Mitochondrial Sirt3 contributes to the bone loss caused by aging or estrogen deficiency

Altered mitochondria activity in osteoblasts and osteoclasts has been implicated in the loss of bone mass associated with aging and estrogen deficiency — the 2 most common causes of osteoporosis. However, the mechanisms that control mitochondrial metabolism in bone cells during health or disease remain unknown. The mitochondrial deacetylase sirtuin-3 (Sirt3) has been earlier implicated in age-related diseases. Here, we show that deletion of Sirt3 had no effect on the skeleton of young mice but attenuated the age-related loss of bone mass in both sexes. This effect was associated with impaired bone resorption. Osteoclast progenitors from aged Sirt3-null mice were able to differentiate into osteoclasts, though the differentiated cells exhibited impaired polykaryon formation and resorptive activity, as well as decreased oxidative phosphorylation and mitophagy. The Sirt3 inhibitor LC-0296 recapitulated the effects of Sirt3 deletion in osteoclast formation and mitochondrial function, and its administration to aging mice increased bone mass. Deletion of Sirt3 also attenuated the increase in bone resorption and loss of bone mass caused by estrogen deficiency. These findings suggest that Sirt3 inhibition and the resulting impairment of osteoclast mitochondrial function could be a novel therapeutic intervention for the 2 most important causes of osteoporosis.


Introduction
A balance between the function of bone-resorbing osteoclasts and bone-building osteoblasts is essential for bone homeostasis. Resorption of the mineralized bone matrix-a physiologic process essential for skeletal and mineral homeostasis-is the function of osteoclasts, multinucleated cells derived from myeloid precursors (1)(2)(3)(4). During osteoclastogenesis, bone marrow macrophages (BMMs) differentiate into tartrate-resistant acid phosphatase (TRAP)positive pre-osteoclasts in response to the receptor activator of nuclear factor kappa-Β ligand (RANKL) (5). Mononuclear pre-osteoclasts then fuse with each other to form multinucleated mature osteoclasts. These highly specialized cells are uniquely capable of dissolving and digesting organic bone matrix (5). Most likely because of the high energy demands of these tasks, a distinct cellular feature of osteoclasts is the high abundance of mitochondria (6,7).
The NAD + -dependent sirtuin-3 (Sirt3) is the primary mitochondrial protein deacetylase and plays a critical role in mitochondrial quality control, including mitochondrial biogenesis, mitochondrial dynamics, and mitophagy-all of which are affected in age-related metabolic diseases (8)(9)(10)(11). Earlier work aiming to elucidate the role of Sirt3 in bone has produced conflicting results with some studies suggesting that Sirt3 is critical for skeletal homeostasis (12,13) while others found no physiological role of Sirt3 in bone (14,15).
Sex steroid deficiency, aging, and inflammation cause an increase in osteoclast numbers and thereby bone loss (1)(2)(3)(4). We and others have shown that RANKL-induced stimulation of mitochondrial function in osteoclasts is critical for osteoclastogenesis under physiologic conditions and contributes to loss of bone mass with estrogen deficiency (16)(17)(18)(19)(20)(21). We have also shown that suppression of osteoclast formation by estrogens is associated with decreased mitochondria OxPhos and ATP production in early osteoclast progenitors (22). Further, increased mitochondrial ROS levels in cells of the osteoblast lineage contributes to the loss of bone mass with age (23). Collectively, these findings strongly implicate altered mitochondrial metabolism in both osteoclasts and osteoblasts as culprit of bone disease. However, the molecular mechanisms that control mitochondrial function in skeletal health and disease remain largely unknown. We show herein, that deletion of Sirt3 or pharmacologic inhibition of Sirt3 in mice impairs osteoclast mitochondrial function and prevents the increased bone resorption and loss of bone mass caused by the two most important causes of osteoporosis in humans -aging and estrogen deficiency.

