Induced Gnas^R201H expression from the endogenous Gnas locus causes fibrous dysplasia by up-regulating Wnt/β-catenin signaling

Significance

Understanding molecular and cellular mechanisms of rare genetic diseases provides invaluable insights into the human biology and pathology of both rare and related common diseases. Fibrous dysplasia (FD) is a mosaic disease resulting from postzygotic activating mutations of GNAS. The mouse models we created allowed us to precisely model FD by expressing the FD Gα_s mutation under the control of its endogenous genetic locus. We found in our FD mouse models that up-regulated Wnt/β-catenin signaling resulted in impaired differentiation and proliferation of bone marrow stem cells, which in turn caused marrow fibrosis. Our work provides a solid new foundation for therapeutic development of FD and understanding the principles whereby Gα_s signaling governs bone formation and maintenance and bone marrow stromal cell differentiation.

Abstract

Fibrous dysplasia (FD; Online Mendelian Inheritance in Man no. 174800) is a crippling skeletal disease caused by activating mutations of the GNAS gene, which encodes the stimulatory G protein Gα_s. FD can lead to severe adverse conditions such as bone deformity, fracture, and severe pain, leading to functional impairment and wheelchair confinement. So far there is no cure, as the underlying molecular and cellular mechanisms remain largely unknown and the lack of appropriate animal models has severely hampered FD research. Here we have investigated the cellular and molecular mechanisms underlying FD and tested its potential treatment by establishing a mouse model in which the human FD mutation (R201H) has been conditionally knocked into the corresponding mouse Gnas locus. We found that the germ-line FD mutant was embryonic lethal, and Cre-induced Gnas FD mutant expression in early osteochondral progenitors, osteoblast cells, or bone marrow stromal cells (BMSCs) recapitulated FD features. In addition, mosaic expression of FD mutant Gα_s in BMSCs induced bone marrow fibrosis both cell autonomously and non-cell autonomously. Furthermore, Wnt/β-catenin signaling was up-regulated in FD mutant mouse bone and BMSCs undergoing osteogenic differentiation, as we have found in FD human tissue previously. Reduction of Wnt/β-catenin signaling by removing one Lrp6 copy in an FD mutant line significantly rescued the phenotypes. We demonstrate that induced expression of the FD Gα_s mutant from the mouse endogenous Gnas locus exhibits human FD phenotypes in vivo, and that inhibitors of Wnt/β-catenin signaling may be repurposed for treating FD and other bone diseases caused by Gα_s activation.

Fibrous dysplasia (FD) of bone (Online Mendelian Inheritance in Man no. 174800) is a severe form of skeletal disorder resulting in deformity, fracture, and pain in the affected bone. FD is well-characterized by bone marrow fibrosis; the bone marrow space is devoid of both hematopoietic tissue and adipocytes and replaced with fibrotic tissue. FD bone also exhibits abnormal architecture (“Chinese writing” pattern), structure, and mineral content of bone trabeculae (1–3). These complex changes result in a mechanically incompetent, brittle, and fracture-prone bone that can cause wheelchair confinement of severely affected individuals. FD is a rare skeletal genetic disorder caused by mosaic activating mutations (R201H or R201C) of the α-subunit of stimulatory G protein (Gα_s) encoded by GNAS (4–6). The activating mutant Gα_s loses inherent GTPase activities and remains in a constitutively active form that stimulates excessive cAMP production (7). FD occurs in isolation or with other clinical features such as skin pigmentation and endocrine dysfunction in McCune–Albright syndrome (MAS) (8, 9). Lack of inheritance of FD/MAS (4, 5, 10, 11) is likely due to embryonic lethality caused by germ line-transmitted activating GNAS mutations, which can only survive through mosaicism (12).

There is no cure for FD, as the molecular and cellular mechanisms of this skeletal disease that can be devastating in some cases remain largely unknown. Mouse models are indispensable tools for elucidating the natural history of human diseases and designing and testing novel treatments. Better understanding of FD is essential to providing new insights into marrow fibrosis and the regulation of osteoblast differentiation and maturation from bone marrow stromal cells (BMSCs, also called bone marrow-derived stem cells), but the lack of appropriate animal models has severely hampered research advancement and therapeutic development for FD. The existing in vivo models are either based on xenotransplantation of GNAS-mutated human skeletal progenitor cells into immunocompromised mice (11) or transgenic mouse models in which either an engineered Gα_s-coupled receptor or a mutated rat Gnas transgene was driven by artificial promoters (13–15). As none of these models were able to accurately recapitulate pathophysiological characteristics of human FD, the development of a “knockin” (KI) mouse model in which the corresponding mouse mutant Gα_s can be expressed from its endogenous locus is absolutely necessary.

