Sexual dimorphism of osteoclast reliance on mitochondrial oxidation of energy substrates in the mouse

Osteoclasts specialize in bone resorption and are critical for bone remodeling. Previous studies have shown that osteoclasts possess abundant mitochondria and derive most energy through oxidative phosphorylation (OXPHOS). However, the energy substrates fueling OXPHOS in osteoclasts remain to be fully defined. Here, we showed that osteoclast differentiation was coupled with increased oxidation of glucose, glutamine, and oleate. Transcriptomic analyses with RNA sequencing revealed marked upregulation of genes participating in OXPHOS and mitochondrial fatty acid oxidation, during osteoclast differentiation. Increased mitochondrial oxidation of long-chain fatty acids was required for osteoclast differentiation in vitro. However, blocking fatty acid oxidation in vivo, by deletion of carnitine palmitoyltransferase 1a (Cpt1a) in osteoclast progenitors, impaired osteoclast formation only in the female mice. The Cpt1a-deficient females were further protected from osteoclast activation by a high-fat diet. The males, on the contrary, exhibited normal bone resorption despite Cpt1a deletion, regardless of the dietary fat content. Moreover, concurrent deletion of mitochondrial pyruvate carrier 1 and Cpt1a, blocking mitochondrial oxidation of both glucose and fatty acids in the osteoclast lineage, failed to impede bone resorption in the males. The study therefore uncovers a female-specific dependence on mitochondrial oxidation of fatty acids and glucose in osteoclasts in vivo.


Introduction
Osteoclasts are multinucleated cells that resorb bone to initiate bone remodeling that is essential for bone health.Osteoclasts form resorption compartments on the bone surface, locally secreting acids as well as proteolytic enzymes to dissolve minerals and digest bone matrix (1).Under healthy conditions, bone resorption is balanced by de novo bone formation to maintain bone mass, but loss of the balance in favor of resorption causes bone loss and increases bone fragility.
Osteoclasts are derived from myeloid lineage cells in response to monocyte/macrophage colony stimulating factor (M-CSF) and receptor activation of NF-κB ligand (RANKL).Whereas M-CSF promotes proliferation and survival of osteoclast precursors, RANKL drives the differentiation of osteoclasts via the induction of key transcription factors such as c-fos and Nfatc1 (2).Therapies targeting osteoclasts, including bisphosphates and the more recent RANKL neutralizing antibody denosumab, have been the mainstay of osteoporosis treatment, but adverse effects including atypical fractures and osteonecrosis of the jaw have been reported (3).Thus, there remains a clinical need for safe and effective antiresorptive agents.
Although much is known about the signals and transcription factors regulating osteoclast differentiation, the bioenergetics of this process is just beginning to be understood.An early study with electron microscopy on bone sections documented that osteoclasts were uniquely abundant with mitochondria, compared to osteoblasts, osteocytes or endosteal lining cells (4).
More recent studies have shown that RANKL upregulates the Pparg coactivator Ppargc1b to stimulate mitochondrial biogenesis in the macrophage progenitors, and that genetic deletion of Ppargc1b impairs osteoclast function but not differentiation (5,6).Knockout of Ndufs4, a subunit of complex I of the mitochondrial electron transport chain (ETC), disrupts osteoclast formation and function in the mouse (7).Moreover, the mitochondrial transcription factor Tfam is required for sustaining intracellular ATP levels in mature osteoclasts to support osteoclast survival (8).Despite the prominent role of mitochondria in osteoclasts, much remains unclear about the energy substrates fueling mitochondrial respiration in those cells.
Studies to date have documented the role of glucose in osteoclast differentiation and function.
Osteoclasts from chickens, mice or humans have been shown to rely on glucose for their resorptive activity (9)(10)(11)(12).Glycolysis, together with OXPHOS, increases with osteoclast differentiation from bone marrow macrophages (13,14).Knockdown of lactate dehydrogenase, or deletion of the main glucose transporter Glut1, in the macrophage progenitors suppresses osteoclast differentiation in vitro (13,14).However, Glut1 knockout in myeloid lineage cells only modestly decreases the osteoclast number in vivo, and the effect is limited to female mice (13).Thus, other energy substrates besides glucose likely contribute to bioenergetics in osteoclasts.
Free fatty acids are a major class of energy substrates for mammalian cells.In particular, long chain fatty acids (C11-20) including oleic acid, palmitic acid and linoleic acid are the most abundant in the blood and tissues (15).Prior to ß-oxidation in the mitochondria, long chain fatty acids are transported into the matrix by a carnitine shuttle.The transport requires coordinated activities of Cpt1 (encoded by Cpt1a, Cpt1b, or Cpt1c) which converts fatty acyl-CoA to acylcarnitine at the outer membrane of mitochondria, and Cpt2 that reverses acylcarnitine to acyl-CoA at the inner membrane (16).A previous study showed that suppression of Cpt1a in macrophages mediated polysaccharide-induced inflammatory osteolysis, but a cell-autonomous role for Cpt1a in osteoclast differentiation was not determined (17).In addition, Cpt1a upregulation was shown to promote fusion among osteoclast progenitors isolated from rheumatoid arthritis patients, but it was unclear whether such a role was unique to the pathological condition (18).Elucidating the role of fatty acid oxidation in osteoclasts is likely to shed light on the mechanistic link between obesity and impaired bone health (19)(20)(21)(22).In mice, high fat diet has been shown to cause bone loss via promotion of osteoclast formation (23).
Although obesity has been linked with increased fatty acids oxidation in skeletal muscle and immune cells, a potential connection in osteoclasts is yet to be explored (24,25).
Here by employing metabolic and genetic techniques, we show that mitochondrial fatty acid oxidation supports osteoclast differentiation both in vitro and in the mouse.Remarkably, the requirement in vivo applies only to the female mice.Osteoclasts in the male mice, on the other hand, are refractory to the blockage of either fatty acid or glucose oxidation or both.The study therefore uncovers unexpected sexual dimorphism of energy substrate dependence in osteoclasts in vivo.