Results
The age-associated loss of bone mass is prevented in Sirt3 knockout mice. Mice heterozygous for the Sirt3 mutant allele were intercrossed to produce homozygous knockout mice. Cohorts of Sirt3 knockout and littermate wild-type mice were aged up to 16 months. Consistent with previous studies (24), Sirt3 knockout mice appeared normal at birth and were indistinguishable from wild type mice with respect to size, body weight (data not shown), and femoral length during adulthood (Supplementary Figure 1 Figure 2) at the femur and spine compared to their age-matched controls. As expected, wild-type mice lost bone mass with age. Specifically, 16-month-old wildtype female mice had lower femoral cortical thickness compared to 6-month-old mice of the same genotype ( Figure 1, A-B). The decrease in cortical thickness in the aged wild-type female mice was due to a disproportionate increase in medullary area as compared to the increase in total bone area ( Figure 1, C-D). Male wild-type mice also exhibited a decrease in cortical thickness with age ( Figure 1E), as previously described (25,26). Cortical thickness in both female and male Sirt3 knockout mice at 16 months of age, was greater than wild-type controls of the same age ( Figure 1, A-B and Figure 1E). More striking, 16-month-old Sirt3 null female mice had thicker cortical bone than 6-month-old mice of the same genotype ( Figure 1, A-B). This was due to the lack of enlargement of the medullary cavity in the face of the expected enlargement of bone (total area) that occurs with age ( Figure 1, C-D). The loss of cortical bone with age was also attenuated in Sirt3 knockout males, but to a lesser degree than in females ( Figure 1E).
In wild-type female mice, trabecular bone volume (BV/TV) and 3D bone mineral density (BMD) decreased between 6 and 16 months of age at both the spine (Figure 2 In contrast to females, the deletion of Sirt3 had no effect on the age-related loss of trabecular bone mass of the femur or spine in male Sirt3 null mice ( . Taken together, these results indicate that Sirt3 contributes to the age-related bone loss in mice and its contribution is greater in female than male mice and more pronounced in cortical than trabecular bone.

Deletion of Sirt3 decreases resorption markers in vivo and osteoclast function in vitro.
Agerelated bone loss is associated with decreased osteoblasts and increased osteoclast numbers. To gain insight into the cellular mechanism(s) of the protective effect of Sirt3 deletion on skeletal aging, we first performed histologic analysis to evaluate osteoclast and osteoblast numbers at the endocortical surface of the femur. Sixteen-month-old Sirt3 knockout female mice showed no differences in osteoclast or osteoblast number (Figure 3, A-C). Strikingly, however, the serum levels of C-terminal telopeptide of type 1 collagen (CTx), osteocalcin, and N-terminal propeptide of type I procollagen (P1NP) -markers of bone resorption and bone formation, respectivelywere dramatically lower in Sirt3 knockout mice ( Figure 3D). This finding suggests that the high bone mass of these mice is, at least in part, due to a decrease in osteoclastic bone resorption and the resorption-driven bone turnover.
To probe into the mechanism of the decreased resorption in the Sirt3 null mice in the face of unchanged osteoclast numbers, we examined osteoclast function. Cultures of bone marrow macrophages from aged mice lacking Sirt3 formed smaller osteoclasts than those from wild-type mice ( Figure 4A). Moreover, the area resorbed by Sirt3-deficient osteoclasts was greatly diminished in these cultures ( Figure 4B). The expression of late/terminal osteoclast differentiation markers was decreased in cultured osteoclasts from aged Sirt3 knockout mice ( Figure 4C). Protein levels of factors known to control early osteoclast differentiation, such as c-Fos and NFATc1 (27), were not affected by Sirt3 deletion ( Figure 4D). Nonetheless, both protein and mRNA levels of Sirt3 were greatly upregulated during the process of osteoclastogenesis, as shown in macrophage cultures from wild-type mice stimulated with RANKL ( Figure 4, D-E).
This result is consistent with the requirement of Sirt3 for optimal mitochondrial biogenesis and function in many different cell types.
Bone marrow-derived stromal cells from aged Sirt3 knockout mice cultured under osteogenic conditions exhibited a mild decrease in osteoblast differentiation ( Figure 4F). The mRNA levels of the early osteoblastogenesis markers Runx2 and Akp were unaffected by the Sirt3 deletion.
