FOXO1 Mediates RANKL Induced Osteoclast Formation and Activity

Abstract

We have previously shown that the transcription factor FOXO1 is elevated in conditions with high levels of bone resorption. To investigate the role of FOXO1 in the formation of osteoclasts we examined mice with lineage specific deletion of FOXO1 in osteoclast precursors and by knockdown of FOXO1 with siRNA. The receptor activator of NF-kappa B ligand (RANKL), a principal bone resorbing factor, induced FOXO1 expression and nuclear localization two days after stimulation in bone marrow macrophages (BMMs) and RAW264.7 osteoclast precursors. RANKL- induced osteoclast formation and osteoclast activity was reduced in half in vivo and in vitro with lineage specific FOXO1 deletion (LyzM.Cre⁺FOXO1^L/L) compared to matched controls (LyzM.Cre⁻FOXO1^L/L). Similar results were obtained by knockdown of FOXO1 in RAW264.7 cells. Moreover, FOXO1-mediated osteoclast formation was linked to regulation of NFATc1 nuclear localization and expression as well as a number of downstream factors including dendritic cell-specific transmembrane protein (DC-STAMP), ATP6vod2, cathepsin K and integrin αν Lastly, FOXO1 deletion reduced M-CSF induced RANK expression and migration of osteoclast precursors. Studies presented here provide the evidence that FOXO1 plays a direct role in osteoclast formation by mediating the effect of RANKL on NFATc1 and several downstream effectors. This is likely to be significant since FOXO1 and RANKL are elevated in osteolytic conditions.

Introduction

Osteoclasts play a critical role in bone development, normal remodeling and pathologic conditions such as periodontitis, rheumatoid arthritis and osteoporosis. Osteoclasts are generated from myeloid progenitors that are produced in bone marrow and released into the vasculature (1). The formation of osteoclasts is governed by ligands, particularly macrophage colony-stimulating factor (M-CSF), receptor activator of nuclear factor-kappa B ligand (RANKL) and tumor necrosis factor-alpha (TNF-α) as well as several other factors that directly or indirectly contribute to their formation and activity (2). Osteoclast formation involves several steps including differentiation of myeloid precursors to pre-osteoclasts, fusion of mononuclear pre-osteoclasts to multinucleated osteoclasts, maturation and activation of osteoclasts to resorbing bone.

Several cytokines have been shown to play a key role in osteoclast formation. M-CSF stimulates progenitors to upregulate RANK expression, the receptor for RANKL (3, 4). RANKL signaling stimulates formation of single cell TRAP⁺ osteoclast precursors and induces cell fusion to form multi-nucleated TRAP⁺ cells. M-CSF and RANKL stimulate several signaling intermediates and transcription factors such as DAP12, TRAF6, NF-κB components (relA, p50, p52) and potential downstream effector molecules needed for cell fusion, osteoclast activity and survival including CD9, DC-STAMP, NFATc1, Atp6v0d2, OC-STAMP, integrin ανβ3, cathepsin K, and others (5, 6).

Cell fusion is a characteristic feature of osteoclasts. It involves cell migration, chemotaxis, intercellular adhesion and fusion of membranes. DC-STAMP, a transmembrane protein plays an important role in cell fusion (7). It plays a central role in cell fusion and is induced by two transcription factors, nuclear factor of activated T cells 1 (NFATc1) and c-Fos (8). NFATc1 is involved in a number of cellular functions in various cell types. RANKL induces NFATc1 in osteoclast precursors through activation of NF-κB, which is further enhanced by Ca⁺ oscillation regulated by Tmem64 (9). Overexpression of NFATc1 can stimulate osteoclast differentiation without RANKL signaling, suggesting that NFATc1 is a master switch that regulates osteoclast differentiation downstream of RANKL (8). Once formed, osteoclasts are polarized so that the cellular constituents needed for resorption are localized to the bone surface. This process involves the formation of actin rings and the integrin alpha-ν beta-3 (ανβ3) (10). Resorption of the bone surface requires the action of vacuolar (V-) ATPase (Atp6v1c1) proton pump located in the osteoclast cell membrane and the release of cathepsin K, a lysosomal cysteine protease that degrades bone matrix proteins including collagen (11, 12).

FOXO1 is a transcription factor that has been linked to protection of osteoblasts from oxidative stress (13, 14). FOXO1 is a member of the forkhead transcription factors in the O-box sub-family. There are four members, FOXO-1, -3, -4 and -6 (15). FOXO1 regulates transcription of many different classes of genes depending upon the cell type and nature of the stimulus (16). We have recently shown that FOXO1 mediates inflammatory gene expression in dendritic cells (17) and that FOXO1 is elevated in conditions with high levels of osteoclastogenesis (18, 19). Similarly FOXO1 mediates gene expression in macrophages (20). A recent report indicates that long-term combined deletion of FOXO1, -3 and -4 decreases physical bone mass by increasing osteoclast numbers and activity (21). While this report examined the indirect effects of FOXO deletion associated with aging and focused on simultaneous deletion of three FOXO members in vivo, we took an alternative approach and examined the direct effect of FOXO1 in mediating short-term RANKL-stimulated osteoclastogenesis. The results indicate that FOXO1 is activated two days after RANKL stimulation in bone marrow macrophages and RAW264.7 cells. To determine whether FOXO1 mediates the effects of RANKL stimulation in vivo, a calvarial model was used (2, 6). FOXO1 deletion significantly reduced osteoclastogenesis and osteoclast activity stimulated by RANKL injection. In vitro experiments agreed well with in vivo experiments and demonstrated that FOXO1 deletion reduced osteoclastogenesis and osteoclast activity induced by RANKL in both bone marrow macrophages derived from experimental mice and in RAW264.7 cells with FOXO1 knockdown by siRNA. Moreover, FOXO1 deletion reduced the expression of several proteins that are involved in osteoclast formation or function. These results point to FOXO1 having a role in mediating RANKL stimulated osteoclast formation, which is distinct from its long-term effect associated with aging (21).

