Anabolic and Antiresorptive Modulation of Bone Homeostasis by the Epigenetic Modulator Sulforaphane, a Naturally Occurring Isothiocyanate*

Bone degenerative pathologies like osteoporosis may be initiated by age-related shifts in anabolic and catabolic responses that control bone homeostasis. Here we show that sulforaphane (SFN), a naturally occurring isothiocyanate, promotes osteoblast differentiation by epigenetic mechanisms. SFN enhances active DNA demethylation via Tet1 and Tet2 and promotes preosteoblast differentiation by enhancing extracellular matrix mineralization and the expression of osteoblastic markers (Runx2, Col1a1, Bglap2, Sp7, Atf4, and Alpl). SFN decreases the expression of the osteoclast activator receptor activator of nuclear factor-κB ligand (RANKL) in osteocytes and mouse calvarial explants and preferentially induces apoptosis in preosteoclastic cells via up-regulation of the Tet1/Fas/Caspase 8 and Caspase 3/7 pathway. These mechanistic effects correlate with higher bone volume (∼20%) in both normal and ovariectomized mice treated with SFN for 5 weeks compared with untreated mice as determined by microcomputed tomography. This effect is due to a higher trabecular number in these mice. Importantly, no shifts in mineral density distribution are observed upon SFN treatment as measured by quantitative backscattered electron imaging. Our data indicate that the food-derived compound SFN epigenetically stimulates osteoblast activity and diminishes osteoclast bone resorption, shifting the balance of bone homeostasis and favoring bone acquisition and/or mitigation of bone resorption in vivo. Thus, SFN is a member of a new class of epigenetic compounds that could be considered for novel strategies to counteract osteoporosis.

Bone degenerative pathologies like osteoporosis may be initiated by age-related shifts in anabolic and catabolic responses that control bone homeostasis. Here we show that sulforaphane (SFN), a naturally occurring isothiocyanate, promotes osteoblast differentiation by epigenetic mechanisms. SFN enhances active DNA demethylation via Tet1 and Tet2 and promotes preosteoblast differentiation by enhancing extracellular matrix mineralization and the expression of osteoblastic markers (Runx2, Col1a1, Bglap2, Sp7, Atf4, and Alpl). SFN decreases the expression of the osteoclast activator receptor activator of nuclear factor-B ligand (RANKL) in osteocytes and mouse calvarial explants and preferentially induces apoptosis in preosteoclastic cells via up-regulation of the Tet1/Fas/Caspase 8 and Caspase 3/7 pathway. These mechanistic effects correlate with higher bone volume (ϳ20%) in both normal and ovariectomized mice treated with SFN for 5 weeks compared with untreated mice as determined by microcomputed tomography. This effect is due to a higher trabecular number in these mice. Importantly, no shifts in mineral density distribution are observed upon SFN treatment as measured by quantitative backscattered electron imaging. Our data indicate that the food-derived compound SFN epigenetically stimulates osteoblast activity and diminishes osteoclast bone resorption, shifting the balance of bone homeostasis and favoring bone acquisition and/or mitigation of bone resorption in vivo. Thus, SFN is a member of a new class of epigenetic compounds that could be considered for novel strategies to counteract osteoporosis.
Bone loss disorders like osteoporosis are often multifactorial diseases due to dysregulation of bone metabolism or homeostasis. To maintain bone density and integrity, complex networks and numerous interactions between different bone cell types and their environment assure that bone remodeling is balanced during growth and adulthood (1)(2)(3). These processes are orchestrated by many distinct regulators of gene expression like specific transcription factors, microRNAs, and epigenetic chromatin modulators (4,5). These gene regulatory processes are controlled by paracrine events involving a variety of secreted ligands that affect both bone-forming osteoblasts (e.g. parathyroid hormone, bone morphogenetic proteins, and winglesstype murine mammary tumor virus integration site family members (WNTs)), bone-resorbing osteoclasts (e.g. RANKL 4 / TNSF11), or both such as members of the serum amyloid A family (6,7). Both cell types are targeted by a number of different therapeutic approaches that counteract bone degeneration in aging patients and other disorders provoking bone loss.
Bisphosphonates, which are potent inhibitors of osteoclastic bone resorption, are widely prescribed and very effective at limiting bone loss. However, there is considerable debate about the efficacy versus the potential adverse effects of these compounds. In mice, the bisphosphonate ibandronate was reported to strongly reduce bone formation by decreasing osteoblast numbers and to disturb the bone marrow plasma cell niche, thus negatively affecting bone remodeling (8). Use of bisphos-phonates is also associated with a higher risk of atypical femoral fractures in women (9,10) with partly elevated bone mineralization patterns (11,12) and with significantly longer union times of distal radius fractures (13). Current therapeutic strategies to promote osteoblastogenesis in osteoporosis aim to increase osteoblast number and/or increase osteoblast activity by the administration of bone anabolic molecules like intermittent parathyroid hormone treatment or bone morphogenetic proteins (14). However, there are only a limited number of bone anabolic strategies that have shown promise in clinical applications.
Sulforaphane (SFN) is an organosulfur compound belonging to the isothiocyanate group. Its precursor molecule, glucoraphanin, naturally occurs at high concentrations in cruciferous vegetables like broccoli and cabbages. Highest concentrations are found in sprouts of broccoli and cauliflower (15). SFN is generated from glucoraphanin by the enzyme myrosinase upon damage to the plant (16) such as from chewing. Due to its antioxidative potential, its ability to selectively induce activating phase II enzymes (17), and inhibit histone deacetylase activity (18 -20), SFN is predominantly studied for its anticarcinogenic and antimicrobial properties. Furthermore, recent studies suggest that SFN has potential beneficial effects for the treatment of diabetes type 1 and type 2 (21)(22)(23) as well as osteoarthritis (24,25) and rheumatoid arthritis. In the latter context, SFN was found to inhibit synovial hyperplasia and T-cell activation (26). In addition, SFN represses matrix-degrading proteases and protects cartilage from destruction in vitro and in vivo (27). SFN also may have effects on bone resorption because it inhibits the RANKL-dependent differentiation of osteoclasts in vitro by suppressing nuclear factor-B (28) and activation of the transcription factor NRF2 (Nfe2l2) (29).
SFN shares molecular properties with the cryopreservant dimethyl sulfoxide (DMSO). As we and others have previously shown, DMSO causes phenotypic changes and supports differentiation of several cell types such as erythroid and embryonic stem cells as well as osteoblasts by epigenetic mechanisms (30 -36). Epigenetic reprogramming of osteoblastic differentiation or of other bone-related cells may represent a novel approach to effectively regulate bone remodeling and homeostasis in certain disorders. Indeed, an increasing number of studies emphasize the central role of epigenetic regulatory mechanisms in bone biology (4). Based on the molecular similarities of SFN and DMSO as well as their bone-related biological properties, we hypothesized that SFN would exhibit bone anabolic effects. Our results show that SFN has beneficial effects on bone and may act through an epigenetic mechanism that promotes teneleven translocation 1 (Tet1)/Tet2-dependent hydroxymethylation of DNA, reactivating gene expression.