However, osteocalcin was decreased in cultured stromal cells from aged Sirt3 knockout mice ( Figure 4G). In cultures derived from wild-type mice the mRNA levels of Sirt3 were upregulated during osteoblastogenesis ( Figure 4H). Taken together, these results suggest that the increase in bone resorption associated with aging depends, at least in part, on direct effects of Sirt3 on osteoclasts.
Deletion of Sirt3 decreases mitochondrial respiration and mitophagy in osteoclasts. We next examined whether Sirt3 alters mitochondrial function in osteoclasts by performing extracellular flux analysis. Mitochondrial respiration was significantly diminished in osteoclasts from aged Sirt3 knockout mice ( Figure 5A). Deletion of Sirt3 also decreased ATP-linked respiration, proton leak, maximum respiration (maximal electron transport chain activity), and non-mitochondrial respiration, but had no impact on reserve respiratory capacity -the difference between the maximum oxygen consumption rate and basal respiratory OCR that provides cells flexibility during increased energy demands ( Figure 5, B-F). However, Sirt3 lacking osteoclasts from young mice exhibited no difference in mitochondrial respiration compared to the cells from agematched controls (Supplementary Figure 3B). Consistent with this, the inhibitory effect of Sirt3 deletion on osteoclast differentiation seen in cells from old mice (-103.5 ± 14.18; Figure 4A We also examined mitochondrial content using MitoBright Green staining and expression levels of two genes encoded by the mitochondria genome, ND2 and Cytb (normalized to genomic DNA content). Both fluorescence imaging and flow cytometry analysis of mitochondria of differentiated osteoclasts revealed no differences in MitoBright Green staining between the two genotypes ( Figure 6, A-C); in fact, the signals were slightly higher in osteoclast progenitors from Sirt3 knockout mice. Likewise, Sirt3 deletion had no effect on the expression of mitochondrial DNA-encoded genes, as well as mRNA levels of Pgc1b or Tfam -two transcription factors that regulate mitochondrial biogenesis (28) (Supplementary Figure 4). The expression of several proteins representative of mitochondria, such as NAD dehydrogenase (NDUFA9; Complex I), succinate dehydrogenase complex subunit A (SDHA; Complex II), cytochrome C oxidase IV (COX IV; Complex V), pyruvate dehydrogenase (PDH, inner mitochondrial membrane), heat-shock protein 60 (HSP60; mitochondrial chaperonin), and voltage-dependent anion channel (VDAC, outer mitochondria membrane) (29)(30)(31)(32)(33)(34) were also unaffected by Sirt3 deletion ( Figure 6D). These results suggest that Sirt3 deletion inhibits mitochondrial respiration independent of changes in mitochondrial content.
We next examined whether Sirt3 deletion altered the abundance of proteins involved in mitochondria dynamics and mitophagy. Mitofusin2 mediates the tethering of adjacent mitochondria as well as the tethering of mitochondria to the ER and promotes osteoclast differentiation (35). Sirt3 deletion had no impact on the protein levels of Mitofusin2 in BMM stimulated with RANKL ( Figure 6E). Sirt3 activates mitophagy via Pink1 and increases expression of Bnip3 and Nix (36)(37)(38). The protein levels of the macroautophagy markers LC3 and p62 were unaffected by Sirt3 deletion. However, Bnip3 and Nix were reduced in cultured osteoclasts from aged Sirt3 knockout mice ( Figure 6E). Interestingly, osteoclasts lacking Sirt3 had increased Pink1 acetylation but no changes in total protein levels ( Figure 6F). These results indicate that Sirt3-mediated Pink1 deacetylation might regulate mitophagy in osteoclasts, which in turn promotes osteoclast differentiation. Indeed, Sirt3 induces mitophagy via direct deacetylation of Pink1 in primary mouse cardiac microvascular endothelial cells (39). Deletion of Sirt3 protects against ovariectomy-induced cortical but not trabecular bone loss. A decline of estrogen levels at menopause or following ovariectomy (OVX) causes bone loss in women and mice (2,41). This loss occurs at both cancellous and cortical sites and is associated with high bone remodeling, as evidenced by increased numbers of both osteoclasts and osteoblasts (42). We have previously shown that 17β-estradiol (E2) inhibits mitochondrial function in osteoclast progenitors (22). Based on this background, we searched for and found that E2 decreased the mRNA levels of Sirt3 in wild-type BMMs cultured with RANKL ( Figure 8A).