Materials and methods

Reagents and Mice

Antibodies were obtained from Santa Cruz Biotechnology (Dallas, Texas) unless noted otherwise. Mice that express Cre recombinase under control of the lysozyme M promoter (LyzM⁺.Cre) were purchased from Jackson Laboratories (Bar Harbor, Maine). FOXO1^L/L mice were generously provided by Dr.Ronald De Pinho (University of Texas MD Anderson Cancer Center, Houston, Texas) (22). FOXO1^L/L mice were bred with LyzM.Cre mice to generate experimental mice (LyzM.Cre⁺FOXO1^L/L) and the control littermates (LyzM.Cre⁻FOXO1^L/L) (23, 24). Homozygous LyzM.Cre mice were crossed with FOXO1^L/L to generate mice that were heterozygous for FOXO1 conditional alleles (LyzM.cre⁺FOXO1^L/W) and then backcrossed with the FOXO1^L/L mice to obtain the experimental LyzM.cre⁺FOXO1^L/L and the FOXO1^L/L control littermates. Genotypes were determined by PCR using primers specific for LyzM.Cre: 5′-ATCCGAAAAGAAAACGTTGA-3′and 5′-ATCCAGGTTACGGATATAGT-3′and specific for FOXO1: 5'-GCTTAGAGCAGAGATGTTCTCACATT-3', 5'-CCAGAGTCTTTGTATCAG GCAAATAA-3', 5'-CAAGTCCATTAATTCAGCACATTG A-3'. We chose both male and female 10-week-old mice for RANKL injection experiment. Anesthesia was achieved with ketamine (80 mg/kg of body weight) and xylazine (10 mg/kg) in sterile phosphate-buffered saline (PBS). All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Pennsylvania. Soluble RANKL (3μg/injection) (Peprotech, Rocky Hill, NJ) or vehicle (PBS) was injected at the midline of the calvaria of LyzM.Cre⁺FOXO1^L/L and littermate control LyzM.Cre⁻FOXO1^L/L mice daily for 5 consecutive days (25). One week after the final injection, mice were euthanized and calvarial bone was examined by microCT-40 (Scanco Medical AG, Bassersdorf, Switzerland). RANKL was injected at the midline between first and second coronal sutures and compared to mice injected with vehicle alone, PBS groups. Specimens were fixed in paraformaldehyde and small pits on the surface of the calvaria were quantified in reconstructed images following microCT scanning with NIS Elements-AR software (Nikon). Specimens were then decalcified in EDTA and osteoclasts were counted as TRAP⁺ multinucleated cells in sutures and bone marrow cavities by TRAP histostain (25) and immunofluorescence with antibody specific for TRAP.

Cell culture

Bone marrow macrophages (BMMs) were cultured from bone marrow cells isolated from femurs and tibiae of experimental and control littermate mice (26). After the marrow was obtained, single-cell suspensions were prepared by mechanically dispersing the bone marrow through a 70-um cell strainer. Then bone marrow cells were cultured in a-minimum Eagle’s media supplemented with 10% FBS, 1% Penicillin, and 0.05% streptomycin (Life Technologies, Grand Island, NY) at 37 °C overnight. Then non-adherent cells were harvested media with 30ng/ml of M-CSF (Peprotech, Rocky Hill, NJ). 72hrs later, floating cells were removed, and attached BMMs were obtained.

Spleens were harvested from experimental and control littermate mice (27), and single cells were obtained with a cell strainer (70um). Erythrocytes were removed by red blood cell lysis buffer (Sigma-Aldrich, St. Louis, MO) and splenocytes were cultured in DMEM supplemented with 10% FBS, 1% penicillin, and 0.05% streptomycin (Life Technologies) at 37 °C overnight. The next day, non-adherent cells were harvested and used for immunofluorescence with antibody specific for FOXO1.

Raw 264.7 cells were cultured in complete DMEM supplemented with 10% FBS, 0.05% β-mercaptoethanol, 1% penicillin and streptomycin. In some experiments RAW264.7 cells were transfected with FOXO1 siRNA, FOXO3 siRNA or control scrambled siRNA (Santa Cruz Biotechnology) with Lipofectin (Invitrogen) or GeneMute siRNA transfection reagents (Signagen, Rockville, MD).

In some experiments freshly isolated single cell suspensions of bone marrow cells (BM) were obtained from bone marrow and centrifuged at 500 rpm for 5min in a cytocentrifuge (Thermo Scientific, Rockford, IL). Attached cells were fixed with 4% paraformaldehyde, washed, and subjected to immunofluorescence with antibody specific for FOXO1 or control IgG. No signal was detected with matched control. Results were obtained by measuring mean fluorescence intensity or counting immunopositive cells.

