The critical role of the TB5 domain of fibrillin-1 in endochondral ossification

Abstract

Mutations in the fibrillin-1 (FBN1) gene are responsible for the autosomal dominant form of geleophysic dysplasia (GD), which is characterized by short stature and extremities, thick skin and cardiovascular disease. All known FBN1 mutations in patients with GD are localized within the region encoding the transforming growth factor-β binding protein-like 5 (TB5) domain of this protein. Herein, we generated a knock-in mouse model, Fbn1^Y1698C by introducing the p.Tyr1696Cys mutation from a patient with GD into the TB5 domain of murine Fbn1 to elucidate the specific role of this domain in endochondral ossification. We found that both Fbn1^Y1698C/+ and Fbn1^{Y1698C/Y1698C} mice exhibited a reduced stature reminiscent of the human GD phenotype. The Fbn1 point mutation introduced in these mice affected the growth plate formation owing to abnormal chondrocyte differentiation such that mutant chondrocytes failed to establish a dense microfibrillar network composed of FBN1. This original Fbn1 mutant mouse model offers new insight into the pathogenic events underlying GD. Our findings suggest that the etiology of GD involves the dysregulation of the extracellular matrix composed of an abnormal FBN1 microfibril network impacting the differentiation of the chondrocytes.

Introduction

The acromelic dysplasia group consists of skeletal disorders characterized by short stature, such as geleophysic dysplasia (GD, OMIM#231050, #614185, #617809), acromicric dysplasia (AD, OMIM#102370), Weill–Marchesani syndrome (WMS, OMIM#277600, #608328, #614819, #613195) and Myhre syndrome (MS, OMIM#139210) (1,2). GD is further associated with short extremities, joint stiffness, facial dysmorphism, thick skin and cardiac defects (3). We have previously evaluated the molecular basis of GD, revealing mutations in ADAMTSL2 to drive the autosomal recessive form of this disease, whereas mutations in fibrillin-1 (FBN1) are responsible for autosomal dominant GD (3). We further demonstrated that ADAMTSL2 and FBN1 are both genetically and biochemically linked (4). The generation of an Adamtsl2^−/− mouse model led us to highlight the impairment of chondrogenesis and microfibrillar network establishment that occur in the context of Adamtsl2 deficiency, suggesting a role for Adamtsl2 in the maintenance of the growth plate extracellular matrix (ECM) owing to its ability to regulate the microfibrillar network (5). Intriguingly, GD and other diseases of the acromelic dysplasia group have been described as ‘mirror images’ of the best-known fibrillinopathy, Marfan syndrome (MFS, OMIM#154700).

MFS is a rare connective tissue disorder with an autosomal dominant mode of transmission. It is a multisystem disease with a characteristic unique combination of skeletal, cardiovascular, and ocular features including long bone overgrowth, thoracic aortic dilatation and dissection, and ectopia lentis (6). Many FBN1 mutations spread throughout the gene have been identified in patients with MFS, all resulting in the same pathogenetic mechanism (7). Interestingly, in patients with GD, all identified FBN1 mutations are restricted to the transforming growth factor-β (TGF-β)-binding protein-like domain 5 (TB5)-coding region of the gene (4). This TB5 domain is not the only TB domain in FBN1 to have been linked to a specific fibrillinopathy. Indeed, TB4, which harbors the only integrin-binding Arginine-Glycine-Asparate (RGD) motif in this protein, has been linked to the development of Stiff skin syndrome (OMIM#184900), which is characterized by joint contractures and thick, hard skin (8).

Figure 1

Skeletal abnormalities in Fbn1^Y1698C mutant mice. (A) The targeted Fbn1 locus, with the introduced point mutation in exon 42 (human tyrosine 1696 replacement with a cysteine) shown in red, corresponding to mouse tyrosine 1698 replacement with a cysteine. (B) Sequencing electropherogram sections showing the point mutation (A > G) in Fbn1^Y1698C mutant mice. Heterozygous mutants exhibit a double peak corresponding to A and G (denoted as an ‘R’), whereas the homozygous mutants exhibit a single peak corresponding to G. (C, D) Quantification of murine naso-anal stature at postnatal stage, P1 and P30. (E, F) Quantification of femur (E) and (F) tibia length at stage P1 (n_WT > 5, n_HT > 5 and n_Ho > 5). (G) Side and (H) front X-ray images of the whole skeleton of WT (from left to right), Fbn1^Y1698C/+ (HT) and Fbn1^{Y1698C/Y1698C} (Ho) mutant mice at P30 (n_WT > 5, n_HT > 5 and n_Ho > 5). ^*P < 0.05, ^***p<0.001 and ^****p<0.0001.

The FBN1 gene encodes FBN1, which is a large and ubiquitous glycoprotein within the ECM and the major component of microfibrillar networks (9). The FBN1 protein is primarily composed of multiple repeating calcium-binding (cb)-epidermal growth factor-like (EGF) domains and TB domains, which contain eight cysteine residues essential for disulfide bond formation within the FBN1 molecule. Functionally, FBN1 confers structural support to tissues and serves as essential scaffolding for elastin deposition. Fibrillin microfibrils are also a niche for growth factors (10). As such, FBN1 microfibrils have been implicated in the bioavailability of TGF-β and the sequestration of bone morphogenetic proteins (BMPs). Moreover, FBN1 can interact with several ECM proteins including A Disintegrin And Metalloproteinase with Thrombospondin Motifs (ADAMTS), integrins, fibronectin, and latent TGF-β binding proteins (LTBPs) (11), with the structures of LTBPs 1–4 being closely related to those of fibrillins (12).