In contrast, BMMs from mice with conditional deletion of estrogen receptor α in cells of the myeloid lineage (22,43) had higher Sirt3 mRNA levels ( Figure 8B), suggesting that Sirt3 might be involved in the actions of estrogens in osteoclasts. To investigate this possibility, we performed OVX on Sirt3 knockout mice and wild-type littermate controls at 5 months of age and examined the impact of estrogen deficiency on bone mass 6 weeks later. OVX caused a decrease of BMD in the femur and spine of wild-type and Sirt3 null mice, as determined by Dual-energy X-ray absorptiometry ( Figure 8, C-D). Nonetheless, the OVX-induced loss of BMD was attenuated in Sirt3 null mice. OVX decreased cortical bone thickness in the femur of control mice ( Figure 8E) due to an increase in the medullary area, while total area was not affected (Figure 8, F-G). Notably, the effects of OVX on cortical bone were greatly attenuated in Sirt3 knockout mice (Figure 8, E and G). Similarly, cortical thickness in the vertebra was reduced with OVX in control mice, but not in Sirt3 knockout mice ( Figure 8H). Two-way ANOVA showed a significant difference between genotypes in the magnitude of the bone loss at both sites (femur: p-int = 0.006 and spine: p-int < 0.0001). In contrast, the loss of trabecular bone and changes in microarchitecture with OVX were indistinguishable between control and Sirt3 knockout mice in both vertebrae and femur ( Figure 8I and Supplementary Figure, [7][8]. Finally, the OVX-induced increase in bone resorption, as determined by serum CTx levels, was attenuated by Sirt3 deletion ( Figure 8J). We found no differences in bone formation parameters between the two genotypes in OVX experiments (Supplementary Figure 9), suggesting that the attenuation of OVX-induced bone loss in Sirt3 null mice is due to a decrease in bone resorption, not an increase in bone formation.

Discussion
The increase of fractures associated with the loss of bone mass with old age or sex steroid deficiency represents a major clinical problem for both women and men (1,25,26,44,45).
Numerous descriptive studies have demonstrated dysregulation of mitochondrial function in agerelated diseases, but it is largely unknown whether altered mitochondrial function in skeletal cells plays a role in the decrease of bone mass in the elderly. Herein we provide evidence in mice that the mitochondrial protein deacetylase Sirt3 is required for the full manifestation of the effects of aging and OVX on bone resorption; and its adverse effects, most likely, result from enhanced mitochondrial function in osteoclasts.
Deletion of Sirt3 decreased bone resorption but had no effect on osteoclast number in trabecular or endocortical bone surfaces. Similar findings have been reported in mice lacking other proteins important for mitochondria. For example, global deletion of PGC1β, a critical transcription factor for mitochondrial oxidative energy metabolism, decreases mitochondrial biogenesis in osteoclasts and increases bone mass (16). Serum levels of CTx are diminished in PGC1β knockout mice but the number of osteoclasts in bone is indistinguishable from that of control mice. Likewise, mice with conditional deletion of PGC1β in myeloid cells have normal osteoclast number in bone but display high bone mass compared to control littermates (46).
These effects of PGC1β are reminiscent of the effect of Sirt3 and support the contention that mitochondria metabolism promotes osteoclast differentiation and bone resorption.