Experimental Procedures
Cell Culture-The following murine cell lines were used: MC3T3-E1, a clonal preosteoblastic cell line derived from newborn mouse calvaria (kindly provided by Dr. Kumegawa, Department of Oral Anatomy, Meikai University, Sakado, Japan); the osteocyte-like MLO-Y4 cell line (kindly provided by Lynda Bonewald, University of Missouri-Kansas City); and the preosteoclastic, macrophage-like RAW 264.7 cell line (ATCC, Manassas, VA).
Differentiation of MC3T3-E1 cells was induced with 50 g/ml ascorbic acid (Sigma-Aldrich) and 5 mM ␤-glycerophosphate (Sigma-Aldrich) in medium containing 5% FBS. RAW 264.7 cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Invitrogen) supplemented with 10% fetal bovine serum (Hyclone) and 2 mM glutamine (Gibco). Cells were seeded in culture dishes at a density of 20,000 cells/cm 2 and cultured overnight unless stated otherwise. Before cells were treated with compounds, the culture medium was changed.
Primary Cells and Tissues-Primary mouse tissues or cells were obtained according to procedures that conform to the regulatory guidelines of the Institutional Animal Care and Use Committee of the Medical University of Vienna. Primary mouse osteoclasts were harvested from 8-week-old C57BL/6 mice that were sacrificed, and bone marrow cells were isolated from tibiae and femora under aseptic conditions and cultured in ␣-MEM (Biochrom) containing 10% FCS (Biochrom), 10 mg/ml gentamycin (Sigma-Aldrich), and macrophage colonystimulating factor 1 (30 ng/ml). After 24 h, non-adherent cells were seeded onto sterile, 300-m-thick dentin slices (elephant ivory) at 700,000 cells/cm 2 in ␣-MEM supplemented with 10% FBS, 2 mM L-glutamine, 30 g/ml gentamycin, 20 ng/ml macrophage colony-stimulating factor, and 2 ng/ml RANKL (R&D Systems, Minneapolis, MN). Culture medium was changed twice per week, and cells were cultured for 14 days.
Mouse bone marrow mesenchymal stem cells were isolated from aseptically dissected long bones of 6-week-old C57BL/6 mice. The marrow cavities were flushed with sterile medium using a 25-gauge needle, and the culture was established in ␣-MEM supplemented with 10% FCS and 10 mg/ml gentamycin. After 48 h of culture at 5% CO 2 and 37°C, the non-adherent cells were removed by gentle rinsing with PBS. At confluence, cells were harvested and seeded at a density of 3,000/cm 2 for cell experiments. For osteogenic differentiation, cells were cultured in ␣-MEM supplemented with 10% FBS, 10 mg/ml gentamycin, 100 nM dexamethasone, 10 mM ␤-glycerophosphate, and 50 g/ml ascorbic acid with or without L-, D-, or DL-SFN. Culture medium was renewed twice a week, and cells were cultured for the periods indicated under "Results." Calvariae from 2-3-day-old and from 7-week-old C57BL/6 mice were dissected aseptically. The calvarial bone explants were cultured in 48-well plates in ␣-MEM (Biochrom) contain-ing 10% FCS (Biochrom), 50 g/ml ascorbic acid (Sigma-Aldrich), 5 mM ␤-glycerophosphate (Sigma-Aldrich), and 10 g/ml gentamicin (Sigma-Aldrich). The day after dissection, medium was changed, and a part of the explants was treated with 3 M L-or DL-SFN for 12 days. Thereafter, one part of the calvariae was fixed for 1 h in 4% paraformaldehyde, and mineralization was measured by Alizarin Red S stain (Sigma-Aldrich). For this purpose, Alizarin Red S dye was extracted using 10% cetylpyridinium chloride in 10 mM sodium phosphate, pH 7.0, for 45 min at room temperature. Alizarin Red S absorbance was measured at 562 nm in a multiplate reader (Tecan, Maennedorf, Switzerland) and normalized to total protein amount measured by bicinchoninic acid assay (Thermo Fisher Scientific, Waltham, MA). From the other part of the calvariae, total RNA was extracted (see below).
Histomorphometric Measurements/Assessment of Osteoclastic Resorption-After culturing the primary mouse osteoclasts for 14 days on dentin slices in the presence or absence of 3 M Lor DL-SFN, substrates were put into water, sonicated for 10 min to remove living cells, and air-dried. Photographs were obtained by reflected light microscopy (objective, 20ϫ) of the entire substrate surface, and resorption trails and pits were analyzed and quantified with standard image analysis software (ImageJ, rsbweb.nih.gov/ij/). Resorption is shown as percent area resorbed to total area of the dentin slice.
Cell Metabolic Activity-To assess cell metabolic activity, a commercially available 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide-like assay (EZ4U, Biomedica, Vienna, Austria) was used. For this purpose, the cell lines were incubated with increasing concentrations of L-or DL-SFN. After a comparable doubling time for all three cell lines, the assay was performed following the protocol of the supplier.
Cell Counts-Cell lines were seeded in 24-well culture dishes at a density of 20,000/cm 2 and either left untreated (controls) or treated with L-, D-, or DL-SFN at the indicated concentrations for up to 48 h. Thereafter, cells were detached with 0.001% Pronase E, and the number of viable cells was assessed with a Casy cell counter (Schaerfe Systems, Reutlingen, Germany). Each experiment was performed in biological quadruplicates, and experiments were repeated twice. For long term experiments, cell number was determined using DNA amount as surrogate. Cell layers were washed with PBS and fixed for 20 min with 4% paraformaldehyde. Thereafter, Hoechst 33258 dye (1 g/ml; Polysciences, Warrington, PA) was added, and after an incubation of 15 min at room temperature, the fluorescence was measured (excitation, 360 nm; emission, 465 nm). The amount of DNA was estimated using a standard curve prepared from calf thymus DNA (Roche Applied Science).
Measurement of Caspase Activity-Caspase 3/7 and Caspase 8 activities were measured using the Caspase-Glo 3/7 and Caspase-Glo 8 assay kits (Promega, Madison, WI) following the manufacturer's instructions. Briefly, after treatments, cells were lysed, and substrate cleavage by Caspases was measured by the generated luminescence signal with a 96-multiwell luminometer (Glomax, Promega). Each experiment was performed in quintuplicate, and experiments were carried out twice.