Here we have successfully created a KI mouse line, Gnas^f(R201H), in which the human FD mutation (R201H) has been conditionally “knocked into” the corresponding mouse Gnas locus to allow FD mutant Gnas expression from its endogenous genetic locus with temporal and tissue specificity upon Cre induction. Using this mouse line, we show that FD mutant Gnas expression in both osteochondral progenitor cells and osteoblast progenitor cells in the Prrx1 and Osx lineages, respectively, replicated human FD phenotypes. Furthermore, mosaic analysis showed that FD mutant Gnas expression in Sox9⁺ BMSCs exhibits both cell-autonomous and non–cell-autonomous activities in causing FD phenotypes. Molecularly, as we have found in human FD bone previously, the FD mutant Gα_s up-regulated Wnt/β-catenin signaling in bone and BMSCs, reduction of which significantly rescued FD phenotypes in Gnas^R201H mice, providing important therapeutic insights.

Results

The Gnas^R201H Mutation Is Embryonic Lethal.

Previously established mouse models have contributed to our current understanding of FD (13–15); however, due to the transgenic nature of the reported FD mouse lines and some inconsistent results obtained from these models, many outstanding questions remain unanswered. To further understand the critical roles of Gnas in multiple aspects of bone formation, maintenance, and resorption, including regulation of BMSCs as well as onset, progression, and cellular origins of FD, we have established a mouse line that allows conditional expression of mouse mutant Gα_s containing a corresponding human FD mutation, R201H (4–6), from the endogenous mouse Gnas locus. This conditional KI Gnas allele [hereafter denoted Gnas^f(R201H)] was generated by homologous recombination in embryonic stem cells (Fig. S1A). The Gnas coding exon sequence is highly conserved between human and mouse (>90% sequence identity) both at nucleotide and amino acid levels (Fig. S2 A and B). The correctly targeted Gnas^f(R201H) allele expressed mutant Gnas, which we denote as Gnas^R201H, only in the presence of Cre expression; otherwise, it expressed a wild-type Gnas containing the minigene cassette (Figs. S1A and S2C).

Mosaicism of the human FD GNAS R201H mutation suggested that GNAS^R201H germ-line transmission may cause embryonic lethality. The Gnas^f(R201H) allele we generated allowed us to test this directly by expressing Gnas^R201H in the early epiblast [before embryonic day (E)6.5] using the Sox2-Cre line that expresses Cre in mouse oocytes (16). Indeed, all Gnas^f(R201H)/+; Sox2-Cre mutants with Gnas^R201H heterozygous expression were found to be much smaller than wild-type littermate controls at E10.5 and dead some time between E12.5 and E15.5 (Fig. 1A). Expression of the mutant allele was confirmed by sequencing the Gnas cDNA made from whole embryos (Fig. S2C). Because Gα_s signaling regulates both Wnt and Hedgehog (Hh) signaling (17, 18), we examined signaling activities of the Wnt and Hh pathways by whole-mount in situ hybridization as well as real-time quantitative PCR (qRT-PCR). Our results showed that while there was a significant up-regulation in the expression of Tcf1 and Axin2 (Fig. 1 B and C), which are Wnt/β-catenin target genes (19), expression of Gli1 and Ptch1 (Fig. 1 D and E), which are Hh signaling target genes (20–22), was down-regulated. Shh expression itself was not altered in the mutant embryos (Fig. S2D). Therefore, Gnas^R201H expression in early embryos enhanced Wnt/β-catenin signaling, while it suppressed Hh signaling. The mutant embryos at E10.5 exhibited severe heart development retardation (Fig. S2E) as well as increased cell death and reduced cell proliferation in the entire embryo (Fig. S2 F and G). As Wnt/β-catenin and Hh signaling are two key signaling pathways required to regulate various aspects of early embryonic development (19, 23–26), it is likely that dysregulation of both by the activating R201H Gα_s has caused embryonic lethality.

