Osteoclast fusion and bone loss are restricted by interferon inducible guanylate binding proteins

Chronic inflammation during many diseases is associated with bone loss. While interferons (IFNs) are often inhibitory to osteoclast formation, the complex role that IFN and interferon-stimulated genes (ISGs) play in osteoimmunology during inflammatory diseases is still poorly understood. We show that mice deficient in IFN signaling components including IFN alpha and beta receptor 1 (IFNAR1), interferon regulatory factor 1 (IRF1), IRF9, and STAT1 each have reduced bone density and increased osteoclastogenesis compared to wild type mice. The IFN-inducible guanylate-binding proteins (GBPs) on mouse chromosome 3 (GBP1, GBP2, GBP3, GBP5, GBP7) are required to negatively regulate age-associated bone loss and osteoclastogenesis. Mechanistically, GBP2 and GBP5 both negatively regulate in vitro osteoclast differentiation, and loss of GBP5, but not GBP2, results in greater age-associated bone loss in mice. Moreover, mice deficient in GBP5 or chromosome 3 GBPs have greater LPS-mediated inflammatory bone loss compared to wild type mice. Overall, we find that GBP5 contributes to restricting age-associated and inflammation-induced bone loss by negatively regulating osteoclastogenesis.

O steoclasts are myeloid-derived multinucleated cells involved in the resorption of bone 1 . Inflammation produced by chronic inflammatory disease or acute infection has been shown to drive the resorption of bone. The role of osteoclasts in this bone destruction is well established, and many inflammatory mediators have been implicated in driving osteoclast-mediated bone destruction, including IL-1, IL-6, IL-17, and TNF [2][3][4] . Autoimmune diseases, such as rheumatoid and psoriatic arthritis, and diseases associated with excessive inflammasome activation, such as cryopyrin-associated periodic syndromes, familial Mediterranean fever, or chronic recurrent multifocal osteomyelitis, are all associated with bone loss [5][6][7] . In addition, inflammation in patients with implanted devices, as commonly observed in total hip replacements, results in osteolysis and subsequent implant failure 8,9 . Infection-associated osteomyelitis, driven by infection with Staphylococcus aureus or Porphyromonas gingivalis, can also lead to localized bone loss 10,11 . Progressive loss of bone mineral density through aging results in osteoporosis, which increases the risk of fractures in the elderly, especially postmenopausal women 12,13 . Overly active osteoclasts contribute to many immune-driven inflammatory bone diseases, but they are also essential for normal formation of the bone marrow cavity, homeostatic bone remodeling and repair. Failure to generate osteoclasts through mutations in key regulators of their differentiation leads to a condition called osteopetrosis, which leads to densely formed bones that are prone to fracture. Given the importance of inflammation in driving destructive bone remodeling, it is important to understand how immune signaling pathways contribute to disease progression, and to identify potential new therapeutic targets for bone diseases.
The IFN pathway plays an inhibitory role in osteoclast generation [14][15][16] . Type I IFNs signal through IFNAR1/2 and the downstream proteins STAT1, STAT2, and IRF9, which together upregulate a number of IFN-stimulated genes (ISGs). Similarly, type II IFN (IFNγ) signals through the IFNGR and STAT1 homodimers, upregulating a distinct but also overlapping set of ISGs 17 . Upon stimulation with RANKL, a cytokine required for osteoclastogenesis, cells produce IFNβ, which acts as a negative feedback signal by inhibiting the expression of c-Fos, an essential transcription factor for osteoclast differentiation 14,18 . Mice lacking IFNAR1, IFNβ, IFNγ, or IFNGR1 have reduced bone density compared to wild type (WT) mice, highlighting the importance of IFN signaling in the negative feedback signaling and homeostasis of osteoclasts 14,15,19 . Similarly, IFNγ also suppresses osteoclastogenesis by suppressing TRAF6-mediated signaling downstream of RANK, the receptor for RANKL 16,20 . STAT1 and IRF1 critically regulate the expression of many ISGs, and osteoclasts deficient in either STAT1 or IRF1 undergo increased osteoclastogenesis in vitro 21,22 . Therapeutically, administration of IFNα to patients with bone loss secondary to mastocytosis, in combination with first-line treatment with bisphosphonates, increases bone density, and reduces the risk of fracture 23,24 . Other studies have shown that administration of IFNγ may prevent bone loss during rheumatoid arthritis and prevent bone loss in patients with postmenopausal osteoporosis 25 .