In vitro formation of TRAP⁺ multinucleated cells and osteoclast activity assays

BMMs were stimulated with 100ng/ml RANKL and 30ng/ml M-CSF or 100ng/ml RANKL alone for 4 days and fixed with 4% paraformaldehyde whereas RAW264.7 cells were stimulated to 100 ng/ml RANKL and then fixed with 4% paraformaldehyde. The number of multinucleated TRAP⁺ (>3 nuclei) cells, average area and average number of nuclei per multinucleated TRAP⁺ cell were counted after staining for TRAP with a kit from Sigma or F-actin with fluorescein isothiocyanate labeled phalloidin from Sigma.

Osteoclast bone resorption activity was measured with the Osteo LyseTM assay kit (Lonza, Allendale, NJ) according to the manufacturer’s protocol. Briefly, BMMs from experimental and control mice or RAW264.7 cells transfected with FOXO1 siRNA or control scrambled siRNA were cultured in 96-well OsteoLyse plates (Lonza) at 50,000 cells/well in complete a-MEM medium. BMM were incubated with 30 ng/ml M-CSF and 100 ng/ml soluble RANKL and RAW265.7 cells incubated with 100ng/ml soluble RANKL. Four days later culture medium was renewed and samples were collected after an additional 24 hours of culture and calcium release was measured by fluorescence.

Treatment by inhibitors

Raw264.7 cells were stimulated with RANKL for 48hrs. Inhibitors such as BML-257 (12.5uM), Triciribine (1uM), SB203580 (10uM), JNK Inhibitor I (10uM), PD098059 (20uM), N-acetyl-L-cysteine (5mM), L-N^G-Nitroarginine-methylester (100uM), Sirtinol (10uM) and DON (50uM) were added for the last 24hrs. Cells were then fixed in 100% methanol at −20°C for 10mins and incubated with primary antibody to FOXO1 or matched non-immune primary antibody overnight at 4°C. Antibody was then detected by a biotinylated secondary antibody and visualized using Alexa 546-conjugated streptavidin, and counterstained with 4'-6-diamidino-2-phenylindole (DAPI). Images were acquired using a Nikon 90i fluorescent microscope and image analysis was performed using NIS-Elements software (Nikon, Melville, NY, USA). Nuclear localization with standard immunofluorescence was confirmed with deconvolution immunofluorescent microscopy. Cells were imaged in five 1.7um stacks. The mid stack representing the mid-center of the cell was then further analyzed for FOXO1 nuclear localization using NIS Elements-AR software (Nikon, Inc, Melville, NY).

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted and real-time PCR was performed with primers designed using the Universal Probe Library Assay Design Center (Roch Applied Science, Indianapolis, IN) and labeled probes (Roche Applied Science). Each value was normalized to ribosomal protein L32 and represents the mean of 3 independent experiments.

Migration assay

Chemotaxis of BMMs and RAW264.7 cells was measured using transwells with polycarbonate filters (8-μm pores) (Corning, Corning, NY) with or without M-CSF added to the bottom chamber for 6 hours at 37 °C. Cells on the top of the filter were removed and those that had migrated to the bottom chamber were counted after DAPI staining by fluorescence microscopy.

Western blot analysis

Cytoplasmic and nuclear extractions were prepared by using NE-PER Nuclear and Cytoplasmic extraction reagents (Thermo Scientific, Rockford, IL). The cytosolic and nuclear fractions were resolved by SDS-PAGE. Primary antibody against FOXO1 (Cell Signaling Technology, Boston, MA); laminine A (Sigma-Aldrich, St Louis, MO); actin (Cell Signaling Technology) were used according to the manufacturer’s directions. Bands were visualized by biotinylated species-specific secondary antibody and chemilluminesnece using a kit from Thermo Scientific (Rockford, IL). No signal was detected with matched control.

Immunofluorescence

For in vitro assays, cells were fixed in 100% methanol (10 min, −20°C) followed by incubation with primary antibody specific for FOXO1, CD11b, RANK, and the integrin αν or matched control antibody. Antibody was then detected by a biotinylated secondary antibody and visualized using Alexa 546-conjugated streptavidin, and counterstained with DAPI. Images were acquired using a Nikon 90i fluorescent microscope and image analysis was performed using NIS-Elements software (Nikon, Melville, NY, USA). No signal was detected with matched control. For most assays data are presented as the mean fluorescence intensity for each group. In some experiments nuclear localization was determined by co-localization of FOXO1 immunostaining and DAPI nuclear staining. For in vivo assays, paraffin sections were subjected to antigen retrieval in citrate buffer at 95°C. Sections were incubated with antibody specific to TRAP, FOXO1 or NFATc1. Antibodies were localized with biotinylated secondary antibody followed by avidin-biotin horseradish peroxidase complex (Vector Laboratories, Burlingame, CA). Antibodies were visualized using streptavidin Alexa-546 (Invitrogen, Carlsbad, CA) for TRAP and Fluorescein Avidin DCS (Vector Laboratories, Burlingame, CA) for FOXO1 or NFATc1, and counterstained with DAPI. Tyramide signal amplification (PerkinElmer, Waltham, MA) was used to enhance the signal. Fluorescent staining of TRAP and FOXO1 or NFATc1 positive cells were observed under 20x magnification of images captured with a Nikon Eclipse 90i microscope (Nikon, Melville, NY, USA) and analyzed with NIS Elements-AR software (Nikon).

