PPARδ-mediated mitochondrial rewiring of osteoblasts determines bone mass

Bone turnover, which is determined by osteoclast-mediated bone resorption and osteoblast-mediated bone formation, represents a highly energy consuming process. The metabolic requirements of osteoblast differentiation and mineralization, both essential for regular bone formation, however, remain incompletely understood. Here we identify the nuclear receptor peroxisome proliferator-activated receptor (PPAR) δ as key regulator of osteoblast metabolism. Induction of PPARδ was essential for the metabolic adaption and increased rate in mitochondrial respiration necessary for the differentiation and mineralization of osteoblasts. Osteoblast-specific deletion of PPARδ in mice, in turn, resulted in an altered energy homeostasis of osteoblasts, impaired mineralization and reduced bone mass. These data show that PPARδ acts as key regulator of osteoblast metabolism and highlight the relevance of cellular metabolic rewiring during osteoblast-mediated bone formation and bone-turnover.


Results
Osteoblast differentiation is dependent on an increase in oxidative phosphorylation. To determine the metabolic requirements for regular osteoblast differentiation, we initially performed an extracellular flux analysis of in vitro cultured primary osteoblasts and their precursors. We compared the metabolic profiles of osteoblast precursors during steady state and upon initiation of osteoblast differentiation. These experiments confirmed a significantly increased oxygen consumption and oxidative phosphorylation of differentiating osteoblasts (Fig. 1A). Although the glycolytic activity slightly increased as well, this metabolic adaption resulted in an increase in the ratio between oxygen consumption rate (OCR, an indicator for mitochondrial respiration) and the extracellular acidification rate (ECAR, an indicator for glycolysis), suggestive of a robust metabolic rewiring of differentiating osteoblasts. In accordance, we observed a time-dependent increase in the expression of multiple genes involved in the control of mitochondrial respiration, mitochondrial biogenesis and oxygen-dependent energy provision such as Peroxisome proliferator-activated receptor gamma coactivator 1 (PGC1)α, mitochondrial transcription factor A (TFAM) or dynamin-1-like protein (DRP)1 in differentiating osteoblasts (Fig. 1B). Pharmacologic inhibition of mitochondrial biogenesis by tigecycline 13 or of oxidative phosphorylation by rotenone, in turn, did not interfere with osteoblast viability, but dramatically diminished their differentiation and mineralization potential (Fig. 1C,D and Suppl. Fig. 1). These data indicated a global shift in the transcriptional program that controlled the cellular metabolic adaption during osteoblast differentiation and mineralization. Oxygen consumption rate (OCR) and the extracellular acidification rate (ECAR) including basal and maximal respiration rate measured with extracellular flux (XF) analyzer in freshly isolated calvarial osteoblasts cultured in regular growth medium (black) and differentiation medium (grey) for 24 hours (n = 9 each). (B) Real-time PCR analysis of mRNA expressions normalized to β-actin in calvarial osteoblasts cultured in regular growth media (black) and differentiation media (grey) for 6, 24 and 72 hours (n = 3). (C) Alizarine Red staining of calvarial osteoblasts cultured in osteoblastic differentiation media supplemented with 10 ng/ml Wnt3a and conditionally supplemented with vehicle (control), 30 µM tigecycline or 20 nM Rotenone at day 3, 19 and 40 of culture (representative for n = 3). (D) Quantitative analysis of mineralized areas of alizarine red staining of calvarial osteoblasts supplemented with 10 ng/ml Wnt3a and conditionally supplemented with 30 µM Tigezycline (green) or 20 nM Rotenone (yellow) cultured in osteoblastic differentiation media for 40 days (n = 3). *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.005 using two-tailed Student's t-test. the nuclear receptor ppARδ mediates metabolic rewiring of osteoblasts. PGC1α is considered a master regulator of mitochondrial biogenesis and was among the most highly induced genes in differentiating osteoblasts. As PGC1alpha serves as co-activator and interacts with various other transcription factors, we next assessed expression levels of genes encoding for proteins that represent known PGC1α interaction partners involved in cellular metabolism and oxidative phosphorylation. This analysis identified peroxisome proliferator-activated receptor (PPAR)δ as one of the prominently expressed transcription factors in differentiating osteoblasts ( Fig. 2A-C). PPARδ belongs to the superfamily of nuclear receptors and acts as a ligand-dependent transcription factor that senses fatty acids and subsequently controls fatty acid oxidation, a process that primarily fuels mitochondrial respiration. We have previously identified PPARδ as important regulator of Wnt signaling and osteoblast/osteoclast crosstalk 14,15 . Our current analysis showed that expression of PPARδ gradually increased during early osteogenic differentiation, whereas expression levels of its family members PPARα and PPARγ remained low ( Fig. 2A-C). We additionally confirmed expression of PPARδ on a protein level, which was induced during osteogenic differentiating of wild-type, but not in PPARδ-deficient MSCs (Fig. 2B). In accordance, induction of PPARδ was paralleled by induction of several known PPARδ target genes such as carnitine palmitoyltransferase (CPT)-1 or pyruvate dehydrogenase kinase (PDK)4, with expression levels likewise increasing during osteoblast differentiation (Fig. 2D). As expected, chromatin immunoprecipitation experiments confirmed direct binding of PPARδ to these known target genes also in osteoblasts (Fig. 2E). Next, we sought to determine whether induction of PPARδ indeed controlled the metabolic rewiring and increase in oxidative phosphorylation we had observed during osteogenic differentiation. We consequently analyzed the metabolic phenotype of PPARδ-deficient osteoblasts. Absence of PPARδ resulted in suppression of the expression of metabolic key genes such as PGC1α (Fig. 2F). This finding indicated that PPARδ controlled a transcriptional program that regulated mitochondrial metabolism and oxidative phosphorylation during osteoblast differentiation. The expression levels of metabolic genes such as glucose transporter (Glut)1 or Glut3, which are involved in glycolysis, in turn, were not suppressed, but even increased in PPARδ-deficient osteoblasts (Fig. 2G). The PPARδ-mediated regulation of genes involved in oxidative phosphorylation accordingly resulted in a reduced oxygen consumption and impaired oxidative phosphorylation in PPARδ-deficient osteoblasts, whereas glycolysis was increased (Fig. 2H).