Sirt3 promotes mitochondria metabolism by deacetylating several target proteins. We found that the acetylation of Pink1 was increased and the levels of Bnip3 and Nix decreased in osteoclasts from aged Sirt3 knockout mice suggesting that Sirt3 promotes osteoclast formation and function in part by stimulating mitophagy. Mitochondrial protein turnover by mitophagy is critical for the maintenance of healthy functional mitochondria. Pink1 and the ubiquitin ligase Parkin normally work together in the same pathway to govern mitochondrial quality control and loss-of-function of the two proteins are involved in many cases of early-onset familial Parkinson's disease (47). Sirt3 activates Pink1 to promote the ubiquitination of mitochondrial outer membrane proteins (48). Mitochondria dysfunction frequently mediates the cellular impairment caused by disruption of autophagy. Interestingly, studies with mice lacking critical proteins for autophagy in osteoclasts have revealed that autophagy promotes bone resorption and mediates the loss of bone mass caused by estrogen deficiency (49). Overall, these findings support the notion that the effects of Sirt3 in osteoclasts are due, at least in part, to mitophagy.
Nonetheless, a broader analysis of lysine acetylome using proteomics will be required to identify the range of target proteins and mitochondria pathways that are responsible for the effects of Sirt3 during osteoclast development.
We have previously reported that the increase in osteoclast number with aging is associated with an increase in RANKL expression by osteocytes; and, this source of RANKL is indispensable for the age-related loss of bone at endocortical and intracortical surfaces (50).
Several studies have shown that RANKL promotes mitochondria activity in osteoclast progenitor cells via multiple mechanisms including increased mitochondria biogenesis and stimulation of OxPhos (16)(17)(18)(19)(20)(21). We found here that RANKL increases the mRNA and protein levels of Sirt3 which in turn stimulates osteoclast formation and bone resorption, most likely by promoting mitochondria metabolism. In the absence of Sirt3, mitochondrial function in osteoclasts declines with age resulting in osteoclasts with reduced function. This preserves bone mass in mice.
Curiously, the skeletal effects of Sirt3 are relevant under stress but not homeostatic conditions. This contention is supported by findings that transgenic mice overexpressing Sirt3 exhibit cortical and trabecular bone loss at 13 month of age but not at 6 month of age (51). Likewise, much of the current knowledge on the functions of Sirt3 in tissues other than bone was obtained from studies with Sirt3 knockout mice under pathological conditions. Specifically, in cells such as neurons, cardiomyocytes, and hepatocytes Sirt3 exerts a protective role against inflammation, oxidative stress, and old age most likely by promoting deacetylation of its target proteins (8)(9)(10)(11)52). While the reasons for the different actions of Sirt3 under homeostatic and disease conditions remain unclear, recent proteomics studies have shed some light on the mechanisms via which Sirt3 exerts different roles in healthy versus stressed mitochondria (10). This study indicates that lower mitochondria matrix pH conditions due to mitochondrial membrane depolarization causes Sirt3 to change protein substrates and deacetylate proteins to promote oxidative metabolism and restore the proton gradient. These findings are in line with our results showing that Sirt3 promotes oxidative phosphorylation in osteoclastic cells from old mice. Furthermore, previous studies have shown that macrophages from old humans or mice exhibit mitochondria dysfunction (53). Together, these findings suggest that age-related changes in mitochondria of osteoclast lineage cells trigger the requirement for Sirt3 to maintain osteoclast formation and resorptive activities.
Estrogens inhibit osteoclast formation and protect against bone loss via direct actions on cells of the osteoclast lineage and indirectly via actions on cells of the mesenchymal lineage and perhaps other cell types, as well as several molecular mechanisms (1,23,43,45). We have shown that mitochondrial H2O2 in osteoclasts plays a role in the loss of bone with estrogen deficiency (23). Furthermore, results from a recent in vitro study of ours have suggested that direct effects on osteoclast mitochondria mediate the antiosteoclastogenic actions of estrogens (22). For reasons that remain unclear, some mechanisms of estrogen action predominantly influence cortical bone while others influence predominately trabecular bone. Here, we found that E2 decreases the expression of Sirt3 in osteoclast progenitors; and that the increase in bone resorption and loss of cortical, but not trabecular, bone mass with estrogen deficiency is attenuated in mice lacking Sirt3. Future studies are needed to further dissect the pathways via which estrogens directly and indirectly impact osteoclast mitochondria and why these predominate in distinct bone compartments. While our results point to a decrease in bone resorption as the culprit of the increased bone mass in aged or OVX Sirt3 knockout mice, a disproportionate effect on bone resorption could potentially mask an effect of Sirt3 on bone formation. Indeed, bone marrow-derived osteoblast progenitors lacking Sirt3 exhibit a small decrease in osteoblastogenesis and the serum levels of osteocalcin and P1NP were diminished in aged Sirt3 knockout mice. Nevertheless, further genetic studies will be required to elucidate whether Sirt3 in osteoclast, osteoblast or other cell types affect age-or estrogen deficiencyrelated bone loss.
