Nocodazole (#M1404) was purchased from Sigma Aldrich and Y27632 from MCE (#HY-10071).

Mice sacrifice and bone marrow harvest were performed in compliance with local animal welfare laws, guidelines and policies, according to the rules of the regional ethical committee.

Primary osteoclasts were obtained from 6- to 8-week-old C57BL/6J mice as previously described (Morel et al., 2018). Briefly, bone marrow macrophages (BMM) were obtained from long bones by growing non-adherent cells for 48 h in αMEM containing 10% heat-inactivated foetal calf serum (Biowest), 2 mM glutamine, 100 U/mL penicillin–streptomycin and 30 ng/ml M-CSF (Miltenyi, #130-101-703). Osteoclasts were then differentiated by culturing BMM in the same medium supplemented with 50 ng/mL RANKL (Miltenyi, #130-094-076) for 4–6 days.

A guide RNA (gRNA) targeting exon 13 of mouse GEF-H1 gene Arhgef2 (5′-AGG​ATA​AGG​CGT​ATC​TCC​GGA​GG-3′) was designed and cloned in lentiCRISPRv2 vector, which also provides the expression of the Streptococcus pyogenes Cas9 and puromycin resistance, a gift from Feng Zhang (Addgene plasmid # 52961; http://n2t.net/addgene:52961; RRID:Addgene_52961) (Sanjana and Shalem, 2014). Empty and GEF-H1 gRNA-containing lentiCRISPRv2 were used to produce lentiviruses and generate respectively control (CTL) and GEF-H1 Knock-Out (KO) RAW264.7 cells. Briefly, growing RAW264.7 cells, a gift from Kevin P McHugh (Gainesville, FL, United States), were infected with lentiviral particles and selected with 3 mg/mL puromycin 48 h later. Puromycin resistant CTL and GEF-H1 KO RAW264.7 clones were individually picked and expanded. GEF-H1 expression was monitored by immunoblot analysis to select clones with reduced GEF-H1 expression.

Wild-type and CRISPR/Cas9 modified RAW264.7 cells were grown in DMEM containing 10% heat-inactivated foetal calf serum (Eurobio) with 2 mM glutamine and 100 U/mL penicillin–streptomycin. For osteoclast differentiation, they were seeded at 5 × 10⁴ cells/well (6-well plate) or 8 × 10³ cells/well (24-well plate) in αMEM containing 10% heat-inactivated fetal calf serum (Biowest), 2 mM glutamine, 100 U/mL penicillin–streptomycin and 50 ng/mL of RANKL (Miltenyi, #130-094-076) for 3–4 days. Since pure GEF-H1 KO clones did not differentiate, 75% GEF-H1 KO clones had to be mixed with 25% CTL RAW264.7 cells to fuse and form GEF-H1 knock-down (KD) osteoclasts. Osteoclasts were imaged with an EVOS FL microscope equipped with a Sony ICX445 CCD camera.

The ACC preparation protocol was simplified from (Saltel et al., 2004). Briefly, 6-well plates or 13 mm diameter glass coverslips were coated with 50 μg/mL calf skin type I collagen (Sigma, #C9791) in 20 mM acetic acid, incubated for 1 h at 37°C and dried overnight. Then supports were successively incubated in (1) 200 mM Tris-buffered saline (TBS) pH 9 containing 0.13 mg/mL egg yolk phosvitin (Sigma, #P1253), 0.13 mg/mL alkaline phosphatase (Sigma, #P764) and 1 mg/mL dimethyl suberimidate dihydrochloride (Sigma, #179523) as a cross-linking reagent during 24 h at 37°C, (2) 6 mM calcium β-glycerophosphate (Sigma, #G6626) for 48 h at 37°C and (3) washed with 200 mM Tris pH 9. The last 3 steps were repeated 3–4 times depending on the amount of precipitated calcium phosphate. Next, the supports were rinsed with distilled water and air dried. Osteoclasts at day 3 of differentiation were detached with Accutase (Sigma, #A6964), scrapped, seeded and grown for 2 more days onto ACC.

