The Sirt1 Activators SRT2183 and SRT3025 Inhibit RANKL-Induced Osteoclastogenesis in Bone Marrow-Derived Macrophages and Down-Regulate Sirt3 in Sirt1 Null Cells

Increased osteoclast-mediated bone resorption is characteristic of osteoporosis, malignant bone disease and inflammatory arthritis. Targeted deletion of Sirtuin1 (Sirt1), a key player in aging and metabolism, in osteoclasts results in increased osteoclast-mediated bone resorption in vivo, making it a potential novel therapeutic target to block bone resorption. Sirt1 activating compounds (STACs) were generated and were investigated in animal disease models and in humans however their mechanism of action was a source of controversy. We studied the effect of SRT2183 and SRT3025 on osteoclastogenesis in bone-marrow derived macrophages (BMMs) in vitro, and discovered that these STACs inhibit RANKL-induced osteoclast differentiation, fusion and resorptive capacity without affecting osteoclast survival. SRT2183 and SRT3025 activated AMPK, increased Sirt1 expression and decreased RelA/p65 lysine310 acetylation, critical for NF-κB activation, and an established Sirt1 target. However, inhibition of osteoclastogenesis by these STACs was also observed in BMMs derived from sirt1 knock out (sirt1-/-) mice lacking the Sirt1 protein, in which neither AMPK nor RelA/p65 lysine 310 acetylation was affected, confirming that these effects require Sirt1, but suggesting that Sirt1 is not essential for inhibition of osteoclastogenesis by these STACs under these conditions. In sirt1 null osteoclasts treated with SRT2183 or SRT3025 Sirt3 was found to be down-regulated. Our findings suggest that SRT2183 and SRT3025 activate Sirt1 and inhibit RANKL-induced osteoclastogenesis in vitro however under conditions of Sirt1 deficiency can affect Sirt3. As aging is associated with reduced Sirt1 level and activity, the influence of STACs on Sirt3 needs to be investigated in vivo in animal and human disease models of aging and osteoporosis.


Introduction
Increased bone resorption by osteoclasts is characteristic of osteoporosis, inflammatory arthritis, hyperparathyroidism, malignant bone disease and other metabolic bone diseases. Currently available therapies to suppress osteoclast-mediated bone resorption include the bisphosphonates which induce osteoclast apoptosis and may result in suppression of bone formation and anti receptor activator of nuclear factor-κB ligand (RANKL)-antibody. These therapeutic agents are precluded from long term use due to side effects. Sirtuin 1 (Sirt1), a nicotinamide adenine dinucleotide (NAD + )-dependent lysine deacetylase, a key player in aging, inflammation and metabolism [1] regulates bone mass, and its targeted deficiency in osteoclasts results in increased bone resorption [2][3][4][5][6]. Enhancing Sirt1 activity is a plausible novel approach to inhibit bone resorption while concurrently ameliorating other age-related pathologies.
Resveratrol, the first Sirt1 activator to be studied, inhibits osteoclast generation and function [7], but this effect may be mediated via its cellular targets beyond Sirt1 such as estrogen receptor alpha, a key regulator of osteoclast generation [8] and influenced by resveratrol [9]. Synthetic Sirtuin 1 activating compounds (STACs), structurally different than resveratrol with a higher potency and bioavailability were generated, however their mechanism of action was a source of ongoing debate [10][11][12][13]. The controversy seemed to have been resolved by a study showing an allosteric activation of Sirt1 by STACs requiring hydrophobic motifs in the substrates and glutamic acid at position 230 of the Sirt1 N-terminal domain [14]. Different STACS were extensively tested in a wide spectrum of disease models in animals and over the past few years in humans in patients with type 2 diabetes mellitus and inflammatory conditions [15][16][17] Osteoclast-mediated bone resorption is a high energy demanding process [18] and sensors of cellular energy are likely to play a role in it. In this study we investigated the effects of second and third generations STACs [19] on osteoclast generation and function in vitro, and discovered that SRT2183 and SRT3025 inhibit RANKL-induced osteoclastogenesis in bone marrowderived macrophages (BMMs) by activating AMPK and deacetylating RelA/p65 lysine 310, critical for activation of the NF-κB signaling pathway. However, inhibition of osteoclastogenesis was also observed in SRT2183 and SRT3025-treated bone marrow macrophages derived from sirt1 knock-out mice in which neither AMPK nor RelA/p65 lysine 310 acetylation was affected but Sirt3 was down-regulated. Our findings suggest that these STACs inhibit osteoclastogenesis and can down-regulate Sirt3 under conditions of Sirt1 deficiency.