Better understanding of the mechanisms by which IFN and other ISGs regulate bone density is important in understanding how this immune pathway regulates bone homeostasis. Within the guanylate-binding protein (GBP) family of proteins, many are highly upregulated by IFN signaling 26,27 . These multifunctional proteins are large GTPases and many are poorly studied to date. Of these GBPs, GBP1, GBP2, and GBP5 are the best studied and are the only GBP family members to contain a C-terminal CaaX domain, which is cleaved and prenylated with farnesyl or geranylgeranyl lipid moieties, targeting them to membranes.
Prenylation is common on small GTPases of the Ras, Rho, and Rab family proteins, including cdc42, Rac1, Rac2, Rap1, and Rab3d, which are each important in promoting osteoclastogenesis [28][29][30][31][32][33][34][35] . One of the front-line treatments for bone loss are nitrogen-containing bisphosphonates, which operate by inhibiting the prenylation of GTPases, resulting in osteoclast apoptosis 36,37 . GBPs have been shown to regulate actin dynamics in cells during bacterial infections and in IFNγ-treated cells, suggesting that they may regulate similar processes in osteoclasts [38][39][40][41] . Together with these observations and the critical role of IFN signaling in regulating osteoclastogenesis, we hypothesized that the IFN-inducible large GTPases (GBPs) regulate osteoclast differentiation and function.
Results and discussion IFN signaling protects from bone loss and restricts osteoclastogenesis. Previous studies have shown that mice lacking key IFN signaling components have homeostatic changes in bone density, but have not been well compared in parallel 14,15,21,22,42,43 . We first reexamined, in age-and sex-matched mice, the role for type I IFN signaling molecules IFNAR1, IRF9, IRF1, and STAT1 in regulating bone density in femurs of 3-month-old mice by micro-computed tomography (μCT). Using μCT, we quantified bone morphometric data obtained from age-and sex-matched mouse femurs. Compared to WT mice, Ifnar1 −/− , Irf9 −/− , Irf1 −/− , and Stat1 −/− mice had reduced bone density, as quantified by reduced bone volume to total volume (BV/TV) ratio, reduced trabecular number, and increased trabecular spacing, while only Stat1 −/− mice showed significantly reduced trabecular thickness (Fig. 1a-e). Our results in Ifnar1 −/− mice are consistent with published findings that Ifnar1 −/− mice have reduced bone density and increased osteoclastogenesis 14 . Earlier studies found that Stat1 −/− and Irf1 −/− mice have increased overall bone density, contrary to what would be expected from increased osteoclastogenesis in vitro 21,22 . In line with our data, the region distal to the growth plate of Irf1 −/− femurs has reduced BV/TV, while the trabecular spacing is increased 21 . These previous studies proposed that increased osteoblast activity in Stat1 −/− and Irf1 −/− mice was responsible for increased bone density. One proposed mechanism is that STAT1 limits nuclear translocation of the Runx2 transcription factor, limiting the proliferation of osteoblasts, and that STAT1 deficiency increases their replication 22 . Another proposed mechanism for increased bone density in Stat1 −/− mice is due to an increase in FGF18-dependent signaling in osteoblasts, and a decrease in the expression of Cdkn1a and Fgfr3, which limit the cell cycle 42 . In Irf1 −/− osteoblasts, the proposed mechanism for increased bone density is due to increased osteoblast proliferation and activity 21 . Because Irf1 −/− and Stat1 −/− mice were previously shown to have increased bone density, in contrast to our findings ( Fig. 1) 21,22 , we utilized an additional independent genetic model and examined femurs from Irf1 fl/fl LysM Cre+ mice, and also found an osteoclast-specific role for IRF1 in restricting bone loss (Supplementary Fig. 1). In the previous studies, Irf1 −/− bones were examined earlier (at 8 weeks) and bones were derived from Stat1 −/− mice on a different genetic background (129S6/SvEv-Stat1 tm1Rds ), suggesting these experimental differences may explain some of the observed differences with our data 21,22,42,44 . To further examine the role for these IFN regulatory proteins, osteoclasts were differentiated in vitro using recombinant M-CSF and RANKL. Consistent with the bone density reduction observed in Ifnar1 −/− , Irf9 −/− , Irf1 −/− , and Stat1 −/− mice, in vitro osteoclasts derived from these mice exhibited enhanced osteoclastogenesis compared to WT (Fig. 1f). Consistent with our findings, patients with Mendelian susceptibility to mycobacterial diseases, defined by loss-of-function mutations in IFNGR1 or STAT1, display reduced bone density and increased osteoclastogenesis 43,45,46 . Together, these data show that IFNAR1, IRF9, IRF1, and STAT1 each negatively regulate in vitro osteoclastogenesis and bone loss in mice.