Fatty acid oxidation is upregulated with osteoclast differentiation
For osteoclast differentiation, bone marrow macrophages (BMM) flushed from murine long bones were expanded in the presence of M-CSF before being induced with M-CSF and RANKL (Supplemental Fig. S1A).TRAP (tartrate acid resistant phosphatase) staining detected numerous multi-nucleated osteoclasts after 4 days of differentiation (Figure 1A).Efficient osteoclast differentiation was confirmed by the robust induction of osteoclast marker genes from day 0 through day 2 and day 4 differentiation (Supplemental Fig. S1B-G).To profile the overall transcriptomic changes, we performed RNA-seq with bone marrow macrophages (BMM) before and after osteoclast differentiation (day 0 vs day 4).Ingenuity Pathway Analysis (IPA) of the RNA-seq data showed that TCA (tricarboxylic acid) cycle and OXPHOS (oxidative phosphorylation) genes were most significantly upregulated whereas oleate biosynthesis was significantly downregulated in osteoclasts compared to BMM (Supplemental Fig. S1H).
Examination of individual genes show that those encoding the TCA enzymes or components of the electron transport chain (ETC) were all induced in osteoclasts compared with BMM (Supplemental Fig. S1I, J).Moreover, genes related to fatty acid oxidation (FAO) in either peroxisomes or mitochondria were oppositely regulated during osteoclast differentiation.
To assess fatty acid oxidation directly, we quantified the CO2 production rate from 14 C-labelled oleate, used as a surrogate for long chain fatty acids.Oleate is physiologically relevant as it is among the most abundant fatty acids in the circulation and tissues of humans or mice (26).As a comparison, CO2 release from 14 C-glucose or 14 C-glutamine was also examined.Oleate oxidation did not change after two days of differentiation but markedly increased after four days (Fig. 1C).On the other hand, glucose and glutamine oxidation progressively increased in both preosteoclasts and osteoclasts over BMM (Fig. 1D, E).Thus, increased oxidation of long-chain fatty acids is selectively associated with mature osteoclast formation.
The increase in substrate oxidation predicts a greater rate of mitochondrial respiration following osteoclast differentiation.In our previous studies, preosteoclasts, upon dissociation and reseeding, displayed a higher oxygen consumption rate (OCR) than BMM according to Seahorse measurements, but mature osteoclasts were not assessed here due to technical difficulties in dissociating them from the culture plates (13).Here, we optimized the protocol to induce osteoclast differentiation directly in the Seahorse plates before measurements without reseeding.

