Bone abnormalities in latent TGF-β binding protein (Ltbp)-3–null mice indicate a role for Ltbp-3 in modulating TGF-β bioavailability

The latent TGF-β binding proteins (LTBPs)^*-1–4 are matrix molecules composed of multiple (14–20) EGF-like domains and four domains containing eight cysteines (8-cys) that are specific for the LTBPs and the fibrillins (for review see Handford et al., 2000; Koli et al., 2001). The modular structure of Ltbp-3 is shown in Fig. 1 A (Yin et al., 1995). LTBP-1, 3, and 4 covalently bind latent TGF-β (Koli et al., 2001). The TGF-βs are 25-kd homodimeric cytokines derived by intracellular proteolytic processing of larger proproteins (Massague, 1998). However, once cleaved from the cytokine, the TGF-β propeptide, called the latency-associated peptide (LAP), remains noncovalently associated with the mature TGF-β after secretion (Koli et al., 2001). This complex of TGF-β and LAP, the small latent complex, is inactive, and the dissociation of TGF-β from LAP is a crucial regulatory step in TGF-β action. The small latent complex can form a large latent complex by the bonding of cysteines in the LAP with a pair of cysteines in the third 8-cys domain of LTBP (Gleizes et al., 1996; Saharinen et al., 1996). All three TGF-β isoforms bind to LTBPs-1, 3, or 4; however, neither LTBP-2 nor the fibrillins bind TGF-β (Saharinen and Keski-Oja, 2000).

The LTBPs have been proposed to have two functions: as structural components of the extracellular matrix (Dallas et al., 1995) and as modulators of TGF-β bioavailability (Koli et al., 2001). Experiments in culture have shown that the association of LTBP with latent TGF-β is important for at least two aspects of TGF-β biology. First, the binding of LTBP-1 to the small latent complex facilitates its folding and secretion (Miyazono et al., 1991). Second, LTBP-1 modulates latent TGF-β activation (Flaumenhaft et al., 1993; Nakajima et al., 1997; Gualandris et al., 2000). However, these studies have not provided direct proof for the physiological role of LTBP in modulating TGF-β activity. To address this point, we have generated Ltbp-3–null mice. These mice have an altered skull shape caused by the premature ossification of the cranial base synchondroses. In addition, there is enhanced accumulation of trabecular bone in the long bones and vertebrae and degeneration of the articular cartilage as the animals age. These phenotypic abnormalities are consistent with postulated roles of TGF-β in bone formation and homeostasis. As such, this report represents the first indication for physiological control of TGF-β activity by an LTBP.

Ltbp-3–null mice were produced by gene targeting using a targeting vector to replace two exons containing a nonunit number of codons (Joyner, 1995), including bases 278–807 of the ORF with the neo^r selectable marker (see online supplemental material, available at http://www.jcb.org/cgi/content/full/200111080/DC1, for details). These two exons code for the first EGF-like repeat, the pro-gly–rich region, and the beginning of the first 8-cys domain (Fig. 1). This deletion causes a frameshift in the ORF and premature termination of translation. Three clones (17, 22, and 25) with homologous recombination in the Ltbp-3 gene were detected by Southern blot analysis (unpublished data), and two clones were used to produce chimeric animals. Ltbp-3–null animals were obtained by crossing heterozygous progeny of chimeric mice. Northern blot hybridization of lung RNA from null animals with a probe mapping 3′ from the deleted region showed the absence of the Ltbp-3 transcript (Fig. 2 A). Therefore, we concluded that these mice were effectively Ltbp-3–null animals. Ltbp-3–null mice were born in the expected Mendelian ratio with no apparent defects. By day 10, null animals displayed a rounded head and shortened snout. X-ray radiography of 2-mo-old mutant mice revealed a domed skull, abnormal apposition of the upper and lower incisors, and curvature of the cervical and thoracic vertebrae (thoracic kyphosis) (Fig. 2 B).

