Discussion
Previous studies suggest that BMP signaling is involved in a large number of developmental events. Many of these events occur early in embryogenesis, and complete inactivation of BMP receptors causes death by E9.5 (Mishina et al. 1995). The Gdf5-Cre recombination system bypasses the early embryonic lethality of Bmpr1a mutations, and provides important new information about the role of this receptor in limb and skeletal development.
The three major limb phenotypes revealed by eliminating Bmpr1a with Gdf5-driven Cre include webbing between digits, lack of joint formation at specific locations in the ankle, and failure to maintain articular cartilage after birth, resulting in severe arthritis. Previous studies have shown that manipulation of BMP signaling alters interdigital apoptosis during development of the limb, but no experiment has identified a specific member of the BMP signaling pathway that is required for this process (Yokouchi et al. 1996; Zou and Niswander 1996; Zou et al. 1997; Guha et al. 2002). Our new loss-of-function data confirm that BMP signaling is required for interdigital apoptosis and suggests that Bmpr1a is a critical component for mediating this signal.
At some sites, loss of Bmpr1a function leads to a defect in the early stages of joint formation, resulting in a complete failure to form a joint and fusion of bones in the ankle. Mutations in two different ligands in the BMP family, Gdf5 and Gdf6, the Bmpr1b receptor, and in the human Noggin locus (Storm and Kingsley 1996; Gong et al. 1999; Baur et al. 2000; Yi et al. 2000; Settle et al. 2003) also produce defects in joint formation at specific locations in the limbs. The joint defects associated with multiple components of the BMP pathway provide strong evidence that BMP signaling is required for early stages of joint formation at some anatomical locations.
Most joints still form normally when Bmpr1a is knocked out in Gdf5 expression domains. The lack of joint fusions outside the ankle region could be due to differences in requirement for BMP signaling in different joints, to compensating expression of other BMP receptors outside the ankles, or to differences in the detailed timing of Gdf5-Cre stimulated gene inactivation in ankles and other joint regions. Comparison of the expression of the HPLAP marker (driven directly by Gdf5 control elements) and the R26R LACZ marker (expressed following Gdf5-Cre recombination) suggests that recombination-stimulated changes in gene expression may be delayed for a 0.5–1 d in the digit region (see Figure 1C). In addition, levels of Bmpr1a mRNA and protein may persist for some time following Gdf5-Cre stimulated recombination, making it possible to bypass an early requirement for Bmpr1a in joint formation at some locations.
Following the decay of Bmpr1a mRNA and protein, the Gdf5-Cre strategy should result in permanent inactivation of Bmpr1a function in recombined cells. This system thus provides one of the first strong genetic tests of Bmpr1a function at later stages of joint development. Despite the normal appearance of articular regions and gene expression immediately after birth, Bmpr1a-deficient animals are unable to maintain the normal differentiated state of articular cartilage as they continue to develop and age. These results suggest that BMP receptor signaling is essential for continued health and integrity of articular cartilage in the postnatal period.
Articular cartilage is a key component of synovial joints and is one of the few regions in the skeleton where cartilage is maintained into adulthood. Despite the importance of articular cartilage in joint health and mobility, little is known about the factors that create and maintain it in thin layers at the ends of long bones. In our experiments, articular cartilage lacking Bmpr1a retains some normal characteristics, in that it maintains a very low proliferation rate, does not express Col1a1, and continues to express SOX9, a major transcription factor regulating expression of structural components of cartilage matrix. However, several of the most prominent structural components of cartilage matrix fail to be maintained in mutant animals, resulting in decreased synthesis of Col2a1, Agg, and proteoglycans. Therefore, BMPR1A appears to maintain articular cartilage primarily through inducing expression of key ECM components.