GBPs limit osteoclast fusion and bone loss. IFNs potently inhibit osteoclastogenesis and lead to the expression of hundreds of ISGs following stimulation 14 . The role of the interferoninducible GBPs in osteoclasts, to our knowledge, has not been explored. Given that GBPs are highly upregulated following IFN stimulation 26,47 , we examined whether osteoclasts deficient in five GBPs on mouse chromosome 3 (GBP1, GBP2, GBP3, GBP5, and GBP7, indicated as Gbp Chr3−/− ) would form larger osteoclasts similar to those differentiated from Irf9 −/− , Irf1 −/− , and Stat1 −/− mice. Consistent with the increased osteoclast fusion observed in IFN signaling-deficient cells (Fig. 1), osteoclasts derived from Gbp Chr3−/− mice fused to a greater extent than WT cells did (Fig. 2a). The extent of multinucleation in osteoclasts which contained at least three nuclei was quantified. Osteoclasts differentiated from Gbp Chr3−/− mice contained more nuclei than WT osteoclasts, suggesting that expression of GBPs negatively ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-20807-8 regulates osteoclast fusion (Fig. 2b). Live cell imaging of WT and Gbp Chr3−/− cells further showed that GBPs negatively regulate osteoclast fusion ( Supplementary Movies 1 and 2). To determine whether GBPs contribute to the negative regulation of osteoclast activity in vivo, femurs were collected from 6-month-old mice, and bone morphometric data were collected by μCT. Consistent with increased in vitro osteoclastogenesis and similar to the bone loss observed in IFN signaling-deficient mice (Fig. 1), femurs from Gbp Chr3−/− mice had reduced bone density (BV/TV) and trabecular number, and increased trabecular spacing compared with WT mice with a trend toward reduced trabecular thickness ( Fig. 2c-h). To determine whether osteoclasts were enhanced in Gbp Chr3−/− femurs, bones were decalcified, sectioned and stained for tartrate-resistant acid phosphatase (TRAP), a marker of mature osteoclasts 48 . Consistent with our μCT findings, bones of Gbp Chr3−/− mice showed more trabeculae-associated osteoclasts compared with WT mice (Fig. 2i). Together, these data reveal that GBPs can negatively regulate osteoclastogenesis under homeostatic aging-dependent bone turnover. Because GBPs have been found to regulate the polymerization of F-actin during infections and Arp2/3-mediated cytoskeletal rearrangements are required for driving osteoclast fusion, the loss of GBPs may mechanistically increase fusion through altered cytoskeletal remodeling [39][40][41][49][50][51] . Consequently, loss of GBPs increased Arp2/3dependent actin polymerization during infection with Burkholderia thailandensis, resulting in significantly increased macrophage cell-cell fusion, consistent with the observed increased osteoclast fusion phenotype, suggesting GBPs regulate actin dynamics across cell types 52 . How GBPs specifically inhibit the polymerization of F-actin via the Arp2/3 complex is an outstanding question in the GBP field 27,53,54 . The mechanisms behind osteoclast fusion are similarly poorly understood, and these data suggest GBP-mediated regulation of cell fusion may be important for inflammatory bone remodeling.