Oleate oxidation in mitochondria promotes osteoclast formation
We next examined the role of fatty acid oxidation in osteoclast differentiation.Addition of BSAconjugated oleate as low as 25 µM to the differentiation media increased both the number and size of osteoclasts (Fig. 2A-D).Molecular analyses confirmed the upregulation of all common osteoclast markers including Acp5 (encoding TRAP), Nfatc1, c-fos and Ctsk by oleate (Fig. 2E).
Addition of oleate to the growth media did not alter the cell number of BMMs after 48 hrs, indicating that the increased differentiation was unlikely due to effects on cell proliferation (Fig. 2F).Furthermore, when charcoal-stripped FBS (cFBS) devoid of lipid was used in lieu of regular FBS in the differentiation media, essentially no osteoclast was formed but differentiation was fully restored when BSA-conjugated oleate was added to the cFBS (Fig. 2G, H).Molecular analyses revealed that with cFBS the mature osteoclast markers Acp5 and Ctsk were suppressed, but they were fully restored by the added oleate (Fig. 2I, J).Interestingly, cFBS also stimulated the expression of early osteoclast markers Nfatc1 and c-fos, but that effect was not reversed by the addition of oleate, indicating that certain yet unknown factors in FBS normally limit the expression of those genes (Fig. 2K, L).Overall, the results support a specific requirement for exogenous fatty acids during the late phase of osteoclast differentiation.
We next sought to verify that fatty acids support osteoclastogenesis by fueling mitochondrial respiration.Cpt2 is essential for long-chain fatty acid oxidation in the mitochondria as it catalyzes the conversion of acylcarnitine to acyl-CoA at the inner mitochondrial membrane.Knockdown of Cpt2 with shRNA not only potently suppressed osteoclast differentiation under regular culture conditions, but also eliminated the pro-osteoclastogenic effect of oleate (Fig. 3A,   B).The requirement for Cpt2 was specific to osteoclast differentiation as the knockdown did not affect the cell number of BMM in the growth media (Fig. 3C).Molecular analyses by RT-qPCR showed that the induction of Ctsk, Nfatc1 and Acp5 by oleate was abrogated by Cpt2 shRNA (Fig. 3D).Thus, mitochondrial ß-oxidation of long-chain fatty acids is necessary for osteoclast differentiation in vitro.