Tet1 and Tet2 siRNA Transfections-For Tet1 and Tet2 depletion by siRNA, cells were seeded at 20,000 cells/cm 2 in 6-well culture plates. Six hours after seeding, cells were transfected with 40 pmol of Tet1 or Tet2 siRNA (Sigma-Aldrich) using X-tremeGENE siRNA Transfection Reagent (Roche Applied Science) as described by the supplier. One day after transfection, a medium change was performed, and one part of the cells was treated with medium containing 3 M DL-SFN, whereas the other part was left untreated. After an incubation time of 24 h, proteins or nucleic acids were isolated as described below and subjected to reverse transcription-quantitative polymerase chain reaction (RT-qPCR), Western blotting, or cell count determination.
Isolation of RNA and RT-qPCR-Total RNA was extracted using the SV Total RNA Isolation kit (Promega) following the supplier's instructions. cDNA was synthesized from 0.5 g of RNA using the First Strand cDNA Synthesis kit (Roche Applied Science) as described by the supplier. The resulting cDNAs were subjected to quantitative PCR amplification with a real time cycler using the QuantiTect SYBR Green PCR kit (Qiagen, Hilden, Germany) for the genes Alpl, Fas, Lox, Tet1, and Tet2 and TaqMan Gene Expression Master Mix (Applied Biosystems, Foster City, CA) for measuring Runx2, Bglap2, Col1a1, Tnfsf11, and 18S rRNA expression (for primers, see Table 2). SYBR Green RT-qPCR was started with an initial denaturation step at 95°C for 10 min and then continued with 45 cycles consisting of 30-s denaturation at 95°C, 30-s annealing at primerspecific temperatures, and extension at 72°C. For measurement of the TaqMan assays, we applied an initial denaturation at 95°C for 10 min followed by 45 cycles alternating 60 s at 60°C and 15 s at 95°C (primers or TaqMan probe references are listed in Table 1). All RT-qPCR assays were performed in triplicates, and expression was evaluated using the comparative quantification method (37).
Protein Isolation and Immunoblotting-Whole cell protein extracts were prepared using SDS sample buffer (2% SDS, 100 mM ␤-mercaptoethanol, and 125 mM Tris-HCl, pH 6.8) and heated at 95°C for 5 min.
For immunoblotting analysis, 15 g of protein extracts were separated on 10% SDS polyacrylamide gels and transferred to nitrocellulose membranes (Millipore), and equal protein loading was confirmed by Ponceau Red staining. Subsequently, membranes were blocked overnight with 10% blocking reagent (Roche Applied Science) in TN buffer (50 mM Tris and 125 mM NaCl, pH 8.0). The following primary antibodies were used: rabbit anti-Runx2 (sc-10758, Santa Cruz Biotechnology, Santa Cruz, CA) rabbit anti-Tet1 (catalog number 61741, Active Motif, Carlsbad, CA), and rabbit anti-Tet2 (sc-136926, Santa Cruz Biotechnology). Washing was performed with TN buffer containing 0.01% Tween. Binding of the secondary antibody (anti-rabbit IgG/anti-mouse IgG horseradish peroxidase-coupled) (Roche Applied Science) diluted 1:10,000 in 10% blocking solution followed by detection with the BM chemiluminescence immunoblotting kit (Roche Applied Science) was carried out as described by the supplier. Chemiluminescence was measured with an image acquisition system (Vilber Lourmat, Marne-la-Vallée, France). Measurements are given as means of three immunoblots normalized to total protein loading, and representative blots are shown.
Immunostaining and Quantifications of Global Cytosine 5-Hydroxymethylation-Cells were fixed with 4% paraformaldehyde in PBS for 20 min at room temperature and permeabilized with 0.2% Triton in 4% paraformaldehyde in PBS. Thereafter cell layers were blocked for 20 min with 10% blocking reagent and incubated for a further 1 h with 0.5 g/ml anti-5hydroxymethylcytosine (5hmC) antibody (Abgent). Afterward, the cells were washed three times with PBS and incubated for a further 1 h with an Alexa Fluor 488-labeled secondary antibody (Invitrogen) diluted 1:300 in blocking buffer. Finally, nuclei were stained with Hoechst 33258 dye.
For immunostaining, cell slides were mounted with Vectashield mounting medium, and immunofluorescence was visualized on a Leica laser-scanning microscope. For better visualization of the small and faint 5hmC spots in the nuclei, gain for nuclear visualization (Hoechst; blue) was strongly reduced to facilitate 5hmC visualization in euchromatin (black spots in low gain representation) and gain for 5hmC signals (Alexa Fluor 488; green) was strongly enhanced. Still, under this conditions, no background signal and no signal was found by using only the Alexa Fluor 488-labeled secondary antibody as a negative control (Invitrogen). For quantification of global cytosine 5hmC, the fluorescence in the plates was measured with a multiwell plate reader (Tecan). The amount of DNA (nuclei signal) was estimated using a standard curve prepared from calf thymus DNA. The signal of the fluorescence of the 5hmC staining was normalized to the amount of DNA. Also in this case, no signals were found when only the Alexa Fluor 488-labeled secondary antibody (Invitrogen) was used.
Specific Promoter Hydroxymethylation Analysis-To analyze the 5hmC content in the Atf4 proximal promoter/first untranslated exon, appropriate fragments of the targeted promoter regions were generated by digestion of 3 g of genomic DNA from cells cultured for 16 h or for 1 day with 3 M DL-SFN with 30 units of BstBI and XmaI for 1 h at 37°C. Subsequently, the enzyme was heat-inactivated at 65°C for 20 min, and DNA was cleaned up with the QIAquick DNA purification kit (Qiagen). Digestion of non-5mC DNA with the mentioned enzymes results in two fragments that overspan a CpG-rich region in the proximal promoter/first untranslated exon of the Atf4 gene (see Fig. 9D; GenBank TM accession number CH466550). After DNA purification, hydroxymethylated DNA fragments were extracted by immunoprecipitation as described previously (30). In brief, hydroxymethylated DNA of three independent biological experiments was captured by a specific hydroxymethylcytosine antibody (Active Motif catalog number 39769) coupled to magnetic beads to separate the 5hmC-containing DNA fragments. DNA was eluted from the magnetic beads with 75 l of 2 M NaCl solution. Finally, the mean methylation or hydroxymethylation status of the fragments was determined by amplifying the fragments by quantitative real time PCR. Amplification ratios of the bound (hydroxymethylated) DNA fraction to unbound (non-hydroxymethylated) DNA fraction were calculated (for primers, see Table 1). For assessment of the specificity of the 5hmC antibody, either 40 ng of non-methylated, fully methylated, or fully hydroxymethylated DNA standards (Zymo Research) were run in parallel and amplified by real time PCR using the supplied PCR Primers.