At present, it remains unclear as to how mutations in the TB5 domain of FBN1 result in short stature rather than the tall stature phenotype often associated with mutations in other domains of FBN1. Moreover, the skeletal phenotype observed in patients with GD (short stature, short extremities, shortened long bones) suggests that the TB5 domain of this protein may play a role in the process of endochondral ossification. To test this hypothesis, we herein generated a Fbn1^Y1698C/+ knock-in (KI) mouse model that was used to explore the role of FBN1 in the pathophysiology of GD and the importance of the FBN1 TB5 domain in the context of skeletal development.

Results

Generation of a new Fbn1 KI mouse model with skeletal abnormalities

To investigate the role of FBN1 and its TB5 domain in short stature, we generated a Fbn1^Y1698C KI mouse model by introducing the GD-associated human p.Tyr1696Cys (mouse p.Tyr1698Cys) mutation into this gene. This mutation in exon 42 is a single base substitution (A > G) resulting in a change in the amino-acid sequence of the resultant protein (tyrosine → cysteine), as described in the Methods section and Figure 1A. This substitution was confirmed in all mice via genomic DNA sequencing (Fig. 1B).

The litters resulting from the breeding of Fbn1^Y1698C/+ mice followed a normal Mendelian distribution. Heterozygous Fbn1^Y1698C/+ (HT) mice were viable and fertile. Both HT and homozygous Fbn1^{Y1698C/Y1698C} (Ho) mice exhibited a skeletal phenotype. Given that short stature, short long bones, and brachydactyly are the primary skeletal features of GD, total naso-anal and long bone length values were measured for these mice, revealing that the stature of both HT and Ho Fbn1 KI mice was reduced relative to corresponding control animals at the postnatal stage 1 (P1) (HT: 3.2%; Ho: 4.1%, P < 0.05) (Fig. 1C). These differences in stature between control and mutant grew with age, reaching 8.9% in Ho animals and 4.2% in HT animals at P30 (Fig 1D). Consistently, femur length was also reduced significantly in Ho mice at P1 compared with control animals (−6%, P < 0.05), and the same was true of tibia length (Fig. 1E and F). X-ray analyses conducted at P30 also confirmed the presence of skeletal abnormalities in these TB5 mutant mice, including short stature and shortened long bones (Fig. 1G and H). No differences in Alcian blue and Alizarin red staining were observed in the extremities of these animals when comparing controls and mutants at P1 or P30 (Supplementary Material, Fig. S1). Our novel Fbn1^Y1698C exhibited the GD skeletal phenotype.

Figure 2

Skin and aortic phenotypes in Fbn1^Y1698C mutant mice. Masson’s trichrome staining of skin from newborn (A, upper panel) and 1-month-old (A, lower panel) Fbn1^Y1698C littermates. Fbn1^{Y1698C/Y1698C} (Ho) mutant mouse skin exhibited increased collagen deposition. Scale bar: 200 μm. (B) Quantification of collagen (Col1a1, Col1a2 and Col3a1) mRNA levels in skin samples from wild-type and mutant littermates as measured by qPCR, with Gapdh, Hprt and Hsp90 being used to normalize gene expression (n = 3). No significant upregulation of collagen mRNA levels was observed in Fbn1^Y1698C/+ (HT) and Fbn1^{Y1698C/Y1698C} (Ho) mutant mice. (C) Quantification of Fbn1 mRNA levels in skin samples from 1-month-old wild-type and mutant littermates as measured by qPCR, with Gapdh, Hprt and Hsp90 being used to normalize gene expression (n = 3). (D) Immunofluorescence against Tgfb1 on skin from newborns Fbn1^Y1698C littermates with associated fluorescence quantification. Scale bar: 200 μm. (E) Aortic root morphology was assessed via the orcein staining of aorta samples from 3-month-old mice. Fbn1^{Y1698C/Y1698C} (Ho) and Fbn1^Y1698C/+ (HT) mice exhibited no elastic fiber alterations (such as fragmentation) as compared with WT mice (n = 3). Scale bar: 100 μm. *P < 0.05.

Fbn1^Y1698C mice do not exhibit thick skin nor aortic disease

A Fbn1 mouse model, WMΔ, that replicates phenotypes associated with WMS including thick skin and shorter long bones has previously been reported (13). As such, we evaluated skin phenotypes to confirm the specificity of our mouse model. Macroscopic inspection of the skin of our mutant and wild-type (WT) mice revealed no differences between these groups concerning skin thickness. Histological evaluation of skin biopsies following Masson’s trichrome staining revealed normal collagen deposition in the dermis of HT mice but excessive collagen in Ho mice as compared with controls (Fig. 2A), with quantification of these results further confirming the increased accumulation of collagen in the skin of Ho mice at P30 but not at P1 (Fig. 2A, right panel). No collagen gene upregulation was detected via quantitative polymerase chain reaction (PCR) in skin samples from Ho mice (Fig. 2B), nor were changes in Fbn1 gene expression observed in these samples, suggesting that the introduced TB5 mutation does not affect the expression of this gene (Fig. 2C). TGF-β1 being a potent stimulator of collagen secretion by fibroblasts, we evaluated the level of TGF-β1 by immunofluorescence on mouse skin sections. No differences were observed between the skin samples at P1 (Fig. 2D). To test whether this point mutation in TB5 causes GD rather than MFS, we analyzed aortic root morphology, which is altered in MFS. Both HT and Ho mice exhibited a normal aortic wall without any elastic fiber fragmentation (Fig. 2E). In addition, no evidence of aortic disease typical of MFS such as thoracic aortic dilation was observed in mutant mice. Taken together, these results confirmed the specificity of our mouse model as one that replicates many of the phenotypes associated with GD but not MFS. As such, our Fbn1 mutant mouse model is distinguished by a skeletal phenotype and increased collagen deposition in the skin of Ho animals.