The appropriate anatomical development of the skull requires the coordinated growth of the membranous and endochondral bones to accommodate the increasing size of the brain. Therefore, the sutures between the bones of the cranial vault as well as growth plates in the skull base remain nonossified for an extensive period after birth. There are reports describing potential roles for TGF-βs in the differentiation of the membranous bones of the skull (Opperman et al., 1999), but there is no information concerning TGF-β function in the biology of the bones of the skull base. Histological studies of the sutures in wild-type and Ltbp-3–null animals revealed no pathological synostosis, i.e. premature fusion of one or more of the cranial vault sutures (unpublished data). However, differential staining for cartilage and bone in whole mount preparations, as well as histological analysis of the cranial base in 21 day old animals, revealed that the cartilaginous growth plates of the synchondroses were absent in mutant animals (Fig. 3, B and D), whereas the synchondroses were nonossified in wild-type littermates (Fig. 3, A and C).

Histologically, a synchondrosis resembles two opposed growth plates with a common zone of resting chondrocytes and separate zones of proliferating and hypertrophic chondrocytes. (Fig. 4 A). The earliest histological changes in the skull base of Ltbp-3–null animals were detected in the basooccipital–basosphenoid synchondrosis 1–2 d after birth. The overall structure of the synchondrosis was altered, as no distinguishable columns of proliferating chondrocytes were visible in the mutant synchondrosis (Fig. 4, compare A and B), and the zones of hypertrophic chondrocytes were wider. Collagen X, a marker for hypertrophic chondrocytes (Elima et al., 1993), was restricted to the ends of the synchondrosis in wild-type animals (Fig. 4 C), but was detected almost throughout the synchondrosis in Ltbp-3–null animals (Fig. 4 D). In addition, collagen type II, a marker for nonhypertrophic chondrocytes (Swalla et al., 1988), was present in the central zone of wild-type synchondrosis, but was absent from that of Ltbp-3– null animals (Fig. 4, E and F). The distance between the cortical bone fronts was smaller in the null animals compared to wild-type littermates as visualized by Masson's trichrome staining, which stains bone blue (Fig. 4, G and H). The ectopic ossification in mutant synchondrosis was also clear from the expression of the bone sialoprotein (Bsp)-1 gene, an osteoblast specific marker (Bianco et al., 1991), in the cells surrounding the synchondrosis (Fig. 4, I and J). Although the basooccipital–basosphenoid synchondrosis was obliterated by 3–5 d after birth, the first changes in the basosphenoid–presphenoid synchondrosis in the null animals were not seen until days 3–10 (unpublished data).

Therefore, premature closure of the synchondroses in Ltbp-3–null mice is responsible for the observed cranio-facial malformations. The failure to extend the presphenoid, basosphenoid, and the basooccipital bones results in a shortened cranial base. To accommodate the growing brain volume, the membranous bones of the vault expand outward and upward creating a domed skull. The anterior displacement of the foramen magnum may cause the development of the kyphosis, as the spinal column must be realigned with the overall body axis. The shortening of the cranial base also causes the alteration in the apposition of the incisors.

The wider zones of hypertrophic chondrocytes, as well as more advanced bone fronts in the Ltbp-3–null mice, resembled certain changes observed in animals deficient for the expression of parathyroid-hormone related protein (PTHrP) or its receptor PTH/PTHrP-R (for review see Karsenty, 2001). Vortkamp et al. (1996) proposed that chondrocyte differentiation is regulated by an inhibitory feedback loop in which Indian hedgehog (Ihh), produced by the most differentiated prehypertrophic and hypertrophic chondrocytes, stimulates the production of PTHrP by the periarticular cartilage and perichondrium, and PTHrP inhibits hypertrophic differentiation by interaction with its receptor on prehypertrophic chondrocytes (Chung et al., 2001). Therefore, we examined the expression of Ihh, PTHrP, and PTH/PTHrP-R in wild-type and mutant mice. In wild-type mice, Ihh was expressed by prehypertrophic and differentiating hypertrophic chondrocytes, whereas in null animals, Ihh expression was broader, indicating that chondrocytes in the central region of the synchondrosis were committed to hypertrophic differentiation (Fig. 4, K and L). In wild-type animals, PTH/PTHrP-R expression was detected in prehypertrophic chondrocytes, whereas in Ltbp-3–null animals, PTH/PTHrP-R transcripts were detected in the central region of the synchondrosis (Fig. 4, M and N). The expression pattern appeared less organized than that observed in wild-type synchondrosis, but the level of expression was similar as judged by the intensity of the staining. In sections probed for PTHrP, there was broad expression throughout the presumptive zone of proliferating chondrocytes, and the level of expression appeared lower in Ltbp-3–null animals compared to wild-type animals (PFig. 4, O and P). This decreased expression of PTHrP was also observed in the synchondroses of younger animals (see online supplemental material, available at http://www.jcb.org/cgi/content/full/200111080/DC1). A decreased level of PTHrP in Ltbp-3–null mice would allow more extensive chondrocyte differentiation and account for the more rapid ossification of the synchondroses.

