|Home | About | Journals | Submit | Contact Us | Français|
The Runx2 gene product is essential for mammalian bone development. In humans, Runx2 haploinsufficiency results in cleidocranial dysplasia, a skeletal disorder characterized by bone and dental abnormalities. At the molecular level, Runx2 acts as a transcription factor for genes expressed in hypertrophic chondrocytes and osteoblasts. Runx2 gene expression and protein function are regulated on multiple levels, including transcription, translation, and post-translational modification. Furthermore, Runx2 is involved in numerous protein-protein interactions, most of which either activate or repress transcription of target genes. In this review, we discuss expression of Runx2 during development as well as the post-translational regulation of Runx2 through modification by phosphorylation, ubiquitination, and acetylation.
The mammalian Runx (runt-related transcription factor x) genes, including Runx1, Runx2, and Runx3, encode a family of transcription factors homologous to Drosophila runt. All form heterodimers with transcriptional co-activator core-binding factor β (Cbfβ) and bind to the same DNA sequence (5′-PyGPyGGTPy-3′) through their highly conserved runt-homologous DNA binding domains (Kamachi et al., 1990; Wang and Speck, 1992; Melnikova et al., 1993; Ogawa et al., 1993; Wang et al., 1993). Each member is essential for proper embryonic development, but is distinctive in expression pattern and function, as demonstrated by the varied outcomes of gene knockout mouse models. Runx1, for instance, is essential for definitive hematopoiesis and is frequently mutated in cells giving rise to acute myelogenous leukemia (Okuda et al., 1996; Wang et al., 1996). Runx2, the focus of this review, is required for both intramembranous and endochondral bone development as well as tooth morphogenesis (Komori et al., 1997; Otto et al., 1997; D’Souza et al., 1999). Finally, Runx3 is necessary for the regulation of gastric epithelial cell growth and neurogenesis of the dorsal root ganglia (Levanon et al., 2002; Li et al., 2002). Additionally, Runx3 has been shown to cooperate with Runx2 in the regulation of chondrocyte proliferation and maturation (Yoshida et al., 2004). Therefore, while these proteins play some redundant functional roles, they each act as master regulators of differentiation for distinct tissues.
Regulation of Runx gene expression and protein function occurs at multiple levels through multiple signaling pathways. During bone development, the temporal and spatial expression patterns of Runx2 are regulated by cytokines, growth factors, and hormones, including TGF-β, BMP, FGF, sonic hedgehog, vitamin D3, and estrogen (Ducy et al., 1997; Tsuji et al., 1998; D’Souza et al., 1999; Lee et al., 1999; Zhou et al., 2000; Tou et al., 2001; Kim et al., 2003; Takamoto et al., 2003). Initiation of Runx2 transcription occurs from two distinct promoters, giving rise to two major isoforms that differ in their amino termini. The distal P1 promoter drives expression of the type II MASNS isoform, while the proximal P2 promoter drives expression of the type I MRIPV isoform (Stewart et al., 1997; Geoffroy et al., 1998; Drissi et al., 2000; Xiao et al., 2001). Type II expression appears to be limited to mature osteoblasts and terminal hypertrophic chondrocytes, while type I is found in less-differentiated osteoblast and chondrocyte precursor cells (Enomoto et al., 2000; Banerjee et al., 2001; Sudhakar et al., 2001). There is some evidence suggesting that the mRNA of Runx2 may also undergo alternative splicing events, producing additional protein isoforms (Stewart et al., 1997; Sun et al., 2004; Terry et al., 2004; Makita et al., 2008). In addition to cap-mediated translation, it has been demonstrated that the mRNA of both type I Runx2 and type II Runx2 can be translated via internal ribosomal entry sites (IRES) found in the 5′-untranslated regions (Xiao et al., 2003). Following translation, Runx2 protein levels and activity are further regulated through post-translational modifications and protein-protein interactions.
In this review, we will first discuss the overall expression pattern and essential nature of Runx2 during development. We will then summarize the post-translational regulation of Runx2, with focus on those interactions and modifications that affect the stability and activity of the protein. We will not cover transcription or translation of Runx2, nor will we discuss in detail the interaction of Runx2 with other transcription factors, co-activators, or co-repressors. The detail of such mechanisms is beyond the scope of this review and is well-summarized elsewhere (Levanon and Groner, 2004; Schroeder et al., 2005; Stock and Otto, 2005; Li and Xiao, 2007).