In Vivo Analysis of SFN Effects on Bone-Analysis of SFN in mice was based on a protocol published previously by Kong et al. (26). In our experimental setting, 8-week-old sham-operated or ovariectomized (OVX) C57BL6/J female mice were purchased from Charles River Laboratories. Mice were injected with 7.5 mM DL-SFN (200 l of SFN at a concentration of 63.8 mg/ml/kg dissolved in ethanol and further diluted in PBS) intraperitoneally every other day for 5 weeks. Control mice received vehicle alone. As described by Kong et al. (26), the injection of this dose of DL-SFN does not show apparent adverse effects, including weight loss, alterations in physical appearance, or changes in behavior in the treated mice.
Microcomputed Tomography (CT) Analysis-Images from proximal tibiae fixed in 4% formaldehyde were acquired in a SkyScan 1174 with a resolution of 6 m (x-ray voltage, 50 kV). Image reconstruction was performed by applying a modified Feldkamp algorithm using the SkyScan NRecon software. Threedimensional and two-dimensional morphometric parameters

T m
GenBank accession no. Forward Reverse°C

SYBR Green
Alpl Mm03413826_mH; Applied Biosystems 60 Col1a1 Mm00801666_g1; Applied Biosystems 60 Lox Mm00495386_m1; Applied Biosystems 60 Runx2 Mm00501578_m1; Applied Biosystems 60 Tnfsf11 Mm00441906_m1; Applied Biosystems 60 were calculated for the trabecular bone (350 consecutive slides; 6 m thick) starting from 300 m from the growth plate. Threshold values were applied for segmenting trabecular bone corresponding to bone mineral density values of 0.6/cm 3 calcium hydroxyapatite. Three-dimensional parameters were based on analysis of a Marching Cubes type model with a rendered surface. Calculation all of two-dimensional areas and perimeters was based on the Pratt algorithm (38). Bone structural variables and nomenclature were those suggested in Bouxsein et al. (39).
Quantitative Backscattered Electron Imaging (qBEI)-Undecalcified distal femora from 13-week-old mice (sham/Ctrl (n ϭ 5), sham/DL-SFN (n ϭ 5), OVX/Ctrl (n ϭ 4), and OVX/DL-SFN (n ϭ 5)) were fixed and dehydrated in alcohol, embedded in polymethylmethacrylate, and prepared for backscattered electron imaging as described previously. Briefly, qBEI is based on backscattering of electrons from the surface layer (i.e. the initial ϳ1.5 m) of a bone section. The rate of these backscattered electrons is proportional to the weight concentration of mineral (hydroxyapatite) and thus of that of calcium in bone. Details of the method have been published elsewhere (11,40,41). A scanning electron microscope (DSM 962, Zeiss, Oberkochen, Germany) equipped with a four-quadrant semiconductor backscattered electron detector was used. The accelerating voltage of the electron beam was adjusted to 20 kV, the probe current was set to 110 pA, and the working distance was 15 mm. The cancellous and cortical bone areas were imaged at 200ϫ nominal magnification (corresponding to a pixel resolution of 1 m/pixel). The backscattered electron signal (grayscale) was calibrated using the "atomic number contrast" between carbon (Z ϭ 6) and aluminum (Z ϭ 13) as reference materials. From the calibrated digital images, frequency distribution of mineral concentration, the so-called bone mineralization density distribution (BMDD), was derived. The BMDD was characterized by five parameters: weighted mean calcium content (CaMean), mode calcium content (CaPeak; peak position), full width at half-maximum of the BMDD peak (CaWidth; reflecting the heterogeneity in matrix mineralization), the percentage of low mineralized bone area (CaLow; below 17.68 weight % calcium), the percentage of highly mineralized bone area (CaHigh; above the 95th percentile value of the corresponding control animal BMDD).
Statistical Analysis-Statistical analyses were performed using analysis of variance or Student's t test in Prism 4.03 (GraphPad Software, La Jolla, CA). Values of p Յ 0.05 were considered significant. Each experiment consisted of at least three biological replicates. For RT-qPCR data, results from technical triplicates were averaged, and the mean value was treated as a single, biological statistical unit. Results are presented as means Ϯ S.D.

SFN Affects
Cell Viability of Bone-related Cells-Our group and others have shown previously that the widely used cryopreservant DMSO triggers differentiation of MC3T3-E1 osteoblasts. Although use of DMSO is restricted in medical applications, there are intriguing analogies in the chemical structures of DMSO and plant secondary metabolites. As deduced from the IUPAC nomenclature, DMSO (methanesulfinylmethane) and the natural food compound SFN (1-isothiocyanato-4methylsulfonylbutane) share striking molecular similarities. SFN carries an additional pentane group with terminal isothiocyanate (Fig. 1A). These structural similarities predict similarities in the biological effects of DMSO and SFN. Therefore, we tested the effects of DMSO and SFN on MC3T3-E1 differentiation by assessing the extent of extracellular matrix (ECM) mineralization. Under our conditions, staining for the late osteoblast biomarker Alizarin Red is not observed in our cultures for the first 2 weeks of culture but typically is evident after 3-4 weeks (42). Our results show that both DMSO and SFN have similar biological effects as revealed by increased matrix mineralization as early as 14 days of treatment (Fig. 1B). Strikingly, SFN is at least 4 orders of magnitude more potent than DMSO in stimulating matrix mineralization, suggesting that the methanesulfonyl group of DMSO is important but not sufficient for full biological activity.
Because SFN has cell growth-suppressive effects (43)(44)(45), we evaluated the impact of SFN on cell proliferation and viability of bone-related immortalized cell lines. MC3T3-E1 osteoblasts and MLO-Y4 osteocytes were treated with increasing concentrations of SFN for up to 48 h (Fig. 2). Because SFN is a chiral molecule, we tested three different preparations: L-, D-, and DL-SFN, which is a mixture of both D and L enantiomers. Both L-SFN and DL-SFN have modest effects (Ͻ50%) on cell number in MC3T3-E1 and in MLO-Y4 after 24 h of treatment at a concentration of 3 M (Fig. 2, A and B). At higher concentrations, both the L enantiomer and the DL mixture compromise cell viability with the greatest decrease in cell number observed for MLO-Y4 cells (Fig. 2B). Furthermore, DL-SFN has a somewhat To assess whether SFN affects cellular metabolic activity after 24 h of treatment, we performed an EZ4U assay, a 3-(4, 5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide-like assay that measures the capability of living cells to reduce tetrazolium salts in the mitochondria into formazan derivatives that absorb at 450 nm. Titration curves for both L-SFN and DL-SFN using either cell line revealed that the half-maximal effective concentration (EC 50 ) of L-SFN is ϳ48 M and of DL-SFN is ϳ13 M in MC3T3-E1 cells (Fig. 2E), whereas lower doses are needed in MLO-Y4 cells (i.e. ϳ11 M for L-SFN and ϳ6 M for DL-SFN) (Fig. 2F). Hence, SFN has fewer metabolic effects on MC3T3-E1 osteoblasts than on MLO-Y4 osteocytes.