Fig. 1.

Germ-line expression of Gnas^R201H causes embryonic lethality. (A) Lateral view of E10.5 and E12.5 littermate embryos with the indicated genotypes. (B and D) Representative images of whole-mount in situ hybridization performed with E9.5 embryos. (C and E) qRT-PCR analysis of the indicated gene expression from RNA isolated from whole embryos at E9.5. *P < 0.05; data are presented as mean ± SD. (Scale bars, 100 mm.)

Gnas^R201H Expression in Osteochondral Progenitor Cells Exhibits FD Features.

As FD-causing GNAS mutations are thought to occur at an early embryonic stage, we hypothesized that Gnas^R201H expression in early mouse osteochondral progenitor cells should phenocopy human FD. To test our hypothesis, Gnas^f(R201H) mice were crossed with the Prrx1-Cre line, in which Cre is expressed in early limb bud mesenchyme cells (27). All Prrx1-Cre; Gnas^f(R201H)/+ pups were born alive at Mendelian ratio and were readily distinguishable from wild-type littermates at birth due to their shorter limbs. The Prrx1-Cre; Gnas^f(R201H)/+ mice manifested marked shortening of the limbs with severe bone deformity (Fig. 2A) throughout their adulthood (Fig. S3 A, B, and D), while the Gnas^f(R201H)/+ littermate controls were normal and used as wild-type control in all analyses. Despite the short and misshapen limbs, bones in the trunk were normal and the mutant mice survived for more than 7 months with no sign of other gross sickness (Fig. S3 C and D). In Prrx1-Cre; Gnas^f(R201H)/+ mice, the bone marrow space was occupied by woven trabecular bone and there was no cortical bone (Fig. 2B and Fig. S3D). Histomorphometric analysis showed that the trabecular thickness and trabecular number were significantly increased, whereas trabecular spacing was decreased in mutant bone (Fig. S3 E and F).

Fig. 2.

Prrx1-Cre; Gnas^f(R201H)/+ mutant mice exhibit FD phenotypes. (A and B) Representative µCT scans of forelimbs from P21 mice with the indicated genotypes. Long bones were deformed in the mutant. Longitudinal and cross-sectional images of the humerus (B) showed that bone marrow space was occupied by trabecular bone in mutant mice. (C) Representative H&E staining of the trabecular region of P21 mouse humerus. Boxed regions are enlarged (Bottom); b, bone; ft, fibrous tissues. (D) von Kossa staining showing mineralization in humerus sections of P21 littermate control and mutant mice. (E) Quantification of mineralization in humerus from P21 mice as the percent of osteoid surface (OS) of bone surface (BS) (n = 3). (F) Western blot analysis for β-catenin and Gα_s from the cell lysates of P6 humerus. IB, immunoblotting. (G) Immunohistochemistry of β-catenin on humerus sections of P6 littermate control and mutant mice. **P < 0.001; data are presented as mean ± SD.

Histological analysis further revealed abnormal architecture, structure, and mineral content of long bones in the Prrx1-Cre; Gnas^f(R201H)/+ mutant mice (Fig. 2 C and D). The growth plate cartilage was disorganized and expanded in the mutant compared with the control (Fig. S3G). The marrow cavity was much reduced and the marrow space was occupied by trabecular bone and fibrous tissue with an appearance of the characteristic Chinese writing pattern (Fig. 2C). Reduced mineralization in the mutant bones was shown by von Kossa staining, and the increased osteoid surface demonstrated poor mineralization of increased bone formation in the Prrx1-Cre; Gnas^f(R201H)/+ mutant mice (Fig. 2 D and E). Increased β-catenin protein levels were found in mutant long bones of postnatal day (P)6 pups (Fig. 2 F and G), supporting our previous findings in human FD bone (17). While poorly mineralized bone formation was increased, differentiation of osteoclasts from the Prrx1-Cre⁻ hematopoietic lineage shown by tartrate-resistant acid phosphatase (TRAP) staining on the bone surface of the mutant mice was also increased (Fig. S3 H and I). This is a characteristic feature of human FD (28). In addition, Rankl expression was increased in Prrx1-Cre; Gnas^f(R201H)/+ mutant bone (Fig. S3J), suggesting that activating Gα_s signaling in osteoblasts has up-regulated Rankl expression, as has been shown for osteoblast-derived parathyroid hormone-related peptide (PTHrP) that signals through Gα_s and promotes Rankl expression directly (29, 30). All these phenotypes of Prrx1-Cre; Gnas^f(R201H)/+ mutant mice closely resemble those found in human FD. Therefore, embryonic expression of the activating R201H mutant Gα_s in limb bud osteochondral progenitor cells recapitulated human FD phenotypes in adult long bones. It is important to note again that Gnas^R201H heterozygous expression was sufficient to produce FD phenotypes in mice and that the phenotypes were the same in male or female mice regardless of whether Gnas^R201H is provided maternally or paternally, as is the case in human FD (31).