Embryonic bone formation occurs through either intramembranous or endochondral ossification. Intramembranous ossification is involved in forming the flat bones of the skeleton, including the bones of the skull and mandible, and is accomplished through the direct differentiation of osteoblasts from mesenchymal condensations. During endochondral ossification, a hyaline cartilage model, or anlage, is first formed, creating a matrix to which osteoblasts then infiltrate. It is through this process that the long bones, ribs, and vertebrae are formed. In both events, mature osteoblasts lay down an extracellular matrix of proteins and hydroxyapatite crystals forming mineralized bone tissue (Olsen et al., 2000; Deng et al., 2008).
Runx2 (also known as Cbfa1, PEBP2A1, and AML3) is expressed in mesenchymal stem cells during early embryonic development and acts as a master regulator in the commitment of these cells to the osteoblastic lineage. Runx2 is necessary for both intramembranous and endochondral bone formation, since mice nullizygous for Runx2 do not form mineralized bone in any part of their skeleton (Komori et al., 1997; Otto et al., 1997). Furthermore, these mice lack mature osteoblasts and show severely decreased or no expression of the osteoblast differentiation marker genes alkaline phosphatase (Alp), Osteopontin, and Osteocalcin. In vitro, calvaria-derived cells from Runx2-deficient mice were not able to differentiate into osteoblasts (Kobayashi et al., 2000). They did differentiate into adipocytes and, when in the presence of BMP-2, to chondrocytes, suggesting that Runx2 is important in lineage determination of multipotential mesenchymal cells. Other reports, however, have shown that Runx2 is required for BMP-2 and TGF-β signaling (Lee et al., 2000; Afzal et al., 2005; Bae et al., 2007; Wang et al., 2007). While some of these data were obtained from a p53-deficient cell line, the requirement of Runx2 for BMP-2 signaling is likely independent of p53, since Runx2-dependent BMP-2 activation of the Smad6 gene was observed in normal C2C12 cells (Wang et al., 2007). Additionally, antisense oligonucleotides against the runt-homology domain blocked the differentiation of primary rat osteoblasts in culture (Banerjee et al., 1997). Finally, forced expression of Runx2 in non-osteoblastic cells led to osteoblast-specific gene expression in vitro (Ducy et al., 1997).
Osteoclasts do not express Runx2 mRNA. TRAP-positive osteoclast formation was normal in E17.5 mouse embryos carrying a Runx2 nuclear matrix-targeting signal (NMTS) mutation (Choi et al., 2001), indicating that Runx2 expression is not essential for osteoclast formation during development. However, several in vitro studies demonstrate that Runx2 may affect osteoclast formation. For example, calvarial osteoblasts derived from Runx2−/− mice produce more osteoprotegerin (OPG) and less RANKL than do those from wild-type mice (Enomoto et al., 2003). Runx2 was also shown to regulate Rankl gene transcription (Mori et al., 2006). Additionally, we reported recently that Runx2 mediates BMP-2-induced RANKL production by chondrocytes and thereby influences osteoclastogenesis (Usui et al., 2008). Thus, in addition to its essential role in osteoblast development and chondrocyte hypertrophy, Runx2 may also regulate osteoclast function indirectly, through RANKL and OPG.
Mice heterozygous for Runx2 show a phenotype similar to that in humans with cleidocranial dysplasia (CCD). CCD is a rare autosomal-dominant disorder characterized by skeletal and dental abnormalities including large fontanelles, hypoplasia or the absence of clavicles, supernumerary teeth, and short stature. Loss-of-function mutations in human Runx2 are responsible for the CCD phenotype (Mundlos et al., 1997). Many of these are missense mutations found in the region encoding the runt domain and resulting in a protein that is unable to bind DNA or co-factor CBFβ (Lee et al., 1997). Translocations, insertions, deletions, nonsense, and splice-site mutations are also found throughout the gene and in the promoter region affecting expression, localization, and transactivation of the Runx2 gene product (Otto et al., 2002; HJ Kim et al., 2006; Napierala et al., 2005). Overall, analysis of the correlative genotype-phenotype data reported in cases of CCD implies that proper gene dosage of Runx2 is crucial for bone and tooth formation (Zhou et al., 1999; Yoshida et al., 2003).
