Neutralization of Receptor activator of nuclear factor-κB ligand reduces fibrosis and promotes osteoblast differentiation in a mouse model of fibrous dysplasia driven by somatic expression of Gnas^R201H

Abstract

Fibrous dysplasia (FD) is a rare disorder caused by somatic activating mutations in GNAS, encoding the alpha subunit of the Gs protein. Activating GNAS mutations result in focal expansile bone lesions, which cause pain, deformity, and increased risk of fracture. Somatic mosaicism in FD leads to both GNAS mutant and genetically WT osteoprogenitor cells, which jointly contribute to the formation of fibrotic lesions within the bone. Additionally, these lesions contain numerous osteoclasts formed in response to robust lesional expression of RANKL. Neutralizing antibody to RANKL is effective in reducing lesion growth in patients with FD and in preclinical models. To determine the effect of RANKL neutralization specifically on mutant cells early after onset of FD, we used a murine model of C57BL/6 Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato mice, which recapitulates the somatic mosaicism of FD bone lesions and in which mutant cells are lineage traced. Analysis of Gnas^(R201H)fl/+ mice showed a diffuse accumulation of SMA⁺ early osteoblastic cells, with contribution from both tdTomato⁺ mutant and tdTomato⁻ WT populations. Anti-RANKL treatment of Gnas^(R201H)fl/+ mice inhibited osteoclast formation and substantially reduced fibrosis, detected by Masson’s trichrome staining within the proximal metaphysis of the femur and the femoral head. Treatment with anti-RANKL decreased the accumulation of both mutant and WT SMA⁺ cells, accompanied by an increased number of mutant cells expressing the mature osteoblast marker osteocalcin, and an increase in overall osteoblast density. To elucidate the role of RANKL expression by mutant cells in the formation of FD lesions, we generated Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato;Rankl^fl/fl mice. Deletion of Rankl in Gnas^(R201H)fl/+ mutant cells did not prevent fibrosis in this model. The results suggest that while anti-RANKL treatment promotes osteoprogenitor differentiation to reduce fibrosis, the loss of RANKL expression from GNAS mutant cells alone is not sufficient to reverse the pathology of FD bone lesions.

Graphical Abstract

Graphical Abstract.

Introduction

Fibrous dysplasia (FD) is a rare disorder caused by somatic activating mutations in GNAS, the gene encoding the alpha subunit of the Gs protein.¹ Mutations predominantly involve the codon corresponding to arginine 201, with arginine to histidine being the most common,² though mutations in other residues, for example, p.Q227H, have also been reported.³ Postzygotic gain-of-function mutations result in focal expansile bone lesions that cause pain, deformity, and have an increased risk of fracture.^4,5 Fibrous dysplasia is rare, affecting approximately 1 in 30 000 people.⁶ Monostotic or polyostotic tumors typically arise during bone growth in children and young adolescents.⁷ Fibrous dysplasia lesions consist of a fibrotic tissue with abundant active stromal cells that form lesions within the bone marrow.⁸ Lesions expand over time, compromising the structural integrity of the bone, and leading to increased risk of fracture. Treatments for FD, most notably bisphosphonates,⁹ have not resulted in consistent reduction of lesional bone nor pain relief. Surgery has historically been the primary approach to managing bone lesions.¹⁰

Fibrous dysplasia lesions are comprised of abundant skeletal stem cells, identified as bone marrow stromal cells.¹¹ Lesions have also been shown to contain high levels of alpha smooth muscle actin (SMA), which is transiently expressed during osteoblast differentiation, as detected in biopsies from patients with FD.¹² Importantly, FD is classified as a mosaic disease; the stromal cells within the lesions either express mutant GNAS or are genetically normal “WT” cells. Furthermore, it has been shown that both mutant and WT osteoprogenitor cells are necessary to produce fibrotic lesions. This was demonstrated in a classic experiment reported by Robey et al., where clones of WT and GNAS mutant bone marrow stromal cells isolated from FD lesions were transplanted into immunocompromised nude mice. The transplanted GNAS mutant cells were incapable of forming fibrotic lesions unless transferred in combination with WT cells.¹¹

Fibrous dysplasia lesions also contain abundant tartrate-resistant acid phosphatase (TRAP) expressing osteoclasts.¹³ Osteoclast differentiation within lesions is hypothesized to be driven by the high local levels of RANKL, the key osteoclastogenic factor that regulates osteoclast formation and activity.¹⁴ RANKL production by osteoblastic cells is stimulated by treatment with PTH^15,16 via cyclic adenosine monophosphate/protein kinase A signaling.¹⁷ PTH signals through PTHR1, a G-protein coupled receptor that couples to Gs, which induces the release of GTP bound Gs alpha, and activates downstream effectors including adenyl cyclases.¹⁸ This signaling suggests aberrant expression of RANKL by GNAS^R201H mutant cells, driven by constitutively active Gs alpha, could play an essential role in initiating FD lesion formation. Patients with FD also have high circulating levels of RANKL,¹⁹ suggesting RANKL could be a possible therapeutic target.

Denosumab, a fully human antibody against RANKL clinically approved for osteoporosis treatment, has shown promising results as a treatment for FD. A number of case reports and several cohort studies reported that denosumab treatment reduced pain scores, improved functionality, reduced tumor growth, and decreased lesional activity, as observed by PET imaging scans in patients with FD.^20–24 A phase 2 trial of high dose denosumab treatment in adults with FD showed denosumab reduced bone lesion activity, assessed by 18F-sodium fluoride uptake and improved symptoms. Histological analysis showed reduced cell proliferation and an increase in the osteocyte marker, sclerostin, in lesional bone after treatment.²⁵ Importantly, a number of studies have shown treatment with anti-RANKL in both murine models of FD and in patients has led to a change in the cellular profile of lesions, increasing the number of osteoblast cells expressing OCN, while reducing the number of SMA⁺ osteoblast precursor cells.²⁶ These important findings show clear evidence that RANKL treatment reduces fibrosis in patients with FD, although the exact mechanism is unknown. The neutralizing antibody against RANKL, by eliminating osteoclasts, likely reduces osteoclast-osteoblast cross talk, indirectly impacting osteoblastic cell fate. Alternatively, the antibody to RANKL could hypothetically act directly on the RANKL expressing lesional cells to induce reverse signaling and promote osteoblast differentiation.²⁷ Moreover, at the cellular level, it is not clear if inhibition of RANKL alters survival or differentiation of GNAS mutant cells, WT cells, or both.