It is interesting that the SOX9 transcription factor continues to be expressed in mutant cartilage despite loss of Col2a1, a direct target of this transcription factor (Bell et al. 1997; Lefebvre et al. 1997). Previous studies suggest that SOX9 activity can be modified by protein kinase A (PKA)-dependent protein phosphorylation, or by coexpression of two related proteins, L-SOX5 and SOX6 (Lefebvre et al. 1998; Huang et al. 2000). In addition, close examination of the order of genes induced during chicken digit formation reveals that Sox9 turns on first, followed by Bmpr1b with L-Sox5, and then Sox6 and the cartilage matrix structural components Col2a1 and Agg (Chimal-Monroy et al. 2003). These results, together with the altered pattern of gene expression seen in our Bmpr1a-deficient mice, suggest that BMPR1A signaling may normally act to stimulate SOX9 by post-translational protein modification, or to induce L-Sox5 or Sox6 in cartilage to maintain expression of ECM components. These models are consistent with the ability of BMP2 to both increase PKA activity and induce expression of Sox6 in tissue culture cells (Lee and Chuong 1997; Fernandez-Lloris et al. 2003). Although we have tried to monitor the expression of L-Sox5 or Sox6 in postnatal articular cartilage, and test the phosphorylation state of SOX9 using previously described reagents (Lefebvre et al. 1998; Huang et al. 2000), we have been unable to obtain specific signal at the late postnatal stages required (unpublished data). Furthermore, null mutations in L-Sox5 or Sox-6 cause lethality at or soon after birth, and no effect on cartilage maintenance has been reported (Smits et al. 2001). However, it seems likely that these or other processes regulated by BMP signaling cooperate with SOX9 to induce target genes in articular cartilage.
Mutation of Smad3 or expression of dominant negative transforming growth factor β (TGF-β) type II receptor also disrupts normal articular cartilage maintenance (Serra et al. 1997; Yang et al. 2001). Both manipulations should disrupt TGFβ rather than BMP signaling, and both manipulations cause articular cartilage to hypertrophy and be replaced by bone. In contrast, our analysis of Bmpr1a mutant articular cartilage showed a loss of ECM components, but no signs of hypertrophy or bone replacement. Therefore, TGFβ and BMP signaling are playing distinct but necessary roles to maintain articular cartilage.
Although BMPs were originally isolated on the basis of their ability to induce ectopic bone formation, their presence in articular cartilage and strong effect on cartilage formation has stimulated interest in using them to repair or regenerate cartilage defects in adult animals (Chang et al. 1994; Erlacher et al. 1998; Edwards and Francis-West 2001; Chubinskaya and Kuettner 2003). The failure to maintain articular cartilage in the absence of normal BMPR1A function suggests that ligands or small molecule agonists that interact specifically with this receptor subtype may be particularly good candidates for designing new approaches to maintain or heal articular cartilage at postnatal stages.
Lack of Bmpr1a function in articular cartilage results in severe fibrillation of the articular surface and loss of joint mobility. The development of severe arthritis symptoms in Bmpr1a-deficient mice raises the possibility that defects in BMP signaling also contribute to human joint disease. Osteoarthritis is known to have a significant genetic component, but it likely involves multiple genetic factors that have been difficult to identify (Spector et al. 1996; Felson et al. 1998; Hirsch et al. 1998). Humans that are heterozygous for loss-of-function mutations in BMPR1A are known to be at risk for juvenile polyposis (Howe et al. 2001; Zhou et al. 2001), but the risk of osteoarthritis for these people has not been reported. However, the control mice used in this study were heterozygous for a null allele of Bmpr1a, and they showed little sign of osteoarthritis even late in life. Several chromosome regions have been previously linked to arthritis phenotypes in humans using either association studies in populations or linkage studies in families. It is interesting to note that several of these chromosome regions contain genes encoding different members of the BMP signaling pathway, including the BMP5 gene on human chromosome 6p12 (Loughlin et al. 2002), the MADH1 gene on human chromosome 4q26–4q31 (Leppavuori et al. 1999; Kent et al. 2002), and the BMPR2 receptor on human chromosome 2q33 (Wright et al. 1996). The complex nature of human osteoarthritis suggests that interactions between multiple genes may be involved in modifying susceptibility to the disease. The inclusion of genetic markers near BMP signaling components may help identify additional osteoarthritis susceptibility loci and facilitate the search for causative mutations.
Development and disease processes in synovial joints have been difficult to study genetically, because synovial joints are generated and function at relatively late stages of vertebrate development. The Gdf5-Cre system provides a new method for restricting gene expression or inactivation primarily to articular regions, thus avoiding the pleiotropic functions of many genes in other tissues. Depending on the configuration of the floxed target gene, this system can be used to either activate the expression of a gene primarily in developing joints (ssee Figure 1B–1D), or to inactivate gene function in articular regions (see Figure 3). Additional studies with this system should greatly enhance our knowledge of the development, function, and disease mechanisms of joints, and may bring us closer to better prevention and treatment of joint diseases.