osteoclasts were identified as multinu- Agostini, M., Schoenmakers, E., Mitchell, C., Szatmari, I., Savage, D., Smith, + cleated (>3 nuclei) TRAP cells. Osteoclast differentiation was quantified by A., Rajanayagam, O., Semple, R., Luan, J., Bath, L., et al. (2006). Non-DNA the RNA expression of RANKL-induced transcription factors and osteoclast binding, dominant-negative, human PPARgamma mutations cause lipody- function genes using RT-QPCR analysis. strophic insulin resistance. Cell Metab. 4, 303–311. Akune, T., Ohba, S., Kamekura, S., Yamaguchi, M., Chung, U.I., Kubota, N., Gene Expression Analyses Terauchi, Y., Harada, Y., Azuma, Y., Nakamura, K., et al. (2004). PPARgamma RNA was reverse transcribed into cDNA using an ABI High Capacity cDNA RT insufficiency enhances osteogenesis through osteoblast formation from bone kit and analyzed using real-time quantitative PCR (SYBR Green) in triplicate. marrow progenitors. J. Clin. Invest. 113, 846–855. All RNA expression was normalized by L19. Antibodies used for western blots Ali, A.A., Weinstein, R.S., Stewart, S.A., Parfitt, A.M., Manolagas, S.C., and were: PGC1b (Santa Cruz), b-catenin (BD Biosciences), and b-actin (Sigma). Jilka, R.L. (2005). Rosiglitazone causes bone loss in mice by suppressing osteoblast differentiation and bone formation. Endocrinology 146, 1226–1235. Promoter Analyses Ash, P., Loutit, J.F., and Townsend, K.M. (1980). Osteoclasts derived from For transient transfection, a luciferase reporter was cotransfected into the haematopoietic stem cells. Nature 283, 669–670. mouse macrophage cell line RAW264.7 cells or HEK293 cells with expression plasmids for b-gal and factors to be tested using FuGENE HD reagent Barak, Y., Nelson, M.C., Ong, E.S., Jones, Y.Z., Ruiz-Lozano, P., Chien, K.R., (Roche). Vector alone served as a negative control. The next day, the cells Koder, A., and Evans, R.M. (1999). PPAR gamma is required for placental, were treated with BRL or DMSO vehicle control overnight. Luciferase activity cardiac, and adipose tissue development. Mol. Cell 4, 585–595. was normalized by b-gal activity. All transfection experiments were performed Barroso, I., Gurnell, M., Crowley, V.E., Agostini, M., Schwabe, J.W., Soos, in triplicates and repeated at least three times. Luciferase reporters PPREx3- M.A., Maslen, G.L., Williams, T.D., Lewis, H., Schafer, A.J., et al. (1999). Domi-

TK-luc, UASGx4-TK-luc, and c-fos1.1kb-luc have been previously described nant negative mutations in human PPARgamma associated with severe insulin (Forman et al., 1995; Wan et al., 2007a). A 1.8 Kb genomic DNA fragment resistance, diabetes mellitus and hypertension. Nature 402, 880–883. upstream of the mouse PGC1b transcription start site was cloned into the Brault, V., Moore, R., Kutsch, S., Ishibashi, M., Rowitch, D.H., McMahon, A.P., pGL3Basic vector to generate the PGC1b-luc reporter. Expression plasmids Sommer, L., Boussadia, O., and Kemler, R. (2001). Inactivation of the beta- for PPARg, RXRa, PGC1b, VP16-PPARg, and PPARg-dAF2 mutant were catenin gene by Wnt1-Cre-mediated deletion results in dramatic brain malfor- previously described (Saez et al., 2004; Sonoda et al., 2007b; Wan et al., mation and failure of craniofacial development. Development 128, 1253–1264. 2007a). An expression plasmid for a constitutively active b-catenin mutant Brown, D., and Breton, S. (1996). Mitochondria-rich, proton-secreting epithe- was generously provided by Dr. Chi Zhang (Texas Scottish Rite Hospital for lial cells. J. Exp. Biol. 199, 2345–2358. Children). Expression plasmids for c-jun,c-fos, NFATc1, and p65 were purchased from Open Biosystems. Promoter sequence alignment was per- Chawla, A., Boisvert, W.A., Lee, C.H., Laffitte, B.A., Barak, Y., Joseph, S.B., formed using Vector NTI Advanced 11 AlignX software (Invitrogen). ChIP Liao, D., Nagy, L., Edwards, P.A., Curtiss, L.K., et al. (2001). A PPAR assays were performed using mouse bone marrow-derived osteoclast gamma-LXR-ABCA1 pathway in macrophages is involved in cholesterol efflux precursors that were treated with BRL/RANKL/MCSF or MCSF alone for and atherogenesis. Mol. Cell 7, 161–171. 