Remarkably, the preparation with both enantiomers (L-and DL-SFN) shows stronger biological effects compared with the sole L enantiomer, suggesting that the biological potency of SFN depends on steric considerations. Indeed, direct comparison of the effects of L-, DL-, and D-SFN enantiomers on cell proliferation in MC3T3-E1 cells shows that the D isoform most strongly affects cell proliferation after 24 (Fig. 2G) or 48 h (Fig. 2H) at the most concentrations tested. Taken together, the results show that 3 M SFN is a relatively non-toxic dose (Fig. 2) that is biologically effective (see Fig. 1).
SFN Induces Extrinsic Apoptosis in Cells of the Osteoblastic Lineage-As we have shown previously, DMSO induces the extrinsic pathway of apoptosis (30), suggesting that the growthsuppressive action of SFN (as reflected by a decrease in cell proliferation and metabolic activity; see Fig. 1) may be caused by programmed cell death in both MC3T3-E1 osteoblasts and MLO-Y4 osteocytes. To test this possibility, we monitored expression levels and activity of apoptotic biomarkers. We found that Fas mRNA expression is not affected by either L-or DL-SFN in MC3T3-E1 and MLO-Y4 cells after 8 h of treatment (Fig. 3A). However, after 16 h of treatment, Fas mRNA expression is significantly increased in both cell types by DL-SFN treatment (Fig. 3B). Corroborating these results, the activity of Caspase 8 is significantly increased by L-SFN and DL-SFN after 24 h (Fig. 3C). Furthermore, we observed a significant increase in the activities of Caspase 3/7 upon DL-SFN treatment of MC3T3-E1 cells and MLO-Y4 cells for 24 h (Fig. 3D). These data indicate that SFN induces the extrinsic apoptotic pathway in osteogenic cells.
Osteogenic Long Term Effects of SFN-Beyond effects on cell growth and survival, SNF may affect osteoblast differentiation and activity. Therefore, we investigated the long term effects (Ͼ20 days) of L-SFN and DL-SFN on mineralization of osteoblasts, bone marrow stromal/stem cell (BMSC) cultures, and neonatal calvarial explants maintained in osteogenic medium. Treatment of MC3T3-E1 cells or BMSCs (Fig. 4A) significantly increases ECM mineralization, the principal marker for mature osteoblast function. Because L-SFN and DL-SFN have mild effects on cell survival (at 3 M) in non-confluent MC3T3-E1 osteoblasts (see Fig. 2, A and C), we assessed cell number of both MC3T3-E1 cells and BMSCs in response to both enantiomer preparations. As anticipated, MC3T3-E1 and BMSC cultures treated with either SFN type exhibit modest lower cell numbers but higher mineral content per DNA unit after long term treatment, indicating that SFN-treated cells are more active in mineralization. To evaluate whether this bone anabolic effect occurs in explants that are rich in postproliferative mature osteoblasts, we measured ECM mineralization of newborn calvarial explants treated with 3 M L-or DL-SFN for 14 days. As in the two-dimensional cell culture models, L-and DL-SFN each significantly increase ECM mineralization in calvarial explants (Fig. 4A). These data indicate that SFN is capable of biologically stimulating mineralization of neonatal bone tissue in culture.   in poor RNA yields for RT-qPCR analysis (Fig. 4B). Furthermore, we were interested in the effects of the different SFN enantiomers on the expression of these genes at day 14 and at day 28 in MC3T3-E1 cells. Therefore, we directly compared the three SFN enantiomer preparations, D-, DL-, and L-SFN. At day 14, Alpl expression is significantly up-regulated by DL-and L-SFN preparations (Fig. 4C), and the racemic DL-SFN mixture significantly increases Lox mRNA expression at this time point (Fig. 4F). Bglap2 (Fig. 4D) and Col1a1 (Fig. 4E) are significantly up-regulated by D-SFN as well as by DL-SFN at day 14. Furthermore, at day 28, Bglap2 is the only gene that is still significantly up-regulated by D-SFN (Fig. 4D). Taken together, our results indicate that SFN promotes the expression of osteoblastic genes and mineralization.

The Bone Anabolic Effects of SFN Are Reflected by Enhanced Expression of Primary Osteoblastic Transcription Factors-SFN
may play a role in cellular differentiation by selectively modulating mRNA expression of osteogenic transcription factors. Therefore, we examined whether SFN affects the mRNA and protein expression of the bone-related master regulator Runx2 in mouse MC3T3-E1 cells or BMSCs. Treatment of cells with DL-SFN stimulates Runx2 mRNA expression in both cell types (Fig. 5A). Immunoblotting analysis of DL-SFN-treated MC3T3-E1 cells shows a significant increase in Runx2 protein expres-  In mouse calvarial explants from neonatal (5 days (d)) and from adult mice (7 weeks (we)), 3 M DL-SFN decreases Tnfsf11 expression after 12 days of treatment (B). Tnfsf11 gene expression is referred to 18S rRNA expression; n ϭ 3. In all graphs Ctrl is set to 1, and treatments are referred to as -fold change to Ctrl. Values are represented as the mean Ϯ S.D.; error bars represent S.D. *, p Յ 0.05. sion after both 24 h (Fig. 5B) and 14 days (Fig. 5C) of treatment, thus validating the pro-osteogenic effects of SFN (see Fig. 4). Representative immunoblots of Runx2 protein expression are shown in Fig. 5D.
After 14 days of treatment, similar to the ECM-related genes (see Fig. 4) as well as for the osteoblastic transcription factors Runx2, Atf4, and Sp7/Osx, 3 M DL-SFN shows the strongest positive effect in MC3T3-E1 cells (Fig. 5E). In regard to the potency of the different SFN isomers, comparison of the three SFN preparations shows that for Runx2 (Fig. 5F), Atf4 (Fig. 5G), and Sp7/Osx (Fig. 5H) the strongest effects are induced by the D-or DL-SFN mixture at day 14. At day 28, significant up-regulations are still observed for Atf4 and Sp7/Osx genes by the D-or L-SFN preparations. In contrast to DMSO (30), SFN does not increase the expression of the osteoblastic transcription factor Dlx5 at the times measured (Fig. 5I).
