The G9a histone methyltransferase represses osteoclastogenesis and bone resorption by regulating NFATc1 function

2.1. Reagents

Human M‐CSF (Leukoprol) and soluble human recombinant RANKL (sRANKL) were purchased from Kyowa Kirin (Tokyo, Japan) and Oriental Yeast (Tokyo, Japan), respectively.

The following antibodies were used: rabbit polyclonal anti‐human NFATc1 (Abcam, 25961) and mouse monoclonal anti‐G9a (PP‐A8602A‐00; Perseus Proteomics, Tokyo, Japan). Mouse monoclonal anti‐H3K9me2 antibodies were prepared by H. Kimura. ¹⁷ Following secondary antibodies were obtained from Invitrogen, Carlsbad, CA, USA: Goat Alexa 488‐labeled anti‐mouse IgG(H+L) (A11001), goat Alexa 405‐labeled anti‐rabbit IgG(H+L) (A31556), and goat Alexa 594‐labeled anti‐mouse IgG(H+L) (A11032). All chemical‐grade reagents were obtained from WAKO Chemicals (Osaka, Japan).

2.2. Mice

To generate G9a ^fl/fl (Ehmt2 ^fl/fl ) mice, G9a deleted in ES cells (TT2 cell line) were injected into ICR mouse embryos. ¹⁸ G9a ^fl/+ mice were sequentially backcrossed with the C57BL/6 background mice, and offspring later than F15 generation were used for further studies. To generate Rank ^Cre/+ (Tnfrsf11a ^Cre/+ ) mice, an eGFP‐fused Cre recombinase cassette was inserted after the STOP codon of Tnfrsf11a (RANK‐encoding gene) followed by an IRES stuffer. ¹⁹ This Rank ^Cre/+ line was backcrossed to C57BL/6 background mice for more than 10 generations. G9a cKO (osteoclast lineage‐specific G9a gene deletion: G9a ^fl/fl ;RANK ^Cre/+ ) and control (G9a ^+/+ ;RANK ^Cre/+ ) mice were generated by crossing G9a ^fl/+ ;RANK ^Cre/+ parents. In vivo experiments were conducted on 10‐week‐old mice. All animal experiments were approved by the Institutional Animal Care Committee, the Recombination Experiment and Biosafety Committee of the Tsurumi University School of Dental Medicine, and the Animal Management Committee of Matsumoto Dental University.

2.3. In vitro osteoclastogenesis

Mouse bone marrow macrophages (BMMs) were prepared as described previously. ²⁰ Briefly, bone marrow cells were isolated from the tibiae of 4‐ to 44‐week‐old male (n = 15) and 16‐ and 44‐week‐old female G9a ^fl/fl mice (n = 2) and from C57BL/6 WT mice of similar age. Bone marrow cells from G9a ^fl/fl and WT mice were cultured for 16 h in alpha‐modified minimal essential medium (αMEM; Life Technologies, Tokyo, Japan) containing 10% fetal bovine serum (FBS), 1% antibiotics (100 mg/mL penicillin and 100 IU/mL streptomycin; Sigma, St. Louis, MO, USA), and 30 ng/mL M‐CSF (Figure 1). Nonadherent cells were collected and used as BMMs. BMMs (1.5 × 10⁵ cells/well) were cultured for 2 or 3 days in the presence of 50 ng/mL M‐CSF in 48‐well culture plates. After removing nonadherent cells by pipetting, adherent BMMs from G9a ^fl/fl and WT mice were used as osteoclast precursors and infected with an adenovirus expression vector‐harboring Cre recombinase at a multiplicity of infection (MOI = 100) for 24 h (Figure 1), as described previously. ²¹ The cells were further cultured for 4 days in the presence of 50 ng/mL M‐CSF and 200 ng/mL sRANKL for the osteoclast formation assay (Figure 1).

2.4. TRAP staining

Cells were fixed with 3.7% neutrally buffered formaldehyde solution for 10 min. The cells were then rinsed with Tris‐buffered saline (TBS) three times, and the fixed cells were permeabilized with a 1:1 solution of acetone and ethanol for 1 min and stained using a TRAP staining kit (Cosmo Bio, Tokyo, Japan). TRAP‐positive multinuclear cells (MNCs) (>3 nuclei/cell) were considered as mature osteoclasts.

2.5. Preparation of adenovirus vectors

The production of adenovirus was performed as described previously ²² according to the manufacturer's protocol (ViraPower Adenoviral Expression System; Thermo Fisher Scientific). An adenovirus vector expressing Cre‐recombinase (AxCANCre; Ad/Cre) was obtained from RIKEN BRC (Tsukuba, Japan). An adenovirus vector expressing enhanced green fluorescent protein (EGFP) (Ad/GFP), which contains CAG‐IRES‐EGFP, was prepared as described. ²² Another adenovirus vector expressing G9a was constructed as follows: a fragment of CAG‐G9a was subcloned into the pENTR vector (Thermo Fisher Scientific). Then, pENTR‐CAG‐G9a‐IRES‐EGFP was subcloned into pAd/PL‐DEST by a site‐specific recombination.