Deletion of Cpt1a reduces bone resorption in female mice
We next examined the role of long-chain fatty acid oxidation in osteoclastogenesis by deleting Cpt1a in the mouse.RNA-seq showed that Cpt1a was expressed at the highest level among the three Cpt1 genes in both BMM and osteoclasts (Supplemental Table II).We chose the LysM-Cre knock-out/knock-in allele for the deletion as it targeted the myeloid cell lineage including osteoclast progenitors.Previous studies have shown that homozygosity for the allele does not cause any bone phenotype by itself but excises the floxed genes more efficiently (27).We therefore generated Cpt1a deficient mice harboring either one (LysM Cre/+ ; Cpt1a f/f ) or two copies (LysM Cre/Cre ; Cpt1a f/f ) of the LysM-Cre allele, hereafter termed CKO Cpt or dCKO Cpt , respectively.
Expression studies confirmed that Cpt1a mRNA in BMMs was reduced by 65% and 90% of the normal level in CKO Cpt and dCKO Cpt , respectively, irrespective of the sex of the mouse (Fig. 4A).Importantly, osteoclast differentiation, with or without oleate supplementation in the culture media, was impaired in both mutant cells but more severely in dCKO Cpt BMMs from either male or female mice (Fig. 4B, C).Molecular analyses showed that all markers for osteoclast differentiation except for except for the earliest transcription factor c-fos were reduced in the Cpt1a-deficient cells (Fig. 4D-G).The results therefore further support an important role for mitochondrial oxidation of fatty acids in osteoclast differentiation in vitro.
To verify the metabolic effect caused by Cpt1a deletion, we performed quantitative studies in BMMs cultures.Oleate oxidation rate was reduced by 50% in CKO Cpt versus control BMM and further down by 50% in dCKO Cpt cells (Fig. 4H).When cultured with oleate added to the media, the CKO Cpt and dCKO Cpt BMM exhibited progressively lower levels of intracellular ATP than normal (Fig. 4I).Seahorse extracellular flux assays detected a decrease in both basal and ATPproduction OCR in CKO Cpt and further reduction in dCKO Cpt when compared to the control BMMs (Fig. 4J-L).Both CKO Cpt and dCKO Cpt similarly reduced the maximal OCR but did not affect the spare OCR (Fig. 4M, N).The mutant BMMs also exhibited lower ECAR than normal but only the dCKO Cpt cells reached statistical significance (Fig. 4O, P).The reduced ECAR likely reflected an increase in pyruvate entering into the TCA cycle, resulting in reduced lactate production.Consistent with this notion, the CKO Cpt BMMs exhibited a higher rate of carbon dioxide production from glucose than normal whereas glutamine oxidation was not affected (Fig. 4Q, R).Thus, deletion of Cpt1a disrupts bioenergetics in osteoclast progenitors.
We next examined the potential effect of Cpt1a deletion on bone in vivo.Among the female mice, µCT analyses detected a significant increase in the trabecular bone fraction (BV/TV) of CKO Cpt compared to the control, and a further increase in dCKO Cpt (Fig. 5A-B).The cortical bone parameters on the other hand were not altered in either of the mutant mice (Supplemental Fig. S2A-E).The increase in trabecular bone mass was accounted for by the increased trabecular number and reduced trabecular spacing without obvious changes in trabecular thickness (Fig. 5C-E).Bone resorption as indicated by serum CTX-I levels was reduced from control in both CKO Cpt and dCKO Cpt mice, whereas the levels of the bone formation marker P1NP were similar across the genotypes (Fig. 5F, G).Similarly, histomorphometric analyses following TRAP staining showed that the osteoclast surface fraction was more severely reduced in dCKO Cpt but that the osteoclast number per bone surface area was similarly reduced in both CKO Cpt and dCKO Cpt compared to the control (Fig. 5H-J).Contrary to the females, the mutant males with either one or two copies of LysM-Cre did not show any defect in either trabecular or cortical bone parameters (Fig. 5K-N, Supplemental Fig. S2F-I).Serum biochemical assays detected no abnormalities with either CTX-1 or PINP levels in the mutant males (Fig. 5O, P).Thus, deletion of Cpt1a impairs osteoclastogenesis and increases trabecular bone mass specifically in the female mice.