Luciferase Reporter Assay

RAW264.7 cells were transfected with FOXO1WT, FOXO1AAA construct or pcDNA empty vector alone as we have described (28) by electroporation (Neon Transfection System, Life Technologies) and co-transfected with an NFATc1-luciferase reporter (kindly provided by Dr. Yihong Wan) (29). Luciferase activity was measured two days later using a dual Luciferase Assay System (Promega) and reported as firefly/Renilla luciferase ratio.

Statistical analysis

Most in vitro experiments were carried out three times to obtain mean and standard error from independent experiments. In some cases experiments were carried out in triplicate to provide a mean and standard error of the mean with experiments was carried out three times to confirm results. In vivo experiments were carried out with 5 to 7 animals per group. Statistical significance was determined by one-way ANOVA with Turkey’s post hoc test for comparisons between multiple groups at the P <0.05 level. In some cases significance between experimental and control groups or directly compared at a given time point by Students t-test to establish significance at the P <0.05 level.

Results

FOXO1 specific deletion in osteoclast precursors

BMMs were isolated from experimental (LyzM.cre⁺FOXO1^L/L) and matched control (LyzM.cre⁻FOXO1^L/L) mice. LyzM driven Cre recombinase reduced FOXO1 mRNA levels (Fig. 1A) by 95% and protein levels by 90% in BMMs (Fig. 1B). The majority of bone marrow cells and splenocytes directly isolated from experimental and control mice did not exhibit FOXO1 deletion as expected since most are not CD11c⁺ (Fig. 1C, D) (Supplemental Fig. 1).

Figure 1. LyzM deletes FOXO1 in bone marrow macrophages.

(A) FOXO1 mRNA levels were measured in BMMs by qRT-PCR. (B) FOXO1 expression was measured by immunofluorescence as mean fluorescence intensity in BMMs. (C) Percent FOXO1 immunopositive cells in single cell suspension of bone marrow cells (BM) was measured by immunofluorescence. (D) FOXO1 expression was measured by immunofluorescence as mean fluorescence intensity in splenic cells. Each value represents the mean ± SEM of triplicate specimens. *Significantly different from matched Cre⁻ group (P < 0.05).

FOXO1 deletion reduces osteoclastogenesis in vivo

Studies were carried out in which RANKL was injected over the calvarial bone of experimental and control mice. The formation or resorption pits was measured on the surface of calvarial bone in 3D microCT images. The number of resorption pits decreased by 60% and total pit area was reduced by 64% in experimental mice compared to control mice (P<0.05) (Fig. 2A, B, C). The number of osteoclasts was measured in the sutures and bone marrow cavities of calvarial bone by TRAP histostain (Fig 2D, E) or by immunofluorescence with antibody to TRAP (Fig 2F). RANKL stimulated a 3 to 10 fold increase in osteoclasts. RANKL induced osteoclast formation was reduced by 67% in sutures of LyzM.cre⁺FOXO1^L/L mice compared with control LyzM.cre⁻FOXO1^L/L mice (p<0.05) and completely blocked in bone marrow cavities (Fig. 2D, E, F). There was no difference in osteoclast numbers in PBS injected groups (p>0.05). The effect of FOXO1 deletion on cortical bone architecture and cancellous bone volume, mineral density, number, thickness and spacing of trabeculae associated with remodeling in young mice was examined (Supplemental Fig. 2B, C, D, E, F). No significant difference was observed between experimental and control mice (P>0.1).

Figure 2. FOXO1 deletion reduces osteoclastogenesis in vivo.

(A) Soluble RANKL (3ug/injection) or vehicle (PBS) was injected at the midline of the calvaria of LyzM.cre⁺FOXO1^L/L and littermate control LyzM.cre⁻FOXO1^L/L mice daily for 5 consecutive days. After 12 days, calvarial bone was examined by microCT. (B, C) Pit number and total pit area per calvaria were analyzed with microCT. (D, E) TRAP⁺ multi-nucleated bone-lining osteoclasts were identified by histostain in calvarial sutures or bone marrow cavities. (F) Osteoclasts were counted as TRAP⁺ multinucleated cells lining bone in bone marrow cavity and detected by immunofluorescence. Each value is the mean ± SEM for n=6 mice per group. *Significantly different compared to matched Cre⁻ group (P < 0.05).

Loss of FOXO1 leads to reduced osteoclast formation and activity in vitro

BMMs from FOXO1 deleted or matched control mice were stimulated with RANKL in vitro to induce osteoclastogenesis (Fig. 3A, Supplemental Fig. 2G). Deletion of FOXO1 resulted in a 65% reduction in the number of osteoclasts formed (p<0.05) (Fig. 3B). In addition there was a 58% reduction in the average area per osteoclast and a 59% reduction in the average number of nuclei per osteoclast (p<0.05) (Fig. 3C, D). Experiments were also carried out with RAW 264.7 cells transfected with FOXO1 siRNA or scrambled siRNA followed by RANKL stimulation (Supplemental Fig. 2H). FOXO1 siRNA reduced FOXO1 levels by 70% (Fig. 3E) and reduced the number of osteoclasts induced by RANKL in RAW264.7 cells by 49%, average area per osteoclast by 27% and average nuclei per osteoclast by 35% (p<0.05) (Fig. 3F, G, H). Scrambled siRNA had no effect (p>0.05).