The assay was performed as previously described (Ito et al., 2017). Cells were lysed with a microtubule stabilizing buffer (100 mM PIPES pH 6.8, 1 mM MgSO4, 1 mM EDTA, 2 M glycerol, 0.1% (w/v) Triton X-100 and protease inhibitor cocktail) for 20 min at 37°C and centrifuged at room temperature for 5 min at 16,000 g. The supernatant was collected as the soluble fraction. The pellet was lysed with a whole cell lysis buffer (10 mM Tris-HCl pH 7.5, 2 mM EDTA, 1% SDS and protease inhibitor cocktail), boiled for 15 min, further centrifuged at room temperature for 5 min at 20,000 g and used as the insoluble fraction.

The RHOA-binding domain of Rhotekin (RBD) pulldown assay was used to detect cellular GTP bound RHOA. In brief, osteoclasts were washed with cold TBS and lysed in a cold buffer containing 50 mM Tris-HCl pH 7.5, 1% triton, 500 mM NaCl, 10 mM MgCl₂, 1 mM DTT and protease inhibitor cocktail. After centrifugation at 13,000 g for 2 min at 4°C, part of the supernatant was stored for total RHOA determination. The remaining supernatant was incubated with 66 µg of GST-RBD-coupled gluthatione sepharose beads (cytoskeleton, #RT02) for 45 min at 4°C. The beads were washed 3 times with a cold buffer containing 50 mM Tris-HCl pH 7.5, 0.5% triton, 150 mM NaCl, 10 mM MgCl₂, 1 mM DTT and protease inhibitor cocktail. Total and active GTP-bound RHOA were detected by Western blotting.

Whole cell extracts were prepared in Laemmli sample buffer, resolved on SDS-PAGE and electrotransferred on PVDF membranes (Millipore, #IPFL00010). Immunoblotting was performed using the following primary antibodies: rabbit anti-GEF-H1 (Abcam, #ab155785, 1/500), mouse anti-alpha tubulin (Sigma, #T6074, 1/1000), rabbit anti-GAPDH (Cell Signaling, #2118, 1/2000), rabbit anti-histone H3 (Abcam, #1791, 1/5000) and mouse anti-RHOA (Santa Cruz, #sc-418, 1/500). Signals were revealed with Dylight 680 or 800 conjugated secondary antibodies (Invitrogen) using the Odyssey Infrared Imaging system and then quantified with Image Studio software (LI-COR).

Osteoclasts on 13-mm diameter glass or ACC-coated coverslips were either fixed for 20 min in 3.2% paraformaldehyde in PHEM (60 mM Pipes, 25 mM Hepes, 10 mM EGTA, 4 mM MgSO4, pH 6.9), permeabilized with 0.1% Triton X100 for 1 min and blocked with 1% BSA in PBS for 15 min or permeabilized and blocked with PBS - 2% BSA - 0.2% triton X100 for 1 h (GEF-H1 staining). Then osteoclasts were incubated for 1 h with primary antibodies: rabbit anti-GEF-H1 (Abcam, #ab155785, 1/200), mouse anti-alpha tubulin (Sigma, #T-5168, 1/2000) and/or Alexa 564-phalloidin (Life Technologies, #A22283, 1/1000). Signal was revealed with the adapted Alexa Fluor 488-, 546 or 647-conjugated secondary antibodies (Life Technologies, 1/1000). Preparations were mounted in Citifluor mounting medium (Biovalley) and imaged with Leica SP5-SMD confocal microscope using 40X HCX Plan Apo CS oil 1.3NA or 63X HCX Plan Apo CS oil 1.4NA objectives. Co-localization was shown using ImageJ pluggin DiAna (Gilles et al., 2017).

Mineral dissolution activity of osteoclasts was measured as described (Morel et al., 2018). Briefly, at day 3 of differentiation, osteoclasts were rinsed once in PBS, detached with Accutase for 5–10 min at 37°C, scrapped, seeded and grown for 2 more days onto inorganic crystalline calcium phosphate (CaP)-coated multiwells (Osteo Assay Surface, Corning). In each experiment, four wells were stained for Tartrate Resistant Acid Phosphatase (TRAP) activity to count osteoclasts and four wells with Von Kossa stain to measure CaP dissolution. They were imaged with an EVOS FL microscope equipped with a Sony ICX445 CCD camera and quantification of osteoclasts and resorbed areas were done with ImageJ 1.53w software. Osteoclast specific activity was expressed as the average area resorbed in the wells stained with von Kossa normalized by the average area of osteoclasts in the wells stained with TRAP.