Progression of FD Phenotypes in Prrx1-Cre; Gnas^f(R201H)/+ Mutant Mice.

In humans, FD lesions start in embryonic development, progress during childhood, and remain static or sometimes improve in later stages of life (32). However, detailed tissue analysis of disease progression was not possible in human patients. The conditional Prrx1-Cre; Gnas^f(R201H)/+ mutant mice therefore allowed us to investigate the onset and progression of FD phenotypes to gain a better understanding of the natural history of FD disease.

Well-formed cortical bone, a marrow cavity with primary spongiosa next to the growth plate, and cartilaginous epiphyses can be seen in control mice starting from P0 (Fig. S4A), but the long bone of P0 Prrx1-Cre; Gnas^f(R201H)/+ mice showed no indication of bone formation (Fig. S4A). By P6, Prrx1-Cre; Gnas^f(R201H)/+ mice showed bone formation, but marrow space was occupied by osseous trabecular tissue and the growth plate was increased in length (Fig. S4B). At 3 wk of age, there was more bone formation in the Prrx1-Cre; Gnas^f(R201H)/+ mice with the same abnormalities as observed at P6: an irregular and thickened growth plate, lack of cortical bone, and expanded woven trabecular bone (Fig. S4C). These phenotypes were also observed in 2- and 4-mo-old Prrx1-Cre; Gnas^f(R201H)/+ mice (Figs. S3D and S4D). Bone marrow fibrosis was observed from P6 onward.

The phenotypes of increased growth plate and delayed bone formation resembled those observed in mice in which either PTHrP or a constitutively active PTH/PTHrP receptor found in Jansen’s metaphyseal chondrodysplasia (JMC) (33) was expressed in chondrocytes, which led to chondrodysplasia and delayed bone formation (34, 35). As the PTH/PTHrP receptor is coupled with Gα_s (36), delayed bone formation and growth plate abnormalities in Prrx1-Cre; Gnas^f(R201H)/+ mice indicated that PTH/PTHrP signaling has been enhanced in the limb. Chondrocyte hypertrophy in long bone development is tightly regulated by a negative feedback loop formed by Indian Hedgehog (Ihh) and PTHrP (37–39), and enhanced PTHrP signaling delays chondrocyte hypertrophy (40, 41). To examine chondrocyte hypertrophy, we analyzed E15.5 and E18.5 embryonic cartilage in the limb both histologically and molecularly (Fig. S5). Marker expression of nonhypertrophic (Col2) (42, 43), prehypertrophic (Ihh) (38, 44), hypertrophic (Col10), and late hypertrophic chondrocytes (Mmp13) (45–47) was examined (Fig. S5C). In the mutant section, Col2 expression was detected everywhere at the expense of Ihh, Col10, and Mmp13 expression. In addition, delay of chondrocyte hypertrophy and the resulting delayed Ihh expression (Fig. S5C) caused delay in osteoblast differentiation in the mutant, shown by delayed osteoblast marker Osterix (Osx) expression (Fig. S5D). Taken together, Gα_s activation delayed chondrocyte hypertrophy, which in turn delayed bone formation in embryos. Abnormal woven bone formation started postnatally and progressed rapidly with marrow fibrosis in Prrx1-Cre; Gnas^f(R201H)/+ mice. In addition, as a PTH/PTHrP receptor mutation was found in a few enchondroma patients and transgenic expression of one mutant PTH/PTHrP receptor led to enchondroma formation, supporting a role of the cAMP/PKA pathway in the development of enchondromas (48), some of the abnormal cartilage in older Prrx1-Cre; Gnas^f(R201H)/+ mice resembled enchondroma-like lesions in some FD patients (49, 50) (Fig. S4D, arrows).