Figure 3. FOXO1 deletion or FOXO1 silencing reduces osteoclastogenesis and osteoclast activity in vitro.

(A-D) BMMs from LyzM.cre⁺FOXO1^L/L and littermate control LyzM.cre⁻FOXO1^L/L mice were incubated with RANKL (100ng/ml) and M-CSF (30ng/ml). After 4 days cells were fixed and stained as indicated. (A) TRAP⁺ multinucleated osteoclasts (10x magnification). (B) TRAP⁺ multinucleated osteoclasts with greater than 30 nuclei per osteoclasts. (C) Mean area per osteoclast. (D) Mean number of nuclei per osteoclast. (E-H) RAW 264.7 cells were transfected with FOXO1 siRNA or scrambled siRNA and incubated with RANKL (100ng/ml). After 4 days cells were fixed and stained as indicated. (E) FOXO1 mRNA levels were measured in RAW264.7 cells by qRT-PCR after siRNA transfection. (F) TRAP⁺ multinucleated osteoclasts with greater than 3 nuclei per osteoclast (10x magnification). (G) Mean area per osteoclast. (H) Mean number of nuclei per osteoclast. Each value represents the mean ± SEM of three independent measurements. (I) BMMs from experimental and control mice were stimulated with RANKL (100ng/ml) and M-CSF (30ng/ml) or incubated with vehicle alone. Osteoclast activity was measured by release of europium-labeled collagen measured by florescence. (J) RAW 264.7 cells were transfected with FOXO1 siRNA or scramble siRNA and were stimulated with RANKL (100ng/ml) for 4 days. Osteoclast activity was measured as in panel I. (K) RAW 264.7 cells were transfected with siFOXO1, siFOXO3 or scrambled siRNA and then incubated with RANKL (100ng/ml) for 4 days, fixed and stained for osteoclasts with TRAP histostain. (L) RAW 264.7 cells were transfected with siFOXO3 or scrambled siRNA and FOXO3 mRNA levels were measured by qRT-PCR after 2 days. Each value represents the mean ± SEM of three or four independent experiments. *Significantly different compared to matched Cre⁻ or scrambled siRNA group (p<0.05). +Significantly different compared to matched Cre⁻ or scrambled siRNA group stimulated with RANKL and M-CSF or RANKL alone (p<0.05).

Osteoclast activity in vitro was measured by release of labeled collagen. RANKL stimulated a 16-fold increase in resorptive activity in osteoclasts from BMMs of control mice, which was blocked by FOXO1 deletion (P<0.05) (Fig. 3I). FOXO1 knockdown by siRNA also blocked RANKL induced resorptive activity from RAW264.7 cells (P<0.05) (Fig. 3J).

To compare the role of FOXO1 and FOXO3, each was knocked down separately in RAW264.7 cells and then stimulated with RANKL. FOXO1 knockdown reduced osteoclast formation by 66% and FOXO3 knockdown had no significant impact on osteoclastogenesis even though FOXO3 siRNA effectively reduced FOXO3 levels (Fig. 3K, L).

FOXO1 is down regulated by RANKL through different pathways

Both BMMs and RAW264.7 cells were examined for RANKL induced FOXO1 by both RT-PCR and immunofluorescence (Fig. 4A, B, C). At the protein level little to no FOXO1 was detected in BMMs or RAW264.7 cells at baseline. RANKL stimulated an increase in FOXO1 levels that could be detected within one day and which increased further on days 2 and 4. RANKL stimulated a maximum 7.9 fold in FOXO1 in BMMs and a 4.5-fold increase in RAW264.7 cells (P<0.05) (Fig. 4C, D, E). To confirm that FOXO1 expression was associated with osteoclastogenesis in vivo, double immunofluorescence was performed to identify FOXO1⁺, TRAP⁺ multinucleated osteoclastic cells (Fig. 4F).

Figure 4. RANKL Induces FOXO1 in BMM and RAW265.7 cells.

(A, B) BMMs from FOXO^L/L mice were incubated with M-CSF (30ng/ml) and RANKL (100ng/ml) or RAW264.7 cells were incubated with RANKL (100ng/ml). Cells were fixed on 0, 1, 2 and 4 days later. FOXO1 (red) was examined by immunofluorescence with specific antibody (20x magnification) and cells were counterstained with DAPI (blue). No signal was detected with matched control antibody (data not shown). (C) FOXO1 mRNA levels were measured by qRT-PCR in BMMs following stimulation with M-CSF (30ng/ml) and RANKL (100ng/ml). (D, E) FOXO1 expression was measured by immunofluorescence as mean fluorescence intensity in BMMs and RAW264.7 cells. (F) RANKL was injected into the calvaria as described in Figure 2. FOXO1⁺,TRAP⁺ double immunopositive cells in close proximity to the bone surface were counted in the suture area. n=6 mice per group. (A-E) Each value represents mean ± SEM of three independent measurements. *Significantly different compared to day 0 (P<0.05) or matched Cre⁻ group (P<0.05); + significantly different compared to BMMs on day 2 (P<0.05).