To understand the cellular impact of RANKL inhibition on both mutant and WT osteoprogenitor cells in the early onset of FD, we utilized an exon switch mouse model of disease, where expression of a mutant mouse allele equivalent to the human GNAS R201H mutation can be induced post-zygotically.²⁸ In the Gnas^(R201H)fl/+ mouse, in the presence of Cre enzyme, floxed WT exons are excised, allowing expression of Gnas^R201H from the endogenous Gnas locus. The Sox9Cre^ERT model was utilized to induce the mutation in osteochondroprogenitor cells that are marked in early postnatal development and has been used previously as a model of FD in mice.^28,29 To recapitulate FD, Sox9^CreERT;Gnas^(R201H)fl/+ animals are treated with tamoxifen shortly after birth to induce Gnas^R201H expression in osteochondral progenitor cells. A Rosa26^LSL-tdTomato transgene was introduced to trace Gnas mutant expressing cells. Importantly, this model generates the presence of mosaic lesions, replicating a key feature of the human disease, and in which WT and mutant cells within lesions can be individually assessed.²⁸

Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato mice were treated with either anti-RANKL antibody or isotype control to determine if RANKL inhibition impacted osteogenic differentiation and proliferation of mutant cells, WT cells, or both in young animals. We also assessed the impact of treatment on fibrosis and osteoclast number to confirm that anti-RANKL inhibits the formation of bone lesions in this model. As osteoblastic RANKL expression is induced by PTH, which signals through Gs alpha, we hypothesized that RANKL production by Gnas^R201H mutant cells might be important for pathogenesis. To test this, we crossed Rankl^fl/fl mice with Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato mice and compared lesion formation with and without Rankl expression by mutant cells. Results from this study describe how inhibition of RANKL, either systemically or specific loss in mutant cells via genetic deletion, affects lesion formation in a post-zygotically induced mosaic model of FD bone lesions in mice.

Materials and methods

Animals

All animal procedures were approved by the Institutional Animal Care and Use Committee, Brigham and Women’s Hospital. Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato and Gnas^+/+;Rosa26^LSL-tdTomato C57BL/6 mice generated by the Yang Lab²⁸ were obtained and mated to generate Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato and Gnas^+/+ WT controls. All mice were gang housed. Male and female pups were genotyped at P5, and the dam was treated with a low dose of tamoxifen (40 mg/kg) at P6. Tamoxifen is then delivered to the pups via lactation, which allows expression of both mutant Gnas and TdTomato as a lineage tracer in the subset of osteoprogenitor cells expressing Sox9 at P6. Tamoxifen treated mice are hereafter referred to as Gnas^R201H;R26^Tom and Gnas⁺;R26^Tom. Pups were then treated with either 5 mg/kg of the anti-RANKL antibody (OYC1; Oriental Yeast) or rat IgG2a isotype control (5 mg/kg) (InVivoPlus, BP0089, BioxCell) by intraperitoneal injection at P6 and P13, and aged to P21 (3 wk of age). Three-week-old mice treated with anti-RANKL (n = 9; 2F and 7M) and IgG2a (n = 8; 1F and 7M) treated mice were litter matched where appropriate with Gnas⁺;R26^Tom control mice (n = 5; 2F and 3M). A second cohort of pups were treated with anti-RANKL antibody or IgG2a isotype control (5 mg/kg) at P6, P13, P28, and P35, and aged until P42 (6 wk of age). Six-week-old mice treated with anti-RANKL (n = 7; 4F and 3M) and IgG2a (n = 6; 3F and 3M) treated mice were litter matched where appropriate with Gnas⁺;R26^Tom control mice (n = 5; 3F and 2M). The experimental treatment times are shown in Figure 1A.

Figure 1.

Schematic depicting experiments utilizing the Sox9^CreERT;Gnas^R201H mouse model of FD. (A) Dams from crosses able to generate Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato offspring were treated with a low dose of tamoxifen (40 mg/kg) at P6. Tamoxifen is then delivered to the pups via lactation, which allows expression of both mutant Gnas^R201H and tdTomato as a lineage tracer in the subset of osteoprogenitor cells expressing Sox9 at P6. Resultant Gnas^R201H;R26^Tom mutant mice were then injected with either IgG2a or anti-RANKL antibody weekly until either P21 or P42 (3 or 6 wk of age). Femurs were collected for analysis. (B) Dams from crosses able to generate Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato mice with either Rankl^fl/fl or Rankl^fl/+ or Rankl^+/+ were genotyped to determine RANKL deletion. Pups were ear notched and genotyped at P5, the dam was treated with a low dose of tamoxifen (40 mg/kg) at P6 to deliver tamoxifen to the pups via lactation. Femurs were collected for analysis at either P21 or P42 (3 or 6 wk of age).

Another cohort of Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato mice were crossed to Rankl^fl/fl mice (Jackson labs strain 018978) to generate Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato;Rankl^fl/fl mice and Rankl^fl/+ controls with intact Rankl expression. Tamoxifen treatment was given as previously stated, and pups were then aged to either P21 (3 wk of age) or P42 (6 wk of age). These mice are hereafter referred to as Gnas⁺;Rankl^D; R26^Tom (n = 9; 3F and 6M), Gnas^R201H;Rankl^D; R26^Tom (n = 8; 6F and 2M), Gnas^R201H;Rankl⁺;R26^Tom (n = 3; 1F and 2M), Gnas⁺;Rankl⁺;R26^Tom (n = 3; 2F and 1M), and Sox9Cre- (n = 8; 1F and 7M). The experimental treatment times are shown in Figure 1B.

Micro-computed tomographic (micro-CT) imaging

All femurs were imaged using the Scanco μCT 35 at a voxel size of 15 μm. Femurs were scanned in 70% ethanol using an X-ray tube potential of 55 kVp, an X-ray intensity of 0.145 mA, and an integration time of 400 ms. 3D reconstructions were generated using Scanco software, with a threshold of 227 mg HA/ccm to distinguish bone from marrow space.

Histology

Femurs were excised and fixed with 4% paraformaldehyde for 48 h and then decalcified with 14% EDTA. One femur was equilibrated in 30% sucrose overnight, frozen embedded in OCT, and sectioned at 10 μm onto cryotape. The other femur was processed for paraffin embedding, sectioned at 10 μm, and stained with either Masson’s Trichrome as per manufacturer’s instructions (Sigma Aldritch, #HT15) or TRAP staining. Sections were analyzed for histologic features in a region of interest (ROI) within the bone marrow cavity. The ROI was located below the distal femoral growth plate, and included all tissue between the endosteal margins, extending 500 μm proximal to the growth plate.

Masson’s trichrome stained sections were analyzed for changes in trabecular bone volume, bone marrow area, and the presence of stromal cells/fibrosis within the bone marrow cavity. Stromal cells were identified as cells with a rounded morphology in close proximity to trabecular bone. Regions of fibrosis (stained red) were identified as more than two layers of stromal cell bodies within a disorganized matrix around the trabecular bone. The percentage of fibrosis and total bone (stained blue) was normalized to the total tissue area (μm²) of the ROI, and was quantified using QuPath analysis software,³⁰ by an observer blinded to both treatment and genotype.

Femur sections were also stained for TRAP, with 0.25% Fast Green counter stain, to detect TRAP⁺ multinucleated osteoclasts, as described.³¹ The number of TRAP⁺ cells, osteoclast surface, total trabecular bone surface (μm), and total tissue area (μm²) were quantified using QuPath in the ROI described above.³⁰

Immunofluorescence

Frozen femoral sections were stained for immunofluorescence for SMA (Abcam, ab5694), Ki67 (Cell Signaling, 12202S), osterix (OSX; Abcam, ab22552), osteocalcin (OCN; Thermo Fisher 23 418-1-AP), RANKL (BE0191; BioXCell), and RUNX2 (Abcam, ab192256). Briefly, the sections were brought to room temperature (RT) and blocked for 1 h with 1X PBS containing 10% normal goat serum, 0.1% bovine serum albumin (BSA), and 0.1% Triton-X-100. Sections were incubated overnight at 4 °C with the primary antibody (1:100 for SMA, Ki67, OSX, OCN, 1 μg/mL for RANKL and 1:100 for RUNX2). Slides were brought to RT, washed twice and incubated at RT for 1 h with the Alexa Fluor 647 secondary antibody (goat anti-rabbit 1:100, (Thermo Fisher, A-21245)) and DAPI nuclear stain (1:400). After washing, sections were mounted with Anti-Fade Fluorescence Mounting Medium (Abcam, ab104135) and imaged using a Zeiss LSM 800 Confocal Microscope at 405 and 647 nm. Rosa26^LSL-tdTomato expression (TOM⁺) was detected by fluorescence at 588 nm.