Figure 3

Impact of Fbn1 KI on growth plate. (A) Safranin O staining of Fbn1^Y1698C/+ (HT), Fbn1^{Y1698C/Y1698C} (Ho) and WT newborn (P1) mouse femur and femoral proximal growth plate sections. Scale bar: 100 μm. (B) Quantification of hypertrophic chondrocyte surface area in WT, Fbn1^Y1698C/+ (HT) and Fbn1^{Y1698C/Y1698C} (Ho) mouse femoral proximal growth plate samples (n_WT > 5, n_HT > 5 and n_Ho > 5; ^**P < 0.01, ^***P < 0.001). (C) Safranin O staining of Fbn1^Y1698C/+ (HT) and Fbn1^{Y1698C/Y1698C} (Ho) and WT mouse femoral proximal growth plate sections at P30. Scale bar: 200 μm. (D) Quantification of growth plate length in WT, HT and Ho femoral proximal growth plate samples at P30 (n_WT > 5, n_HT > 5 and n_Ho > 5, ^**P < 0.01).

FBN1 is involved in the regulation of growth plate cytoarchitecture

Given that our HT and Ho mutant mice presented with a shorter stature as well as shorter long bones, we next analyzed these animals to investigate the femoral growth plate, which is a cartilaginous tissue that plays a critical role in bone elongation. In the femoral head of WT animals, chondrocytes proliferated and exhibited a discoidal shape, and were aligned in a columnar conformation (Fig. 3A). Chondrocytes that had reached the hypertrophic stage of maturation were identified by their enlarged volume, spherical shape, and decreased surrounding matrix (Fig. 3A, magnification). Although the chondrocyte zones within growth plates of WT and mutant femurs at P1 were clearly defined, the organization and shape of the hypertrophic chondrocytes were altered in mutant mice as compared with controls, with the size of the hypertrophic chondrocyte surface area being decreased in both Fbn1^Y1698C mutant mice at P1 and P30 (Fig. 3B and Supplementary Material, Fig. S2A). Moreover, the Ho mice exhibited impaired hypertrophic chondrocyte columnar organization (Fig. 3A, magnification), with cartilage exhibiting more severe disorganization than that observed in HT mice. At P30, global reductions in growth plate length were also observed in both types of Fbn1^Y1698C mutant mice (Fig. 3C and D). These results led us to conclude that growth plate defects may be responsible for the short stature observed in both Ho and HT Fbn1^Y1698C mutant animals.

Figure 4

Impact of Fbn1 KI on chondrocyte function. (A) For in situ hybridization experiments, newborn WT, Fbn1^Y1698C/+ (HT) and Fbn1^{Y1698C/Y1698C} (Ho) femoral growth plates were hybridized with an antisense riboprobe specific for Col10. A sense riboprobe was used as a negative control. Scale bar: 100 μm. Quantification of Col10 mRNA levels in femoral samples from wild-type and mutant littermates as measured by qPCR, with Gapdh (glyceraldehyde 3-phosphate dehydrogenase), Hprt (hypoxanthine-guanine phosphoribosyltransferase), and Hsp90 (mouse heat shock protein 90 kDa) being used to normalize gene expression (n > 3). (B) Immunofluorescence on newborn WT, HT, Ho growth plates against hypertrophic marker, Col10, associated with quantification of mean fluorescence intensity of Col10 and the length of hypertrophic zone (n = 3). Scale bar: 100 μm. (C) Immunohistochemical staining of femoral growth plate sections using antibody anti-Sox9 and the quantification of Sox9 staining. Scale bar: 100 μm. (D, E) Immunofluorescence analysis of RankL and RunX2 on newborn growth plates. No significant differences were observed. (F) Proliferation analysis using Ki67 immunostaining performed on newborn growth plates showed no significant differences in mutant mice compared with WT mice (n = 3). ^*P < 0.05, ^**P < 0.01 and ^***P < 0.001. PZ, proliferative zone; PHZ, prehypertrophic zone; HZ, hypertrophic zone (n = 3).

Fbn1^Y1698C mice exhibit alterations in key chondrocyte markers

To assess whether the changes observed in mutant growth plates are associated with differences in the expression patterns of key growth plate marker genes, mRNA in situ hybridization and immunolabeling were performed on serial femur sections to detect markers of hypertrophic chondrocytes [type X collagen (Col10)] and proliferative chondrocytes [type II collagen (Col2) and SRY box 9 (Sox9)]. WT femurs exhibited robust Col10a1 expression restricted to hypertrophic chondrocytes (Fig. 4A). However, the mRNA level of Col10a1 in HT and Ho mutants was drastically decreased near the chondro-osseous junction, especially in Ho mice (Fig. 4A). These in situ hybridization results were confirmed using a specific antibody raised against Col10 (Fig. 4B). The quantification of Col10 mean fluorescence intensity revealed a significant decrease in such expression in Ho mice (Fig. 4B, right panel). The data of Fig. 4A and B permit us to measure the hypertrophic zone length. At P1, the hypertrophic zone length was reduced in HT and Ho mice (Fig. 4B right panel). The same results were obtained at P30 (Supplementary Material, Fig. S2B). The expression pattern of Col2a1, a common cartilage marker expressed by all committed chondrocytes, was not significantly altered by the TB5 mutation, despite the observed variations in chondrocyte morphology (data not shown). Sox9, a critical transcription factor that controls chondrocyte differentiation, is expressed concomitantly with Col2a1 during cartilage development. Sox9 in the WT growth plate was detected throughout the resting zone and followed by an abrupt decline in its expression in the hypertrophic zone (Fig. 4C). Conversely, mutant growth plates exhibited an increase in Sox9 expression in the prehypertrophic zone (Fig. 4C). To test the impact of this increased Sox9 expression level on chondrocyte fate, we evaluated the expression level of Rankl and RunX2 on femoral growth plate sections. These growth plates did not display any difference in Rankl and RunX2 expression levels (Fig. 4D and E and Supplementary Material, Fig. S3). Ki67-based analyses of proliferative activity indicated that chondrocyte proliferation rates were unaffected at P1 in the Fbn1 mutant growth plates as compared with controls (Fig. 4F and Supplementary Material, Fig. S3). These results showed that two major actors of chondrocyte differentiation, Sox9, and Col10, were dysregulated in mutant growth plates.