FOXO1 is active in the nucleus and FOXO1 activation can be followed by nuclear localization. FOXO1 nuclear localization induced by RANKL could also be detected after 1 day and peaked with a 10-fold increase on day 2 in both BMMs and RAW264.7 cells by immunofluorescence (Fig 5A, B). Interestingly the pattern of FOXO1 nuclear localization differed from the expression pattern by declining substantially between day 2 and day 4 with a 70% reduction in FOXO1 nuclear localization in BMMs and a 64% decline in RAW264.7 cells (P<0.05) (Fig. 5A, B). Nuclear localization was confirmed by deconvolution immunofluorescent microscopy, which examined FOXO1 co-localization with DAPI at the mid center of each cell. By this analysis RANKL stimulated a 7.4-fold increase in FOXO1 nuclear localization in BMMs from wild-type mice and a 5.6-fold increase in RAW264.7 cells (P<0.05) (Fig. 5C, D), consistent with results obtained by standard immunofluorescence microscopy (Fig. 4A, B). RANKL induced FOXO1 nuclear localization was also examined by Western blot analysis. FOXO1 protein accumulated in the cytoplasm within four hours of RANKL stimulation (Fig. 5E). Nuclear localization was strongly detected two days after RANKL stimulation (Fig. 5E), which was consistent with the immunofluorescence data.

Figure 5. RANKL stimulates FOXO1 nuclear localization.

(A, B) BMMs were incubated with M-CSF (30ng/ml) and RANKL (100ng/ml) or RAW264.7 cells were incubated with RANKL (100ng/ml) for 0 – 4 days. FOXO1 nuclear localization was determined by FOXO1 immunofluorescence and co-localization with DAPI nuclear stain. (C, D) Immunofluoresence was repeated and co-localization of FOXO1 and DAPI nuclear stain was determined by Z-stack analysis in the center plane of BMMs and RAW264.7 cells. (E) RAW264.7 cells were incubated with RANKL (100ng/ml). Cells were fixed after 0, 4, and 48hrs. Cytoplasmic and nuclear extracts were prepared and FOXO1 cytoplasmic and nuclear levels were measured by immunoblot analysis. (F) Raw264.7 cells were stimulated with RANKL (100ng/ml) for 48hrs. Inhibitors of Akt, BML-257 (12.5uM), Triciribine (1uM); P38, SB203580 (10uM); JNK Inhibitor I (10uM); ERK, PD098059 (20uM); ROS, N-acetyl-L-cysteine (5mM); nitric oxide synthase, L-N^G-Nitroarginine-methyl-ester (100uM); sirtuin-1, sirtinol (10uM); phosphate- activated glutaminase, DON (50uM) were added for the last 24hrs. Cells were then fixed and FOXO1 was measured as described in Panel A. Each value represents mean ± SEM of three independent measurements. *Significantly different compared to BMMs on day 0 (P<0.05) or matched control group (p<0.05). +Significantly different compared to BMMs on day 2 after RANKL and M-CSF stimulation (P<0.05) or matched control group (p<0.05).

To investigate RANKL stimulated FOXO1 nuclear localization further, experiments were performed examining pathways that may mediate RANKL induced FOXO1 nuclear localization. Inhibitors of p38 and JNK but not Erk MAP kinase inhibitors reduced FOXO1 nuclear localization by ~50%. A ROS inhibitor, sirtuin-1 inhibitor and nitric oxide synthase inhibitor also reduced FOXO1 nuclear localization by 39% to 65% (P<0.05) (Fig. 5F). Thus, a number of known pathways stimulated by RANKL induce FOXO1 nuclear localization. However, an Akt inhibitor did not have a significant effect on RANKL induced FOXO1 nuclear localization.

Effect of FOXO1 on expression of target genes

Several potential target genes were examined at the mRNA level to investigate the role of FOXO1 in mediating RANKL stimulation. This was done in BMMs obtained from experimental mice with FOXO1 deletion compared to matched controls. RANKL increased FOXO1 mRNA levels which was blocked by Cre recombinase (Figure 6A) (p<0.05). Deletion of FOXO1 reduced by ~60% the increase in NFATc1 mRNA levels stimulated by RANKL in vitro (Fig 6B, C)(Supplemental Fig. 3A) and reduced RANKL stimulated NFATc1 nuclear localization by 53% (p<0.05) (Fig. 6D). In vivo RANKL simulated a 2.1-fold increase in NFATc1⁺TRAP⁺ multinucleated cells measured by double immunofluorescence staining. This was reduced 70% in LyzMcre⁺FOXO1^L/L mice compared with LyzMcre⁻FOXO1^L/L mice (p<0.05) (Fig. 6E). However, co-transfection of wild-type (FOXO1 WT) or a constitutively active FOXO1 (FOXO1AAA) expression vector with an NFATc1 reporter did not enhance NFATc1 transcriptional activity compared to co-transfection with empty vector (PcDNS) (Fig. 6F). However, deletion of FOXO1 reduced the capacity of RANKL to stimulate increased mRNA levels of factors that are involved in osteoclast formation and activity including ATP6v0d2 by 49%, cathepsin K by 63%, and DC-STAMP by 55% (p<0.05) (Fig. 6G, H, I).

Figure 6. FOXO1 deletion inhibits expression of genes involved in osteoclastogenesis.