Femur sections were analyzed within the marrow ROI described above for total number of SMA⁺, OSX⁺, OCN⁺, or Ki67⁺ cells, as well as positive expression within the subset of TOM⁺  Gnas^R201H mutant cells. Additionally, the number of TOM⁺ positive cells were quantified per total tissue area (μm²) and per trabecular area (μm²), and the number of SMA⁺, OSX⁺, OCN⁺, and Ki67⁺ cells were quantified per total tissue area (μm²). Analysis was performed using ImageJ.³² A second ROI focusing on lesional bone was defined as fibrotic areas adjacent to the cortical bone and enriched for TOM+ Gnas^R201H mutant cells within the bone marrow cavity. These areas were quantified as a second standardized ROI within the bone marrow cavity of the femur, by selectively determining an area with an increased density of TOM⁺  Gnas^R201H mutant cells within a 500 μm by 500 μm region. Cellular content within the lesion was quantified using ImageJ analysis software.³²

The efficiency of Rankl deletion in Gnas^R201H;Rankl^Δ;R26^Tom mice was analyzed within a defined 500 μm by 500 μm ROI, starting 50 μm below femoral growth plate and centered on lesional tissue within the bone marrow cavity. TOM⁺  Gnas^R201H mutant cells were quantified within the ROI, and immunostaining for SMA⁺, RANKL⁺, and RUNX2⁺, was performed and normalized to the total number of DAPI+ cells within the ROI. Analysis was performed using QuPath.³⁰

Measurement of serum RANKL

Blood was obtained by cardiac puncture from 3-wk-old Gnas^R201H;Rankl⁺;R26^Tom and Gnas^R201H;Rankl^Δ;R26^Tom (n = 4/group) and was allowed to clot for 2 h at RT. Serum was collected by centrifugation for 20 min at 2000 × g at RT. ELISA was performed on serum diluted 2-fold with Calibrator Diluent RD6-12 using mouse RANKL immunoassay kit (MTR00; R&D System) according to the manufacturer’s directions.

Single-cell transcriptome profiling of cell isolates from FD lesion

The GSE263294 dataset was downloaded and re-analyzed according to the original paper.³³ Briefly, after importation of the matrices in Seurat using the “Read10X()” function, cells with fewer than 200 or more than 6000 expressed genes, or with a percentage of mitochondrial DNA genes over 15% were filtered. The integrated joint embedding from Harmony was used to cluster the cells using the Seurat package “FindClusters()” function, using the Louvain algorithm to perform the graph based clustering. The non-linear dimensional reduction “uniform manifold approximation and projection” (UMAP) technique was used to visualize the identified clusters as a map. Dotplot was used to show gene expression in different cell clusters.

Statistical analysis

Statistical analysis was performed using GraphPad Prism v.9 for Windows (GraphPad Software). The Mann–Whitney non-parametric t-test was used to calculate significance for comparisons of fibrosis, osteoclast number and surface, and immunofluorescent cellular subsets in anti-RANKL and IgG2a treated animals. Mann–Whitney non-parametric t-test was used to calculate significance for osteoclast number and osteoclast surface in 6-wk-old Gnas^R201H;Rankl⁺;R26^Tom and Gnas^R201H;Rankl^D;R26^Tom mice. One-way ANOVA with Tukey’s post-test was used to compare fibrosis in the 6-wk-old Cre-, Gnas⁺;Rankl⁺;R26^Tom, Gnas⁺;Rankl^D;R26^Tom, Gnas^R201H;Rankl⁺;R26^Tom, and Gnas^R201H;Rankl^D;R26^Tom mice. Analysis of 3-wk-old Gnas^R201H;Rankl^D;R26^Tom and Gnas^R201H;Rankl⁺;R26^Tom mice and serum RANKL was performed using an unpaired Student’s t-test. Significance was considered when p < .05 for all analyses.

Results

Gnas ^R201H mutant mice treated with anti-RANKL showed changes in bone architecture, reduced osteoclast number, and increased trabecular bone volume

Comparison of micro-CT images of femurs from 3-wk-old Gnas^R201H;R26^Tom and Gnas⁺;R26^Tom mice demonstrated visible pitting of the cortex of Gnas^R201H;R26^Tom mice compared to Gnas⁺;R26^Tom control (Figure 2A, black arrow). Gnas^R201H;R26^Tom mice treated with anti-RANKL resulted in widened epiphyses and increased trabecular bone below the growth plate compared to IgG2a isotype treated mice (Figure 2A), consistent with an osteoclast-deficient phenotype.

Figure 2.

Anti-RANKL treatment decreased osteoclasts and increases trabecular bone in the Gnas^R201H;R26^Tom mouse model of fibrous dysplasia. (A) 2D and 3D images of representative micro-CT scanned femurs show the bone architecture of 3-wk-old control Gnas⁺;R26^Tom (left) and Gnas^R201H;R26^Tom mutant mice (right) treated with either anti-RANKL (aRANKL) or IgG2a isotype. Arrow denotes cortical irregularity. (B) Representative images of TRAP and fast green stained femur sections from Gnas^R201H;R26^Tom mice treated with aRANKL and IgG2a. (C) Quantification of trabecular bone area (BA/TA (%) from histologic sections. (D-H) A region of interest (ROI) extending 500 μm below the growth plate from IgG2a and aRANKL treated Gnas^R201H;R26^Tom mutant mice was analyzed for number of osteoclasts per bone surface (OC.N/BS (/mm)) and osteoclast surface per bone surface (OC.S/BS (%)) at either 3 wk (D-E) or 6 wk (F-I). (D) aRANKL reduced OC.N/BS and (E) OC.S/BS in 3-wk-old Gnas^R201H;R26^Tom mice. (F) Representative images of 6-wk-old IgG2a and aRANKL treated Gnas^R201H;R26^Tom mutant mice top, with higher magnification views of inset, bottom. (G) Quantification of the total tissue area of the ROI (T.Ar (mm²)). (H) The number of osteoclasts per total tissue area (OC.N/T.Ar (/mm²)). (I) The osteoclast surface per tissue area (OC.S/T.Ar (%)). Statistical analysis was performed using non-parametric Mann–Whitney t-test, where p < .05 was considered significant.

Paraffin sections from the femoral bone of Gnas^R201H;R26^Tom mice were TRAP stained to confirm inhibition of osteoclast formation by anti-RANKL treatment, as shown in representative images (Figure 2B). Anti-RANKL treatment of Gnas^R201H;R26^Tom mice clearly reduced but did not completely eliminate TRAP⁺ cells, compared to IgG2a treatment. Trabecular bone area per tissue area was increased (Figure 2C), while both the number of TRAP⁺ osteoclasts per bone surface and osteoclast surface per bone surface were significantly decreased by anti-RANKL treatment compared to IgG2a treated mutant mice (Figure 2D-E).