Animals
8-week-old female 129/Sv mice were used for this study. Inbred 129/Sv Sirt1 +/Δ mice [20] were a generous gift (see Acknowledgments), and were used for generating Sirt1 Δ/Δ (Sirt1 -/-) mice and the littermates of the parental WT strain. Genotyping of mice was performed at 4 weeks of age using ear genomic DNA as templates. All mice were maintained under specific pathogenfree conditions. Mice were housed in a constant temperature room with a 12-hour dark/ 12-hour light circle and were allowed free access to standard chow and water. Mice were sacrificed by isoflurane inhalation (Minrad INC, USA). All experiments were performed with the approval of the Animal Study Committee of the Hebrew University-Hadassah Medical School (MD-12-13154-3).

In vitro assays of osteoclast differentiation
Bone marrow-derived macrophages (BMMs) from femurs and tibias were collected, plated, and non-adherent cells were re-plated 24-hrs later in a 96-well plate at a concentration of 20,000 cells/well unless otherwise specified. The cells were cultured for 3 days in 5% CMG14-12 culture supernatant as a source of macrophage-colony stimulating factor (M-CSF) [21] in minimum essential medium α (α-MEM) containing 15% FBS. The plated cells were then induced to differentiation with 10% M-CSF and 10 ng/ml RANKL (PeproTech, Rocky Hill, New Jersey) for 4 days with a medium change every 3 days. Cells were TRAP-stained using a commercial kit (Sigma-Aldrich product 387-A, St. Louis, MO). Four non-overlapping images representing 80% of the area of each well were photographed with the Nikon DS Fi1 camera attached to Nikon Eclipse 80i microscope. Octeoclasts, defined as TRAP-positive multi-nucleated (!3 nuclei) cells, were manually counted.
Compounds SRT2183 ( Fig 1A) and SRT3025, kindly provided by Sirtris-GSK (see Acknowledgments), were dissolved in DMSO and were co-administrated with RANKL, unless otherwise specified. The compounds or the vehicle (0.01% DMSO) were added upon each medium exchange. All experiments were conducted with SRT2183 and some key experiments were repeated with SRT3025. Initial dose-response experiments with 0.5, 1, 2μM SRT2183 and 1, 2, 5μM SRT3025 were conducted based on the manufacturer recommendation, and TRAP staining suggested that the 2μM and 5μM concentrations are toxic for SRT2183 and SRT3025, respectively. All experiments were therefore conducted with 1μM SRT2183 and 2μM SRT3025. For the time course studies SRT2183, SRT3025 or a vehicle was added in the proliferation phase (co-administrated with M-CSF on day of plating and removed 3 days post plating), differentiation phase (coadministrated with RANKL on day 4 and removed on day 7), maturation phase only (day 7 post plating for 24 hrs) or differentiation and maturation phases (co-administrated with RANKL on day 4).

Pit formation assay
BMMs were harvested, plated and 24-hrs later non-adherent cells were re-plated on Osteo Assay Plate with an inorganic crystalline calcium phosphate coating (Corning, NY-cat no CI-3988) [22] at a density of 20,000 cells/well in α-MEM/15%FBS/5% M-CSF. On day 4 cells were induced to differentiation with 10% M-CSF/20ng/ml RANKL in the presence of SRT2183, SRT3025 or a vehicle for 7 days. Higher doses of RANKL compared to the differentiation assays were used for this experiment as cells were maintained for a longer period of time in culture. On day 11 adherent osteoclasts were removed using sodium hypochlorite solution (Sigma-Aldrich, St. Louis, MO) and the resorption area was determined by a Nikon ecplise 80i microscope coupled color camera Nikon DS (Digital sight)-Fi1 and IMAGE-PRO EXPRESS 4.0 software (Media Cybernetics, Silver Spring, MD). Resorption area was quantified by MATLAB Image Processing Toolbox (MATLAB R2013a) and is presented as percentage of well area.

Gene expression analyses
RNA was extracted from osteoclasts using peqGOLD TriFast (PeqLab, Erlangen, Germany) at the indicated time points, reverse transcribed into cDNA and analyzed with SYBR Greenbased quantitative Real-Time PCR in triplicates. mRNA expression level was normalized to Gapdh, β-actin or RNA polymerase II Polr2a. Gapdh is commonly used as a reference gene in osteoclast studies and indeed was stable in our experiments. To confirm that it does not affect the results when analyzing genes involved in energy metabolism we also used the mentioned above genes.