In addition to its role in osteoblast differentiation, Runx2 also promotes chondrocyte maturation. Runx2 knockout (Runx2−/−) mice exhibit delayed or, in some skeletal elements, an absence of chondrocyte hypertrophy (Inada et al., 1999; Kim et al., 1999). Targeted expression of Runx2 in non-hypertrophic col2a1-expressing chondrocytes accelerates chondrocyte differentiation, leads to maturation of chondrocytes that normally do not differentiate into hypertrophic chondrocytes, and rescues the defect of chondrocyte maturation in Runx2−/− mice (Takeda et al., 2001; Ueta et al., 2001). Over-expression of a dominant-negative Runx2 in non-hypertrophic col2a1-expressing chondrocytes suppresses chondrocyte maturation, vascular invasion, and periosteal bone formation (Ueta et al., 2001; Stricker et al., 2002). These results clearly indicate that Runx2 plays an important role in chondrocyte maturation and suggest that Runx2 acts not only in hypertrophic chondrocytes, but also in col2a1-expressing proliferating chondrocytes (Fig. 1).
Although chondrocyte maturation is delayed in Runx2-deficient mice, terminal differentiation of chondrocytes does occur, suggesting that other Runx family members may play a redundant role in the regulation of chondrocyte maturation (Inada et al., 1999; Kim et al., 1999). Further evidence demonstrates that chondrocyte hypertrophy is completely absent in Runx2/Runx3 double-knockout (dKO) mice. At E18.5, all of the chondrocytes in embryos of the dKO mice express col2a1, but not col10a1, a marker of hypertrophy. Mineralization is observed in restricted regions of the limbs, vertebrae, and ribs of Runx2−/− newborns, but is completely absent throughout the entire skeletons of the Runx2/Runx3 dKO newborns (Yoshida et al., 2004). These findings indicate that Runx2 and Runx3 are both required for chondrocyte hypertrophy and maturation (Fig. 1).
During mouse embryogenesis, Runx2 mRNA expression was first observed at E11.5 in limb buds and the condensation of the humerus, respectively (Inada et al., 1999; Kim et al., 1999). Weak expression was reported as early as E9.5 in the notochord underlying the mid- and hindbrain (Otto et al., 1997). Others reported only diffuse, if any, expression prior to E12.5 (Ducy et al., 1997; Komori et al., 1997). All groups observed strong expression of Runx2 by E12.5 or E13.5 in the anlagen of the skull, axial, and appendicular skeletons. Additionally, Runx2 expression in the perichondrium and pre-hypertrophic chondrocytes was also noted at this time (Inada et al., 1999; Kim et al., 1999). By E14.5, Runx2 expression was evident in hypertrophic chondrocytes as well as osteoblasts. This expression was concomitant with the generation of bone marrow and vascular invasion into the cartilage. By E16.5, Runx2 was highly expressed in terminal hypertrophic chondrocytes. Although Runx2 mRNA expression was extensively studied by in situ hybridization, Runx2 protein expression patterns during skeletal development have not been carefully investigated. Chondrocyte hypertrophy and formation of the primary ossification center do not occur until E14.5 during mouse skeletal development. Because Runx2 mRNA expression precedes these events, it is likely that Runx2 protein levels or function may be suppressed in the early stages of development.
In summary, the expression of Runx2 mRNA begins in the perichondrium around the central regions of the anlagen, and, subsequently, appears in pre-hypertrophic and hypertrophic chondrocytes, with patchy expression in proliferating chondrocytes (Kim et al., 1999). In the perichondrium, Runx2 activates the expression of fgf-18, which in turn inhibits chondrocyte proliferation (Hinoi et al., 2006) (Fig. 1). These findings suggest that Runx2 inhibits chondrocyte proliferation while it promotes chondrocyte maturation. Consistent with this idea, we found that Runx2 stimulates expression of p27, a cyclin-dependent kinase inhibitor, in mesenchymal progenitor C3H10T1/2 cells (Shen et al., 2006b).