Mutations in the TB5 domain of FBN1 impair microfibrillar network assembly

As FBN1 is an ECM protein and the major component within microfibrillar networks, we analyzed the impact of TB5 disruption on the structure of the ECM in the growth plate. Fibronectin was also studied given that it plays a role in regulating microfibril assembly and was affected by the absence of Adamtsl2 in a mouse model of GD (5). Therefore, we evaluated microfibrillar network formation in cultures of primary chondrocytes isolated from the ribs of Fbn1^Y1698C mutant and WT mice. Before that, we investigated Fbn1 expression in the control and mutant growth plates. However, no differences in Fbn1 protein or mRNA levels were observed following TB5 mutation in tissue or chondrocyte cultures (Fig. 5A–C). Visualization of fibril networks via confocal microscopy indicated that fibronectin networks in Fbn1^Y1698C/+ and Fbn1^{Y1698C/Y1698C} cultures were as dense and organized as in WT cultures (Fig. 5D). In contrast, FBN1 fibrils were significantly much thicker and present within a less dense and organized microfibrillar network in Fbn1^Y1698C/+ and Fbn1^{Y1698C/Y1698C} chondrocyte cultures as compared with WT cultures (Fig. 5E–G). Taken together, these results indicated that the mutation in the TB5 domain leads to the impairment of fibril network assembly without affecting Fbn1 expression levels.

Figure 5

Composition of the ECM in the femur growth plate. (A) Immunohistochemistry against Fbn1 on P30 WT, HT and Ho growth plates (scale bar: 100 μm) and (B) Fbn1 mRNA level on primary chondrocytes and (C) Fbn1 mRNA level from the whole hindlimb from newborn were analyzed by quantitative PCR showing no differences between the genotypes. (D, E) Primary chondrocytes isolated from P1 WT and Fbn1^Y1698C KI ribs and labeled with antibodies specific for fibronectin (D) and fibrillin-1 (E) (negative control can be found in Supplementary Material, Fig. S2C). These data revealed the impairment of the Fbn1 microfibrillar network in Fbn1^Y1698C KI chondrocytes upon fluorescent secondary antibody staining. The nuclei were labeled with the 4′,6-diamidino-2-phenylindole (DAPI). Scale bar: 20 μm. (F) Microfibrillar network density and (G) thickness of Fbn1 microfibrils were observed (^****P < 0.0001). Scale bar: 20 μm. (n > 3) ^**P < 0.01.

The impact of TB5 mutation on TGF-β signaling

Given the link between TGF-β and FBN1, we next investigated whether this mutation in the TB5 domain of FBN1 could impair TGF-β signaling in primary chondrocytes. Immunofluorescence analysis revealed comparable levels of TGF-β expression in mutant and WT growth plates, and this was confirmed via an enzyme-linked immunosorbent assay (ELISA) used to detect TGFβ-1 present in chondrocyte growth media (Fig. 6A and B). Western blotting analyses of the activation of the TGF-β signaling pathway revealed no differences in phosphorylated Smad2 levels in mutant chondrocytes as compared with controls (Fig. 6C). Finally, mRNA levels of downstream TGFβ-1 target genes including Ctgf (connective tissue growth factor) and Pai1 (Plasminogen activator inhibitor 1) were evaluated by real-time quantitative PCR. Their expression levels were not altered in mutant mice (Fig. 6D), confirming that this TB5 mutation did not affect TGF-β activity in growth plates. We, therefore, tested two other signaling pathways, BMP and phospho-p38/mitogen-activated protein kinase involved in WMS (14). No differences in these pathways were observed between mutants and controls (Supplementary Material, Fig. S4).

Figure 6

Analysis of the TGFβ signaling pathway. (A) Immunostaining for TGF-β1 in proximal femoral sections from WT, HT and Ho mice at P1. PZ, proliferative zone; PHZ, prehypertrophic zone; HZ, hypertrophic zone. Scale bar: 100 μm. (B) Total and active TGF-β1 levels in culture media from WT, HT and Ho primary chondrocytes were measured via ELISA (n = 3). (C) Western blotting analysis and quantification of phosphorylated Smad2 levels in extracts from P1 WT, HT and Ho primary chondrocyte samples, with total Smad2 being used for normalization, actin for loading control (n = 3). (D) mRNA expression level of target genes of TGF-β signaling pathway, Ctgf and Pai-1 on the whole hindlimb from newborns WT, HT and Ho (n > 3).

Discussion

In this study, we reported for the first time the crucial role of FBN1 in endochondral ossification. The skeletal anomalies observed in Fbn1^Y1698C/+ and Fbn1^{Y1698C/Y1698C} mice highlight the importance of FBN1 as a positive regulator of chondrogenesis. HT Fbn1^Y1698C/+ mice recapitulated some characteristics of GD such as short stature and short long bones, whereas the Ho Fbn1^{Y1698C/Y1698C} mice exhibited a more severe phenotype. More importantly, this novel mouse model does not exhibit the aortic disease observed in MFS mouse models harboring Fbn1 mutations. Overall, this report offers further insight regarding the involvement of the TB5 domain of FBN1 in growth plate abnormalities.