Proliferation assay
BMMs were harvested, plated and 24-hrs later non-adherent cells were re-plated at a density of 20,000 cells/well in α-MEM/15%FBS/5% M-CSF. For determining cell proliferation, cells were treated with SRT2183 or a vehicle for 72 hours post plating and BrdU reagent (Abcam, UK) was added 48hrs post SRT2183 or vehicle administration according to the manufacturer's instructions to determine cell proliferation 3 days post plating.

Cell viability assay
BMMs were harvested, plated and 24-hrs later non-adherent cells were re-plated at a density of 20,000 cells/well in α-MEM/15%FBS/5% M-CSF. For determining cell viability during the proliferation phase, cells were treated with SRT2183 or a vehicle for 72 hours post plating and cell-Titer-Blue reagent (Promega, Madison, Wis) was added according to the manufacturer's instructions to determine cell survival on day 4. For determining cell viability during the differentiation and maturation phases, cells were induced to differentiation on day 4 with 10% M-CSF/10ng/ml RANKL in the presence of SRT2183 or a vehicle, and cellTiter-Blue reagent was added to determine cell survival on day 8 post plating.

Apoptosis assay by Caspase 3 activity
BMMs were harvested, plated and 24-hrs later non-adherent cells were re-plated at a density of 20,000 cells/well in α-MEM/15%FBS/5% M-CSF. For apoptosis determination during the proliferation phase, cells were treated with SRT2183 or a vehicle for 72 hours and Caspase 3 activity within the cells was assessed by using the Apo-ONE Homogeneous Caspase 3/7 Assay Kit (Promega) on day 4.
For apoptosis determination during the differentiation and maturation phases, cells were induced to differentiation on day 4 with 10% M-CSF/10ng/ml RANKL in the presence of SRT2183 or a vehicle. Caspase 3 activity within the cells was assessed with the same kit on day 8. Experiments were carried out 3 times in parallel with cell viability assays.

Statistical analysis
Results are presented as Mean ± SEM. Data was analyzed by 2-way ANOVA, one or two sample Student's t-test as appropriate to compare treated versus untreated cells using graphPad Prism version 6 (software San Diego CA). Each experiment was repeated at least 3 times. P-values less than 0.05 were considered significant.

SRT2183 inhibits RANKL-induced osteoclast generation and resorptive capacity in bone marrow macrophages
The generation of multi-nucleated osteoclasts as determined by TRAP staining was significantly hampered when BMMs were induced to osteoclastogenesis in the presence of SRT2183 in a dose-dependent manner (Fig 1B). Total osteoclast number was not affected by SRT2183 administration, but the generation of large osteoclasts with a high nuclei number was markedly decreased ( Fig 1C). Importantly, SRT2183 dramatically inhibited osteoclast resorptive capacity as indicated by a marked reduction in the area eroded by treated cells (Fig 1D).
Time course experiments revealed that SRT2183 treatment at the proliferation phase did not affect osteoclast generation, however osteoclast differentiation and maturation were markedly reduced when SRT2183 was administered at the differentiation stages (Fig 1E a-d).
The proliferation of osteoclast precursors was not altered by SRT2183 treatment (Fig 1F). To understand if SRT2183 affects cell survival, viability and apoptosis studies were conducted at the proliferation and differentiation phases. The administration of SRT2183 at the proliferation and differentiation phases did not decrease cell viability or increased apoptosis (Fig 1G and  1H). Accordingly, total protein was unchanged in SRT2183 versus vehicle-treated cells (S1 Fig). These results suggest that SRT2183 inhibits osteoclast differentiation and function but not precursors' proliferation or cell survival. Consistently, nuclear factor of activated T-cell cytoplasmic 1 (NFATc1), a master transcription factor in osteoclast differentiation [23] as well as its downstream target, dendritic cell-specific transmembrane protein (DC-STAMP), necessary for osteoclast multi-nucleation [24] were reduced in SRT2183-treated cells (Fig 2A and  2B). Similarly, mRNA expression of osteoclast markers and specifically key osteoclast fusionrelated genes, Tm7sf4 (encoding for DC-STAMP) and the gene encoding for osteoclast stimulatory trans-membrane protein (OC-STAMP) were significantly decreased in SRT2183-treated osteoclasts (Fig 2C).