3 days as previously described (Wan et al., 2007a). Antibodies used were: Cock, T.A., Back, J., Elefteriou, F., Karsenty, G., Kastner, P., Chan, S., and c-jun (Cell Signaling), PGC1b (Santa Cruz), acetyl-Histone H3 (Upstate/Milli- Auwerx, J. (2004). Enhanced bone formation in lipodystrophic PPARgamma pore), PPARg (Santa Cruz), and IgG negative control (BD Biosciences). (hyp/hyp) mice relocates haematopoiesis to the spleen. EMBO Rep. 5, ChIP output was quantified by real-time PCR in triplicates and normalized 1007–1012. by 10% input. Constien, R., Forde, A., Liliensiek, B., Gro¨ ne, H.J., Nawroth, P., Ha¨ mmerling, G., and Arnold, B. (2001). Characterization of a novel EGFP reporter mouse Statistical Analyses to monitor Cre recombination as demonstrated by a Tie2 Cre mouse line. All statistical analyses were performed with Student’s t test and represented Genesis 30, 36–44. as mean ± standard deviation (SD). The p values were designated as: Delhon, I., Gutzwiller, S., Morvan, F., Rangwala, S., Wyder, L., Evans, G., *p < 0.05; **p < 0.01; ***p < 0.005; ****p < 0.001; *****p < 0.0005; ******p < Studer, A., Kneissel, M., and Fournier, B. (2009). Absence of estrogen 0.0001; ns, nonsignificant (p > 0.05). receptor-related-alpha increases osteoblastic differentiation and cancellous bone mineral density. Endocrinology 150, 4463–4472. SUPPLEMENTAL INFORMATION Edwards, C.M., and Mundy, G.R. (2008). Eph receptors and ephrin signaling pathways: a role in bone homeostasis. Int. J. Med. Sci. 5, 263–272. Supplemental Information includes five figures and can be found with this Evans, R.M., Barish, G.D., and Wang, Y.X. (2004). PPARs and the complex article online at doi:10.1016/j.cmet.2010.04.015. journey to obesity. Nat. Med. 10, 355–361. Forman, B.M., Tontonoz, P., Chen, J., Brun, R.P., Spiegelman, B.M., and ACKNOWLEDGMENTS Evans, R.M. (1995). 15-Deoxy-delta 12, 14-prostaglandin J2 is a ligand for the adipocyte determination factor PPAR gamma. Cell 83, 803–812. We would like to thank R. Evans (Salk Institute) for providing the PGC1bflox/flox Grey, A. (2009). Thiazolidinedione-induced skeletal fragility—mechanisms and mice; V. Gigue` re (McGill University) for providing the ERRa KO mice; D. Man- implications. Diabetes Obes. Metab. 11, 275–284. gelsdorf, S. Kliewer, J. Zerwekh, O. Oz (University of Texas Southwestern Medical Center), and J. Sonoda (Salk Institute) for helpful discussion; and Grigoriadis, A.E., Wang, Z.Q., Cecchini, M.G., Hofstetter, W., Felix, R., Fleisch, A. Gray for administrative assistance. Y.W. is a Virginia Murchison Linthicum H.A., and Wagner, E.F. (1994). c-Fos: a key regulator of osteoclast-macro- Scholar in Medical Research. This work was supported by the University of phage lineage determination and bone remodeling. Science 266, 443–448. Texas Southwestern Medical Center Endowed Scholar Startup Fund, a BD He, W., Barak, Y., Hevener, A., Olson, P., Liao, D., Le, J., Nelson, M., Ong, E., Biosciences Research Grant Award, and CPRIT funding (RP100841). Olefsky, J.M., and Evans, R.M. (2003). Adipose-specific peroxisome

514 Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