Blocking mitochondrial fatty acid oxidation eliminates HFD-induced bone loss in female mice
As previous studies have shown that high fat diet activates bone resorption in mice, we next investigated the potential role of mitochondrial fatty acid ß-oxidation in mediating the catabolic effect.To this end, we subjected both control and CKO Cpt mice to a high-fat diet (HFD) or a corresponding low-fat diet (LFD) for 8 weeks starting from weaning at 3 weeks of age.The two diets were strictly matched nutritionally except for fat contents (10 % vs 60% kcal) and the corresponding carbohydrate contents.The male mice, like those fed the regular chow diet, did not exhibit any difference in trabecular bone between the two genotypes, regardless of LFD or HFD (Fig. 6A-D).Moreover, HFD did not alter the trabecular bone parameters in the control males.Thus, mitochondrial fatty acid oxidation in osteoclasts appears to be dispensable for trabecular bone homeostasis in the male mice irrespective of the fat content in the diet.
In contrast to the males, the female mice, depending on their genotypes, responded differently to HFD.The control mice exhibited a marked reduction in trabecular bone fraction (BV/TV) in response to HFD, but the CKO Cpt mice showed no such change (Fig. 6E, F).Similarly, HFD decreased trabecular number and increased trabecular spacing only in the control but not the CKO Cpt females (Fig. 6G-I).Unexpectedly, the CKO Cpt females on LFD, unlike those fed the regular chow (containing 25% kcal in fat) as shown earlier, did not exhibit any bone defect when compared to the controls (Fig. 6E-I).Therefore, it appeared that the reliance of osteoclast bioenergetics on fatty acid oxidation was influenced by fat contents in the diet.When the dietary fat supply was minimal as in the LFD (10% kcal), the contribution of fatty acid oxidation was either negligible or easily compensated by the other energy sources when Cpt1a was deleted.
When the control mice were fed HFD, the serum CTX-I levels were significantly elevated, but the effect was eliminated in the CKO Cpt females (Fig. 6J).On the other hand, HFD did not affect P1NP levels in mice with either genotype (Fig. 6K).TRAP staining on femur sections revealed that HFD increased osteoclast surface areas without altering osteoclast numbers in the control mice but had no effect in the CKO Cpt mice (Fig. 6L, M).Thus, HFD promotes bone resorption in female mice mainly by increasing mitochondrial oxidation of fatty acids in osteoclasts.

Mitochondrial oxidation of fatty acids and glucose is dispensable in osteoclasts of male mice
The lack of bone phenotype in the male mice with Cpt1a deletion prompted us to examine the contribution of glucose oxidation in the mitochondria to osteoclast differentiation.To this end, we deleted Mpc1 which is required for the glycolytic product pyruvate to enter the mitochondria for oxidation.Here again the male mutant mice harboring two alleles of LysM-Cre (LysM Cre/Cre; Mpc1 f/f ) (termed dCKO Mpc ) did not exhibit any significant change in the fractional bone mass or any other trabecular bone parameters (Fig. 7A-E).The female mutants on the other hand showed an overall increase in the trabecular bone mass even though the individual contribution of trabecular thickness, number or spacing was not statistically significant (Fig. 7F-J).
Histomorphometry showed that the osteoclast fractional surface but not the relative number was significantly reduced in the Mpc1 mutant females (Fig. 7K, L).Thus, like fatty acids, mitochondrial oxidation of pyruvate, derived from glucose or other sources, appears to be dispensable for bone mass homeostasis in male mice.Potential compensation between fatty acid and pyruvate oxidation could mask their individual contribution to osteoclast bioenergetics in the male mice.To investigate this possibility, we generated double deletion mice with two alleles of LysM-Cre (LysM Cre/Cre ; Cpt1a f/f ; Mpc1 f/f ), designated herein dCKO CptMpc .Quantification by RT-qPCR confirmed that both Cpt1a and Mpc1 were deleted with >90% efficiency in BMM in either sex (Fig. 7M, N).However, no changes were observed with the bone parameters in the male mutants (Fig. 7O-R).In contrast, the female mutant mice showed a clear increase in the fractional bone mass, attributable to increased trabecular number and reduced trabecular spacing (Fig. 7S-V).The excess bone mass was accounted for by a nearly 50% decrease in bone resorption as indicated by the serum CTX-I level, without any change in the bone formation marker P1NP (Fig. 7W, X).Histomorphometry confirmed that the fractional surface of osteoclasts was significantly reduced in the mutant females, whereas the relative osteoclast number showed a trend of reduction without reaching statistical significance (Fig. 7Y, Z).The results therefore underscore the remarkable resilience of osteoclasts in the male mice when mitochondrial oxidation of both fatty acids and pyruvate is blocked.