2.6. Bone resorption assay (pit formation assay)

BMMs were further cultured with M‐CSF and sRANKL for 4–6 days in 60 mm RepCell plates (temperature‐responsive polymer‐coated plates; CellSeed Inc., Tokyo, Japan) to allow the development of multinucleated osteoclasts. Osteoclasts were harvested and seeded at approximately 100 osteoclasts per 4‐mm‐square dentin slice in 48‐well plates with 0.5 mL medium supplemented with M‐CSF and sRANKL for 48 h. Dentin slices were stained with TRAP to confirm the presence of osteoclasts. After the removal of cells from the surface, these slices were re‐stained with toluidine blue to detect resorption pits. The stained pit images were binarized, and the resorption area was measured using the ImageJ software (NIH).

2.7. Reverse transcription and real‐time PCR

RNA was isolated using an RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. RNA was reverse transcribed using oligo‐dT primers and reverse transcriptase (Superscript III; Invitrogen) to synthesize cDNA. After cDNA synthesis, qPCR was performed as previously described. ²² Three or more independent cultures were subjected to qPCR and representative results are shown. The primer sequences used for each PCR experiment are listed in Table 1.

2.8. Immunofluorescence staining

BMMs were cultured on coverslips or chamber slides for osteoclast formation as described above. The cells were fixed with 3.7% neutrally buffered formaldehyde solution for 10 min. After rinsing with TBS three times, the fixed cells were permeabilized with a 1:1 solution of acetone and ethanol for 1 min and then blocked with 5% goat serum in TBS. The cells were then incubated with mouse anti‐G9a, rabbit anti‐NFATc1, or mouse anti‐H3K9me2 ¹⁷ antibodies, with the appropriate Alexa Fluor‐conjugated secondary antibody, and counterstained with DAPI.

NFATc1‐positive nucleus was defined as being its positive immunofluorescent nuclear area occupying more than 50% of its nucleus area. The image thresholding method of Image J (version 1.53k, NIH, USA) was used to count NFATc1‐positive nuclei in fluorescent images and the total number of nuclei in DIC images, respectively. Then, the percentage of NFATc1‐positive nuclei to total nuclei of multinucleated osteoclasts was calculated.

2.9. Chromatin immunoprecipitation (ChIP) assay

The ChIP assay was performed as previously described. ⁷ BMMs were cultured at 4 × 10⁶ cells per 6 cm tissue culture dish and treated with 1% formaldehyde for 10 min at room temperature. The crosslinked chromatin was sheared to an average length of 200–1000 bp by sonication with a sonicator (Bioruptor UCD‐200; Cosmobio, Tokyo, Japan). The precleaned chromatin was used as the input control. Immunoprecipitation was carried out overnight at 4°C against the following antibodies: NFATc1 (4 μg, sc‐7294X, Santa Cruz, USA), G9a (1 μg, 68851, Cell Signaling Technology, MA, USA), normal mouse IgG (4 μg, 015‐000‐003, Jackson ImmunoResearch, PA, USA), and normal rabbit IgG (1 μg, ab46540, Abcam, Cambridge, UK). Immunoprecipitation was performed using Dynabeads M‐280 sheep anti‐mouse IgG (11202D, Thermo Fisher Scientific) and Dynabeads M‐280 sheep anti‐rabbit IgG (11204D, Thermo Fisher Scientific). After repeated rinses, protein–DNA cross‐linking was reversed, and the precipitated DNA was treated with RNAse at 37°C and with proteinase K at 55°C, and then purified by phenol–chloroform extraction. DNA was analyzed by real‐time PCR (qPCR). Data obtained for each primer pair were expressed as a percentage of the input DNA. Primer sets used for the mouse Ctsk promoters ²³ are as follows:

N1 site, 5′‐CCCCCAAAGTCAGTCAGATG‐3′ and 5′‐GGTAAGGATTGCGGAAGTCA‐3′; N3 site, 5′‐CTAGCCTTTCCTCCCCTCTC‐3′ and 5′‐CCCTAGTTGTCTCCATTCTTCC‐3′.

2.10. Microcomputed tomography (μCT) measurement

X‐ray images of the distal end of the femora were obtained using the μCT scanner, ScanXmate‐A080 (Comscantecno, Kanagawa, Japan), with the following parameters: 22 kV and 250 μA. Three‐dimensional images reconstructed from the multi‐slice images, which were taken at the ROI of 1–3 mm (20 μm slices) away from the growth plate, were analyzed using the TRI/3D‐BON software (Ratoc System Engineering, Tokyo, Japan). For cortical bone analysis, images of horizontal cross‐sections 3 mm away from the femoral growth plate were analyzed by ImageJ.