All analyses were performed using GraphPad Prism 5.01 (GraphPad Software, Inc., La Jolla, CA). All data are presented as the mean ±sem; p values <0.05 were considered statistically significant.

In osteoclasts, stabilization of the unique podosomes organization by microtubules is instrumental for bone resorption function (Blangy et al., 2020). Microtubule depolymerization with nocodazole causes the rapid collapse of the adhesion structure (Destaing et al., 2003), but the underlying mechanisms linking actin cytoskeleton and microtubules in osteoclasts remain largely unknown. It was shown that a tight control of RHOA activity is required for the unique patterning of podosomes in osteoclast (Touaitahuata et al., 2014) and that RHOA inhibition by TAT-C3 prevented the destabilization of the podosome belt by nocodazole (Destaing et al., 2005). Since the depolymerization of microtubules increased RHOA activity in various cell types (Kee et al., 2002) (Chang et al., 2006; Chang et al., 2008; Ito et al., 2017; Takesono et al., 2010), we tested whether it was also the case in osteoclasts. To do so, we treated mouse bone marrow macrophages (BMM)-derived osteoclasts with nocodazole and performed RHOA pull-down assay to monitor the activity of the GTPase. When osteoclasts were seeded on plastic, we observed a significant 1.5-fold increase in RHOA activity after 1-hour nocodazole treatment, as compared to the control (Figure 1A). Moreover, when osteoclasts were seeded on mineralized apatite collage complex (ACC), RHOA basal activity was higher and increased by 3-fold after the same nocodazole treatment (Figure 1B). Thus, the depolymerization of microtubules leads to the activation of RHOA in osteoclasts.

To confirm that the nocodazole-induced disorganization of osteoclast actin cytoskeleton was mediated by RHOA activation, we used Y-27632, an inhibitor of the ROCK kinase, a major effector of RHOA downstream signaling. Osteoclasts plated on glass do not form a bona fide sealing zone but a thinner peripheral structure called the podosome belt, which is characterized by a lower density of podosomes (Luxenburg et al., 2007). Treatment of osteoclasts with Y-27632 alone did not affect podosome organization (Figures 2A, B). More interestingly, Y-27632 was able to prevent the nocodazole-induced disorganization of the podosome belt (Figures 2A, B). Thus, RHOA activation upon microtubule depolymerization in osteoclasts contributes to the disorganization of the podosomes, via the activation of ROCK kinase.

Overall, these results suggest that microtubules have a key role in osteoclasts for the maintenance of actin cytoskeleton organization thought the modulation of the activity of the GTPase RHOA.

We reported previously SILAC proteomic and RNAseq transcriptomic data showing that the mRNA of the RHOA exchange factor GEF-H1 was expressed at high levels in osteoclasts, with comparable levels of mRNA and proteins in BMM and osteoclasts (Guérit et al., 2020). We also reported proteomic data indicating that GEF-H1 is present in microtubule-associated protein enriched fraction of osteoclasts derived from RAW264.7 cells (Maurin et al., 2021). In various epithelial cancer cell lines, it was shown that nocodazole treatment induced the activation of RHOA due to the release GEF-H1 from the microtubules to which it is basally anchored and held in an inactive state (Meiri et al., 2012). This suggests that GEF-H1 could be associated with microtubules in BMM-derived osteoclasts and mediate the activation of RHOA upon microtubule depolymerization.

We confirmed by Western blot on total cell extracts that GEF-H1 was expressed at comparable levels in BMM and osteoclasts (Figure 3A). We then examined whether GEF-H1 was indeed associated with osteoclast microtubules. First, we performed a microtubule sedimentation assays on BMM-derived osteoclasts in the absence or presence of nocodazole. In control conditions, GEF-H1 was predominantly in the microtubule-containing insoluble fraction of osteoclasts; in contrast, the protein mostly shifted in the soluble fraction, when the fractionation was performed in the presence of nocodazole, similar to the behavior of β-tubulin used as a marker of microtubule depolymerization (Figures 3B, C). These results are consistent with the hypothesis that GEF-H1 associates with osteoclast microtubules, and thus could be released upon nocodazole treatment. By confocal fluorescent microscopy, we confirmed that GEF-H1 colocalizes with microtubules (Figures 3D–F).