proliferator-activated receptor gamma knockout causes insulin resistance in Lai, L., Leone, T.C., Zechner, C., Schaeffer, P.J., Kelly, S.M., Flanagan, D.P., fat and liver but not in muscle. Proc. Natl. Acad. Sci. USA 100, 15712–15717. Medeiros, D.M., Kovacs, A., and Kelly, D.P. (2008). Transcriptional coactiva- Hildebrand, T., and Ru¨ egsegger, P. (1997). Quantification of bone microarch- tors PGC-1alpha and PGC-lbeta control overlapping programs required for itecture with the structure model index. Comput. Methods Biomech. Biomed. perinatal maturation of the heart. Genes Dev. 22, 1948–1961. Engin. 1, 15–23. Lazarenko, O.P., Rzonca, S.O., Hogue, W.R., Swain, F.L., Suva, L.J., and Lecka-Czernik, B. (2007). Rosiglitazone induces decreases in bone mass Home, P.D., Pocock, S.J., Beck-Nielsen, H., Curtis, P.S., Gomis, R., Hanefeld, and strength that are reminiscent of aged bone. Endocrinology 148, 2669– M., Jones, N.P., Komajda, M., and McMurray, J.J. (2009). Rosiglitazone eval- 2680. uated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD): a multicentre, randomised, open-label trial. Lancet 373, Lehmann, J.M., Moore, L.B., Smith-Oliver, T.A., Wilkison, W.O., Willson, T.M., 2125–2035. and Kliewer, S.A. (1995). An antidiabetic thiazolidinedione is a high affinity ligand for peroxisome proliferator-activated receptor gamma (PPAR gamma). Ishii, K.A., Fumoto, T., Iwai, K., Takeshita, S., Ito, M., Shimohata, N., Aburatani, J. Biol. Chem. 270, 12953–12956. H., Taketani, S., Lelliott, C.J., Vidal-Puig, A., and Ikeda, K. (2009). Coordination of PGC-1beta and iron uptake in mitochondrial biogenesis and osteoclast Lehrke, M., and Lazar, M.A. (2005). The many faces of PPARgamma. Cell 123, activation. Nat. Med. 15, 259–266. 993–999. Lelliott, C.J., Medina-Gomez, G., Petrovic, N., Kis, A., Feldmann, H.M., Bjur- Jiang, C., Ting, A.T., and Seed, B. (1998). PPAR-gamma agonists inhibit sell, M., Parker, N., Curtis, K., Campbell, M., Hu, P., et al. (2006). Ablation of production of monocyte inflammatory cytokines. Nature 391, 82–86. PGC-1beta results in defective mitochondrial activity, thermogenesis, hepatic Kahn, S.E., Haffner, S.M., Heise, M.A., Herman, W.H., Holman, R.R., Jones, function, and cardiac performance. PLoS Biol. 4, e369. N.P., Kravitz, B.G., Lachin, J.M., O’Neill, M.C., Zinman, B., and Viberti, G.; Li, M., Pan, L.C., Simmons, H.A., Li, Y., Healy, D.R., Robinson, B.S., Ke, H.Z., ADOPT Study Group. (2006). Glycemic durability of rosiglitazone, metformin, and Brown, T.A. (2006). Surface-specific effects of a PPARgamma agonist, or glyburide monotherapy. N. Engl. J. Med. 355, 2427–2443. darglitazone, on bone in mice. Bone 39, 796–806. Kahn, S.E., Zinman, B., Lachin, J.M., Haffner, S.M., Herman, W.H., Holman, Lin, J., Yang, R., Tarr, P.T., Wu, P.H., Handschin, C., Li, S., Yang, W., Pei, L., R.R., Kravitz, B.G., Yu, D., Heise, M.A., Aftring, R.P., and Viberti, G.; Diabetes Uldry, M., Tontonoz, P., et al. (2005). Hyperlipidemic effects of dietary satu- Outcome Progression Trial (ADOPT) Study Group. (2008). Rosiglitazone-asso- rated fats mediated through PGC-1beta coactivation of SREBP. Cell 120, ciated fractures in type 2 diabetes: an Analysis from A Diabetes Outcome 261–273. Progression Trial (ADOPT). Diabetes Care 31, 845–851. Luo, J., Sladek, R., Carrier, J., Bader, J.A., Richard, D., and Gigue` re, V. (2003). Kamei, Y., Ohizumi, H., Fujitani, Y., Nemoto, T., Tanaka, T., Takahashi, N., Reduced fat mass in mice lacking orphan nuclear receptor estrogen-related Kawada, T., Miyoshi, M., Ezaki, O., and Kakizuka, A. (2003). PPARgamma receptor alpha. Mol. Cell. Biol. 23, 7947–7956. coactivator 1beta/ERR ligand 1 is an ERR protein ligand, whose expression Mo¨ dder, U.I., Monroe, D.G., Fraser, D.G., Spelsberg, T.C., Rosen, C.J., Ge´ hin, induces a high-energy expenditure and antagonizes