SRT2183 activates AMPK in osteoclasts
To gain insight into the mechanism by which SRT2183 inhibits RANKL-induced osteoclast differentiation and function major signaling pathways downstream of RANK were screened. Early phosphorylation of: c-jun N-terminal kinase (JNK), mitogen-activated protein kinase 14 (p38), ERK and p65 were not affected by SRT2183 administration (S2 Fig).
Previous work has shown that Sirt1 is closely coupled to AMP-activated protein kinase (AMPK) activity in a mutually enforcing mechanism [25]. Moreover, AMPK regulates osteoclast differentiation and function, and AMPKα1 deficiency in mice causes enhanced osteoclast differentiation and fusion [26]. We therefore investigated AMPK activation and discovered increased phosphorylation of AMPKα and its target acetyl CoA carboxylase (ACC) in SRT2183-treated cells, indicating AMPK stimulation (Fig 3A and 3B). Of note, increased Sirt1 level in SRT2183-treated cells was also observed (Fig 3C), and can result from AMPK activation, as AMPK was shown to positively regulate Sirt1 level [27].

SRT2183 regulates factors of the NF-κB signaling pathway in osteoclasts
Both Sirt1 and AMPK were shown to inhibit NF-κB signaling, a key pathway in RANKLinduced osteoclastogenesis [28]. Sirt1 represses NF-κB transcriptional activity by deacetylating RelA/p65 at lysine 310, critical for NF-κB activation [29]. AMPK was shown to inactivate the  NF-κB pathway via inhibition of IκB kinase and IκBα degradation. IκBα is an inhibitory subunit complexed to NF-κB /Rel proteins in the cytoplasm, preventing the release and movement of NF-κB into the nucleus [30]. Along these lines, decreased IκBα was reported in osteoclasts derived from AMPKα1 -/mice [26]. Indeed, IκBα was markedly increased and RelA/p65 K310 acetylation was significantly decreased in SRT2183-treated osteoclasts (Fig 3D and 3E). These results suggest that SRT2183 activates Sirt1 and AMPK in bone marrow-derived osteoclasts leading to inhibition of RANKL-induced NF-κB activation and NFATc1 expression.

SRT2183 inhibits osteoclastogenesis in sirt1 -/--derived bone marrow macrophages
To understand the role of Sirt1, the influence of SRT2183 on osteoclastogenesis was evaluated in bone marrow cells derived from sirt1 -/mice. Sirt1 Δ/Δ (Sirt1 -/-) mice lacking Sirt1 protein were generated from inbred 129/Sv Sirt1 +/Δ mice, whereas their littermates WT served as the controls. These KO mice are lacking sirt1 exons 5-7 resulting in no sirt1 protein production ( Fig 4A) and have a birth rate lower than 3% [20]. Strikingly, SRT2183 abolished the generation of large multi-nucleated osteoclasts and their resorptive capacity in sirt1 -/-BMMs similar to the effect observed in WT-derived osteoclasts, indicating that Sirt1 is not essential for inhibition of osteoclast generation and function under these conditions (Fig 4B and 4C). As expected, RelA/p65 K310 acetylation was not changed in SRT2183-treated sirt1 -/osteoclasts (Fig 4D), as this is a direct Sirt1 target. Furthermore, AMPKα phosphorylation was not affected by SRT2183 administration in sirt1 -/--derived osteoclasts (Fig 4E), suggesting that Sirt1 is upstream of and is required for AMPK activation by SRT2183 under these conditions. Consistently IκBα level was unchanged (Fig 4F).

SRT2183 down-regulates Sirt3 in sirt1 -/--derived bone marrow macrophages
To better understand the mechanism by which osteoclastogenesis is inhibited in SRT2183-treated sirt1 -/osteoclasts, we first asked if Sirtuins 2-7 protein level is changed in sirt1 -/--derived bone marrow macrophages or osteoclasts. No change was detected ( S3 Fig). Next, we asked if Sirt2-7 protein level is modulated by SRT2183 administration in WT or sirt1 KO cells. Strikingly, while no change was observed in treated WT osteoclasts (S4 Fig), Sirt3 protein but not mRNA expression was significantly reduced in sirt1 -/--treated osteoclasts ( Fig  4G and 4H). Moreover, decreased Sirt3 activity was detected in SRT2183-treated sirt1 null cells as indicated by increased acetylation of its target manganese superoxide dismutase (MnSOD, Sod2) [31] (Fig 4I). Finally, we asked if inhibition of osteoclastogenesis occurs also with a more advanced STAC, such as SRT3025, a third generation STAC [19]. Similar effects of SRT3025 on inhibition of osteoclastogenesis were observed in both WT (S5 Fig) and in sirt1 null cells (S6 Fig). Consistently, Sirt3 was reduced in sirt1 null cells treated with SRT3025. Thus, both SRT2183 and SRT3025 inhibited RANKLinduced osteoclastogenesis independently of Sirt1 and down-regulated Sirt3 in sirt1 null cells.