2.11. Bone histomorphometry

Left femora of 10‐week‐old mice were fixed overnight in 4% paraformaldehyde and soaked in a series of sucrose concentrations up to a maximum of 100% prior to mounting in OTC compound in a hexane cooling bath at −70°. Ten μm serial sections were prepared using a cryotome (CM‐300S, Leica, Germany) and stained for tartrate‐resistant acid phosphatase (TRAP) to identify osteoclast cells. The sections were counterstained with methyl green. Images were captured using a microscope (BZ‐X710, Keyence, Osaka, Japan). Osteoclast surface (Oc.S/BS) and osteoclast number (N.Oc/BS) were measured on the trabecular bone in the ROI area (0.75 mm square), 100 μm away from the growth plate cartilage, by tracing cell bodies and contact surfaces on the digitizer in images of tissue sections using ImageJ. Bone measurement parameters were described according to the report of the ASBMR Histomorphometry Nomenclature Committee. ²⁴

2.12. Measurement of serum CTX and osteocalcin

The cross‐linked C‐telopeptide of type I collagen (CTX) in the serum, which reflects degraded collagen derived from resorbed bone, was measured using a RatLap ELISA kit (AC‐06F1, Immunodiagnostic Systems, UK) according to the manufacturer's instructions. Serum Gla‐type osteocalcin levels, an indicator of bone formation, were measured by EIA method according to the manual for the Mouse Gla‐Osteocalcin High Sensitive EIA Kit (MK127, Takara‐Bio, Kyoto, Japan).

2.13. Statistics

The mean values of measures in the two groups of gene expressions, TRAP‐positive MNCs numbers, ratio of NFATC1‐positive nuclei number/total nuclei number of MNOC, and CHIP assay were compared using Student's t‐test with equal variances or Welch test with unequal variances. The mean values in the three or more groups of gene expressions were examined using a one way of variance (ANOVA) and a post hoc test (Scheffé method). Those measures are expressed as the mean and standard deviations (SD).

The statistical differences between the two groups of bone morphometric and pit size data were examined using Mann–Whitney U test. The former and latter measures were expressed as box‐and‐whisker diagrams and mean ± 1 SD, respectively.

3.1. G9a deletion in BMMs promoted osteoclast differentiation and G9a overexpression reversed it

First, we examined G9a expression during osteoclastic differentiation. We induced osteoclast differentiation in wild‐type (WT) BMMs in vitro by treatment with M‐CSF and RANKL. G9a mRNA expression slightly diminished after the addition of RANKL, while mRNA for osteoclast‐specific genes, Ctsk, Acp5, and Nfatc1, increased during osteoclastogenesis in vitro (Figure 1A).

To investigate the function of G9a in osteoclastic differentiation, we deleted G9a in BMMs in vitro. BMMs derived from G9a ^fl/fl mice were cultured with M‐CSF, and an adenovirus expressing Cre (Ad/Cre) was used to infect these cells 24 h before the addition of RANKL (see the experimental outline in Figure S1). We confirmed that G9a expression levels were low in BMMs at 2 days (mRNA) and 4 days (mRNA and protein) after RANKL treatment (Figure 1B,D) in G9a‐deleted cells. TRAP staining revealed that the numbers of TRAP‐positive MNCs increased after G9a deletion (Figure 1C). Furthermore, the mRNA expression of Acp5 and Ctsk, and to a lesser extent, Nfatc1, was increased in G9a KO cells (Figure 1D). Thus, G9a deletion in BMMs generated more mature osteoclasts in vitro. These effects were observed in BMMs derived from female mice (Figure S2) and male mice.

We reintroduced G9a cDNA using a recombinant adenovirus into G9a‐deleted BMMs. Ad/GFP‐expressing cells were used as the control. To create G9a‐deleted BMMs, Ad/Cre was infected into BMMs isolated from G9a ^fl/fl mice. Then, Ad/G9a was reintroduced into G9a‐deleted BMMs (Ad/Cre and Ad/G9a) to rescue G9a deletion. The reintroduction of G9a suppressed the increase in TRAP‐positive MNCs in G9a‐deleted BMMs (Figure 1E(a)). We confirmed that G9a expression was inhibited by Ad/Cre and restored by Ad/Cre and Ad/G9a infection (Figure 1E(b)).