The changes in Ltbp-3–null mice are consistent with previous reports describing a role for TGF-β; in regulating PTHrP expression, although we have been unable to detect differences with antibodies that recognize either active TGF-β or phosphorylated Smads. Pateder et al. (2001) and Serra et al. (1999) demonstrated in cell and organ culture that TGF-β stimulates PTHrP expression and inhibits hypertrophic differentiation. We infer that in Ltbp-3–null mice there is a deficit in TGF-β that results in decreased PTHrP expression and early differentiation of the synchondroses, ectopic ossification, and synostosis. The decrease in PTHrP coincident with Ihh expression in the central zone of the synchondroses suggests that TGF-β acts before the induction of PTHrP expression and after Ihh expression.

Because TGF-β has been shown to be important for long bone physiology, we examined the structure of the long bones of Ltbp-3–null mice to determine if other TGF-β-mediated functions were affected. By day 8 after birth, null animals showed growth retardation. The weight of adult Ltbp-3–null animals was 30–80% of sex-matched littermates, and the endochondral bones were shorter by ∼10–25%. Histological analysis of the growth plates of tibiae, femora, and vertebrae from 1 d to 2-mo-old animals revealed no obvious differences between Ltbp-3–null and wild-type littermates (unpublished data). Consistent with the report of Filvaroff et al. (1999), who expressed a type II TGF-β receptor with a truncated cytoplasmic domain (RIIDN) in osteoblasts and found age-dependent increases in trabecular bone mass, increased trabecular mass in long bones was observed at 3 mo of age in Ltbp-3–null mice with a body weight <60% of sex-matched wild-type littermates (unpublished data). This osteosclerosis was more pronounced in 6- and 9-mo-old mutant animals (Fig. 5, compare A and B). Staining for proteoglycans revealed unmineralized cores within the trabecular bones close to the growth plates, suggesting rapid extracellular matrix deposition and initiation of trabecular bone formation. The increased number of metaphysial trabeculae also suggested slow turnover. Similar changes occurred in the vertebrae (5, C–F). Interestingly, mutations in LAP cause Camurati–Engelmann syndrome, which is characterized by sclerosis and hyperosteosis (Janssens et al., 2000; Kinoshita et al., 2000), whereas transgenic animals overexpressing TGF-β2 in osteoblasts develop osteoporosis (Erlebacher and Derynck, 1996). Hence, we conclude that an Ltbp-3 defect mirrors long bone phenotypes caused by impaired TGF-β signaling.

The inhibition of TGF-β signaling has been shown to lead to periarticular cartilage terminal differentiation and ossification. Mice either expressing a TGF-β RIIDN in articular cartilage, synovium and periosteum/perichondrium (Serra et al., 1997), or deficient in Smad-3 (Yang et al., 2001) develop degenerative joint disease. Ltbp-3–null animals also develop progressive degeneration of articular cartilage resembling osteoarthritis. Histological analysis of the knee joints of mutant and wild-type littermates revealed pathological changes in the articular cartilage of 6-mo-old Ltbp-3–null animals: proteoglycan staining was decreased (compare Fig. 5, G and H) and hypertrophic chondrocytes were detected in the superficial layers of the articular cartilage (Fig. 5 H). In wild-type mice, the articular cartilage consisted almost exclusively of mature nonhypertrophic chondrocytes (Fig. 5 G). In mutant mice, at 9 mo, articular cartilage was absent, the articular surface appeared ossified and fibrotic (Fig. 5 J), and osteophytes were present (unpublished data). Similar changes were observed in the vertebral joints (Fig. 5, C–F). The degenerative changes in Ltbp-3–null joints present later (6–9 mo) than those in RIIDN transgenic and Smad-3–deficient animals. We believe that Ltbp-3 deficiency causes a more moderate decrease in TGF-β levels compared to the inhibition of TGF-β signaling in DNIIR and Smad-3–null animals.