SFN Attenuates RANKL/Tnfsf11 Expression in Mouse Calvarial Explants-Runx2 expressed in osteoblastic cells mediates biological feedback by promoting osteoclast differentiation via up-regulation of Tnfsf11 (RANKL) (46). Tnfsf11 promotes osteoclast formation and survival by binding to the osteoclastic receptor Tnfrsf11a (alias RANK), which in turn activates nuclear factor-B (47). Furthermore, osteocyte-derived Tnfsf11 controls bone remodeling during postnatal development and in adult mammals. In osteocyte-like MLO-Y4 cells, DL-SFN reduces Tnfsf11 mRNA expression at both 3 and 8 days after treatment, although changes in expression are only significant at the latter time point (Fig. 6A). More importantly, SFN decreases Tnfsf11 expression in mouse calvarial explants from neonatal mice and to a greater extent in explants from adult mice after 12 days of treatment with DL-SFN ex vivo (Fig. 6B). These results suggest that SFN may potentially promote net bone accumulation by suppressing RANKL/Tnfsf11-induced osteoclast formation.
SFN Induces Global DNA Hydroxymethylation Changes-As we have shown previously, DMSO dramatically induces active DNA demethylation in preosteoblastic MC3T3-E1 cells within less than 1 day of treatment (Ͻ16 h) through formation of 5hmC (30). Due to the structural and biological similarities between DMSO and SFN, we analyzed whether SFN alters the level of active DNA demethylation in cells of the osteogenic linage. Indeed, we measured a significant global increase in 5hmC as a marker for ongoing active DNA demethylation upon treatment with either L-or DL-SFN in MC3T3-E1 cells at 16 h after treatment based on immunofluorescence laser-scanning microscopy (Fig. 7A) and spectrophotometry (Fig. 7B). As expected, 5hmC spots are observed in the looser euchromatic regions (dark nuclear parts in the low gain picture version in Fig. 7A), demonstrating opening of chromatin by active DNA demethylation after SFN treatment. Interestingly, in MLO-Y4 osteocytes that are differentiated beyond the osteoblastic stage, both L-and DL-SFN do not elevate 5hmC levels (Fig. 7B). These findings collectively suggest that SFN selectively induces active DNA demethylation in osteoblastic cells that are at early rather than at late stages of osteogenic differentiation.
SFN Transiently Enhances Expression of Tet1 in Osteoblastic Cells-Global changes in 5hmC levels are mediated by enzymes that control DNA hydroxymethylation and are encoded by Tet genes. All three Tet genes (Tet1-3) are expressed in MC3T3-E1 cells, and expression of Tet1 and Tet2 decreases whereas expression of Tet3 increases within the first 8 days of MC3T3-E1 osteoblast differentiation (Fig. 8A). However, the levels of the Tet genes are distinctly regulated during early stages in MLO-Y4 cells. Tet1 is much more rapidly down-regulated in MLO-Y4 cells compared with MC3T3-E1 cells, whereas Tet2 and Tet3 levels are steadily up-regulated (Fig. 8B). We next examined mRNA expression of Tet genes in the absence or presence of DL-SFN in both MC3T3-E1 and MLO-Y4 cells. Beyond the modulations in Tet1 and Tet2 levels in response to biological induction of differentiation, we found that expression of Tet1 and Tet2 increases to a limited degree (between 20 and 40%) by DL-SFN at 8 and 16 h after treatment in MC3T3-E1 cells; albeit these values reach statistical significance only for Tet1 at the 16-h time point (Fig. 8C). DL-SFN does not positively change mRNA expression of Tet1 or Tet2 in MLO-Y4 cells (Fig.  8D) or Tet3 in both cell lines (Fig. 8, C and D). We conclude that mRNA levels of the hydroxylation enzymes Tet1 and Tet2 are modestly regulated by SFN during bone cell differentiation and may perhaps contribute to DL-SFN-induced changes in global hydroxymethylation.
Effects of SFN-induced Active DNA Demethylation in MC3T3-E1 Osteoblasts-The results presented here are consistent with the idea that Tet-dependent hydroxymethylation may compromise cell survival as shown previously by our group (30). To test this hypothesis, we treated MC3T3-E1 cells with non-silencing RNA (negative control) or siRNA against either Tet1 or Tet2 The early effect of SFN on osteoblastic gene expression (e.g. Runx2; Fig. 5B) suggests that SFN, similarly to DMSO (30), promotes osteoblastic differentiation by active DNA demethylation. The Runx2 P1 promoter, which is active in osteoblasts (48,49), is not CpG methylation-sensitive (not shown), which may indicate an indirect or active DNA demethylation-independent effect of SFN on Runx2 expression. However, a CpGrich region is present in the Atf4 proximal promoter/first exon region of the gene (Fig. 9D). As shown in Fig. 9, E and F, already after 16 h and partially after 1 day of treatment, 3 M DL-SFN treatment significantly increases the 5hmC levels in the selected fragments of the proximal promoter region/first-second exon of the Atf4 gene by 3-4-fold. In summary, these results indicate that DL-SFN-induced active DNA demethylation affects primary cellular functions in MC3T3-E1 cells.
Effects of SFN on Osteoclast Resorption and Apoptosis-Because SFN has been shown to inhibit osteoclastogenesis by suppressing nuclear factor-B (28) and it also has an antiprolif-erative effect in bone anabolic MC3T3-E1 and MLO-Y4 cells, we examined whether SFN affects proliferation and activity of RAW 264.7 cells. Treatment of these cells with increasing concentrations of SFN for 24 h with the two different preparations of SFN (i.e. L-SFN and DL-SFN) reduces the number of viable cells at doses of 3 M and higher (Fig. 10A). This inhibitory effect appears more pronounced compared with observations made with osteogenic cell lines (see Fig. 2). Furthermore, SFN strongly reduces the cellular metabolic activity of RAW 264.7 cells in EZ4U assays (Fig. 10B). Also for this parameter, a stronger effect in the RAW 264.7 cells is observed when compared with the osteogenic cell lines (Fig. 10C). We next assessed whether these negative effects alter the functional capacity of osteoclasts in bone resorption. Treatment with SFN significantly reduces the area resorbed by primary mouse osteoclast cultures on ivory discs (Fig. 10, D and E). We also investigated whether SFN alters progression of osteoclastic differentiation in RAW 264.7 cells upon induction with RANKL and macrophage colony-stimulating factor by monitoring the temporal expression of classical osteoclastic markers (e.g. Acp5, Clcr, and Ctsk). However, we did not observe appreciable changes in the expression of these genes upon co-treatment with SFN (data not shown). Hence, SFN primarily affects activity of osteoclastic cells via its cytostatic effects.