Sirt3 protects against the loss of cortical bone with age in females, but has a much milder effects in males suggesting gender-related differences of Sirt3 actions in bone. During our previous studies aimed at characterizing skeletal aging in mice, we noticed that female and male C57BL/6 mice exhibit different features of bone aging (26). For example, the increase in intracortical bone remodeling is of greater magnitude in aged female than male mice. Others have also noted osteoclast related differences between males and females. Specifically, female mice have more osteoclasts and a lower bone mass in the trabecular compartment than males at any given age (54) and cultures of bone marrow macrophages from female mice form more osteoclasts than those from male mice (55). In addition, Veis and colleagues have reported that female but not male mice lacking mitochondrial fission protein Mitofusin2 in myeloid cells are protected from bone loss with age and from RANKL-induced osteolysis (35). The authors noted that this might be a threshold effect for mitochondrial function on osteoclast differentiation, and that females have more sensitivity to Mitofusin2 loss than males. Along with this earlier evidence, our findings here support the notion that the mechanisms of skeletal aging are in part different between males and females and that sexual dimorphic responses of osteoclasts are the likely reason. In addition, our findings suggest that osteoclast mitochondria and Sirt3 might underlie the sex different mechanisms in these cells. Evidently in humans, the sex-related differences in skeletal aging are attributed, in part, to hormonal changes at the menopause. Thus, in women it is very difficult to dissect effects of aging from the ones of estrogen loss. In contrast, female mice maintain functional levels of estrogen until old age (23) and, thereby, represent a very useful model to discern mechanisms of aging from mechanisms of estrogen deficiency and thus elucidate gender dimorphic effects on age-related bone loss.
We found no changes of bone mass in 6-month-old Sirt3 knockout female or male mice, in line with earlier reports by Ho et al. and Busse et al. (15,51). As in bone, Sirt3 has no effect under unstressed conditions in several other tissues (56)(57)(58). In difference to the lack of an effect of Sirt3 deletion on bone mass, Gao et al. (13) had suggested that Sirt3 knockout mice have low bone mass. However, due to the lack of specific information about the mice (e.g., sex and age) used by Gao et al., it is unclear whether the seeming discrepancy can be explained by differences in age or genetic background. While our paper was in review, Li et al reported that deletion of Sirt3 increases bone mass in 3-and 6-month-old female but not male mice. In addition, these authors showed that Sirt3 knockout mice are protected against ovariectomy-induced bone loss (59), in line with our present findings. The different impact of Sirt3 deletion on the bone phenotype of young mice might be due to the different genetic background of the mice.
In conclusion, the results of the present work reveal that mitochondrial deacetylase Sirt3 is a major contributor of the increased bone resorption in the pathogenesis of osteoporosis. We also provide proof of principle that pharmacological inhibition of Sirt3 with LC-0296 is sufficient to mimic the skeletal effects of the genetic deletion and attenuate the loss of bone mass in aging mice. Nonetheless, these findings are at odds with previous suggestions that stimulation of Sirt3 represents a good strategy to ameliorate diseases of aging (8)(9)(10)(11)52). The usefulness of targeting Sirt3 in age-related diseases notwithstanding, our study highlights the importance of the regulation of mitochondria in bone cells and its contribution to diseases of the skeleton.