GBP5 but not GBP2 restricts bone loss in mice. Because Gbp Chr3−/− mice lack five GBPs (GBP1, GBP2, GBP3, GBP5, and GBP7), we sought to determine which GBPs within this locus contribute to the regulation of osteoclastogenesis and bone loss. Among the best studied murine GBPs are GBP2 and GBP5, and previous studies found that both were required for cellautonomous immunity against the intracellular bacterium Francisella novicida 55,56 . Few studies have shown distinct functions for GBP2 and GBP5, though the proteins are thought to differ in their localization, suggesting that each may have unique functions 57 . While their individual functions are poorly understood, GBP2 has been shown to play a larger role in recognition of Escherichia coli outer membrane vesicles, while GBP5 has been shown to play a dominant role in mediating Brucella abortus LPSdependent activation of caspase-11 (refs. 58,59 ). Both GBP2 and GBP5 are prenylated to facilitate membrane localization, similar to other small GTPases that regulate osteoclastogenesis 29,60,61 . To determine the contribution of each of these GBPs to bone density, femurs from the respective gene deficient mice were collected, and bone morphometric data were collected by μCT. Both Gbp5 −/− and Gbp Chr3−/− mice, but not Gbp2 −/− mice had reduced bone density as measured by BV/TV, trabecular number, and trabecular spacing (Fig. 3a-e). In vitro osteoclasts derived from Gbp2 −/− , Gbp5 −/− , or Gbp Chr3−/− mice all have increased osteoclastogenesis compared with WT mice, suggesting that in vivo, GBP2 is not necessary for normal bone maintenance (Fig. 3a-e and Supplementary Fig. 2a, b). Mechanistically, increased cell-cell fusion in GBP-deficient osteoclasts is likely due to changes in actin cytoskeleton dynamics, as expression of required osteoclast fusion mediators DC-STAMP and ATP6V0D2 are similar to WT osteoclasts ( Supplementary  Fig. 2c). How GBPs regulate actin cytoskeleton dynamics is an outstanding question in the field 27,54,62 . While it is surprising that GBP5 alone regulated bone density in mice, it is possible that the loss of GBP2 causes a compensatory response that counteracts any changes in osteoclastogenesis in vivo. Indeed, previous work has shown that human GBP1, which most closely phenocopies mouse GBP2, is upregulated in mesenchymal stromal cells which give rise to osteoblasts and increased expression is associated with osteoporosis 63,64 . Together, these data show a distinct role for GBP5 in negatively regulating bone loss in mice during homeostasis.
GBPs limit acute inflammatory bone loss. To further examine the role of GBPs in regulating bone turnover in mice, we utilized an acute bone loss model where LPS injection induces rapid bone loss and osteoclastogenesis 2,14,65,66 . Mice were injected intraperitoneally with 5 mg/kg LPS at days 0 and 4, and bones were collected at day 8, as previously described 66,67 . Compared with untreated control mice, LPS injection resulted in significantly greater bone loss in Gbp5 −/− and Gbp Chr3−/− mice (Fig. 4a-f).
To determine the effect of LPS on osteoclastogenesis in vivo in GBP-deficient mice, bone sections were stained for osteoclasts using a TRAP stain. Consistent with the observed increase in bone loss in Gbp5 −/− and Gbp Chr3−/− mice compared with WT mice, osteoclasts associated with trabecular bone were increased in both Gbp5 −/− and Gbp Chr3−/− femurs (Fig. 4g). Together these data establish a critical role for GBP5 in negatively regulating inflammatory bone loss. Future studies examining GBP expression during inflammatory bone loss may reveal important regulatory networks that were previously not appreciated. Because GBPs contribute to both age-related and acute inflammation-related bone loss, they likely function during chronic low-grade inflammation driven by microbiota and during acute infection-induced osteomyelitis. This regulation of inflammatory bone loss may lead to new strategies for treating inflammation-driven bone diseases.  