We next examined 6-wk-old Gnas^R201H;R26^Tom mice treated with either anti-RANKL antibody or IgG2a isotype from P6. We observed a visible reduction in TRAP⁺ osteoclasts (Figure 2F) in comparison to IgG2a treatment. Due to the increase in cartilaginous bone in the 6-wk-old Gnas^R201H;R26^Tom mice treated with anti-RANKL, we were unable to accurately quantify the bone surface, therefore TRAP⁺ osteoclasts were normalized to the total tissue area in both anti-RANKL and IgG2a treated mice. There was no change in the total tissue area of the ROI (Figure 2G). The number of osteoclasts per total tissue area was reduced, although not significantly (Figure 2H), while the osteoclast surface per total tissue area was significantly reduced in anti-RANKL treated Gnas^R201H;R26^Tom mutant mice (Figure 2I). Presence of the Sox9^CreERT and Rosa26^LSL-tdTomato alleles did not result in fibrosis or excess osteoclastogenesis, as shown in a representative image of Gnas⁺ stained with TRAP (Figure S1A).

Anti-RANKL treatment reduced fibrosis in Gnas^R201H mutant mice

Fibrosis was observed in Gnas^R201H;R26^Tom mice treated with IgG2a, and reduced by anti-RANKL treatment (Figure 3A). Fibrosis was quantified from Masson’s trichrome stained sections, from the distal femur ROI of 3-wk-old Gnas^R201H;R26^Tom mice treated with either IgG2a or anti-RANKL antibody. Representative images are shown of the ROI of the proximal metaphysis, below the growth plate, at 4x (Figure 3A) and 8x magnification (Figure 3B). A representative image of a Gnas WT mouse is shown for comparison (Figure 3C). Fibrosis was also observed in the femoral head (Figure 3D). Quantification of fibrosis within the ROI demonstrated that anti-RANKL treatment significantly reduced fibrosis (fibrosis area/total tissue area (%)) with no effect on the total tissue area (Figure 3E and F). Stromal cells are abundant in young animals, leading to measurable fibrosis in 3-wk-old Gnas⁺;R26^Tom mice (Figure 3E).

Figure 3.

Anti-RANKL treatment significantly reduced fibrosis in both 3- and 6-wk-old Gnas^R201H;R26^Tom FD mouse femurs. (A-C) Masson’s trichrome stained femur sections from 3-wk-old Gnas^R201H;R26^Tom mice treated with IgG2a isotype demonstrate fibrosis (arrows), which is absent in Sox9Cre^ERT;Gnas⁺;R26^TOM WT mice (C) and reduced in aRANKL treated animals. (A and B) Fibrosis in the area below the growth plate and (D) in the femoral head. Scale bar = 200 μm (A, C), 100 μm (B). (E and F) Quantification of fibrosis in the femur ROI extending 500 μm below the growth plate from IgG2a and aRANKL treated 3-wk-old Gnas^R201H;R26^Tom and Gnas⁺;R26^Tom control animals. (E) Fibrosis area per tissue area (Stromal.Ar/T.Ar (%)) and (F) tissue area (T.Ar (μm²)) for aRANKL and IgG2a treated 3-wk-old animals. (G and H) Masson’s trichrome stained femur sections from 6-wk-old Gnas^R201H;R26^Tom mice treated with IgG2a or aRANKL. Scale bar = 200 μm (G), 100 μm (H). (I and J) Fibrosis quantified in the femur ROI extending 500 μm below the growth plate in 6-wk-old Gnas^R201H;R26^Tom mice treated with IgG2a or aRANKL and Gnas⁺ littermates. (I) Stromal.Ar/T.Ar (%) is reduced by aRANKL without changing (J) T.Ar (μm²). Statistical analysis was performed using non-parametric Mann–Whitney t-test, where p < .05 was considered significant.

We observed a similar reduction in fibrosis in the 6-wk-old Gnas^R201H;R26^Tom mice with anti-RANKL treatment, compared to IgG2a. Representative images are shown from the femur of 6-wk-old mice at 4x and 8x magnification (Figure 3G and H). Anti-RANKL treatment significantly reduced fibrosis with no effect on the total tissue area (Figure 3I and J). In contrast to 3-wk-old animals, 6-wk-old WT Gnas⁺;R26^Tom animals have minimal fibrosis (Figure 3I). Representative images of Gnas⁺ femur sections are shown in Figure S1B.

Anti-RANKL treatment altered the phenotype of Gnas^R201H mutant cells but had no effect on mutant cell density

In the presence of Cre enzyme, both the Rosa26^LSL-tdTomato transgene and expression of the Gnas^R201H allele are activated. Thus, Gnas^R201H mutant cells expressing tdTomato can be detected by endogenous fluorescence. In order to determine if anti-RANKL treatment affected TOM⁺ mutant cell density, we quantified TOM⁺ cells in the bone marrow ROI of 3-wk-old mice treated with either anti-RANKL or IgG2a. TOM⁺ cells were quantified from four sections of varying depths within the coronal plane of the femur. Representative images show the presence of TOM⁺ cells below the growth plate in both anti-RANKL and IgG2a treated mutant mice (Figure 4A). There was a trend toward increased TOM⁺ cells per total tissue area in anti-RANKL treated mice, but this was not significant (Figure 4B). As the percentage of bone was increased in the anti-RANKL treated mice as mentioned above, we quantified TOM⁺ per total trabecular area and per bone marrow area (Figure 4C). We observed no change in TOM⁺ density for either parameter. These results suggest anti-RANKL treatment does not reduce the viability of Gnas^R201H mutant cells in this model.

Figure 4.

Anti-RANKL treatment of Gnas^R201H;R26^Tom mice shifts both WT and mutant cells to a more mature osteoblast phenotype, demonstrated by increased OCN and decreased SMA expression. (A-F) Representative confocal images of femur sections from 3-wk-old Gnas^R201H;R26^Tom treated with IgG2a or aRANKL. (A) The proximal metaphyseal region of the femur from 3-wk-old IgG2a and aRANKL treated Gnas^R201H;R26^Tom mutant mice shows endogenous TdTomato (TOM⁺) signal in Gnas^R201H mutant cells with DAPI nuclear staining. Scale bar = 100 μm. (B and C) Cells in the ROI extending 500 μm below the growth plate were quantified to determine (B) TOM⁺ cells per tissue area (TOM⁺/T.Ar (/μm²)) and (C) TOM⁺ cells per trabecular bone area (TOM⁺/B.Ar (/μm²)) and TOM⁺ cells per bone marrow area (TOM⁺/bone marrow area (/μm²)) from IgG2a and aRANKL treated Gnas^R201H;R26^Tom mice. (D-F) Images of Gnas^R201H;R26^Tom femur sections immunostained for SMA, OSX, or OCN, with DAPI nuclear stain. Endogenous tdTomato signal identifies Gnas^R201H mutant cells. (G-L) Cells in the ROI extending 500 μm below the growth plate were quantified to determine (G). Percentage of TOM⁺  Gnas^R201H mutant cells that co-expressed SMA (SMA⁺TOM⁺/TOM⁺ (%)); (H) Total SMA⁺ cells in the ROI tissue area (SMA⁺/T.Ar (/μm²)); (I) Percentage of TOM⁺  Gnas^R201H mutant cells that co-expressed OSX (OSX⁺TOM⁺/TOM⁺ (%)); (J) Total OSX⁺ cells in the ROI tissue area (OSX⁺/T.Ar (/μm²)); (K) Percentage of TOM⁺  Gnas^R201H mutant cells that co-expressed OCN (OCN⁺TOM⁺/TOM⁺ (%)); and (L) Total OCN⁺ cells per ROI tissue area (OCN⁺/T.Ar (/μm²)). Statistical analysis was performed using non-parametric Mann–Whitney t-test, where p < .05 was considered significant.