Gnas^R201H Expression Impairs Osteoblastic Differentiation of Bone Marrow Mesenchymal Stem Cells.

It has previously been shown that Gα_s signaling plays a critical role in osteoblast differentiation from mesenchymal progenitor cells by modulating activities of the Wnt/β-catenin and Hh signaling pathways; both play fundamental roles in skeletal development, and defects in these signaling pathways cause skeletal diseases (18, 51, 52). In FD patients, there is extensive marrow fibrosis, and fibrotic cells in the marrow are arrested at early osteoblast differentiation stages (53, 54), suggesting that in adults, FD mutations cause aberrant osteoblast differentiation of BMSCs that contain stem cells. To test whether Gnas^R201H expression in mouse BMSCs affects their osteogenic potential, we isolated BMSCs from Gnas^f(R201H)/+ mice and infected them with adenoviruses carrying Cre or GFP (Ad-Cre or Ad-GFP). Cre expression in BMSCs induced Gnas^R201H expression and thus activated PKA, shown by increased CREB phosphorylation (55, 56) (Fig. 3A). The Gnas^R201H-expressing BMSCs cultured in osteogenic medium showed marked reduction in alkaline phosphatase (ALP) activity and mineralization (Fig. 3B). In addition, expression of osteogenic genes, such as Osx and Osteocalcin (Ocn), were significantly down-regulated (Fig. 3C). Importantly, Gnas^R201H expression in BMSCs also reduced colony-forming units, suggesting that activated Gα_s reduced stemness and/or proliferation of BMSCs (Fig. 3D). Furthermore, β-catenin protein levels were increased and expression of Wnt/β-catenin target genes Axin2 and Tcf1 were significantly up-regulated (Fig. 3 E and F) while Hh signaling was down-regulated in mutant BMSCs (Fig. 3G), consistent with our previous findings (17, 18, 51) that activated Gα_s promotes Wnt/β-catenin signaling and inhibits Hh signaling.

Fig. 3.

Gnas^R201H expression impairs osteoblastic differentiation of BMSCs. (A) Western blot analysis of phosphorylated CREB and Gα_s in cell lysates of BMSCs 7 d after being transduced with Ad-Cre and Ad-GFP. (B) ALP and alizarin red staining of BMSCs after osteogenic differentiation for 7 and 14 d, respectively. (C) qRT-PCR analysis of osteogenic marker expression from the RNA isolated from BMSCs 7 d post differentiation after being transduced with Ad-Cre or Ad-GFP. (D) Colony-forming assay of BMSCs transduced with Ad-Cre or Ad-GFP. (E) Western blot analysis of β-catenin in BMSCs 7 d after being transduced with Ad-Cre or Ad-GFP. (F and G) qRT-PCR analysis of signaling target genes of the Wnt (Axin2 and Lef1) and Hedgehog (Hip1 and Gli1) pathways from RNA isolated from BMSCs 7 d after being transduced with Ad-Cre and Ad-GFP. *P < 0.05, **P < 0.001; data are presented as mean ± SD.

Mosaic Analysis Reveals Non–Cell-Autonomous Roles of the Gα_s FD Mutation in BMSCs.

As FD in humans is caused by somatic mosaicism of Gα_s mutations, to model this disease more accurately, we created a mosaic state of the Gnas^R201H mutation in vivo in bone by crossing the Gnas^f(R201H) line with Sox9-CreER; Rosa26Td-Tomato mice to generate Gnas^f(R201H); Sox9-CreER; Rosa26Td-Tomato mice. The Sox9-CreER line was chosen, as it has been shown to be expressed in BMSCs after birth apart from its expression in chondrocytes (57). Tamoxifen (TM)-injected mosaic mutant mice exhibited increased bone volume as well as trabecular number and thickness and displayed an irregular growth plate architecture (Fig. S6 A and B) similar to, but less severe than, the phenotypes of Prrx1-Cre; Gnas^f(R201H) mice. Tdtomato marked the Gnas^R201H mutant cells, and increased CREB phosphorylation was found in Tdtomato⁺ cells, confirming Gα_s activation (Fig. S6C). Interestingly, Osx⁺ osteoprogenitor cells were increased in the Tdtomato⁺ population while Osteopontin (Opn⁺) cells were reduced (Fig. 4 A, A′, B, and B′), indicating that Gnas^R201H expression promotes osteoblast differentiation from BMSCs but inhibits osteoblast maturation in vivo. Consistent with findings in Prrx1-Cre; Gnas^f(R201H) mice, β-catenin expression was increased in the Tdtomato⁺ population (Fig. S6E).