FBN1 has previously been reported to be linked to MFS-related phenotypes (15,16). Multiple mouse models have been generated to decipher the pathophysiological mechanisms underlying MFS, including the hypomorphic Fbn1^mgR/mgR model and the Fbn1^C1039G/+ model with a cbEGF domain substitution (17,18). However, these prior studies focused on the role of FBN1 in MFS-related aortic phenotypes, and its role in skeletal phenotypes was not fully explored.

By controlling TGF-β and BMP signaling, FBN1 is involved in bone remodeling. A study of bone phenotypes in the Fbn1^mgR/mgR MFS mouse model revealed a decrease in bone mineral density and trabecular bone anomalies that mimicked skeletal manifestations observed in patients with MFS (19). FBN1 can thus play a negative role in osteoclastogenesis by sequestering Receptor Activator of Nuclear factor Kappa-B Ligand (RANKL), an osteoclastogenic factor, in the ECM (20). Osteopenia was observed in Fbn1^Prx1−/− and Fbn1^Osx−/− mice, putatively due to premature osteoprogenitor differentiation suggesting that FBN1 modulates the osteoblast/osteoclast balance (21). The fate of mesenchymal stem cells depends upon the partition of adult bone marrow tissue into functionally distinct microenvironments called niches (22). FBN1 can shape the fate of skeletal stem cells by modulating the bioavailability of TGF-β in these bone marrow niches. Our identification of mutations specifically within the TB5 domain of FBN1 in patients with GD with short stature also suggests a possible link between FBN1 and endochondral ossification. The skeletal anomalies exhibited by our novel Fbn1^Y1698C mouse model conclusively demonstrate that the introduced TB5 domain mutation contributes to the incidence of short stature. Reductions in long bone growth mimicking human WMS have previously been observed in WMS model mice harboring a Fbn1 deletion, although this phenotype was transient (13). Here, we show that FBN1 is a key factor involved in the differentiation of chondrocytes owing to its ability to impact Col10a1, a hypertrophic chondrocyte marker that was downregulated in both Fbn1^Y1698C/+ and Fbn1^{Y1698C/Y1698C} mice. As a result, disruption of the TB5 domain in Fbn1 profoundly altered the normal differentiation of chondrocytes. The observed decreases in the size of chondrocytes in the growth plates of mutant mice may explain or exemplify such abnormal differentiation. Sox9 also exhibits a distinct expression profile in mutant prehypertrophic chondrocytes. Sox9 negatively regulates Col10 (23), so we posit that mutations in the TB5 domain are conducive to the upregulation of Sox9, in turn leading to the downregulation of Col10. Indeed, the ectopic expression of Sox9 in a transgenic mouse model has been reported to repress Col10 expression. Sox9 can also activate the expression of ECM components in cartilage. This may suggest the existence of a regulatory feedback loop that maintains the FBN1 microenvironment within the cartilage as we observed previously in an Adamtsl2-knock-out (KO) mouse model of GD (5).

Our study on FBN1, the major component of the microfibrillar network, demonstrated that defects in mutant growth plates were associated with a loss of FBN1-rich network density. Although fibrillins compose just 1–3% of total ECM proteins in the skeleton (22), mutations in the Fbn1 gene lead to severe skeletal anomalies. As a structural ECM glycoprotein, FBN1 may be essential for chondrocyte column formation. Indeed, the absence of appropriate chondrocyte column formation observed in Fbn1 mutant mice in this study underscores the importance of ECM in this process.

Mutations in the FBN1 TB5 domain identified in GD have been shown to disrupt heparan sulfate interactions in cells (24). Indeed, these findings have been reported in co-cultures of fibroblasts and human embryonic kidney cells 293 cells overexpressing GD mutants. Heparan sulfate permits interaction between cells and the ECM. More importantly, heparan sulfate is involved in the regulation of the early stages of microfibril assembly (25,26). This may explain the fact that despite the mutation of the TB5 domain, we observed FBN1 fibrils in our mutant mice, suggesting that this mutation does not impact the expression of FBN1 but rather alters its capacity to form an abundant fibril network. A defect in the fibril network formed by FBN1 and fibronectin has also been reported in an Adamtsl2 deficient mouse model (5). In a separate GD mouse model with a specific deletion of Adamtsl2 in tendinous tissues, Hubmacher et al. (27) observed a disrupted network and an impaired microfibril assembly in the ECM. In prior studies, we determined that FBN1 deposition in GD patient-derived fibroblasts was disturbed (4). Moreover, our previous human genetic data have established a link between ADAMTSL2 and FBN1 that was reinforced by biochemical results exhibiting a direct interaction between these two proteins (4). ADAMTSL2 co-localizes with FBN1 fibrils in cell culture and in vivo (3,5,27,28). An analysis of our Cytomegalovirus-CreAdamtsl2 KO mouse model revealed that the complex formed by a minimum of Adamtsl2 and Fbn1 creates microenvironmental conditions that are critical for appropriate chondrocyte column formation (5). In summary, abnormal fibril assembly in distinct tissues has been observed in all human and mouse GD models, suggesting that the underlying mechanistic basis for GD involves the dysregulation of fibrillin microfibril deposition through mechanisms associated with Adamtsl2 and the Fbn1 TB5 domain.