As RANKL inhibition has been suggested to alter the cellular profile in FD lesions, reducing SMA⁺ osteoblast precursor cells and increasing OCN⁺ osteoblasts,²⁶ we investigated the effect of anti-RANKL on the cellular phenotype of mutant Gnas^R201H cells. Immunofluorescence staining was performed for key markers involved in osteogenic differentiation. Cells expressing these markers were quantified in the ROI extending 500 mm below the growth plate, within the bone marrow cavity. Immunofluorescence staining showed a distribution of TOM⁺  Gnas^R201H mutant cells which stained positive for SMA (osteoblast precursor cells), OSX (immature osteoblasts), and OCN (mature osteoblasts) (Figure 4D-F). Treatment with anti-RANKL reduced the percentage of TOM⁺  Gnas^R201H mutant cells expressing SMA (Figure 4G), without changing the number of SMA⁺ cells within the ROI (Figure 4F). While anti-RANKL did not change the percentage of mutant cells nor total OSX⁺ cells expressing the intermediate marker OSX (Figure 4I and J), the percentage of TOM⁺  Gnas^R201H mutant cells that expressed the mature osteoblast marker OCN, as well as the number of OCN⁺ cells per tissue area were increased with treatment (Figure 4K and L). As expected, anti-RANKL increased the total bone surface within the ROI (Figure S2A). Total tissue area of the ROI from the sections from varying depths of the femur was unchanged (Figure S2B), consistent with the analysis of TRAP-stained sections (Figure 2G).

Anti-RANKL treatment altered the cellular phenotype of FD lesions

Fibrotic lesions were defined in this model as areas with significant numbers of TOM⁺ mutant cells within the bone marrow of femurs from Gnas^R201H mutant mice. Serial sections were analyzed and cells within a 500 μm × 500 μm region of a lesional ROI were quantified. Consistent with the findings in the larger ROI described above, there was no change in the number of TOM⁺  Gnas^R201H mutant cells in the lesional ROI with treatment (Figure S2C). Representative images show TOM⁺  Gnas^R201H mutant cells and positive immunofluorescence staining for SMA, OSX, and OCN within the lesional ROI of 3-wk-old mice (Figure 5A-C). The number of positive SMA, OSX, and OCN cells were quantified within the ROI, as well as the percentage of Gnas^R201H mutant cells (TOM⁺) co-stained with each marker. Treatment with anti-RANKL significantly reduced the number of SMA⁺ WT (TOM⁻) cells within the lesional ROI, with a trend toward a decrease in SMA⁺ TOM⁺ mutant cells (Figure 5D). Consistent with Figure 4E, anti-RANKL reduced the percentage of TOM⁺  Gnas^R201H mutant cells, which expressed SMA (Figure 5E), suggesting anti-RANKL treatment reduces the number of TOM⁺ mutant cells with an immature osteoblast phenotype. Similar to the results from the larger ROI described previously, anti-RANKL did not affect the number of OSX⁺ cells or the percentage of TOM⁺  Gnas^R201H mutant cells expressing OSX⁺ within the lesional ROI (Figure 5F and G). Additionally, the number of OCN⁺ cells, as well as the percentage of OCN⁺ TOM⁺  Gnas^R201H mutant cells was significantly increased by anti-RANKL treatment in the lesional ROI (Figure 5H and I). These results support the hypothesis that inhibition of RANKL stimulated the differentiation of both WT and Gnas^R201H mutant osteogenic cells.

Figure 5.

Lesional cells in aRANKL treated Gnas^R201H;R26^Tom fibrous dysplasia mice demonstrate increased expression of mature osteoblast marker OCN and decreased SMA expression. Representative images of fibrotic lesional areas in femur sections from 3-wk-old (A-I) and 6-wk-old (J-O) IgG2a and aRANKL treated Gnas^R201H;R26^Tom mice. A ROI restricted to fibrotic lesions was analyzed and the percentage of cells expressing specific osteoblast markers was quantified in IgG2a and aRANKL treated mice. (A-C) Immunofluorescence staining of 3-wk-old femur lesions for (A) SMA, (B) OSX, and (C) OCN with endogenous TdTomato signal identifying Gnas^R201H mutant cells and DAPI nuclear stain (scale bar = 50 μM). (D-I) (D) Total SMA⁺ cells per lesion area (SMA⁺/T.Ar (/μm²)), SMA⁺ mutant cells per lesion area (SMA⁺TOM⁺/T.Ar (/μm²)), and SMA⁺ WT cells per lesion area (SMA⁺TOM⁻/T.Ar (/μm²)) (E). Percentage of all TOM⁺  Gnas^R201H mutant cells that co-expressed SMA within the lesion (SMA⁺TOM⁺/TOM⁺ (%)); (F) Total OSX⁺ cells per lesion area (OSX⁺/T.Ar (/μm²)) and OSX⁺ mutant cells per lesion area (OSX⁺TOM⁺/T.Ar (/μm²)). (G) Percentage of all TOM⁺  Gnas^R201H mutant cells that co-expressed OSX within the lesion (OSX⁺TOM⁺/TOM⁺ (%)). (H) Total OCN⁺ cells per lesion area (OCN⁺/T.Ar (/μm²)) and OCN⁺ mutant cells per lesion area (OCN⁺TOM⁺/T.Ar (/μm²)). (I) Percentage of TOM⁺  Gnas^R201H mutant cells that co-expressed OCN within the lesion (OCN⁺TOM⁺/TOM⁺ (%)). (J and K) Immunofluorescence staining of 6-wk-old femur lesions for (J) SMA and (K) Ki67 (scale bar = 50 μM). (L) Total SMA⁺ cells per lesion area (SMA⁺/T.Ar (/μm²)), SMA⁺ mutant cells per lesion area (SMA⁺TOM⁺/T.Ar (/μm²)), and SMA⁺ WT cells per lesion area (SMA⁺TOM⁻/T.Ar (/μm²)). (M) Percentage of TOM⁺  Gnas^R201H mutant cells that co-expressed SMA within the lesion (SMA⁺TOM⁺/TOM⁺ (%)). (N) Total Ki67⁺ cells per lesion area (Ki67⁺/T.Ar (/μm²)) and Ki67⁺ mutant cells per lesion area (Ki67⁺TOM⁺/T.Ar (/μm²)). (O) Percentage of TOM⁺  Gnas^R201H mutant cells that co-expressed Ki67 within the lesion (Ki67⁺TOM⁺/TOM⁺ (%)). Statistical analysis was performed using non-parametric Mann-Whitney t-test, where p < .05 was considered significant.