Proper formation of the mammalian dentition involves interactions between the epithelium of the developing oral cavity and the neural-crest-derived mesenchyme. Signaling between these two tissues, initiated in the epithelium, leads to distinct stages of tooth morphogenesis that can be observed histologically. First, the epithelium thickens to form a bud around which mesenchymal cells condense. This is referred to as the bud stage. The cap stage is characterized by an initial fold in the tip of the epithelial bud, establishing the cusp of the tooth and the dental papilla from which the odontoblasts and pulp arise. Additional cusps for the molars are formed through additional folds during the subsequent bell stage. Finally, mesenchymal-derived odontoblasts and epithelial-derived ameloblasts differentiate during the late bell stage to produce the dentin and enamel matrices, respectively (Thesleff and Nieminen, 1996; Miletich and Sharpe, 2004).
The importance of Runx2 in tooth development is highlighted by the dental anomalies observed in individuals with CCD. These persons have supernumerary teeth as well as delayed or ectopic eruption of permanent teeth. In the mouse, an absence of Runx2 expression leads to arrest of tooth morphogenesis at the late bud stage, suggesting that Runx2 is necessary for the bud-to-cap-stage transition (Aberg et al., 2004). As a consequence, these mice did not develop differentiated odontoblasts, ameloblasts, normal dentin, or enamel matrices. In vitro, Runx2 was necessary for the differentiation of odontogenic cells into both odontoblasts and ameloblasts (Kobayashi et al., 2006). While tooth development was globally affected in the Runx2−/− mice, the incisors were less affected than the upper molars, and the upper molars were less affected than the lower molars, suggesting that the importance of Runx2 function during development varies between these organs. Additionally, accessory epithelial buds developed in the Runx2−/− mice, suggesting that a normal function of Runx2 is to inhibit successional tooth formation (Aberg et al., 2004; Wang et al., 2005). This may help to explain the presence of supernumerary teeth in humans with CCD. Interestingly, unlike the compensatory role in chondrocyte hypertrophy, Runx3 appears not to affect tooth development in Runx2−/− mice. Runx2/Runx3 dKO and Runx2−/− mice have a similar molar development abnormality at the late bud stage, with the lower molars more severely affected than the upper molars (Wanget al., 2005).
Runx2 mRNA is normally expressed in the dental papilla mesenchyme at E12 of development (D’Souza et al., 1999). It is at this time that the inductive potential driving odontogenesis is transferred from the epithelium to the mesenchyme. Around E16, however, after completion of tooth morphogenesis, Runx2 expression in the mesenchyme is down-regulated. Expression remains high, however, in mature epithelial-derived ameloblasts (Jiang et al., 1999).
Within the cell, Runx2 is predominantly localized to distinct subnuclear foci associated with the nuclear matrix. Import into the nucleus is governed by a nuclear localization signal (NLS) that is adjacent to the carboxy-terminus of the runt-homologous DNA-binding domain (Thirunavukkarasu et al., 1998) (Fig. 2). This sequence is 9 amino acids in length and contains 5 basic residues, including the motif, RRHR, known to be responsible for the nuclear localization of numerous proteins. Within the nucleus, Runx2 is further localized to stationary foci that are associated with the nuclear matrix. This localization is dependent upon a nuclear matrix-targeting signal (NMTS) that consists of 38 amino acids residing in the carboxy-terminal portion of the protein (Zeng et al., 1997; Zaidi et al., 2001) (Fig. 2). Runx2 NMTS mutants no longer possess transactivation potential and no longer localize to stationary nuclear foci within the nuclear matrix (Zaidi et al., 2001, 2006; Harrington et al., 2002). Furthermore, mice lacking the NMTS and remaining carboxy-terminus do not develop bone, due to maturational arrest of osteoblasts, supporting the idea that this region is essential for Runx2 function in vivo (Choi et al., 2001).
During mitosis, Runx2 subnuclear localization is initially disrupted, but is then restored in telophase, allowing for equal partitioning of Runx2 to each daughter cell (Zaidi et al., 2003). More recently, Runx2 was shown to localize with components of RNA polymerase I to nucleolar organizing regions and to repress the expression of rRNA genes during both interphase and mitosis (Young et al., 2007; Ali et al., 2008). The equal distribution of Runx2 protein to mitotic progeny and, further, the stable association of Runx2 with mitotic chromatin at sites of rRNA synthesis may function to maintain the lineage specificity of Runx2-expressing cells. Supporting this hypothesis, transcription of rRNA genes is down-regulated during osteogenesis (Ali et al., 2008). Furthermore, by inhibiting ribosome biogenesis, Runx2 may also inhibit cell growth. Interestingly, primary dental pulp cells from humans with CCD appear larger in culture than their normal counterparts (Chen et al., 2005).