obesity. Proc. Natl. M., Chambon, P., O’Malley, B.W., and Khosla, S. (2009). Skeletal conse- Acad. Sci. USA 100, 12378–12383. quences of deletion of steroid receptor coactivator-2/transcription interme- Kang, S., Bennett, C.N., Gerin, I., Rapp, L.A., Hankenson, K.D., and Macdou- diary factor-2. J. Biol. Chem. 284, 18767–18777. gald, O.A. (2007). Wnt signaling stimulates osteoblastogenesis of mesen- Nolan, J.J., Ludvik, B., Beerdsen, P., Joyce, M., and Olefsky, J. (1994). chymal precursors by suppressing CCAAT/enhancer-binding protein alpha Improvement in glucose tolerance and insulin resistance in obese subjects and peroxisome proliferator-activated receptor gamma. J. Biol. Chem. 282, treated with troglitazone. N. Engl. J. Med. 331, 1188–1193. 14515–14524. Novack, D.V., and Teitelbaum, S.L. (2008). The osteoclast: friend or foe? Annu. Kawano, H., Sato, T., Yamada, T., Matsumoto, T., Sekine, K., Watanabe, T., Rev. Pathol. 3, 457–484. Nakamura, T., Fukuda, T., Yoshimura, K., Yoshizawa, T., et al. (2003). Pittenger, M.F., Mackay, A.M., Beck, S.C., Jaiswal, R.K., Douglas, R., Mosca, Suppressive function of androgen receptor in bone resorption. Proc. Natl. J.D., Moorman, M.A., Simonetti, D.W., Craig, S., and Marshak, D.R. (1999). Acad. Sci. USA 100, 9416–9421. Multilineage potential of adult human mesenchymal stem cells. Science 284, Kim, M.S., Yang, Y.M., Son, A., Tian, Y.S., Lee, S.I., Kang, S.W., Muallem, S., 143–147. and Shin, D.M. (2010). RANKL-mediated reactive oxygen species pathway Ricote, M., Li, A.C., Willson, T.M., Kelly, C.J., and Glass, C.K. (1998). The that induces long lasting Ca2+ oscillations essential for osteoclastogenesis. peroxisome proliferator-activated receptor-gamma is a negative regulator of J. Biol. Chem. 285, 6913–6921. macrophage activation. Nature 391, 79–82. Kisanuki, Y.Y., Hammer, R.E., Miyazaki, J., Williams, S.C., Richardson, J.A., Rosen, E.D., Sarraf, P., Troy, A.E., Bradwin, G., Moore, K., Milstone, D.S., and Yanagisawa, M. (2001). Tie2-Cre transgenic mice: a new model for endo- Spiegelman, B.M., and Mortensen, R.M. (1999). PPAR gamma is required thelial cell-lineage analysis in vivo. Dev. Biol. 230, 230–242. for the differentiation of adipose tissue in vivo and in vitro. Mol. Cell 4, 611–617. Kubota, N., Terauchi, Y., Miki, H., Tamemoto, H., Yamauchi, T., Komeda, K., Saez, E., Rosenfeld, J., Livolsi, A., Olson, P., Lombardo, E., Nelson, M., Satoh, S., Nakano, R., Ishii, C., Sugiyama, T., et al. (1999). PPAR gamma medi- Banayo, E., Cardiff, R.D., Izpisua-Belmonte, J.C., and Evans, R.M. (2004). ates high-fat diet-induced adipocyte hypertrophy and insulin resistance. Mol. PPAR gamma signaling exacerbates mammary gland tumor development. Cell 4, 597–609. Genes Dev. 18, 528–540. Lacey, D.L., Timms, E., Tan, H.L., Kelley, M.J., Dunstan, C.R., Burgess, T., Scheven, B.A., Visser, J.W., and Nijweide, P.J. (1986). In vitro osteoclast Elliott, R., Colombero, A., Elliott, G., Scully, S., et al. (1998). Osteoprotegerin generation from different bone marrow fractions, including a highly enriched ligand is a cytokine that regulates osteoclast differentiation and activation. haematopoietic stem cell population. Nature 321, 79–81. Cell 93, 165–176. Sonoda, J., Laganie` re, J., Mehl, I.R., Barish, G.D., Chong, L.W., Li, X., Schef- Laflamme, N., Giroux, S., Loredo-Osti, J.C., Elfassihi, L., Dodin, S., Blanchet, fler, I.E., Mock, D.C., Bataille, A.R., Robert, F., et al. (2007a). Nuclear receptor C., Morgan, K., Gigue` re, V., and Rousseau, F. (2005). A frequent regulatory ERR alpha and coactivator PGC-1 beta are effectors of IFN-gamma-induced variant of the estrogen-related receptor alpha gene associated with BMD in host