We have previously reported abnormal TGF-β signaling in GD patient-derived fibroblasts harboring mutations in the ADAMTSL2 or FBN1 genes (4). Increased TGF-β signaling activity was also detected in Adamtsl2 deficient mice in previous studies (5,28). However, we did not observe any comparable increase in TGF-β signaling activity in chondrocytes harboring a mutation in the FBN1 TB5 domain. These findings raise the question as to whether the observed alterations in TGF-β signaling are a cause or a consequence of GD pathogenesis. A lack of TGF-β signaling overactivation has similarly been reported in GD caused by mutations in the LTBP3 gene (29). Moreover, treatment with TGF-β neutralizing antibodies was not sufficient to rescue severe phenotypes in Adamtsl2 KO lungs (28). Altogether, these data suggest that TGF-β signaling may not be a direct pathogenic driver of GD but may instead contribute to the regulation of an abnormal ECM.

This original Fbn1^Y1698C mutant mouse model provides new insight into the molecular events underlying GD. Our study confirmed that FBN1 microfibril assembly plays a major role in the pathophysiological etiology of GD and demonstrated that the FBN1 TB5 domain is a structural domain involved in the assembly of growth plate ECM scaffolds, thereby impacting column structure. Importantly, this study brought to light the pivotal role of the FBN1-associated ECM in the control of chondrocyte differentiation. Overall, FBN1 TB5 domain may act as a sensor to gauge the state of the fibril network that is necessary for appropriate ECM regulation and organization in the growth plate.

Materials and Methods

Generation of Fbn1^Y1698C mice

To generate Fbn1^Y1698C/+ mice, a Fbn1^Y1698C targeting vector with a substitution of one base was generated as described in Figure 1 with the collaboration of the Clinique de la Souris (Strasbourg, France). The Homo sapiens p.Tyr1696Cys variant inserted in the genome of the mice correspond to the Mus musculus p.Tyr1698Cys variant. The germ-line transmission of the mutated allele was achieved by crossing the chimeric mice with C57BL/6 mice. The Fbn1^Y1698C/+ mice were intercrossed to generate (HT, Fbn1^Y1698C/+) and (Ho, Fbn1^{Y1698C/Y1698C}) mice carrying the mutations in the TB5 domain of FBN1. Genotyping was performed using tail samples and a Direct PCR Lysis Reagent (Viagen Biotech Cat # 101-T, 102-T) followed by sequencing. All research protocols were approved by the respective Institutional Animal Care and Use committees of IMAGINE Institute/Necker Hospital and Paris Descartes University.

Bone histology

All mice were analyzed via X-Ray (Faxitron MX-20 DC12 Edimex) and measured after euthanasia. For naso-anal, femur and tibia length acquisition, after dissection of the mice using binocular magnifier, we measured these lengths with a digital caliper. For histological analyses, bone tissues were fixed in 4% paraformaldehyde at 4°C for 24 h, decalcified in 0.5 M ethylenediaminetetraacetic acid (EDTA) (pH 7.4) for one week, and embedded in paraffin. Bone sections (5 μm) were prepared with a microtome (Thermo Fisher Scientific) and used for Safranin O staining, immunohistological analyses, and in situ hybridization.

For Safranin O staining, tissue sections were deparaffinized and rehydrated with ethanol. Slides were stained with Weigert’s iron hematoxylin working solution for 5 min then rinsed in running tap water for 4–5 min. Slides were then stained with 0.02% fast green (FCF) solution for 30 s, rinsed with 1% acetic acid for 30 s, and stained in 0.1% safranin O solution for 45 min. Slides were then mounted by using Euparal Mounting Medium.

Chondrocyte surface area analysis

For the surface area chondrocyte analysis, using Fiji® software, approximately 200 hypertrophic chondrocytes were measured for each section. To not introduce a bias in the analysis, all measurements were done at the same depth in the growth plates.

Skin and aorta staining

For orcein staining, aortic tissue sections were deparaffinized and rehydrated with ethanol. Slides were stained with an orcein 1% working solution for 15 min then rinsed under running tap water. Slides were differentiated with lithium carbonate 1% solution for 15 min with constant agitation, and were then mounted in xylene-based Eukitt® medium (Sigma-Aldrich).

For Masson’s trichrome staining, skin tissue sections were deparaffinized and rehydrated with ethanol. Slides were stained with Mayer’s hematoxylin for 10 min then rinsed under running tap water. Samples were then subjected to fuchsin/ponceau S staining for 5 min and rinsed with distilled water. Slides were differentiated in phosphomolybdic acid for 5 min, directly stained in aniline blue for 2 min, and incubated in 1% acetic acid for 5 min.

In all cases, aorta and skin sections were mounted in xylene-based Eukitt® medium (Sigma-Aldrich) and observed using Nanozoomer 2.0. Slide Scanner (HAMAMATSU).

In situ hybridization

PCR products were used to generate antisense and sense cRNA probes (Table 1) with a T7 DIG RNA Labeling kit (Roche) and digoxigenin-11-uridine-5′-triphosphate (DIG-11-UTP) (Roche) being used according to the manufacturer’s specifications. Paraformaldehyde-fixed paraffin-embedded sections obtained from Fbn1 model mice were hybridized to 1 μg/ml of DIG-11-UTP-labeled Col2a1 and Col10 cRNA probes as described previously. After staining with 5-bromo-4-chloro-3′-indolyphosphate/nitro blue tetrazolium in the dark (Roche) based on provided directions, the slides were imaged with an Olympus PD70-1X2-UCB microscope and analyzed with the CellSens® and Fiji® software.