Post-translational modification, especially phosphorylation, is one important regulatory mechanism of Runx2 activity. Phosphorylation of Runx2 by different kinases and signaling pathways results in differential effects on Runx2 function and osteoblast differentiation. In human bone marrow stromal cells, osteoblast differentiation is associated with an increase in Runx2 activity through the phosphorylation of key residues, without altering Runx2 protein levels (Shui et al., 2003). Several groups have demonstrated that Runx2 is phosphorylated and activated by mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) inducers. For instance, Insulin-like growth factor-1 (IGF-1), which activates PI3K and, subsequently, MAPK/ERK pathways, increases Runx2 phosphorylation and binding to OSE2 (osteoblast-specific enhancer 2) in vascular endothelial cells, as well as during differentiation of bone marrow stromal cells (Qiao et al., 2004; Celil and Campbell, 2005; Celil et al., 2005). Mechanical loading, which is well-documented to increase osteoblast activity and bone formation both in vitro and in vivo, also increases Runx2 phosphorylation and expression of osteoblast marker genes. Specific inhibitors of the MAPK/ERK pathway block these effects (Wang et al., 2002; Ziros et al., 2002; Kanno et al., 2007). Likewise, the loading of periodontal ligament cells (osteoprogenitor-like cells) stimulates Runx2 phosphorylation and binding to OSE2 DNA via a MAPK/ERK-dependent process. Basic fibroblast growth factor-2 (FGF-2), a well-known stimulator of bone formation in vivo, stimulates osteocalcin mRNA expression and promoter activity by dramatically increasing the phosphorylation of ERK1/2 that precedes Runx2 phosphorylation. The addition of U0126, an ERK1/2 phosphorylation inhibitor, completely blocks FGF-2-induced Runx2 phosphorylation. A Runx2 deletion mutant lacking the entire C-terminal PST domain (proline/serine/threonine-rich domain) (Δ258-528) loses its response to FGF-2 stimulation, indicating that FGF-2 activates Runx2 by phosphorylating certain residues in the C-terminal PST domain through the MAPK pathway (Xiao et al., 2002). It has recently been shown that Runx2 phosphorylation and transcriptional activity are elevated in calvarial osteoblasts by transgenic over-expression of Mek-sp (a constitutively active form of MAPK/ERK1) driven by an osteoblast-selective osteocalcin promoter (TgMek-sp) (Ge et al., 2007). In vitro differentiation of calvarial osteoblasts and in vivo bone development are significantly enhanced in these mice. Interestingly, crossing the TgMek-sp transgenic mice with the Runx2+/− mice partially restored the phenotypes of hypoplastic clavicles and undermineralized calvaria that are associated with the Runx2 haploinsufficiency seen in persons with cleidocranial dysplasia (CCD). In contrast, Runx2 phosphorylation, osteoblast differentiation parameters, and bone development are all down-regulated by transgenic over-expression of a dominant-negative form of MAPK/ERK1 (Mek-dn), driven by the same promoter (TgMek-dn). As expected, TgMek-dn;Runx2+/− hybrid mice had a more severe skeletal phenotype. Collectively, these studies indicate that the MAPK/ERK pathway is involved in Runx2 phosphorylation and activation during osteoblast differentiation (Ge et al., 2007). One molecular consequence of MAPK-dependent Runx2 phosphorylation may be the association of Runx2 with Smads. It has been shown that the Runx2/Smad1 protein-protein interaction was dependent upon MAPK signaling, since the association of the two proteins was absent in lysates of cells treated with MAPK inhibitor PD98059 (Afzal et al., 2005). Although the stimulatory effect of MAPK/ERK on Runx2 phosphorylation and activity, as well as osteoblast activity, is well-established, the site(s) phosphorylated by this kinase within the Runx2 molecule remain to be identified.
In addition to activation by MAPK, FGF-2 also increases Runx2 activity through the PKCδ pathway (BG Kim et al., 2006). The authors showed that FGF-2 stimulated phosphorylation of Runx2 at S247 (S226 of human type I Runx2) (Fig. 2). Additionally, a serine-to-alanine mutation at this site resulted in decreased Runx2 transcriptional activation compared with that of wild-type Runx2 in response to FGF-2 treatment. Parathyroid hormone (PTH) induces collagenase-3 by activating Runx2 through a PKA-dependent pathway. During this process, S347 (T341 of human type I Runx2) (Fig. 2), a specific consensus site of Runx2 within the transactivation domain, is phosphorylated by PKA (Selvamurugan et al., 2000).