SFN Induces FAS-dependent Extrinsic Apoptosis in Preosteoclastic Cells-Because SFN induces apoptotic markers in MC3T3-E1 osteoblasts and MLO-Y4 osteocytes (see Fig. 3), we examined whether RAW 264.7 are sensitive to programmed cell death. Similar to MC3T3-E1 and MLO-Y4 cells, the results show that treatment of RAW 264.7 preosteoclastic cells with 3 M L-or DL-SFN up-regulates expression of the proapoptotic gene Fas at 16 h but not at 8 h (Fig. 11A). In addition, either SFN preparation stimulates Caspase 8 and Caspase 3/7 activity (Fig. 11B), indicating that SFN induces the Fas-Caspase 8-Caspase 3/7 pathway that defines extrinsic apoptosis. Similar to findings with MC3T3-E1 osteoblasts and MLO-Y4 osteocytes (see Fig. 3), the DL-SFN mixture is more potent than the pure L-SFN enantiomer in inducing apoptosis (Fig.  11, A and B). More importantly from a biological perspective, we observed that bone-resorbing RAW 264.7 osteoclasts are twice as sensitive to apoptotic induction as the bone anabolic MC3T3-E1 osteoblasts and MLO-Y4 osteocytes (Fig. 11, C-E). These findings suggest that SFN may positively affect bone homeostasis in vivo by preferentially inducing apoptosis in osteoclasts.
SFN Induces Global DNA Hydroxymethylation and Controls Tet1-dependent Cell Death in Osteoclasts-Similar to MC3T3-E1 preosteoblasts, we tested whether SFN alters active DNA demethylation through formation of 5hmC in RAW 264.7 preosteoclasts. Treatment of these cells with 3 M DL-SFN for 16 h strongly increases the global levels of 5hmC (ϳ3-fold) as measured by immunofluorescence microscopy (Fig. 12A) and spectrophotometry (Fig. 12B). The L-SFN enantiomer that has lower biological activity exhibits a less pronounced effect (Ͻ2fold) (Fig. 12B). The observed changes in 5hmC levels are most prominent in RAW 264.7 cells compared with MC3T3-E1 and MLO-Y4 cells (Fig. 12C). The mRNA levels of the hydroxymethylation-related gene Tet1 but not of Tet2 or Tet3 are up- regulated after DL-SFN treatment in RAW 264.7 cells, which becomes evident at 8 h but is more pronounced at 16 h (Fig.  12D). Therefore, to test whether the strong cytostatic effects induced by SFN are dependent on the up-regulation of Tet1 similar to MC3T3-E1 preosteoblasts, we pretreated RAW 264.7 cells with non-silencing RNA (negative control) or siRNA for either Tet1 or Tet2 (Fig. 12E). Again depletion of Tet1 but not Tet2 reduced the cytostatic effect of DL-SFN (Fig. 12F), concurring with the data from the MC3T3-E1 cells and with our previous results (30).
Taken together, our results show that the two organosulfur compounds DMSO (30) and SFN induce active DNA demethylation via up-regulation of the Tet genes in vitro. This epigenetic reprogramming of the chromatin leads to apoptosis of preosteoclasts but only to a lower extent of preosteoblasts. More importantly, SFN leads to enhanced osteoblast differentiation without promoting osteoclast differentiation. Furthermore, in osteocytes and bone tissue explants, SFN decreases the expression of RANKL/Tnfsf11, a major activator of osteoclastogenesis. Although the mechanism of this modulation remains to be investigated, it could be associated with reduced reactive oxygen species required for osteoclast differentiation (50) (Fig. 13).

SFN Has Anabolic and Antiresorptive Effects on Bone Homeostasis in Vivo-Our in vitro results
show that SFN stimulates osteogenic differentiation as well as acts as an antiresorptive agent by blocking osteoclastogenesis and increasing osteoclast apoptosis. Therefore, the attractive possibility arises that SFN may affect bone homeostasis in vivo. As shown by Kong et al. (26), SFN treatment of young mice for 5 weeks with 7.5 mM DL-SFN (which results in a serum concentration of 24 M and lower 4 h postinjection) every other day has no side effects on the animals. We investigated the biological effects of SFN in OVX mice that exhibit bone loss due to estrogen deficiency (Fig. 14). Treatment of sham-operated young adult mice (8 weeks old) with DL-SFN for 5 weeks shows significantly higher metaphyseal cancellous bone volume per tissue volume (BV/ TV) (Fig. 14A) and trabecular number (Tb.N) (Fig. 14B) in proximal tibiae, decreased trabecular separation (Tb.Sp) (Fig.  14C), and no effect on trabecular thickness (Tb.Th) (Fig. 14D) in these same samples. Furthermore, DL-SFN treatment for 5 weeks does not affect cortical thickness in any of the treated groups (data not shown). Thus, the relatively higher BV/TV values found in the DL-SFN-treated animal groups are reflected by changes in trabecular number and not due to a thickening of the trabeculae by enhanced bone apposition. Importantly, SFN treatment of OVX mice reduces bone loss due to estrogen deficiency as reflected by mitigation of the observed OVX-dependent decrease of trabecular number (Fig. 14, A and B).
To complement the CT results on bone microstructural indices, we measured the mineralization status of the bone matrix by assessing local mineral content and distribution using qBEI. Mice were examined for changes in BMDD parameters CaMean, CaPeak, CaWidth, CaLow, and CaHigh reflecting bone turnover, mineralization kinetics, and average bone matrix age (40,51) (Table 2 and Fig. 15). Two-way analysis of variance of CaPeak and CaWidth of trabecular bone revealed significant differences between control (sham) and estrogendepleted (OVX) mice but no effects for treatment with DL-SFN. The parameters CaPeak and CaWidth correlate significantly with the corresponding CT parameters BV/TV, Tb.N, and Tb.Sp (Table 2 and Fig. 15, A-F). Independently of the treatment, it appears that mice with a higher trabecular number have higher matrix mineralization (CaPeak; Fig. 15C) that is less heterogeneous (CaWidth; Fig. 15D), whereas mice with a low trabecular number have a less mineralized matrix that tends to be more heterogenous. Notably, the correlations between structural indices and BMDD parameters show a clear separation between the sham and OVX groups. Consistent with the expected lower bone turnover rates in sham versus OVX mice, the sham group has higher CaPeak and lower CaWidth values than the OVX group. Taken together, our data suggest that DL-SFN has a beneficial effect on bone volume by mitigating loss in trabecular bone number due to both anabolic and antiresorptive cellular effects without altering bone matrix mineralization content and distribution.

Discussion
In this study, we show that the broccoli-derived natural compound SFN has very similar osteoblast stimulatory effects as the structurally related organic solvent DMSO, which is known to promote osteoblast differentiation (30,36). Compared with DMSO, SFN contains an aliphatic isothiocyanate moiety instead of a methyl group, and this modification apparently increases its biological potency in stimulating osteoblast differentiation by Ͼ10,000-fold. We tested whether SFN has anabolic effects on bone in vivo and found that SFN increases bone volume by increasing trabecular number albeit not cortical thickness. In addition, SFN mitigates bone loss induced by estrogen depletion. Thus, SFN exhibits promising efficacy as a possible treatment modality for bone loss-related pathologies like osteoporosis.