Table 1

Riboprobe sequences used for in situ hybridization

Probes	Target genes	Sequences (5′ → 3′)
Col10a1 S forward	Col10 (sense probe)	TAATACGACTCACTATAGGGAGACAAACGGCCTCTACTCCTCTGA
Col10a1 S reverse	Col10 (sense probe)	CGATGGAATTGGGTGGAAAG
Col10a1 AS forward	Col10 (antisense probe)	CAAACGGCCTCTACTCCTCTGA
Col10a1 AS reverse	Col10 (antisense probe)	TAATACGACTCACTATAGGGAGACGATGGAATTGGGTGGAAAG

Table 2

Sequences of primers used in qPCR analyses

Target genes	Forward sequences (5′—3′)	Reverse sequences (5′—3′)
Col1a1	GCTCCTCTTAGGGGCCACT	CCACGTCTCACCATTGGGG
Col1a2	CTGGTCCTGTTGGAAGTCGT	CAGATGCACCTGTTTCTCCA
Col3a1	CAATGTAAAGAAGTCTCTGAAG	CAAACAGGGCCAATGTCCAC
Pai-1	AGGATCGAGGTAAACGAGGC	GCGGGCTGAGATGACAAA
Ctgf	TGACCTGGAGGAAAACATTAAGA	AGCCCTGTATGTCTTCACACTG
Smad6	GTTGCAACCCCTACCACTTC	GGAGGAGACAGCCGAGAATA
Id1	CCTAGCTGTTCGCTGAAGGCT	CTCCGACAGACCAAGTACCAC
Fbn1	CCTGTGCTATGATGGGTTCA	AGGTCCCACTAAGGCAGATGT
Col10a1	TGCAATCATGGAGCTCACAGA	CAGAGGAGTAGAGGCCGTTTGA
Gapdh	TGTCCGTCGTGGATCTGAC	CCTGCTTCACCACCTTCTTG
Hprt	GTTGGGCTTACCTCACTGCT	TCATCGCTAATCACGACGCT
Hsp90	CCAGAAACCCGGATGACA	TGACCTCTACAGAGAAGTGCTTG

Immunohistological analysis

Growth plates were analyzed using antibodies against Collagen 10 (1:200; Abcam ab58632), Sox9 (1:50; Santa Cruz Biotechnology sc-20 095), FBN1 (1:100; LS-Bio LS-C358981), TGFβ-1 (1:200; Abcam ab92486), RankL (1:500; R&D AF462), RunX2 (1:200; Cell Signaling Technology #12556) and phospho-p38 (1:500; Abcam ab4822). Slides were deparaffinized and rehydrated using ethanol, followed by antigen retrieval with Target Retrieval Solution Citrate pH 6 (DAKO) for 30 min at 95°C. Enzymatic digestion was also performed, using hyaluronidase 0.25% for 1 h at room temperature. Blocking was performed using Protein Block, Serum-Free (DAKO) for 1 h at room temperature, and antibodies were diluted in Antibody Diluent, Background Reducing buffer (DAKO). Primary antibodies were incubated overnight at 4°C. After several washes using phosphate-buffered saline, Highly Cross-Adsorbed Alexa Fluor 647 (Thermo Fisher Scientific) or secondary antibodies with HRP (DAKO) diluted in Antibody Diluent, Background Reducing buffer (DAKO) was incubated for 2 h at room temperature. Finally, sections were washed using phosphate-buffered saline and mounted using ProLong™ Gold Antifade Mountant with DAPI (Invitrogen™). Fluorescence was acquired using Leica DMi8 microscope at specific wavelength and analyzed using Leica software and Fiji®. 3,3′-diaminobenzidine (DAB) staining was acquired using Olympus PD70-1X2-UCB microscope and analyzed with CellSens and Fiji®. Negative controls have no primary antibodies, but undergo the same different steps.

Primary chondrocyte isolation and microfibrils network immunofluorescence

Mice were sacrificed, and chondrocytes were isolated from the rib cage and were put in cultures as previously described. Chondrocytes were grown in cell culture chambers with DMEM/HamF12 (with 10% FBS and antibiotics) until confluence. Then, cells were fixed with 3% formaldehyde for 20 min. After a treatment of glycine 0.2 M, cells were blocked with 3% Bovide Serum Albumine (BSA) for 1 h, then incubated with primary antibody diluted in 3% BSA overnight. Fibronectin (1:100; Abcam ab2413) and FBN1 (1:100; LS-Bio LS-C358981) antibodies were used. The day after, cells were rinsed 3 times with PBS/MgCL₂/CaCl₂, incubated with secondary antibody diluted in BSA 3% for 45 min. The cell culture chambers were mounted in ProLong™ Gold Antifade Mountant with DAPI (Invitrogen™) and analyzed with confocal microscope LSM-700 from a platform of Cell Imagery of Imagine Institute. Acquisitions were analyzed with Fiji.

Fbn1 microfibrils thickness distribution frequency and density of the microfibrillar network

For the analysis of the Fbn1 microfibrils thickness, at least four primary chondrocyte cultures per genotype were used. To determine the population’s distribution of the Fbn1 microfibrils, using Fiji® software, from the high-quality confocal acquisition of the networks, each microfibril was measured in diameter. The results were presented as the number of values on the thickness of microfibrils with a 5 μm bound. For the analysis of the microfibrillar network’s density, using Fiji® software, the fluorescence mean intensity of Fbn1 network was reported to the area of the acquisition which is proportionate to the network’s density.

Western blotting

Primary chondrocytes were used to obtain cell lysates and protein extracts in RIPA buffer (Thermo Scientific) containing a protease and phosphatase inhibitor cocktail (Thermo Scientific). Protein extractions were dosed with Pierce™ BCA Protein Assay Kit (Thermo Scientific). Antibodies against Smad2 (1:500, Abcam ab63672) and phospho-Smad2 (Ser467) (1:500; Abcam ab53100) were used on a nitrocellulose membrane. Revelation was made using secondary antibodies linked to horseradish-peroxidase. The levels of proteins were normalized to Smad2 using Fiji®.