Runx2 transcriptional activity is also negatively regulated by phosphorylation of its serine residues, S104 and S451 (corresponding to S125 and S472 in the murine Runx2 MASN isoform) (Wee et al., 2002) (Fig. 2). Phosphorylation of S104 abolishes the interaction of Runx2 with its heterodimerization partner CBFβ, thereby decreasing Runx2 protein stability. Interestingly, glucocorticoids induce osteogenesis partially by attenuating the phosphorylation status of Runx2 at S104 via MKP-1 (MAPK phosphatase 1) (Fig. 2). Mutation of S104 to glutamic acid, mimicking constitutive phosphorylation, inhibits Runx2-mediated osteoblastic differentiation, an effect that could not be rescued by glucocorticoid supplementation (Phillips et al., 2006). Similarly, the GSK3β-dependent phosphorylation of Runx2 at S369-S373-S377 (S280-S284-S288 of human type I Runx2) (Fig. 2) was shown to inhibit Runx2 transactivation activity (Kugimiya et al., 2007). Mice heterozygous for GSK-3β, therefore, have increased bone formation. Our recent findings indicate that cell-cycle proteins cyclin D1/CDK4 mediate the phosphorylation of S472 of murine Runx2 (Fig. 2), leading to subsequent ubiquitination and proteasomal degradation of Runx2 (Shen et al., 2006b) (Fig. 3). Other reports demonstrate that cell-cycle proteins cdc2 and cyclin B/CDK1 can also phosphorylate Runx2 (Qiao et al., 2006; Rajgopal et al., 2007). Specifically, cyclin B/CDK1 was shown to regulate phosphorylation of Runx2 at S472 to modulate DNA binding during mitosis. These findings suggest that Runx2 activity is regulated in multiple phases of the cell cycle by, presumably, different signaling pathways.
In addition to gene expression, Runx2 protein levels are regulated through post-translational mechanisms involving the ubiquitin-proteasome pathway. Evidence of this was first reported when treatment of osteoblastic MC3T3-E1 cells with the camp elevating agent, forskolin, resulted in down-regulation of Runx2 protein levels via ubiquitin-dependent proteasomal degradation (Tintut et al., 1999). Since this time, several mechanisms leading to the degradation of Runx2 in an ubiquitin-dependent manner have been discovered.
One such mechanism involves Hect domain E3 ligase Smad ubiquitin regulatory factor (Smurf) family members. Smurf1 was originally identified as an E3 ligase that interacts with both Smad1 and Smad5 to induce their degradation in an ubiquitin-dependent manner (Zhu et al., 1999). In addition to the Hect domain, which catalyzes the ubiquitination reaction, members of the Hect family of E3 ligases also possess WW domains. These domains bind to short proline-rich sequences, known as PY motifs (PPXY), in target proteins, to facilitate protein-protein interactions. Different WW domains, therefore, confer different substrate specificities for these E3 ligases. Sequence analysis of Runx family members shows that a conserved PY motif (PPPY) is located in the carboxy-terminal region of Runx family proteins. Our studies demonstrated that Smurf1 interacts with Runx2 and induces Runx2 ubiquitination and proteasomal degradation in myoblast/osteoblast precursor C2C12 cells (Zhao et al., 2003). Smurf1 appears to act as an important regulator of BMP signaling, Runx2 activity, and osteoblast differentiation. Over-expression of Smurf1 in 2.3Col1-Smurf1 transgenic mice inhibits post-natal bone formation (Zhao et al., 2004). Additionally, osteoblast proliferation and differentiation were significantly reduced in these mice. Consistent with this, adult Smurf1−/− mice showed increased bone mass and increased osteoblast proliferation and differentiation (Yamashita et al., 2005). Recently, we found that tumor necrosis factor (TNF) causes degradation of Runx2 and inhibition of osteoblast differentiation through up-regulation of Smurf1 expression levels in osteoblast precursor cells (Kaneki et al., 2006). Crossing Smurf1−/− mice with TNF transgenic mice that have joint arthritis and systemic osteoporosis partially rescues the bone loss phenotype observed in TNF transgenic mice, suggesting that Smurf1 may also play an important role in TNF-induced pathological bone loss (Guo et al., 2008). Moreover, continuous parathyroid hormone (PTH) treatment was reported to induce Runx2 proteasomal degradation via Smurf1, offering an explanation as to why intermittent administration of PTH is needed for bone formation in osteoporotic individuals (Bellido et al., 2003).