defense. Genes Dev. 21, 1909–1920. French-Canadian premenopausal women. J. Bone Miner. Res. 20, 938–944. Sonoda, J., Mehl, I.R., Chong, L.W., Nofsinger, R.R., and Evans, R.M. (2007b). Laganie` re, J., Tremblay, G.B., Dufour, C.R., Giroux, S., Rousseau, F., and PGC-1beta controls mitochondrial metabolism to modulate circadian activity, Gigue` re, V. (2004). A polymorphic autoregulatory hormone response element adaptive thermogenesis, and hepatic steatosis. Proc. Natl. Acad. Sci. USA in the human estrogen-related receptor alpha (ERRalpha) promoter dictates 104, 5223–5228. peroxisome proliferator-activated receptor gamma coactivator-1alpha control Sottile, V., Seuwen, K., and Kneissel, M. (2004). Enhanced marrow adipogen- of ERRalpha expression. J. Biol. Chem. 279, 18504–18510. esis and bone resorption in estrogen-deprived rats treated with the

Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc. 515 Cell Metabolism PGC1b Is Required for TZD-Induced Bone Loss

PPARgamma agonist BRL49653 (rosiglitazone). Calcif. Tissue Int. 75, tion of PGC-1beta causes mitochondrial dysfunction and liver insulin resis- 329–337. tance. Cell Metab. 4, 453–464. Strotmeyer, E.S., and Cauley, J.A. (2007). Diabetes mellitus, bone mineral Wahli, W. (2008). PPAR gamma: ally and foe in bone metabolism. Cell Metab. density, and fracture risk. Curr. Opin. Endocrinol. Diabetes Obes. 14, 429–435. 7, 188–190. Takada, I., Mihara, M., Suzawa, M., Ohtake, F., Kobayashi, S., Igarashi, M., Wan, Y., Chong, L.W., and Evans, R.M. (2007a). PPAR-gamma regulates Youn, M.Y., Takeyama, K., Nakamura, T., Mezaki, Y., et al. (2007). A histone osteoclastogenesis in mice. Nat. Med. 13, 1496–1503. lysine methyltransferase activated by non-canonical Wnt signalling suppresses PPAR-gamma transactivation. Nat. Cell Biol. 9, 1273–1285. Wan, Y., Saghatelian, A., Chong, L.W., Zhang, C.L., Cravatt, B.F., and Evans, R.M. (2007b). Maternal PPAR gamma protects nursing neonates by suppress- Tondravi, M.M., McKercher, S.R., Anderson, K., Erdmann, J.M., Quiroz, M., ing the production of inﬂammatory milk. Genes Dev. 21, 1895–1908. Maki, R., and Teitelbaum, S.L. (1997). Osteopetrosis in mice lacking haematopoietic transcription factor PU.1. Nature 386, 81–84. Yasuda, H., Shima, N., Nakagawa, N., Yamaguchi, K., Kinosaki, M., Mochi- Tontonoz, P., and Spiegelman, B.M. (2008). Fat and beyond: the diverse zuki, S., Tomoyasu, A., Yano, K., Goto, M., Murakami, A., et al. (1998). Osteo- biology of PPARgamma. Annu. Rev. Biochem. 77, 289–312. clast differentiation factor is a ligand for osteoprotegerin/osteoclastogenesis- inhibitory factor and is identical to TRANCE/RANKL. Proc. Natl. Acad. Sci. Tontonoz, P., Hu, E., and Spiegelman, B.M. (1994). Stimulation of adipogene- USA 95, 3597–3602. sis in ﬁbroblasts by PPAR gamma 2, a lipid-activated transcription factor. Cell 79, 1147–1156. Yu, S., and Reddy, J.K. (2007). Transcription coactivators for peroxisome Tontonoz, P., Nagy, L., Alvarez, J.G., Thomazy, V.A., and Evans, R.M. (1998). proliferator-activated receptors. Biochim. Biophys. Acta 1771, 936–951. PPARgamma promotes monocyte/macrophage differentiation and uptake of Zinman, B., Haffner, S.M., Herman, W.H., Holman, R.R., Lachin, J.M., Kravitz, oxidized LDL. Cell 93, 241–252. B.G., Paul, G., Jones, N.P., Aftring, R.P., Viberti, G., and Kahn, S.E. (2009). Vianna, C.R., Huntgeburth, M., Coppari, R., Choi, C.S., Lin, J., Krauss, S., Effect of rosiglitazone, metformin, and glyburide on bone biomarkers in Barbatelli, G., Tzameli, I., Kim, Y.B., Cinti, S., et al. (2006). Hypomorphic muta- patients with type 2 diabetes. J. Clin. Endocrinol. Metab. 95, 134–142.

516 Cell Metabolism 11, 503–516, June 9, 2010 ª2010 Elsevier Inc.