Fig. 4.

Mosaic expression of Gnas^R201H causes enhanced bone fibrosis but reduced bone maturation. Immunostaining (A–D and A′–D′) and quantification (A″–D″) of marker expression as indicated in TdTomato-labeled mutant [Sox9-CreER; Rosa26TdTomato; Gnas^f(R201H)] in wild-type (Sox9-CreER; Rosa26TdTomato) cells. TM was injected at P5 and the humerus was sectioned at P21 for analysis. *P < 0.05, **P < 0.001; data are presented as mean ± SD. Opn⁺ and Osx⁺ cells refer to Opn⁺; TdTomato⁺ and Osx⁺; TdTomato⁺ double-positive cells.

The growth plate was very disorganized in the TM-injected Sox9-CreER; Rosa26Td-Tomato; Gnas^f(R201H) mutant mice. As chondrocyte hypertrophy is synchronized in wild-type mice, we asked whether disorganization of the mutant growth plate was caused by asynchronous hypertrophy due to mosaic expression of Gnas^R201H. Indeed, Col10 immunostaining showed that the Gnas^R201H-expressing Tdtomato⁺ chondrocyte column was much increased in length compared with the wild-type Tdtomato⁺ chondrocyte column or Tdtomato⁻ one (Fig. 4 C–C′′), demonstrating that Gnas^R201H expression significantly delayed chondrocyte hypertrophy cell autonomously.

As bone marrow fibrosis is a prominent feature of FD, we also found extensive fibrotic tissue in bone marrow of TM-injected Gnas^f(R201H); Sox9-CreER; Rosa26Td-Tomato mutant mice compared with the TM-injected wild-type control (Fig. S6B). When analyzed with a fibrosis marker, αSMA (58), there was an increase of αSMA⁺ cells in the mutant mice (Fig. 4 D–D′′). This was further confirmed by much increased αSMA expression in Gnas^R201H-expressing BMSCs (Fig. S6D). Interestingly, many of the αSMA⁺ cells were Tdtomato⁻, suggesting that they were wild-type cells with no Gnas^R201H expression (Fig. 4D′). In addition, these αSMA⁺; Tdtomato⁻ cells were not associated with CD31⁺ endothelial cells (Fig. S6 F and G), indicating they were not smooth muscle cells associated with blood vessels. Therefore, Gnas^R201H expression in BMSCs also caused FD, and mosaic analysis allowed uncovering non–cell-autonomous activities of Gnas^R201H in bone marrow fibrosis.

Wnt/β-Catenin Signaling Plays a Key Role in FD Pathogenesis.