This report describes the phenotype of the Ltbp-3–null mouse: premature obliteration of synchondroses, osteosclerosis, and osteoarthritis. These manifestations are consistent with published evidence suggesting TGF-β involvement in bone remodeling and homeostasis. Therefore, we propose that Ltbp-3 regulates TGF-β bioavailability either by enhancing secretion of the TGF-β small latent complex or by participating in the activation of the latent TGF-β, as suggested from results obtained for LTBP-1 in cell culture (Miyazono et al., 1991; Flaumenhaft et al., 1993). Irrespective of the mechanism, the phenotypic changes in Ltbp-3–null mice are consistent with previous results describing a role for TGF-β in regulating PTHrP expression, as well as with the effects of impaired TGF-β signaling on bone physiology in vivo. Hence, this represents the first report providing genetic evidence in support of the role for LTBP in regulating TGF-β bioavailability. Experiments to determine the precise mechanism underlying these phenotypes are underway.

Ltbp-3 gene targeting strategy, as well as the production and characterization of mice with a disrupted Ltbp-3 gene, are described in the online supplemental material (available at http://www.jcb.org/cgi/content/full/200111080/DC1). X-ray radiography was performed on anaesthetized animals using a Micro 50 (Microfocus Imaging) at 30 KV for 10 s.

Whole-skull staining with Alizarin red S and Alcian blue was performed as described (Lufkin et al., 1992). For histological analysis, samples were fixed overnight in 4% paraformaldehyde in PBS at 4°C, decalcified in 10% EDTA/2.5% paraformaldehyde in PBS for 7–14 d at 4°C, dehydrated through an ethanol series, cleared in xylene, and embedded in paraffin. 5-μm sections were stained with either Masson's Trichrome Stain or Weingart's hematoxilin/Fast Green/Safranin O (Luna, 1992) for bone and cartilage. Collagen II was detected using mouse monoclonal 2B1.5 antibody (NeoMarkers) and M.O.M. Kit (Vector) and collagen X using rabbit antiserum pXNC1-8, a gift from G. Lunstrum (Shriners Hospital for Children, Portland, OR) and the Vector Elite Kit.

RNA probes were prepared using DIG RNA labeling kit (Roche). The probes used were Bsp-1, a gift from I. Thesleff (University of Helsinki, Helsinki, Finland), Ihh, a gift from A. McMahon (Harvard University, Cambridge, MA), PTHrP-R, gift of A. Broadus (Yale University Medical School, New Haven, CT), and PTHrP, a gift of H. Kronenberg (Massachusetts General Hospital, Boston, MA). In situ hybridization was performed as described (Wassarman et al., 1997).

Included in the online supplemental materials (available at http://www.jcb.org/cgi/content/full/200111080/DC1) are details of the targeting strategy for production of Ltbp-3 disruption and characterization of the genotypes of the mutant mice (Fig. S1), a picture of in situ hybridization of embryos probed for Ltbp-3 expression (Fig. S2), and an illustration of PTHrP expression in the synchondroses of newborn animals (Fig. S3).

The authors would like to thank J. Annes, B. Arias, E. Arteaga-Solis, A. Auerbach, A. Joyner, S. Ortega, and M. Vassallo for assistance.

This work was supported by grants CA78422, CA34282, and DE13742 from the National Institutes of Health, and from the National Marfan Foundation to D.B. Rifkin. B. Dabovic was supported by Department of Defense grant DAMD 17-99-1-9319. H. Obata was supported by a Bausch and Lomb Overseas Research Fellowship.