3.2. Osteoclasts lacking G9a showed reduced H3K9me2 levels

One of the major substrates of G9a is H3K9me2. Next, we investigated the levels of H3K9me2 in G9a‐deleted and non‐deleted BMMs. Immunofluorescent staining of H3K9me2 was uniformly present in macrophage nuclei before induction of osteoclast differentiation (Figure 2A, rightmost panel). On day 2 when osteoclast differentiation was induced by addition of RANKL, H3K9me2 staining intensity was reduced in G9a‐deleted BMMs compared to WT BMMs (Figure 2B, right‐most panels). On day 4 after induction of differentiation, H3K9me2 was localized in the nuclei of WT mononuclear and multinuclear cells, whereas its localization was markedly reduced in G9a‐deleted osteoclasts (Figure 2C, rightmost panels). Quantification of H3K9me2 fluorescence intensity per DAPI‐positive nucleus showed significant reduction in G9a‐deleted cells on days 2 and 4 after differentiation (Figure 2D). Thus, the increased osteoclast differentiation due to G9a deletion was accompanied by a decrease in H3K9me2 level.

3.3. G9a suppressed NFATc1 activity

To address the molecular mechanisms underlying the suppressive roles of G9a in osteoclasts, we investigated the effect of G9a on the function of a key transcription factor of osteoclastogenesis, NFATc1. Because nuclear import is critical for NFATc1 function, we examined NFATc1 localization using immunofluorescent staining. We found that the nuclear localization of NFATc1 protein in multinucleated osteoclasts increased in G9a‐deleted osteoclasts (Figure 3A). We further analyzed the binding of NFATc1 to target regions in the genome. The ChIP experiment confirmed that the binding of NFATc1 to two Ctsk promoter regions, N1 and N3, increased after RANKL‐induced osteoclastic differentiation (data not shown). The binding of NFATc1 to N1 and N3 was further enhanced by G9a deletion (Figure 3B). Enhanced binding to the promoter was more significant in N3 than in N1. These data suggest that G9a suppressed NFATc1 function in the nucleus.

3.4. G9a cKO mice showed increased bone resorption

Since in vitro experiments revealed promotion of osteoclast differentiation by G9a deletion, we investigated the effects of G9a deletion on bone resorption in vivo. G9a cKO mice were generated by crossing G9a ^fl/fl mice with RANK‐Cre (Rank ^Cre/+ ) mice, in which G9a was deleted from the osteoclast lineage. Micro‐CT analysis of the femur of 10‐week‐old female mice revealed reduced trabecular bone volume in G9a cKO mice as compared to that in WT mice (Figure 4A). G9a cKO mice also displayed reduced bone volume per tissue volume (BV/TV), bone surface per tissue volume (BS/TV), and number of trabecular bones (Tb.N) (Figure 4B). On the other hand, cortical bone area (Ct. Ar) and bone marrow area (Ma. Ar) were unchanged in horizontal cross sections of the femur from control and cKO female mice (Figure S3). No significant differences were evident in the outer and inner circumference length of the cortical bone (Ct.Pm and Ec.Pm, respectively) between the control and cKO mice (Figure S3). Thus, G9a cKO mice demonstrated accelerated bone loss in trabecular bone.

Serum levels of cross‐linked C‐telopeptide of type I collagen (CTx), a bone resorption marker, were increased twofold in G9a cKO mice as compared to that in control mice (Figure 4C, upper panel). In contrast, no significant change was observed in serum levels of osteocalcin, a marker of bone formation (Figure 4C, lower panel). We further analyzed the number of osteoclasts using sagittal sections of the distal femur from 10‐week‐old control and G9a cKO female mice. The number of osteoclasts per bone surface (N.Oc/BS) and the surface area of osteoclasts per bone surface (Oc.S/BS) were not significantly different between the two mouse groups (Figure 4D). These results suggest that bone resorption is enhanced in G9a cKO mice, resulting in bone loss. Increased bone resorption in the trabecular bone was observed in female mice but not in male mice (Figure S4A). Additionally, serum levels of cross‐linked C‐telopeptide of type I collagen (CTx) and osteocalcin were also not elevated in male mice (Figure S4B).

3.5. Osteoclasts from G9a cKO mice showed increased bone‐resorbing activity

In accordance with unchanged number of osteoclasts in G9a cKO mice, there was no difference in the number of TRAP⁺ MNCs differentiated from BMMs in G9a‐cKO and control mice (Figure S5). However, more heavily stained TRAP‐positive cells were observed in G9a cKO mice BMMs than in control mice BMM. Therefore, we examined the bone‐resorbing activity of RANKL‐induced osteoclasts in G9a cKO mice. To investigate the bone‐resorbing activity, we performed a pit formation assay by seeding the same number of TRAP‐positive osteoclasts onto dentin slices (Figure 5A). The pit area per slice and the number of pits increased in osteoclasts derived from G9a cKO than in control mice (Figure 5B,C). These results suggest that G9a cKO BMMs treated with RANKL generated multinucleated TRAP‐positive osteoclasts with increased bone‐resorbing activity.