Discussion
We have examined in detail the role of mitochondrial fatty acid oxidation in osteoclast formation in vitro and in the mouse.In keeping with the broad upregulation of genes associated with mitochondrial respiration and fatty acid oxidation, mature osteoclasts in vitro markedly increased the oxidation of long-chain fatty acids compared to their progenitors.Genetic disruption of mitochondrial fatty acid oxidation in the osteoclast lineage suppressed bone resorption in female mice fed with either regular chow or a high fat diet.A similar effect, however, was not observed in the males, even when both fatty acids and pyruvate oxidation in the mitochondria were blocked in osteoclasts.The study therefore not only identifies an important role for fatty acid oxidation in osteoclast formation, but also reveals unexpected sexual dimorphism in energy substrate dependence of osteoclasts in vivo.
The sex-dependent effect in vivo following Cpt1a or Mpc1 deletion or both in osteoclasts warrants further investigation.Our finding is consistent with the recent report that deletion of Cpt2 in osteoclasts also specifically affects the female mice (28).Our in vitro studies show that osteoclast differentiation from BMM of either sex is equally impaired by the loss of Cpt1a or Cpt2.Thus, the female-specific effect is unlikely due to cell-intrinsic differences caused by the sex chromosomes.Previously, we have found that restricting glucose metabolism in osteoclasts by deletion of Glut1 impairs bone resorption also only in the female mice (13).Collectively, the data indicate that sex-specific endocrine or paracrine factors likely play a major role in controlling energy substrate utilization by osteoclasts in vivo.Future experiments are necessary to determine whether the male hormones confer a greater level of plasticity in metabolic switching to either alternative substrates or pathways in osteoclasts.
The in vitro studies have revealed stage-specific preferences for the different energy substrates during osteoclast differentiation.The increase in fatty acid oxidation occurred only in the large multinucleated osteoclasts which formed after four days of differentiation, but not in the preosteoclasts formed after two days in our culture system.In contrast, oxidation of glucose and glutamine oxidation increased in both preosteoclasts and osteoclasts compared to BMM.The surge in fatty acid oxidation coincides with the fusion of mononuclear preosteoclasts to form multinuclear osteoclasts but the underlying mechanism is unclear at present.Apart from the obvious contribution to energy production, increased fatty acid oxidation could alter the levels of certain metabolites to influence either cell membrane dynamics or the epigenetic regulation in support of mature osteoclast formation.
Our finding about the male mice fed with HFD is at odds with a previous study that used the same LFD and HFD as ours but detected an increase in bone resorption in the male mice (23).
One obvious difference was in the starting age for LFD or HFD feeding, namely 5 weeks in the previous study versus 3 weeks here.The different starting age could be significant as the bone resorption rate can vary considerably with age.Furthermore, the regular chow (LabDiet 5015) that our mice consumed before switching to matched HFD/LFD was different from that used in the previous study (LabDiet 5010), with ours containing a significantly higher amount of fat (25% vs 12.7% kcal).It is possible that the different fat contents consumed by the mice before switching to the HFD led to different responses.We did not monitor whole-body metabolism in our studies, but the previous study reported notable hyperglycemia and metabolic dysfunction in their mice, which could have separately contributed to the reported bone resorption defect, independent of fatty acid fueling of osteoclasts.Overall, the different outcomes highlight the highly variable metabolic responses by osteoclasts to different diets, and further studies are needed for a mechanistic understanding.

Metabolic assays
For Seahorse assays, 2 x 10 4 BMMs were seeded in a Seahorse plate and incubated in growth or differentiation media for 2 or 4 days, with media replaced every 12 hours.Osteoclast differentiation in Seahorse plates was verified with TRAP staining on day 4. Before Seahorse measurements, the cell culture media was replaced with Seahorse XF DMEM media supplemented with 5.5 mM glucose, 2 mM glutamine, 1 mM pyruvate and 100 mM BSAconjugated oleate, and placed in a CO2-free incubator for 1 hr.For Mito Stress tests, 2 μM Oligomycin, 3 μM FCCP and 1 μM Rotenone/Antimycin A were used.The Seahorse data was normalized to genomic DNA as measured in lysed cells with BioTek Cytation 5 (Agilent Technologies) following staining with Hoechst 33342 (Thermo Fisher Scientific, USA).
For steady-state ATP measurements, BMM were seeded in 96 well plates at 7.5 x 10 4 cells/cm 2 and cultured in growth media overnight.100 μL fresh media per well was replaced 4 hrs before intracellular ATP levels were measured with the CellTiter-Glo kit (Promega, USA).The ATP levels were normalized to the cell numbers.
Substrates oxidation assays were conducted as previously described (32).In brief, 14 C-labelled oleate, glucose or glutamine were mixed as tracers with the bulk unlabeled substrates.Full media containing the different tracers were incubated with cells for 4 hours before an aliquot of the media was transferred to a 1.5 ml Eppendorf tube containing 200 µL perchloric acid.Released CO2 was captured by a filter paper fitted inside the tube cap and saturated with sodium hydroxide.The radiation counts on the filter papers were used to compute the substrate oxidation rates.