Next, we examined the effect of anti-RANKL treatment on the cellular phenotype of Gnas^R201H mutant cells in lesions from 6-wk-old mice. Representative images of the lesional area ROI show positive staining for SMA (Figure 5J), along with endogenous tdTomato signal marking Gnas^R201H mutant cells. There was no change in the number of TOM⁺ cells within the lesional ROI with treatment (Figure S2D). Similar to 3-wk-old animals, anti-RANKL treatment decreased the percentage of TOM⁺  Gnas^R201H mutant cells that were SMA⁺ within the lesional ROI (Figure 5K), although there was no significant change in the number of SMA⁺ TOM⁺  Gnas mutant cells or WT SMA⁺ TOM⁻ cells (Figure 5L). These data support the hypothesis that anti-RANKL treatment is stimulating the differentiation of the Gnas^R201H mutant cell population.

As denosumab treatment was recently shown to reduce proliferation in FD lesions,²⁵ we tested if anti-RANKL treatment affected the proliferation of Gnas^R201H mutant cells in this model. Immunofluorescence staining was performed to detect Ki67, a marker expressed by actively proliferating cells, in femur sections from 6-wk-old mice (Figure 5M). Quantification of Ki67 staining in the 3-wk-old mice showed no change in either the percentage of TOM⁺ mutant cells expressing Ki67⁺ or the number of Ki67⁺ cells per tissue area in anti-RANKL treated animals (data not shown), possibly due to the active modeling still occurring in these younger mice. In contrast, anti-RANKL treatment in the 6-wk-old cohort significantly reduced the total number of Ki67⁺ cells within the lesional ROI (Figure 5N) due primarily to a decrease in TOM⁻ Ki67⁺ cells, as there was no change in both the number of TOM⁺ Ki67⁺ mutant cells, and the percentage of TOM⁺ mutant cells that were Ki67⁺ (Figure 5O). These results are consistent with the lack of effect of anti-RANKL treatment on TOM⁺  Gnas^R201H mutant cell density within lesions in either 3-wk-old or 6-wk-old cohorts (Figure S2C and D). This data suggests that RANKL inhibition reduces fibrosis primarily by reducing the proliferation of WT cells within lesions.

Loss of RANKL production by Gnas^R201H mutant osteoprogenitor cells does not prevent lesion formation and is not sufficient to inhibit the accumulation of SMA⁺ cells in fibrotic bone lesions

Fibrous dysplasia lesions express abundant RANKL, as shown by immunohistochemistry.^13,19 Osteoblastic expression of RANKL is stimulated by activation of the Gs coupled PTHR1,^15–17 and we hypothesized that aberrant expression of RANKL by Gnas^R201H mutant cells could be required for FD lesion formation. To test this hypothesis, we generated mice deficient for Rankl in Gnas^R201H mutant cells by inducing Cre activity in Sox9^CreERT;Gnas^(R201H)fl/+;Rosa26^LSL-tdTomato;Rankl^fl/fl mice and Rankl^fl/+ controls with tamoxifen treatment. The resultant Gnas^R201H;Rankl^Δ;R26^Tom and Gnas^R201H;Rankl⁺;R26^Tom mice were analyzed for development of fibrotic lesions at 3 and 6 wk of age.

To confirm successful deletion of Rankl driven by Sox9^CreERT, immunofluorescence staining for RANKL was performed on the 3-wk-old cohort of Gnas^R201H;Rankl^Δ;R26^Tom mice and Gnas^R201H;Rankl⁺;R26^Tom with RUNX2 co-staining used to identify osteoblasts (Figure S3A). The percentage of TOM⁺  Gnas^R201H mutant cells positive for RANKL was significantly reduced in Gnas^R201H;Rankl^Δ;R26^Tom mice, suggesting that Rankl was successfully deleted in Gnas^R201H mutant cells (Figure S3B). This is not due to a decrease in RUNX2⁺ cells, as there was no significant change in the percentage of RUNX2⁺ cells within the ROI between genotypes (Figure S3C).

In order to determine if Rankl deletion in Gnas^R201H;Rankl^Δ;R26^Tom mice altered the production of osteoclasts in this model of FD, histological analysis was performed using TRAP staining. TRAP⁺ osteoclasts were detected within the femoral bone marrow cavity of both Gnas^R201H;Rankl^Δ;R26^Tom mice and Gnas^R201H;Rankl⁺;R26^Tom 3-wk-old mice (Figure 6A). There was no significant difference in osteoclast number or osteoclast surface per bone surface in Gnas^R201H;Rankl^Δ;R26^Tom compared to Gnas^R201H;Rankl⁺;R26^Tom mice (Figure 6B and C). Furthermore, analysis of Masson’s Trichrome stained sections showed no difference in fibrosis in the ROI below the growth plate of 3-wk-old Gnas^R201H;Rankl^Δ;R26^Tom mice, compared to Gnas^R201H;Rankl⁺;R26^Tom mice with intact Rankl (Figure 6D and E). We next analyzed the femurs of older mice to confirm this unexpected lack of effect of Rankl deletion in mutant cells. As seen in younger mice, loss of Rankl in Gnas^R201H mutant cells did not alter osteoclast number or surface in 6-wk-old mice (Figure 6F-H), nor reduce fibrosis (Figure 6I and L). In this cohort, we also assessed fibrosis on femoral sections from Gnas⁺;Rankl^Δ;R26^Tom and Gnas⁺;Rankl⁺;R26^Tom mice to confirm deletion of Rankl had no effect in Gnas⁺ WT mice. Sox9Cre- were included as an additional control. Loss of Rankl on the Gnas⁺ WT background did not qualitatively affect bone phenotype, as shown in representative images of Gnas⁺ femurs stained with Masson’s Trichrome (Figure S3D) or in osteoclastogenesis by TRAP staining (data not shown). Altogether, these data suggest that genetic deletion of Rankl in Gnas^R201H mutant cells did not affect the development of fibrotic bone lesions in this model of FD.

Figure 6.