Alteration of the epigenome by nutritional habits or specific food compounds is associated with promoting or repressing   specific gene expression and pathologic conditions (52)(53)(54)(55). SFN has previously been shown to inhibit histone deacetylase activity (18,45,56). In human volunteers, a single ingestion of a small amount of broccoli sprouts (ϳ70 g), which contain particularly high amounts of SFN, inhibits histone deacetylase activity in circulating peripheral blood mononuclear cells and consequently induces histone H3 and H4 acetylation a few hours after intake (56). The epigenetic landscape of cells can also be modified by other food compounds and secondary plant metabolites, including the short-chain fatty acid butyrate, which is abundant in rancid butter; resveratrol (57), a natural phenol that is enriched in grapes and peanuts; and organosulfur compounds (58,59), which are enriched in garlic. Our results define SFN as an epigenetic bone anabolic agent.
A number of studies that assessed the role of SFN as a tumor cell inhibitor have recognized its epigenetic potential (60 -64). The results of this work show that SFN, besides possible effects on DNA demethylation patterns from ascorbic acid occurring in the culture medium (65), promotes Tet-mediated hydroxymethylation, which is a key mechanistic step in the derepression of gene transcription silenced by CpG methylation, and demethylation of FAS promoter frequently activates the expression of this gene as shown in tumor treatment (66). Because loss of CpG methylation decreases the gene-repressive histone mark H3K27me3 (67), SFN may indirectly increase formation of transcriptionally active acetylated histones H3 and H4. Hence, our findings suggest that SFN may be an indirect histone deacetylase inhibitor via direct effects on hydroxymethylation.
Our findings indicate that SFN potentially has a two-pronged effect on bone formation and homeostasis by acting differently on osteoblasts and osteoclasts. Although matrix mineralization is remarkably enhanced during osteoblast differentiation, SFN does not promote osteoclastogenesis, reflecting lineage-specific effects in cellular differentiation. Furthermore, osteoclast viability is rather sensitive to escalating doses of SFN; however, osteoblasts are more refractory to the cytotoxic effects of SFN, and they are fairly resistant to apoptosis (via the Fas-Caspase 8-Caspase 3/7 pathway) at a bone anabolic dose (3 M) that induces apoptosis in osteoclasts. These divergent and poten-  tially cooperating biological effects of SFN may be of immediate clinical interest because of adverse effects observed for some bisphosphonate-based therapies, which are a first line of defense against bone loss (68). The bone-stimulatory effects of SFN render this natural food compound a viable candidate for alternative therapeutic strategies to promote bone formation.
SFN-dependent enhancement of Tet1-mediated apoptosis may exert a net bone anabolic effect by selective attrition of osteoclasts relative to osteoblasts. Furthermore, SFN has a positive effect on osteoblast activity by inducing bone matrix mineralization in in vitro cell culture (e.g. primary osteocytes) and ex vivo tissue culture (e.g. calvarial explants of both newborn and adult mice). These biological effects are linked to the up-regulation of genes involved in matrix mineralization and enhanced expression of osteoblastic transcription factors like Runx2, Sp7, and Atf4. Although the Runx2 P1 promoter or the Sp7 promoter are not CpG methylation-sensitive by SFN treatment, we measured a significant increase in 5hmC levels in the Atf4 promoter with a concomitant increase in Atf4 gene expression, demonstrating a mechanistic link between active DNA demethylation and enhanced osteoblastic differentiation. Interestingly, as we have shown before, early modulation of epigenetic marks like inhibition of histone acetylation or methylation (69,70) and, as shown here, an increase of active DNA demethylation show long term effects on osteoblastic differentiation, suggesting that SFN may specifi- TABLE 2 Examination of changes in BMDD parameters CaMean, CaPeak, CaWidth, CaLow, and CaHigh, which reflect bone turnover, mineralization kinetics, and average bone matrix age by qBEI Two-way analysis of variance of CaPeak and CaWidth of trabecular bone between all treatment groups is shown (n ϭ 5). cally support epigenetic chromatin reprogramming at early differentiation stages.
SFN significantly decreases the expression of the osteoclast activator Tnfsf11 (RANKL) in osteocytes. This diminished expression cannot be directly accredited to SFN-dependent mechanisms affecting hydroxymethylation. However, because osteocytes represent a main systemic source for Tnfsf11/ RANKL, changes in the expression of this osteoclastogenic ligand may perhaps contribute to the antiresorptive effects of SFN. Furthermore, SFN was shown to activate the transcription factor NRF2 (Nfe2l2), which promotes the expression of the two key antioxidant genes peroxiredoxin-1 and NAD(P)H dehydrogenase quinone 1, thus activating a sustained antioxidant response in osteoclast progenitors. The inhibition of osteoclast differentiation might thus also be associated with a down-regulation in RANKL-dependent intracellular reactive oxygen species levels in preosteoclastic cells (50). Thus, the effects observed by SFN in osteoclasts and in osteoblasts might also partially be attributed to its antioxidative proprieties.
From a clinical perspective, bisphosphonates (e.g. ibandronate) not only block osteoclast formation but may also inhibit osteoblastic differentiation and negatively affect bone remodeling (8). Patients with distal radius fractures administered bisphosphonates show a significantly longer union time compared with controls (13). Furthermore, use of bisphosphonates is associated with a higher risk of atypical femoral fractures in women (9,10,68). Widely used bisphosphonates may reduce bone turnover rates and increase the degree of mineralization beyond normal physiological values (11,12). Our studies show that SFN inhibits osteoclastogenesis but additionally promotes osteoblastic differentiation and bone matrix mineralization, which may account for the net anabolic effect of SFN we observed in vivo in which SFN augments bone volume by increasing trabecular number but not shifting the local, physiological distribution of mineral in the bone matrix. A limitation of our study is the relatively young age of the mice (8 weeks old) used. To better address the potential beneficial effects of SFN for the treatment of e.g. osteoporosis, the use of older/adult mice might be more appropriate. Nevertheless, these encouraging findings provide the impetus for considering SFN and related compounds as candidates for new therapies with food-derived epigenetic compounds.
Author Contributions-R. T., F. V., H. K., N. R., A. T., J. Z., K. K., and A. J. v. W. designed the study. R. T., A. M., P. R., I. S., F. K., S. S., M. R., and A. D. performed the experiments. R. T., A. J. v. W., and F. V. wrote the paper with comments and contributions from all authors.