These data show that the RHOA exchange factor GEF-H1 associates with osteoclast microtubules; they also suggest that the activation of RHOA we observed upon osteoclast microtubule depolymerization could be mediated by GEF-H1 release from its inhibitory association with microtubules.

To be able to examine the role of GEF-H1 in osteoclasts and its regulatory action on the activity of the GTPase RHOA, we generated a GEF-H1 knock-out model by CRISPR/Cas9-mediated genome editing of mouse monocytic RAW264.7 cell line, using a guide RNA to target Arhgef2 exon 12. As a control, RAW264.7 cells were treated the same with an empty guide RNA vector. We selected two clones based on GEF-H1 expression level: on one hand, clone KO#5 presented a one-nucleotide frameshift insertion or deletion in each allele, resulting in no expression of GEF-H1; on the other hand, clone KO#26 presented a 169 out of frame deletion on one allele, resulting in reduced levels of GEF-H1, as compared to CTL cells (Supplementary Figure S1). Incubation of the cells with RANKL revealed that neither clone #5 nor clone #26 was able to form multinucleated cells, in contrast with CTL RAW264.7 cells. Mixing 75% of KO#5 or KO#26 with 25% CTL cells rescued multinucleated cell formation (Supplementary Figures S2A, B), leading to normal-sized osteoclasts as compared to CTL (Supplementary Figure S2C). These data show that minimal levels of GEF-H1 are required for the differentiation of osteoclasts. From now on, knock-down osteoclasts expressing lower levels of GEF-H1 compared to CTL, KD#5 and KD#26, were studied after differentiation of a mixture of 75% KO and 25% CTL RAW264.7 cells.

To investigate the implication of GEF-H1 in the activation of RHOA upon osteoclast microtubule depolymerization, we compared the activation of the GTPase in response to nocodazole in CTL and GEF-H1 KD osteoclasts. Similar to BMM-derived osteoclasts, treatment of RAW264.7-derived CTL osteoclasts seeded on ACC with nocodazole led to a strong increase in the levels of active RHOA (Figure 4A). In contrast, the levels of active RHOA did not change in response to nocodazole in KD#5 and in KD#26 osteoclasts (Figure 4A). This shows that GEF-H1 plays a key role in the activation of RHOA upon osteoclast microtubule depolymerization. Since fine tuning of RHOA activity is critical for the organization of osteoclast actin cytoskeleton, we also examined the role of GEF-H1 in the formation and maintenance of sealing zones. Interestingly, in basal conditions, the proportion of osteoclasts presenting a sealing zone was comparable between CTL, KD#5 and KD#26 osteoclasts (Figure 4B). But in contrast with CTL osteoclasts, the sealing zones in KD#5 and KD#26 osteoclasts were not destabilized upon microtubule depolymerization by nocodazole (Figure 4B).

Overall, these data suggest that whereas GEF-H1 is required to the formation of osteoclasts, its sequestering on microtubules is necessary to ensure the optimal activity levels of RHOA and the maintenance of the sealing zone.

The above data show that GEF-H1 contributes to osteoclast differentiation and to the regulation of their actin cytoskeleton. As osteoclast cytoskeleton is key for their bone resorption activity (Blangy et al., 2020), we further investigated how GEF-H1 influences this function. First, we observed that KD#5 and KD#26 osteoclasts on ACC exhibited smaller sealing zone size as compared to CTL osteoclasts (Figures 5A, B), despite the fact that the proportion of osteoclasts with a sealing zone was similar in CTL and KD osteoclasts (Figure 4B). Thus, notwithstanding it destabilizes the sealing zone when it is released from microtubules (Figure 4B), GEF-H1 is also required for the optimal assembly and/or stability of this structure. To confirm the functional relevance of this observation, we examined the effect of GEF-H1 on the resorption activity of osteoclasts. Consistent with the reduction in sealing zone size, we found that the resorption activity of GEF-H1 KD osteoclast was markedly reduced as compared to CTL (Figure 5C; Supplementary Figure S3).

These results suggest that GEF-H1 has to be finely tuned for optimal osteoclast activity.