BMMs from LyzM.cre⁺FOXO1^L/L or littermate control LyzM.cre⁻FOXO1^L/L mice were stimulated with M-CSF (30ng/ml) and RANKL (100ng/ml) for 0 or 2 days. (A, B) FOXO1 or NFATc1 mRNA levels were assessed by real-time PCR. (C) NFATc1 was examined by immunofluorescence with antibody specific for FOXO1. Mean fluorescence intensity was measured. (D) NFATc1 was detected by immunofluorescence and nuclear co-localization with DAPI measured. (E) RANKL was applied to experimental and matched control mice described in Figure 2. NFATc1 and TRAP were examined by double immunofluorescence and sections were counterstained with DAPI. The percent NFATc1⁺,TRAP⁺ double immunopositive cells in close proximity to bone were measured in the calvarial sutures. (F) Luciferase assay were performed in RAW264.7 cells co-transfected with FOXO1 WT expression vector, FOXO1AAA constitutively active or PcDNS empty vector alone and an NFATc1 promoter-luciferase construct. Luciferase activity was measured as described in Methods. (G, H, I) mRNA levels of ATP6v0d2, cathepsin K and DC-STAMP were measured in BMMs from LyzM.cre⁺FOXO1^L/L and littermate control LyzM.cre⁻FOXO1^L/L mice stimulated with M-CSF and RANKL as described in Panel A. Each value represents the mean ± SEM of three independent measurements. *Significantly different compared to matched control group (P < 0.05).

FOXO1 regulates RANK and osteoclast precursor migration

Osteoclast precursors (CD11b⁺ bone marrow macrophages) are induced by M-CSF from myeloid progenitor cells (6). The effect of FOXO1 deletion on production of CD11b⁺ BMMs was examined in vitro by immunofluorescence. Neither the percent CD11b⁺ bone marrow cells or the number of M-CSF induced BMMs in vitro was affected by FOXO1 deletion indicating that the effect of FOXO1 on osteoclastogenesis are downstream of BMM formation (P>0.05) (Fig. 7A, B)(Supplemental Fig. 4). RANK expression was examined in BM cells and BMMs from LyzMcre⁺FOXO1^L/L mice and matched control LyzMcre⁻FOXO1^L/L mice following stimulation with M-CSF (Supplemental Fig. 3B). BMMs from FOXO1 deleted mice had a ~40% reduction in RANK⁺ cells compared to BMMs from matched control mice (p<0.05) (Fig. 7C). We also examined M-CSF stimulation of BMM migration (30, 31). M-CSF induced a dose-dependent 16-fold increase in BMM migration. This increase was reduced by up to 72% when FOXO1 was deleted (p<0.05) (Fig. 7D). Similarly, FOXO1 knockdown in RAW 264.7 cells reduced migration in these cells up to 56% (p<0.05) (Fig. 7E). To determine how FOXO1 may affect BMMs migration, integrin αν was examined in BMMs based on reports that it plays an important role in migration of these cells by immunofluorescence (32, 33). FOXO1 deletion resulted in 73% decrease of integrin αν at protein level (P<0.05) (Fig. 7F).

Figure 7. FOXO1 deletion or knockdown reduces M-CSF induced RANK, integrin expression and BMM migration.

(A) BM and BMMs were prepared from LyzM.cre⁺FOXO1^L/L and littermate control LyzM.cre⁻FOXO1^L/L mice was described in Methods. The cells characterized as BMM was determined by immunofluorescence with antibody specific for CD11b and DAPI counterstain. (B) CD11b expression was measured by immunofluorescence as mean fluorescence intensity. (C) BM and BMMs from LyzM.cre⁺FOXO1^L/L and littermate control LyzM.cre⁻FOXO1^L/L mice were fixed and RANK was visualized by immunofluorescence and DAPI counterstained. (D) BMM cell migration was measured in a transwell assay in response to M-CSF (0-30ng/ml) placed in the lower chamber. (E) RAW 264.7 cells were transfected with scrambled or FOXO1 siRNA and cell migration was measured in response to M-CSF (0-30ng/ml) placed in the lower chamber. (F) The protein level of the integrin αν was measured in BMMs from LyzM.cre⁺FOXO1^L/L or littermate control LyzM.cre⁻FOXO1^L/L mice by immunofluorescence with specific antibody and the results expressed as mean fluorescence intensity. Each value represents the mean ± SEM of three independent measurements. *Significantly different compared to matched Cre⁻ or matched control group (p<0.05).

Discussion

The role of FOXO1 in regulating monocytic cells has received relatively little attention. FOXO1 has been shown to modulate TLR expression and cytokine production in monocytes and dendritic cells, which share a common precursor with osteoclasts (17, 20). We report here that FOXO1 has an important role in mediating induced osteoclastogenesis. RANKL stimulated FOXO1 expression and FOXO1 nuclear localization in vivo and in vitro. FOXO1 deletion or knockdown reduced RANKL-stimulated osteoclast formation in vivo and reduced osteoclast formation from BMMs and RAW 264.7 cells in vitro. In addition, FOXO1 deletion in myeloid cells reduced M-CSF induced RANK expression and migration of BMMs. However, FOXO1 deletion did not affect formation of BMMs in vivo or in vitro.

To investigate the role of FOXO1 on RANKL stimulated osteoclastogenesis, mice that expressed lysozyme M promoter-driven Cre recombinase (34) were bred with floxed FOXO1 mice (35). This led to substantial deletion of FOXO1 in osteoclast precursor cells, BMMs. BMMs from experimental mice had a 95% reduction of FOXO1 mRNA and 90% reduction of FOXO1 at the protein level compared to control mice. This was not a general reduction of FOXO1 in hematopoietic cells since there was no significant difference in FOXO1 levels in bone marrow cells or splenocytes from experimental and control mice.