Rankl deletion in Gnas^R201H;R26^Tom mutant cells does not reduce fibrosis, alter SMA expression, nor reduce osteoclast formation in 6-wk-old mice. Formation of fibrotic lesions in Gnas^R201H;Rankl⁺;R26^Tom and Gnas^R201H;Rankl^Δ;R26^Tom femurs was assessed by histology at (A-E, M-O) 3-wk and (F-L) 6-wk of age. Analysis was performed on the ROI 500 μm below the growth plate (A-L) or restricted to lesional area (M-O). (A) Representative images of TRAP stained femur sections from 3-wk-old animals counter stained with fast green. Scale bar = 200 μm. (B) Number of osteoclasts per bone surface (OC.N/BS (/mm)) and (C) osteoclast surface per bone surface (OC.S/BS (%)) within the ROI. (D) Representative images of Masson’s trichrome stained proximal metaphyseal region of the femur showed fibrosis (arrows) below the growth plate in both genotypes. Scale bar = 200 μm. (E) Fibrosis area per tissue area (Stromal.Ar/T.Ar (%)) for Gnas^R201H;Rankl^Δ;R26^Tom and Gnas^R201H;Rankl⁺;R26^Tom mice. (F) Representative images of TRAP stained (pink) femur sections from 6-wk-old animals counter stained with fast green. Scale bar = 200 μm. (G) Number of osteoclasts per bone surface (OC.N/BS (/mm)), and (H) osteoclast surface per bone surface (OC.S/BS (%)) within the ROI. (I and J) Representative images of Masson’s trichrome stained proximal metaphyseal region of the femur showed fibrosis below the growth plate in both genotypes. Scale bar = 200 μm (I) and 100 μm (J). (K) Fibrosis area per tissue area (Stromal.Ar/T.Ar (%)), and (L) total tissue area (Stromal.Ar./T.AR(%)) for Gnas^R201H;Rankl^Δ;R26^Tom and Gnas^R201H;Rankl⁺;R26^Tom mice, compared to Gnas⁺ and Cre negative control littermates. (M) Representative images of SMA immunostained femur sections of 3-wk-old Gnas^R201H;Rankl⁺;R26^Tom and Gnas^R201H;Rankl^Δ;R26^Tom mice. Scale bar = 50 μm. (N) Percentage of TOM⁺  Gnas^R201H mutant cells that co-expressed SMA within the lesion (SMA⁺TOM⁺/TOM⁺ (%)). (O) Total SMA⁺ cells per total cells (SMA⁺/DAPI⁺ (%)), SMA⁺ WT cells per total cells (SMA⁺TOM⁻/DAPI⁺ (%)), and SMA⁺ mutant cells per total cells (SMA⁺TOM⁺/DAPI⁺ (%)). Statistical analysis of osteoclast parameters at both ages was performed using non-parametric Mann–Whitney t-test. Analysis of fibrosis was performed using an unpaired Student’s t-test for 3-wk-old animals and a one-way ANOVA with Tukey’s post-test for 6-wk-old animals. Analysis of SMA⁺ cells was performed using an unpaired Student’s t-test. For all tests p < .05 was considered significant.

In order to determine if Rankl deletion in Gnas^R201H mutant cells altered the cellular phenotype of lesions, we performed immunofluorescence staining for SMA, and analyzed the ROI below the growth plate in the femurs of 3- and 6-wk-old Gnas^R201H;R26^Tom mice with and without Rankl. In the 3-wk-old cohort, there was no difference in the percentage of TOM⁺  Gnas^R201H mutant cells expressing SMA within the ROI between the two genotypes (Figure 6N). Furthermore, total SMA⁺ cells, the number of SMA⁺ TOM⁻ WT cells and SMA⁺ TOM⁺  Gnas^R201H mutant cells were unchanged in Gnas^R201H;Rankl^Δ;R26^Tom mice compared to Gnas^R201H;Rankl⁺;R26^Tom (Figure 6O). Representative images of the 6-wk-old mice similarly show no qualitative change in SMA staining by genotype (Figure S3E).

Thus, a significant reduction in RANKL expression solely in Gnas^R201H mutant cells was not sufficient to inhibit the accumulation of SMA⁺ cells and formation of fibrotic bone lesions, or the formation of abundant lesional osteoclasts in this model of FD. This suggests that other cells within the bone microenvironment produce sufficient RANKL to stimulate osteoclast formation. Consistent with this, circulating RANKL levels were similar in Gnas^R201H;Rankl^Δ;R26^Tom and Rankl WT mice (Figure S3F). Re-analysis of scRNA sequencing data from human FD organoid cultures,³³ which contain a heterogenous population of cells, demonstrated that osteoblasts, rather than SMA⁺ early osteoprogenitors, are the most significant source of RANKL (Figure S3G and H). As a minority of Gnas^R201H mutant cells express mature osteoblast markers, this is supportive of our finding that loss of Rankl in Gnas^R201H mutant cells did not alter circulating RANKL nor osteoclast formation. While our data suggests that RANKL expression by Gnas^R201H mutant cells is not required for lesion formation, our data and that of others supports a critical role for RANKL in the pathogenesis of FD bone lesions. Further investigation is needed to understand how Gnas^R201H mutant cells induce RANKL expression in other cells present within fibrotic lesions.

Discussion

Fibrous dysplasia is a rare disease, where postzygotic gain-of-function mutations in Gnas, encoding the Gs alpha subunit, lead to the formation of fibro-osseous lesions with abundant osteoclasts in the skeleton. These lesions compromise the integrity of the skeletal structure; the failure to produce mature lamellar bone leads to bone deformity and/or fractures. Fibrous dysplasia bone lesions exhibit high levels of RANKL expression, and the skeletal burden of FD has been shown to correlate with serum RANKL levels.^19,20 Treatment with denosumab, a neutralizing antibody to RANKL, has shown promising results in reducing lesional activity in patients with FD.^20–23 Similarly, pre-clinical models have also shown a reduction in fibrosis with anti-RANKL treatment.^34,35 Denosumab actively prevents RANKL binding to its receptor RANK,³⁶ preventing osteoclast differentiation. As expected, denosumab reduces osteoclastic bone resorption measured by CTX in FD patients¹⁹ and decreases lesional osteoclasts,²⁶ yet how this translates into reducing fibrosis is unknown.

RANKL inhibition could work indirectly by reducing osteoclasts and thus removing osteoclast-derived factors that promote the accumulation of WT cells and fibrosis. However, bisphosphonates, which also inhibit osteoclast activity, have not demonstrated efficacy in a placebo controlled randomized study and were less effective than RANKL inhibition in a pre-clinical model.^37,38 Alternatively, antibodies to RANKL could act directly on RANKL-expressing osteoblastic cells. For instance, they might induce antibody-dependent cell cytotoxicity or inhibit the RANK/RANKL pathway in bone marrow stromal cells, promoting osteogenesis.³⁹ Furthermore, RANKL expression is not confined to skeletal tissue, and it cannot be ruled out that the relevant target of anti-RANKL could be RANKL-dependent specialized epithelial tissues.^40–42

We utilized a mouse model of FD that mimics the mosaicism of the human disease. This model post-zygotically induces expression of the activating R201H mutation of GNAS into a subset of lineage traced osteochondral progenitor cells, to gain insight into the effect of anti-RANKL on GNAS mutant and WT cells independently. In these young Gnas^R201H;R26^Tom FD mice (3 and 6 wk of age), fibrotic lesions formed within the long bones, as detected by Masson’s trichrome staining. Regions where fibrotic lesions were identified also showed abundant SMA⁺ stromal cells—a characteristic of FD bone lesions¹¹—as well as TRAP⁺ osteoclasts. Treatment with anti-RANKL antibody reduced osteoclast density and the formation of fibrotic lesions, with increases in bone area fraction as expected with inhibition of osteoclast activity. A similar anti-fibrotic effect was observed in other mouse models of FD. In a study by de Castro et al., where FD was induced in a conditional tetracycline inducible model in mice (Tet-Gα_s^R201C/Prrx1-Cre/LSL-rtTA-IRES-GFP), mice were treated with a longer dose regimen of a different anti-RANKL antibody (BE0191, Bioxcell Lebanon) and showed both reduced fibrosis and increased number of OCN⁺ osteoblasts.²⁶ Similar results were observed in a tetracycline inducible FD mouse model (Tet-Gα_s^R201C/Prrx1-Cre/LSL-rtTA-IRES-GFP), where mice were treated with a chemical inhibitor of RANKL.³⁵