As Prrx1-Cre and Sox9-CreER are also expressed in chondrocytes, to investigate whether Gnas^R201H expression in early osteoblast progenitors would lead to FD, we crossed the Gnas^f(R201H) line with the Osx-GFP::Cre line, a tetracycline responsive element (tetO)-controlled GFP::Cre fusion gene expression driven by the Osx promoter (59). To our surprise, mutant mice died immediately after birth, possibly due to general bone defects throughout the body. To get around this, we suppressed GFP::Cre expression by feeding pregnant female mice with doxycycline in water starting at E11.5 until birth. The pups were born with grossly normal morphology but, at weaning, mutant mice showed strong FD phenotypes, though less severe than the Prrx1-Cre; Gnas^f(R201H) mice (Fig. 5 A and B). As we found that Wnt/β-catenin signaling was highly up-regulated in both Prrx1-Cre– and Sox9-CreER–driven mutants, we decided to genetically down-regulate it in this less severe FD model to see if reducing Wnt/β-catenin signaling can rescue the FD phenotype by generating Osx-GFP::Cre; Gnas^f(R201H); Lrp6^c/+ mice. Lrp6 is a Wnt coreceptor that transmits canonical Wnt/β-catenin signaling (60). Interestingly, removing one Lrp6 copy in Osx-GFP::Cre; Gnas^f(R201H) mutant mice resulted in significant rescue of FD phenotypes. Microcomputed tomography (μCT) as well as histological analysis showed obvious rescue of bone marrow space and cortical bone as well as reduction of bone marrow fibrosis (Fig. 5 A and B). Wnt/β-catenin target gene expression was reduced by Lrp6 removal (Fig. 5C). Furthermore, expression of early osteoblast differentiation markers such as Runx2 and Alp was reduced, while expression of mature osteoblast markers Ocn and Opn was increased and fibrosis marker Timp1 was decreased in the rescued mutants (Fig. 5C). Therefore, up-regulated canonical Wnt/β-catenin signaling plays a key role in mediating the effects of Gnas^f(R201H) expression in FD. Osteoclast differentiation was not altered by removing Lrp6 (Fig. S7). To further test the therapeutic value of small-molecule inhibitors of Wnt/β-catenin signaling in improving bone mineralization inhibited by Gnas^R201H, we treated Gnas^R201H-expressing BMSCs at the beginning of osteogenic differentiation with LGK-974, a potent small-molecule inhibitor of Wnt secretion (61). LGK-974 treatment dose dependently improved mineralization in Gnas^R201H mutant BMSCs (Fig. 5 D and E), which was confirmed by reduced Wnt signaling and increased osteogenic marker expression (Fig. 5 F and G). Taken together, both in vivo and in vitro experiments demonstrate that up-regulated canonical Wnt/β-catenin signaling mediates the function of Gnas^R201H in causing FD, and that Wnt/β-catenin signaling may be a critical therapeutic target for treating FD.

Fig. 5.

Up-regulated Wnt/β-catenin signaling plays a key role in FD. (A) Representative µCT scans of the humerus of littermate control [Gnas^f(R201H)/+] and OsxCre; Gnas^f(R201H)/+ and OsxCre; Gnas^f(R201H)/+; Lrp6^C/+ mutant mice. Sagittal and transverse section views of humerus are also shown. (B) H&E staining of humerus sections from P21 mice of the indicated genotypes. (C) qRT-PCR analysis of humerus bones from mice with the indicated genotypes. (D) Alizarin red staining of BMSCs transduced with Ad-Cre and Ad-GFP viruses 7 d after osteogenic differentiation. At day 0 of osteogenic differentiation, BMSCs were treated with DMSO or LGK-974, a small molecule that inhibits Wnt secretion. (E) Alizarin red staining was quantified. (F and G) qRT-PCR analysis of gene expression in BMSCs at day 7 post differentiation. *P < 0.05, **P < 0.001; ns, not significant; data are presented as mean ± SD.

Discussion

Here we report the generation of a mouse model that allows the closest possible modeling of FD bone disease, a skeletal disorder originally described in 1942 (62). The conditional knockin approach allowed expression of Gnas^R201H corresponding to the human FD mutant from the Gnas locus with spatial and temporal control. This model can be used to study not only FD but also MAS, which has severe symptoms outside of the skeletal system. Germ-line expression of Gnas^R201H caused embryonic lethality, but Cre-induced Gnas^R201H expression in osteochondral progenitor cells, early osteoblast cells, or sporadically in postnatal BMSCs led to typical FD phenotypes. Embryonic lethality due to germ-line Gnas^R201H expression confirmed the longstanding hypothesis postulated by Happle that the disease genotype would be lethal if germ line‐transmitted, and is only able to survive through mosaicism (12). Successful development of the Gnas^f(R201H) line has opened a door to elucidating many still poorly understood aspects of FD and MAS and to designing and testing novel FD treatment strategies.