Discussion
This study demonstrates that Sirt1 activating compounds (STACs), SRT2183 and SRT3025, inhibit RANKL-induced osteoclast differentiation, multi-nucleation and resorptive capacity in bone marrow derived-macrophages in vitro without hampering cell survival. SRT2183 and SRT3025 inhibited RANKL-induced osteoclast differentiation by promoting deacetylation of RelA/p65 at lysine 310, a well recognized direct Sirt1 target [29] critical for NF-κB activation [32], and by activating AMPK. RelA/p65 was previously shown to promote osteoclast differentiation by blocking a RANKL-induced apoptotic JNK pathway, leading to enhanced OC differentiation [33]. Moreover, mice lacking RelA/p65 in the hematopoietic compartment have a deficient osteoclastogenic response to RANKL [33]. RelA/p65 contains seven lysine acetylation sites of which lysine 310 acetylation is required for NF-κB transcription activation and was shown to be deacetylated by Sirt1 [34]. In agreement with our results, targeted sirt1 deletion in osteoclasts leads to increased osteoclastogenesis and bone resorption accompanied by elevated osteoclast RelA/p65 lysine 310 acetylation [6]. Consistently, deacetylation of RelA/p65 K310 by STACs was previously described in U20S or HEK293 cells [35]. SRT2183 and SRT3025 activated AMPK via Sirt1 stimulation, as indicated by lack of effect on AMPK in sirt1 null cells. An intimate interplay between Sirt1 and AMPK was previously shown in a number of cell types. Sirt1 deacetylases and activates serine-threonine liver kinase B1 (LKB1), the primary AMPK kinase and activator [36]. On the other hand, AMPK increases Sirt1 expression and function via increasing its co-factor NAD + [27,37]. AMPK inhibits osteoclastogenesis by inhibiting the NF-κB pathway in part by preventing the degradation of IκBα a repressor of NF-κB which holds it quiescent in the cytoplasm [38]. The physiologic significance of this effect is illustrated by the phenotype in IκBα haplo-insufficient mice which display increased osteoclastogenesis [6].
Suppression of osteoclastogenesis by SRT2183 and SRT3025 occurred also in BMMs derived from sirt1 -/mice. These findings are in disagreement with some previously published studies which reported no effect of STACs in sirt1 deficient cells [14,39]. However, other previously published studies reported Sirt1-independent effects of STACs [10,40]. The discrepancy may be explained at least in part by the fact that in those studies claiming no off target effects only sirt1 exons 4-5, encoding for the enzyme catalytic domain, were deleted, allowing for STACs binding to inactive existing Sirt1 protein and precluding their binding to other targets, whereas in our study cells with complete deletion of sirt1 were investigated. It is also plausible that sirt1 specificity of these compounds is cell-dependent and osteoclasts were never studied before.
We discovered Sirt3 to be a target of SRT2183 and SRT3025 in sirt1 null cells resulting in its down-regulation, thus these STACs had an inhibitory action rather than being activators under these conditions. Supporting our findings, SRT1720 was previously shown to inhibit mouse and human Sirt3 by partially occupying the acetyl-lysine binding site, thus competing with the peptide substrate [41,42]. Furthermore, resveratrol, the first described Sirt1 activator, was also shown to inhibit Sirt3 [43]. The mechanism by which reduced Sirt3 in sirt1 -/osteoclasts leads to inhibition of osteoclastogenesis is not completely understood, however an increase in the acetylated inactive form of MnSod2 was found in SRT2183-treated sirt1 null cells,. The role of Sod2 in osteoclasts is unknown. Sod2was identified as a susceptibility gene for osteoporosis in humans. SNPs in the Sod2 gene were to be translated into changes in mRNA transcription and protein expression, and Sod2 protein expression is inversely associated with BMD in the Chinese population, suggesting that low Sod2 may be bone protective [44,45].
In summary, this study demonstrates that the STACs SRT2183 and SRT3025 inhibit osteoclast generation and function in vitro. These compounds did not cause osteoclast apoptosis and therefore are unlikely to impair the coupling between osteoclasts and osteoblasts. Whether these STACs or other STACs inhibit osteoclast-mediated bone resorption or influence other Sirtuins in vivo remains to be investigated in disease models of osteoporosis, aging and impaired metabolism as these conditions are associated with reduced Sirt1 level and function [46,47]. Importantly, our findings may have implications beyond osteoclast biology as they shed novel light on the STACs mechanism of action.