Methods
Animal experiments. The Sirt3 knockout mice were maintained and bred in our facility using pairs of mice heterozygous for the Sirt3 mutant allele (B6/Sv129 mixed background) (The Jackson Laboratory, 012755) donated by David Gius (Northwestern University, Chicago, IL) (24). To determine whether Sirt3 plays a role in age-associated bone loss, mice were aged up to 16 months. Six-month-old female and 4-month-old male mice of the same genotypes were used as young adult controls. To test the effects of estrogen deficiency, 5-month-old female Sirt3 knockout mice and wild-type littermate controls were randomized into sham-or OVX-surgical groups according to their femoral Dual-energy X-ray absorptiometry (DXA) BMD. BMD measurements were performed 1 day prior to surgery and before euthanasia. Mice were injected with tetracycline (15 mg/kg body weight) 10 and 3 days before euthanasia to quantify boneformation rates. Animals were sacrificed 6 weeks after surgery and the tissues were dissected for further analyses.
To test the in vivo effects of Sirt3 antagonizing compounds, 12-month-old female C57BL/6 mice (The Jackson Laboratory) were randomly assigned to one of two treatment groups based on body weight and received i.p. injections of either the vehicle (100 µL/each injection; 1:2 DMSO/PBS mixture) or LC-0296 (5 µg/g body weight per day; AOBIOUS INC) 3 times per week for 16 weeks. Mice were euthanized and tissues were harvested 2 days after the last LC-0296 administration. ERα ΔLysM mice (C57BL/6 background) were generated as described before (22).
Genomic DNA extracted from tail samples was used for PCR-based genotyping following the protocols from the Jackson Laboratory. All mice used in this study were housed under standard laboratory conditions with a 12 hour dark, 12 hour light cycle, a constant temperature of 23 °C, and humidity of 48%. A standard rodent diet (Envigo, Teklad 22/5) containing 22% protein, 1.13% calcium, and 0.94% phosphorus was provided to mice ad-libitum. For aging study, the mice were switched to Teklad 2014 rodent diet (Envigo), containing 14% protein and 4% fat, and acidified water ad libitum at 8 months of age. Investigators were blinded to study groups during animal handling and endpoint measurements. DXA and micro-CT. BMD measurements were performed by DXA using a PIXImus densitometer (GE Lunar) on mice sedated with 2% isoflurane and data were analyzed as previously described (43). Scans of the entire left femur or lumbar spine were used for the measurement of BMD. Micro-CT analyses of the distal end of the femora and the vertebrae (L5) were performed after the bones were dissected, cleaned, fixed in Millonig's phosphate buffer (Leica Biosystems), and gradually dehydrated in 100% ethanol. Bones were scanned with a MicroCT40 (Scanco Medical) at medium resolution (12 μm isotropic voxel size) for quantitative determinations and integrated into 3-D voxel images (1024 x 1024 pixel matrices for each planar stack). For the latter, a Gaussian filter (sigma = 0.8, support = 1) was applied to all analyzed scans. Scanco Eval Program v.6.0 was used for measuring bone volume. Scan settings included X-ray tube potential (55 kVp), X-ray intensity (145 μA), and integration time (220 ms). The nomenclature used conforms to recommendations of the American Society for Bone and Mineral Research (60). At a threshold of 200 mg/cm 3 , cortical dimensions were determined using 18 to 23 slices at the femoral mid-diaphysis and using 50 slices between slices 300 and 350 at the distal metaphysis. Total and medullary area and circumference measurements were calculated from these slices. For cortical porosity measurements, slices were analyzed from a point immediately distal to the third trochanter to the primary spongiosa. After defining endosteal and periosteal boundaries, an additional image processing script ("peel-iter =2") was used to eliminate false voids caused by an imperfect wrap of the contours to the bone surface. Images were binarized with a threshold of 365 mg/cm 3 , and overall porosity was determined with the "cl_image" script to obtain bone volume and void volume. To avoid the inclusion of osteocyte lacunae and canalicular space, void volumes < 31,104 µm 3 (18 voxels) were excluded in the determination of porosity. All trabecular measurements were made by drawing contours every 10 to 20 slices and using voxel counting for bone volume per tissue volume (BV/TV) and sphere filling distance transformation indices, without pre-assumptions about the bone shape as a rod or plate. Vertebral cortical bone thickness was determined on the ventral cortical wall using contours of cross-sectional images drawn to exclude trabecular bone, as described for femoral cortical bone.