Our finding, though consistent with the human FD etiology that is linked to missense activating GNAS mutations that occur after fertilization in somatic cells (5, 63), is in contrast with a previous finding in a transgenic model in which the human FD mutant transgene could survive through germ-line transmission (13). This is likely due to the transgenic nature of the previous model in which an artificial promoter was driving the mutant GNAS cDNA expression from a heterologous genomic locus, highlighting the necessity to express the FD mutant from the endogenous Gnas locus to precisely model the disease. Several other animal models have been generated in the past that contributed to our current understanding of FD. These models, though insightful in certain ways, are limited in others by their transgenic and interspecies tissue transplantation nature, which does not change the Gα_s expressed from its endogenous locus as in FD human patients. Furthermore, these models could not be used to study other associated phenotypes found in MAS, where Gα_s activation occurs outside of the skeleton.

Interestingly, we found that induced Gnas^R201H expression in osteochondral progenitor cells resulted in much delayed chondrocyte hypertrophy, a thickened growth plate, and a condition similar to enchondroma (Fig. S4D). It is well-recognized that an FD lesion may contain cartilage, though the amount is quite variable. Jaffe and Lichtenstein (64) in their original article on FD recognized that cartilage was “an integral part of the dysplastic process”; FD case reports have shown that multifocal FD has been found together with enchondroma-like areas in conditions of fibrocartilaginous dysplasia (FCD) (65, 66). The animal models we have generated provided invaluable insight and tools in understanding FD and FCD. The PTH/PTHrP receptor couples to Gα_s and Gα_q/11 (67–69). Patients with JMC (characterized by short-limbed dwarfism, severe growth plate abnormalities, and hypercalcemia) and also a few patients with enchondroma alone (48) carry activating mutations in the PTH/PTHrP receptor that cause constitutive receptor activation (33, 70). Delayed chondrocyte hypertrophy and FCD-like lesions in our FD models resemble those caused by activated PTH/PTHrP receptor. It is interesting to observe that chondrocyte hypertrophy eventually occurred in Gnas^f(R201H); Prrx1-Cre newborn pups (Fig. S4). As other signaling pathways can also regulate chondrocyte hypertrophy, it is possible that postnatally, the relative contribution of the PTH/PTHrP receptor in regulating chondrocyte hypertrophy is reduced.

The cellular origin of FD has remained largely unknown. Our data demonstrate that expression of FD mutation in early osteochondral progenitor cells or osteoblast lineage cells, or BMSCs, can establish similar human FD phenotypes. In all these FD models, consistent with our previous finding in human tissues (17), Wnt/β-catenin signaling was up-regulated; this is critical for FD phenotypes, as removal of one Lrp6 gene copy resulted in significant rescue of FD phenotypes, demonstrating that activated Wnt/β-catenin signaling is a key mechanism that induces FD. The function of Wnt signaling in bone formation has been shown to be dose- and stage-dependent. Sustained activation of Wnt/β-catenin signaling in mesenchymal progenitor cells and osteoblast lineages results in reduced mineralization and bone formation (ref. 17 and references therein). Both in vivo genetic rescue and in vitro chemical treatment showed that inhibition of Wnt/β-catenin signaling reduced FD phenotypes by promoting osteoblast maturation and reducing bone marrow fibrosis. Therefore, Wnt/β-catenin signaling inhibitors such as LGK-974 could be repurposed to treat FD, and our FD models are well-suited to test potential treatments for FD and McCune–Albright syndrome.

Materials and Methods

Mouse.

Animal care and experiments were performed in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines at the National Institutes of Health and the Harvard Medical School.

Immunohistochemistry and Western Blotting.

Immunohistochemistry was performed according to previously described methods (18). Western blotting was performed using standard techniques. Antibody information is provided in Supporting Information.

RNA in Situ Hybridization.

RNA in situ hybridization was performed using DIG-labeled antisense RNA probes as described before (71).

BMSC Isolation and Osteogenic Differentiation.

BMSCs were isolated by flushing the bone marrow cavity of 6-wk-old mice and plating cells in alpha-MEM, 20% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin. Equal numbers of cells were then seeded in 12-well plates and infected with Ad-Cre or Ad-GFP before confluence. Cells were switched to osteogenic medium (17, 18) after confluence for the indicated time before analysis.

Adenovirus Cell Culture Treatment.

The Cre and GFP adenoviruses were made by SAIC-Frederick (∼1 × 10¹⁰ pfu/mL) and diluted 1:500 for infection.

Statistical Analysis.

Statistical significance was tested by two-tailed Student’s t test between two groups. P < 0.05 was considered significant. Data are presented as mean ± SD unless otherwise indicated.