Absence of ppARδ in osteoblasts alters their differentiation and mineralization.
To understand whether osteoblast differentiation and/or mineralization were dependent on this PPARδ-mediated metabolic rewiring, we studied osteoblast differentiation in the absence of PPARδ. In comparison to wild-type osteoblast, PPARδ-deficient cells showed a defective upregulation of osteoblast differentiation markers such as Runx2, osteocalcin (OCN), osterix (Osx), and alkaline phosphatase (ALP) (Fig. 3A). PPARδ-deficient MSCs accordingly displayed a defective mineralization upon induction of osteogenic differentiation, demonstrating an important intrinsic role of the PPARδ-controlled metabolic adaption during osteoblast differentiation and function (Fig. 3B,C). To determine the in vivo relevance of the observed phenotype, we generated mice carrying an osteoblast-specific deletion of PPARδ. Analysis of these Runx2 cre PPARd fl/fl mice showed that these animals displayed a reduced bone mass in comparison to their wild type littermates with a significantly decreased bone volume/total volume, lower trabecular number and decreased bone mineral density in both tibia and spine ( Fig. 3D-F). This low bone mass phenotype was associated with a reduced number of osteoblasts and a tendency towards a reduced mineral apposition and bone formation rate (Fig. 3G).

Discussion
Nuclear receptors including the members of the PPAR family control pleotropic processes throughout the body, including control of fatty acid and glucose metabolism in liver, adipose tissue and muscle 16 . Previous data have identified PPARδ as important regulator of bone turnover and musculoskeletal homeostasis 14,17 . Although these studies determined a contribution of this nuclear receptor during muscle and bone biology as well as during osteoblast/osteoclast crosstalk, its exact molecular role during bone formation remained unclear. Our current data expand these insights on the involvement of this nuclear receptor in osteoblast biology and reveal an osteoblast-intrinsic role of this nuclear receptor during bone homeostasis in vivo. We show that PPARδ regulates the metabolic adaption and increased mitochondrial respiration, events that are required for efficient osteoblast differentiation and mineralization.
It remains to be determined to which extend the PPARδ-mediated control of energy homeostasis accounts for the reduced mineralization and bone mass observed in Runx2 cre PPARd fl/fl mice and whether additional effects exerted by this transcription factor contribute to this phenotype. Control of osteoblast-intrinsic mitochondrial respiration, however, seems to critically influence differentiation and mineralization of this cell type. Although mice that carry a Sox2-mediated complete depletion of PPARδ also showed an altered OPG/RANKL ratio and an increased osteoclast differentiation 14 , the current Runx-2-meidated osteoblast-specific deletion of this nuclear receptor primarily resulted in osteoblast-intrinsic effects and reduced mineralization without increasing osteoclast differentiation, suggesting predominant effects of PPARδ on osteoblast metabolism, differentiation and mineralization. The exact reason for this discrepancy is unclear, but might be due an additional PPARδ-mediated control of the OPG/RANKL system in cells others than osteoblasts.
PPARδ belongs to the nuclear receptor superfamily of transcription factors. Its close family member PPARγ acts as key factor during adipocyte differentiation. Although both PPARγ and PPARδ sense a repertoire of lipophilic compounds and subsequently act as ligand-activated transcription factors, they dictate partially opposing cellular pathways such as fatty acid synthesis and fatty acid oxidation. Moreover, PPARγ and PPARδ are differentially expressed in distinct cell types such as adipocytes and osteoblasts. Ligand-induced activation of these different PPAR family members accordingly exerts contrasting effects on both systemic energy and bone metabolism 15 . These findings highlight the importance of such metabolic sensors during cellular fate decision processes and cell culture. Isolation of calvarial osteoblasts was previously described 18 . For osteoblast monocultures, we differentiated freshly isolated calvarial osteoblast