RNA extraction, reverse-transcription and quantitative PCR

Skin and femoral head from bone tissues were cryopulverized in liquid nitrogen. Followed by RNA extraction using the NucleoSpin® RNA kit (Macherey-Nagel) according to the manufacturer’s instructions, and 1 μg of RNA was used for reverse transcription using the RNA-to-cDNA™ kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. reverse transcription (RT)-qPCR was performed using a Step One Plus RT-qPCR system (Applied Biosystems, Thermo Scientific) according to the manufacturer’s instructions. Absolute Blue qPCR SYBR Green supermix (Thermo Scientific) was used to quantify Col1a1, Col1a2, Col3a1, Fbn1, Col10, Pai-1, Ctgf, Smad6 and Id1 according to manufacturer’s instructions (see the primers in the Table 2). Quantification of each gene was performed in triplicate. The results were normalized to the expression of three housekeeping genes (glyceraldehyde 3-phosphate dehydrogenase, hypoxanthine-guanine phosphoribosyltransferase and mouse heat shock protein 90 kDa).

ELISA detection of active and total TGF-β1

TGF-β1 levels in 100 μl samples of culture medium from confluent primary chondrocytes from the ribs of mutant and WT mice were quantified with a Quantikine mouse TGF-β1 ELISA kit (R&D systems #MB100B). Using approximatively 80% of confluence primary chondrocyte cultures, cells at the same passage underwent several washes using PBS 1X and then starved in DMEM media without FBS for 2 days before retrieving the conditioned media. The samples were acidified for measurements of total (active plus latent) TGF-β1. TGF-β1 standard curves were generated for each assay. All experiments were performed in triplicate.

Statistical analysis

All statistical analyses were performed using PRISM 8 software (GraphPad, La Jolla, CA, USA). All values are shown as means ± SD. All statistical tests were the ordinary one-way analysis of variance (ANOVA) or the Brown–Forsythe and Welch ANOVA tests depending on the SD’s differences.

Abbreviations
 
• ad

  Acromicric dysplasia

 
• ADAMTS(L)

  A Disintegrin And Metalloproteinase with Thrombospondin Motifs (Like)

 
• BCIP/NBT

  5-bromo-4-chloro-3′-indolyphosphate/nitro blue tetrazolium

 
• BMP

  Bone morphogenetic protein

 
• cbEGF

  calcium-binding epidermal growth factor

 
• CMV

  Cytomegalovirus

 
• Col1a1

  type 1a1 collagen

 
• Col1a2

  type 1a2 collagen

 
• Col2a1

  type 2a1 collagen

 
• Col3a1

  type 3a1 collagen

 
• Col10a1

  type 10a1 collagen

 
• Ctgf

  Connective tissue growth factor

 
• DIG-11-UTP

  Digoxigenin-11-uridine-5′-triphosphate

 
• DNA

  Deoxyribonucleotide acid

 
• ECM

  Extracellular matrix

 
• ELISA

  Enzyme-linked immunosorbent assay

 
• FBN1

  Fibrillin-1

 
• FBS

  Fetal bovine serum

 
• GAPDH

  Glyceraldehyde 3-phosphate dehydrogenase

 
• GD

  Geleophysic dysplasia

 
• HEK293

  Human embryonic kidney cells 293

 
• Ho

  Homozygote

 
• Hprt

  Hypoxanthine-guanine phosphoribosyltransferase

 
• Hsp90

  mouse heat shock protein 90 kDa

 
• HT

  Heterozygote

 
• HZ

  Hypertrophic zone

 
• Id1

  Inhibitor of DNA-binding 1

 
• KI

  Knock-in

 
• KO

  Knock-out

 
• LTBP

  Latent TGF-β binding protein

 
• MAPK

  Mitogen-activated protein kinase

 
• MFS

  Marfan syndrome

 
• mRNA

  messenger ribonucleotide acid

 
• MS

  Myhre syndrome

 
• OMIM

  Online Mendelian inheritance in man

 
• P1

  Post-natal day 1

 
• P30

  Post-natal day 30

 
• Pai1

  Plasminogen activator inhibitor-1

 
• PBS

  Phosphate buffer saline

 
• PCR

  Polymerase chain reaction

 
• PHZ

  Pre-hypertrophic zone

 
• PZ

  Proliferative zone

 
• RankL

  Receptor activator of nuclear factor kappa-B ligand

 
• Runx2

  Runt-related transcription factor 2

 
• SSKS

  Stiff skin syndrome

 
• (p-)Smad2

  (phospho-)Mothers against decapentaplegic homolog 2

 
• Smad6

  Mothers against decapentaplegic homolog 6

 
• Sox9

  SRY-box transcription factor 9

 
• TB(5)

  TGF-β binding protein-like (5)

 
• TGF-β

  Transforming growth factor-β

 
• WMS

  Weill–Marchesani syndrome

 
• WT

  Wild-type

Acknowledgements

This program has received a state subsidy managed by the National Research Agency under the ‘Investments for the future’ program (ANR-A0-IAHU-01). We thank the animal facilities and histology platform from Imagine institute. We also thank the mouse clinic at Strasbourg, (IGBMC).

Conflict of Interest statement

The authors declare no conflict of interest.

Funding

Agence Nationale de la Recherche (R09183KS to C.L.G. and V.C.-D.); Seventh Framework Programme SYBIL (to V.C.-D.); Fondation pour la Recherche Médicale prix de la Fondation Line-Pomaret-Delalande (to L.D. and C.L.G.).

Author notes

Laure Delhon, Zakaria Mougin have equally contributed to this work.