RNA-Seq and RT-qPCR
Total RNA was extracted with RNeasy kits (Qiagen, USA) before being subjected to highthroughput RNA sequencing (Genewiz, USA), or RT-qPCR.For RNA sequencing, RNA samples were quantified using Qubit 2.0 Fluorometer (Life Technologies, USA) and RNA integrity was checked using Agilent TapeStation 4200 (Agilent Technologies, USA).The RNA sequencing libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina following manufacturer's instructions (New England Biolabs, USA).The sequencing libraries were clustered on one flowcell lane.After clustering, the flowcell was loaded on the Illumina HiSeq instrument (4000 or equivalent) according to manufacturer's instructions.The samples were sequenced using a 2x150bp Paired End (PE) configuration.Image analysis and base calling were conducted by the Control software.Raw sequence data (.bcl files) generated from the sequencer were converted into fastq files and de-multiplexed using Illumina's bcl2fastq 2.17 software.After investigating the quality of the raw data, sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality using Trimmomatic v.0.36.
The trimmed reads were mapped to the reference genome available on ENSEMBL using the STAR aligner v.2.5.2b.After extraction of gene hit counts, the gene hit counts table was used for downstream differential expression analysis.Using DESeq2, a comparison of gene expression between the groups of samples was performed.The Wald test was used to generate p-values and Log2 fold changes.
For RT-qPCR, cDNA was produced with High Capacity RNA-to-cDNA kit (ThermoFisher Scientific) and then used for qPCR with SYBR green PCR master mix (Applied Biosystems) in a Quantstudio3 machine (Applied Biosystems).Gene-specific qPCR primers are listed in Supplemental Table I.

Oleate-BSA conjugation
To make 4 mM BSA, 5.5 g BSA was mixed with 11.5 mL Dulbecco's phosphate-buffered saline (DPBS) followed by gentle shaking at room temperature for 3 hrs.To conjugate oleate with BSA, 24.36 mg of sodium oleate was added to 2 mL of H2O in a water bath at 70 °C for ~20 min before mixing quickly with 6 ml of the 4 mM BSA stock solution prewarmed to 70 °C.

Bone histology and µCT analyses
Femurs were fixed in 10 % formalin for 48 hrs at room temperature and then decalcified in 14% EDTA solution (pH 7.2) for 14 days with the solution changed every 4 days.The bones were then processed, embedded in paraffin and sectioned at 6 µm thickness.The sections were subjected to TRAP staining for osteoclast detection.Osteoclasts were quantified with BIOQUANT OSTEO 2021 V21.5.60 (Bioquant, USA).
For µCT analyses, femurs fixed with 10% formalin were scanned using µCT 45 (Scanco Medical AG, Switzerland) at 4.5 µm isotropic voxel size according to the guidelines of American Society of Bone and Mineral Research (34).For trabecular bone analysis, 200 slices starting at 70 slices under the growth plate of the distal femur were analyzed with the threshold set at 330.For cortical bone analysis, 35 slices before and after the middle slice of the femur diaphysis were analyzed with the threshold set at 370.

Figure 4 .
Figure 4. Mitochondrial fatty acid oxidation supports BMM bioenergetics and

Figure 5 .
Figure 5. Deletion of Cpt1a suppresses osteoclast formation and increases bone mass in

Figure 6 .
Figure 6.Cpt1a mediates bone loss caused by HFD in female mice only.(A-D)

Figure 7 .
Figure 7. Deletion of Mpc1 alone or both Mpc1 and Cpt1a suppresses osteoclasts in female