precursors in the presence of 5 mM ß-glycerolphosphate (CALBIOCHEM #35675) and 100 mg/ml ascorbic acid (Sigma-Aldrich #A0278). MSCs were isolated and fully characterized from PPARδ wildtype and PPARδ-deficient mice bone-marrow as described previously 19 confirming the mesenchymal stem cell minimal criteria 20 . We plated both MSC and osteoblast cell suspensions (0.5 × 10 6 cells cm −2 ) in α-MEM (GIBCO ® by life technologies #32571-028) supplemented with 10% FBS (Biowest #S1580), 100 U/ml penicillin and 10 mg/ml streptomycin (PAN Biotech GmbH #P06-07100), at 37 °C in humidified atmosphere containing 5% CO2 in air. Culture media were changed every 2 days. At subconfluence, we plated cells at a density of 5 000 cells cm −2 . MSC were used between passages 7 and 12.
immunohistochemistry. PPARdelta immunohistochemical expression was evaluated MSC at day 7 after induction of osteoblast differentiation We used polyclonal anti-PPARβ from rabbit (Santa Cruz Biotechnology, #sc-7197) in a dilution of 1:100 in 2% BSA/PBS as previously described 21 . Slides were viewed under the fluorescence microscope ECLIPSE Ni-Series microscope, Nikon, USA.
Cells were stimulated with Tigecycline 30 µM. At indicated time points, cells were briefly washed with PBS and then fixed with 95% Ethanol for 30 minutes at room temperature. After briefly washing mineralized spots were dyed with 2% Alizarin Red S (Merck #A5533) in H 2 O (pH adjusted between 4 and 4,3) for 3 to 5 minutes. Then wells were washed with H 2 O for 5 to 8? times. Stained plates were dried and pictures were taken using the HP Scanjet G4050 and HP Scansoftware. Calcification areas were quantified with help of adobe photoshop CS6 software. (2020) 10:8428 | https://doi.org/10.1038/s41598-020-65305-5 www.nature.com/scientificreports www.nature.com/scientificreports/ (antisense). Normalized gene expression values for each sample were calculated as the ratio of expression of mRNA of the gene of interest to the expression of mRNA for β-actin.

Metabolic analysis. The cells' bioenergetics were assessed using an XFe96 Extracellular Flux Analyzer
(Seahorse Bioscience, North Billerica, MA) as well as corresponding kits (Agilent, Santa Clara, California, USA). Cells were seeded in an optimized concentration of 15 000 cells per well. One hour before performance of seahorse experiment, cells were incubated at 37 °C in a CO 2 -free atmosphere. XF Mitochondrial Stress Test Kits (Agilent, #103015) and XF Glycolysis Stress Test Kits (Agilent, #103020) were utilized according to the user guide (Seahorse Bioscience) and as described before 23 . For the Mitochondrial Stress Test, basal oxygen consumption rate (OCR) (an indicator for mitochondrial respiration) and extracellular acidification rate (ECAR) (an indicator for lactic acid production or glycolysis) were analyzed. Next, OCR and ECAR responses toward the application of oligomycin (1 µM), FCCP (2.5 µM), and the combination of antimycin (3 µM) and rotenone (3 µM) were evaluated. For Glycolysis Stress Test, OCR and ECAR were detected. Then, OCR and ECAR responses toward the application of Glucose (100 mM), Oligomycin (100 µM) and 2-DG (500 mM) were evaluated. chromatin immunoprecipitation (chip) assays. To verify the binding of PPARd on selected promoters, ChIP assays were performed according to manufacturer instructions using the ChIP-IT ® Express Chromatin Immunoprecipitation Kit (#53008, Active Motif, CA, USA). 25 μg of sonicated chromatin extract were incubated with 3.4 µg specific antibodies against PPARdelta (#ab178866, Abcam, Cambridge, MA, USA) or normal rabbit IgG antibodies (#sc-2027, Santa Cruz Biotechnology, CA, USA). After purification of DNA using Chromatin IP DNA Purification Kit (Active Motif, #58002) qPCR reactions were carried out on specific genomic regions using SYBR Green (Bio-Rad, USA). The resulting signals were normalized for primer efficiency by carrying out qPCR for each primer pair using the genomic input DNA. For positive control ChIP H3K36me3 antibody was used.

Statistical analyses.
All data are presented as mean ± SEM. Tests for statistical significance were performed with Student's t test using GraphPad Prism Version 5 (GraphPad Prism Software Inc. La Jolla, California, USA). P < 0.05 was considered as significant.