Several lines of evidence establish that RANKL stimulates FOXO1 in osteoclast precursors. RANKL induced a time-dependent increase in FOXO1 mRNA and protein expression in both BMMs and the osteoclast precursor cell line, RAW264.7, which increased 5 to 8 fold after stimulation. In vivo RANKL stimulated an increase in FOXO1 in osteoclasts as determined by the number of FOXO1⁺TRAP⁺ double positive multi-nucleated cells. Interestingly RANKL also stimulated an increase in FOXO1 nuclear localization, which peaked at day 2 and declined on day 4, although expression was still high on day 4. This pattern was observed in both BMMs and RAW264.7 cells. FOXO1 nuclear localization two days after RANKL stimulation was shown by both immunofluorescence and western blot assays. Each of these assays was carried out with primary antibodies to FOXO1 from different sources and gave consistent results. The time frame for induction of FOXO1 nuclear localization by RANKL is considerably slower than the 1 to 2 hour induction that is observed for TNF stimulation. However, it is similar to the time frame of RANKL induced NFATc1 (8). Interestingly, both the protein expression and nuclear localization of NFATc1 are continuously increased during RANKL-induced differentiation of BMMs to osteoclasts. Since FOXO1 activity is directly related to its nuclear localization, the functional significance of high FOXO1 expression but exclusion from the nucleus on day 4 requires further investigation. Pathways that are known to mediate RANKL induced NFATc1 also were shown to in involved in FOXO1 nuclear localization (36-39). Thus, FOXO1 nuclear localization was significantly reduced by inhibitors of MAP kinase, ROS, NOS and deacetylase. In contrast, Akt which is known to negatively regulate FOXO1 did not have a significant effect on RANKL induced FOXO1 nuclear localization two days after stimulation.

That FOXO1 had a significant effect in mediating RANKL stimulation was shown by significantly reduced osteoclast formation in vivo and reduced total resorption pit area in experimental mice compared to matched control mice. The effect of FOXO1 deletion on reducing RANKL stimulated osteoclast formation from BMMs or RAW264.7 cells in vitro was similar to in vivo results. In addition the size and number of nuclei per osteoclast was significantly smaller when FOXO1 was deleted or silenced suggesting that the amount of cell fusion was diminished. Osteoclast resorbing activity was also reduced when FOXO1 was deleted in BMMs or silenced in RAW264.7 cells in vitro. We also compared cancellous bone volume and several aspects of bone architecture in vertebrae of LyzM.Cre⁺FOXO1^L/L mice versus control littermates. However there was no significant difference between the two groups suggesting that FOXO1 deletion alone does not have a major impact in physiologic bone remodeling in young mice.

Several downstream targets were identified to investigate mechanisms through which FOXO1 may affect osteoclastogenesis. These studies investigated whether FOXO1 mediated the expression of gene targets that are induced by RANKL stimulation (1, 40). NFATc1, ATP6v0d2 and DC-STAMP expression and nuclear localization of NFATc1 were reduced by FOXO1 deletion. NFATc1 is a master regulator of osteoclastogenesis transducing the effects of RANKL induced fusion of precursor cells (1, 40). Thus, FOXO1 may impact cell fusion through downstream regulation of NFATc1, which in turn regulates DC-STAMP and ATP6v0d2 (7). FOXO1 also modulated mRNA levels of genes that are needed for osteoclast function including cathepsin K and ATP6v0d2, which are responsible for dissolution of the mineral phase and proteolytic degradation of bone matrix proteins (6). The latter are consistent with our findings that FOXO1 deletion or knockdown reduced osteoclast activity stimulated by RANKL in vitro. Moreover, the linkage between NFATc1 and FOXO1 is supported by in vivo results that FOXO1 deletion reduces NFATc1 expression in osteoclasts following RANKL injection. However a luciferase reporter assay suggests that NFATc1 promoter activity is not directly regulated by FOXO1, which may indicate that other mechanisms that are involved.

FOXO1 deletion did not affect the formation of osteoclast precursors but did affect M-CSF induced BMM migration and M-CSF induced integrin and RANK expression. RANK, integrin αν and BMM migration are important in the formation of osteoclasts from BMM osteoclast precursors (32, 33). Thus, FOXO1 may also influence the formation of osteoclasts by mediating M-CSF stimulation of BMM.

Bartell and co-workers reported that RANKL reduced FOXO1 promoter activity at 16 hours in vitro and that simultaneous FOXO1, -3, and -4 deletion in vivo increased osteoclastogenesis and bone resorption (21). This report concluded that FOXO1, -3 and 4 acted to inhibit bone loss by inducing catalase production, which acted to reduce formation of H₂O₂. In contrast we focused on FOXO1 alone and downstream events that occur in a later time frame, two days after stimulation when osteoclast precursors fuse to form multi-nucleated cells. Our results demonstrate that FOXO1 mediates several events associated with RANKL induced osteoclastogenesis in vivo and in vitro and also affects events downstream of M-CSF including RANK and integrin αv expression as well as M-CSF induced migration of bone marrow macrophages. Thus, we have shown here that under direct stimulation FOXO1 promotes M-CSF/RANKL induced osteoclastogenesis whereas the net long term indirect effect of FOXO1, -3 and -4, as shown by Bartell and colleagues is to reduce osteoclastogenesis associated with aging by attenuating oxidative stress.