We assessed the impact of anti-RANKL antibody treatment on Gnas^R201H mutant cells using the lineage tracer tdTomato, in Gnas^R201H;R26^Tom mice treated from P6 to either 3 or 6 wk of age. Anti-RANKL did not reduce the density of mutant cells normalized either to bone area or tissue area, suggesting that treatment does not induce mutant cell death. We next examined SMA expression as a marker of osteoblastic/stromal precursors that characterize FD lesions. Studies have shown that the abundance of SMA⁺ stromal cells present in FD directly contributes to the formation of the fibrotic tissue.¹² Smooth muscle actin is expressed by both Gnas mutant and genetically WT stromal cells that act in concert to produce a disorganized collagen matrix.^43,44 Anti-RANKL treatment reduced SMA⁺ stromal cells within the bone marrow of the femur and reduced the percentage of mutant cells expressing SMA. In contrast, treatment increased the number of mature OCN expressing osteoblasts and increased the percentage of OCN⁺ mutant cells. The concordant increase in mature osteoblasts and decrease in immature SMA⁺ osteoblastic/stromal cells suggests that anti-RANKL promotes osteoblast differentiation in FD bone lesions. We observed that the percentage of OCN⁺ cells increased, while percentage of SMA⁺ cells decreased in both Gnas^R201H mutant cells (TOM⁺) and in WT stromal cells (TOM⁻), suggesting inhibition of RANKL is stimulating the differentiation of both mutant and WT stromal cell populations. Both reverse signaling through RANKL, and signaling through the receptor RANK on osteoblasts have been described to modulate osteoblast differentiation,^27,39 so it remains unclear whether anti-RANKL promotes osteoblast differentiation directly or indirectly through its effect on osteoclasts.

Aberrant proliferation of Gnas mutant and WT stromal cells have also been shown to contribute to fibrosis in patients with FD.^45,46 Anti-RANKL treatment did not decrease Ki67⁺ cells in 3-wk-old mice, likely due to the active modeling occurring in these young animals. However, there was a significant reduction in the density of Ki67⁺ cells in FD lesions of 6-wk-old mice treated with anti-RANKL. The anti-proliferative effect was primarily due to the WT cells within the lesions, as there was no change in the percentage of mutant cells expressing Ki67, or the number of Ki67⁺  Gnas^R201H mutant cells within the lesions in these mice.

We observed similar changes with anti-RANKL treatment as previously reported, including an increase in mature OCN⁺ osteoblastic cells, suggesting that treatment drives osteoblast differentiation,³⁵ as well as the reduction in stromal cell proliferation.²⁶ While our results are consistent with prior reports in both mouse models of FD and in human FD bone biopsies, importantly, here we show anti-RANKL promotes osteoblast differentiation of both WT and mutant cell populations in this model of FD.

We also tested the hypothesis that RANKL induced by constitutively active Gs alpha in Gnas^R201H mutant cells, initiates lesion formation by driving differentiation of osteoclasts, which then could recruit WT osteoblastic cells into lesions through the action of coupling factors.^47,48 We generated Sox9^CreERT;Gnas^R201Hfl/+;Rosa26^Tdtomato;Rankl^fl/fl mice, in which tamoxifen drives both Gnas^R201H expression and Rankl deletion in Sox9 expressing osteochondral progenitors. Unexpectedly, we found that deletion of Rankl in Gnas^R201Hfl/+ mutant cells neither reduced the number of TRAP⁺ osteoclasts or ameliorated fibrosis compared to mice with intact Rankl. This suggests in this model of FD that the production of RANKL by Sox9^CreERT;Gnas^R201H mutant cells is not critical for initiating FD bone lesions and that other cells within the bone microenvironment are able to produce the aberrant RANKL observed in this disease. Analysis of RNA-seq data from FD patient derived organoids containing a heterogenous population of cells, including stromal cells and osteoblasts, showed osteoblasts express higher levels of RANKL in FD than SMA⁺ fibroblastic cells.³³

Limitations of this study include the use of the Sox9Cre^ERT that labels only a small subset of the cells present within the bone microenvironment. While this Cre model has been used previously and is known to label early postnatal osteoblast progenitor cells,²⁹ the precise cell lineage in which post-natal mutation causes FD in humans in not defined, and the phenotype of animal models clearly depends on the lineage in which mutant Gnas is expressed. For example, mice expressing Col1a1-GsαR201C do not develop a FD phenotype.⁴⁹ Another limitation of this study is the variable induction of Gnas^R201H recombination with administration of tamoxifen through breast milk. We observed a wide variation in the number of lineage traced Gnas^R201H mutant (TOM⁺) cells across samples, regardless of treatment, which likely contributed to the variation in fibrosis and SMA⁺ cell density quantified in this model. Additionally, the age of the mice in this study, while reflecting a similar biological age to children and young adolescents with FD, are undergoing active modeling that occurs with development and inhibition of RANKL at this stage of skeletal development can impair bone structure. Reoccurrence of FD lesions after withdrawal of RANKL inhibition in a mouse model has previously been observed,³⁴ and examining the consequence of withdrawal of treatment in this model should be investigated in future.

In conclusion, RANKL inhibition prevents the formation of fibrotic lesions in a mouse model of FD by promoting the differentiation of both mutant and WT stromal cells and reducing proliferation of WT cells. Deletion of Rankl in Gnas^R201H mutant cells alone did not prevent the formation of lesions in this model of FD, suggesting other cells within the bone microenvironment, including the WT stromal cells, osteoblasts, and osteocytes, are sufficient to produce the aberrant RANKL observed in this disease. Further research is required to determine how Gnas mutant cells induce RANKL expression in bone lesions, the cellular origin of RANKL in these lesions, and the relative contribution of osteoclast inhibition to the efficacy of anti-RANKL treatment in FD.

Author contributions

Renee T. Ormsby (Conceptualization, Data curation, Formal analysis, Investigation, Writing—original draft, Writing—review & editing), Yongxing Zhang (Data curation, Formal analysis), Cole Hodys (Data curation, Formal analysis), Lella A. Wake (Data curation, Formal analysis, Writing—review & editing), Samantha Menendez Perez (Data curation, Formal analysis), Kelly Tsang (Data curation, Formal analysis), Yingzi Yang (Conceptualization, Methodology), and Julia F. Charles (Conceptualization, Funding acquisition, Project administration, Supervision, Writing—review & editing)

Funding

This project was supported by Million Dollar Bike Ride (FD/MAS Alliance) grant number MDBR-20-116-FD/MAS 2020 and R21 AR080849 from the National Institutes of Health (NIH) (J.F.C.), a pilot grant from the P30 AR070253 funded Joint Biology Consortium, and Orthopaedic Surgery Department Funds.

Conflicts of interest

Alexion/AstraZeneca Rare Diseases has provided investigator-initiated research funding to J.F.C. The remaining authors have nothing to disclose.

Data availability

The unprocessed images from which this data were generated will be deposited to Harvard Dataverse and will be available at https://dataverse.harvard.edu/dataverse/GnasR201H within 6 months of publication.