CTx, P1NP, and Osteocalcin ELISA. Circulating CTx, P1NP, and osteocalcin in serum was measured using a mouse RatLaps (CTx-I) EIA kit (Immunodiagnostic Systems), Rat/Mouse PINP EIA kit (Immunodiagnostic Systems), and an Osteocalcin enzyme immunoassay kit (Thermo Fisher), respectively, according to the manufacturer's directions. Blood was collected into 1.7 mL microcentrifuge tubes by retro-orbital bleeding. Blood was then kept on ice for 1 hour and centrifuged at 10,000 x rpm for 10 min to separate serum from cells.  For osteoblast cultures, total bone marrow cells were obtained as described above. Cells from 3 to 4 mice of each group were pooled and cultured with 20% FBS, 1% PSG, and 50 µg/mL ascorbic acid (Sigma-Aldrich) in 10-cm culture dishes for 7 days. Adherent bone marrow stromal cells were trypsinized and re-plated in 12-well tissue culture plates at 0.15-0.25 × 10 6 cells per well with 10% FBS, 1% PSG, 50 µg/mL of ascorbic acid, and 10 mM β-glycerophosphate (Sigma-Aldrich) for 3 days, to perform qPCR assays and western blotting, or for 21 days to assess mineralization. The mineralized matrix was stained with 40 mM alizarin red solution, following the manufacturer's instructions (Sigma-Aldrich).
Bone resorption assay. BMMs were isolated as described above and stimulated with RANKL to form osteoclasts on Osteo Assay Surface 24-well plates (Corning Life Sciences) coated with an inorganic bone biomaterial surface. Cells were removed using a 2% hypochlorite solution for 5 min, washed with distilled water, and dried at room temperature. For Von Kossa staining, wells were treated in darkness with 150 µL/well of 5% (w/v) aqueous silver nitrate solution for 20 min.
Plates were then washed for 5 min with distilled water and incubated in darkness with 150 µL/well of 5% (w/v) sodium carbonate in a 10% formalin solution. Wells were then washed twice with PBS, rinsed with distilled water, and dried in a 50 °C oven for 30 min. Three wells per group were assessed microscopically. In this assay, the resorbed areas appear white, and the unresorbed mineralized surface appears black.
Seahorse mitochondrial flux analysis. The BMMs were plated and treated with 30 ng/mL RANKL for 3 days with or without 10 nM LC-0296. The media in the wells was replaced with XF assay media and the plate was kept in a non-CO2 incubator for 20 minutes at 37 °C. After recording three total cellular respiration measurements with the XF96 analyzer, 10 µg/mL oligomycin was added to inhibit mitochondrial ATP synthase and measure the decrease in the oxygen-consumption rate that is linked to ATP turnover. To determine the maximal respiration potential of the cells, 10 µM FCCP (an uncoupler of oxidative phosphorilation) was used. The amount of non-mitochondrial oxygen consumption was determined by inhibiting the respiratory chain activity with an antimycin A and 10 µM rotenone cocktail. These data were used to calculate the mitochondrial basal respiration, ATP-linked respiration, reserve respiratory capacity, and proton leak as we previously described (22,62).
Quantitative RT-PCR. Total RNA was purified from cultured BMMs or bone marrow stromal cells using TRIzol reagent (ThermoFisher Scientific) according to the manufacturer's directions.
RNA was quantified using a Nanodrop instrument (ThermoFisher Scientific) and 2 µg (cultured cells) of RNA was then used to synthesize cDNA using a High-Capacity cDNA Reverse Transcription kit (Applied Biosystems) according to the manufacturer's instructions. Transcript abundance in the cDNA was measured by quantitative PCR using TaqMan Universal PCR Master Mix (ThermoFisher Scientific). The primers and probes for murine Sirt3 Statistics. All data were analyzed using GraphPad Prism 8 or 9 (GraphPad Software).
Statistically significant treatment effects were detected with a 2-way ANOVA after determining that the data were normally distributed and exhibited equivalent variances. In some cases, log or rank transformations were used to obtain normally distributed data and equal variance.