Fig. 2 Guanylate-binding proteins limit osteoclast fusion and bone loss. a Representative images of in vitro osteoclasts from wild type (WT) or Gbp Chr3−/− mice stained for F-actin (red) and nuclei (green). b The number of nuclei per osteoclast (>3 nuclei/osteoclast) was quantified by counting 12 images at 10× magnification per genotype (WT vs Gbp Chr3−/− 21-30 nuclei osteoclasts p = 0.0013). c-h Representative 3D and 2D micro-computed tomography (μCT) sections were obtained from at least six femurs from 6-month-old c, e-h WT (n = 6) and d, e-h Gbp Chr3−/− (n = 9) mice, and quantitative data of e bone volume to total volume (BV/TV; p = 0.0004), f trabecular thickness (mm), g trabecular number (1/mm; p = 0.0004), and h trabecular spacing (mm; p = 0.0004) were obtained. i Femur osteoclasts in 3-month-old mice were stained for tartrate-resistant acid phosphatase (TRAP) activity (magenta; green arrow), counterstained with hematoxylin, and imaged at the indicated magnification. Significance was determined by b two-way ANOVA with Sidak's multiple comparison test or e−h two-tailed Student's t test, **p < 0.01, ***p < 0.001. Scale bars (black) indicate 200 μm (a, i). Data are presented as mean ± SEM (b, e-h). M-CSF macrophage colony-stimulating factor, RANKL receptor activator of nuclear factor-κΒ ligand. Osteoclast culture. Primary osteoclasts were derived from mouse bone marrow differentiated in alpha-minimum essential medium (ThermoFisher Scientific 12571063) supplemented with 1× nonessential amino acids (ThermoFisher Scientific, 11140076), 10% FBS, and 1× penicillin-streptomycin (15070063, Ther-moFisher Scientific) 74 . Bone marrow cells were flushed from the femur and tibia, and plated on plastic petri dishes with 25 ng/mL recombinant mouse macrophage colony-stimulating factor (M-CSF) for 3 days. Cells were then seeded in tissue culture plates at 1.2 × 10 4 cells per well in 150 μL media (96-well plate) or 1.2 × 10 5 cells per well in 1.5 mL media (12-well plate) in differentiation medium containing 25 ng/mL M-CSF and 50 ng/mL RANKL. Media was replenished after 48 h.
Bone morphometry. Bone morphometric parameters were measured by µCT scans from age-and sex-matched mouse femurs at 6 months or 3 months of age in a blinded fashion (JK) by the Center for In Vivo Imaging and Therapeutics (CIVIT) core facility at St. Jude Children's Research Hospital. Untreated control morphometric data was pooled from multiple independent scans, and used to compare appropriate age-matched untreated and treated mice throughout this study. µCT images were obtained using a Siemens Inveon µCT scanner (Siemens Healthcare). Mouse femurs were imaged using a 1024 × 2304 mm matrix with field of view 18.29 × 41.15 mm using one bed position. Projections were acquired at 80 kVp and 500 µA (3900 ms exposure and 3500 ms settle time) over full rotation (480 steps) providing an isotropic resolution of 17.86 µm. Data were post-processed using the segmentation tool in Inveon Research Workplace (IRW version 4.2) software to obtain morphometric measurements.
Microscopy. Images and videos (hourly image acquisition) of osteoclasts were automatically collected using an IncuCyte S3 (EssenBiosciences). Cells were fixed in 4% PFA, permeabilized with 0.5% Triton X-100, and stained for F-actin (phalloidin-iFluor555, ab176756, Abcam, 1:2000) and nuclei (25 nM Sytox Green,  S7020, ThermoFisher Scientific), according to manufacturer's protocols. Similarly, permeabilized cells were stained for TRAP following the manufacturer's protocol (MK301, Takara) and automatically imaged using a Nikon C2 microscope. TRAPstained femur sections were stained by the St. Jude Children's Research Hospital Veterinary Pathology Core and images were collected by a trained pathologist (PV).
In vivo LPS-induced bone loss. Male mice, age-matched at 3 months, were treated with LPS (5 mg/kg) by intraperitoneal injection at days 0 and 4, and bones were collected at day 8, fixed in formalin and transferred to 70% ethanol for scanning by µCT. LPS treatment group data were compared to the pooled control age-and sexmatched scan data used throughout this study.
Quantification and statistical analysis. GraphPad Prism 6.0 software was used for data analysis. Data are shown as mean ± SEM. Statistical significance was determined by Student's t test for two groups or one-way analysis of variance (ANOVA) for three or more groups, and two-way ANOVA for comparison between multiple groups. The specific statistical testing for each experiment is indicated in the figure legends.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data generated and analyzed during the current study are contained within the manuscript, and/or are available from the corresponding author on reasonable request.