As a Hect domain E3 ligase, Smurf1 normally interacts with the PY motif of substrate proteins through its WW domain (Zhu et al., 1999). Interestingly, Smurf1 partially maintains its ability to induce Runx2 degradation even when the PY motif of Runx2 is deleted. Our more recent findings indicate that the BMP signaling inhibitor Smad6 binds to Runx2 and serves as an adaptor to mediate Smurf1- and WWP1-induced Runx2 degradation in a PY motif-independent manner (Shen et al., 2006a). Casein kinase-2 interacting protein-1 (CKIP-1) was also identified as an adaptor for Smurf1 (Lu et al., 2008). CKIP-1 binds the linker region between the Smurf1 WW domains, enhancing the binding and ubiquitination of target substrates by Smurf1 (Lu et al., 2008). CKIP-1-deficient adult mice have increased bone mass, enhanced osteoblast activity, and decreased Smurf1 activity. Whether Runx2 protein levels were affected in these mice was not reported.
In addition to Smad6, Hect domain E3 ligase WWP1 also utilizes Schnurri-3 (Shn3) as an adaptor to induce Runx2 degradation. Shn3 interacts with WWP1 and induces Runx2 ubiquitination and proteasomal degradation. In Shn3−/− mice, Runx2 protein levels are increased, while Runx2 mRNA expression is not altered. Osteoblast function and bone mass are significantly increased in these mice. In contrast, osteoclast formation and function are not altered (Jones et al., 2006). These results suggest that, under physiological conditions, Runx2 protein levels are suppressed in osteoblasts.
Finally, a third E3 ligase, CHIP (C terminus of Hsc70-interacting protein), was shown to regulate the ubiquitination and degradation of Runx2. CHIP is a U-box E3 ubiquitin ligase (Jiang et al., 2001). Over-expression of CHIP in MC3T3-E1 cells leads to Runx2 degradation and inhibition of osteoblast differentiation. In contrast, depletion of CHIP in these cells enhanced osteoblast differentiation. Furthermore, endogenous CHIP protein levels gradually decline while Runx2 protein levels increase during MC3T3-E1 osteoblast differentiation. Runx2 mRNA levels, however, do not change (Li et al., 2008). This trend was not observed for Smurf1, Schnurri-3, or WWP1. The authors suggest that perhaps CHIP is involved in keeping protein levels of Runx2 low during development prior to cellular differentiation, while Smurf1 and WWP1 may be involved in the maintenance of Runx2 protein levels after differentiation. Thus, the higher levels of CHIP prior to differentiation may serve to keep Runx2 protein levels low until cells receive the necessary cues required for the initiation of differentiation.
We recently showed that cyclin D1 induces Runx2 ubiquitination and degradation in a phosphorylation-dependent manner that leads to the inhibition of Runx2 transcriptional activity. In C3H10T1/2 cells, Runx2 stimulates osteoblast differentiation and up-regulation of p27 mRNA levels. These effects are significantly inhibited by cyclin D1-induced Runx2 phosphorylation (Shen et al., 2006b). Consistent with these findings, recent reports demonstrated that other cell-cycle proteins, such as cdc2 or cyclin B/CDK1, also phosphorylate Runx2 and reduce Runx2 activity (Qiao et al., 2006; Rajgopal et al., 2007). More recently, we further demonstrated that cyclin D1 induces Runx3 phosphorylation, ubiquitination, and proteasome degradation in a manner similar to that for Runx2 (Zhang et al., 2009) (Fig. 3). These findings suggest that Runx2 and Runx3 activities are regulated in coordination with the cell-cycle machinery. Indeed, Runx2 levels were found to oscillate during the cell cycle (Galindo et al., 2005).
PTHrP is a critical factor regulating chondrocyte function in the growth plate. It maintains chondrocytes in a proliferative state and prevents premature chondrocyte hypertrophy. Targeted disruption of the pthrp or pthrp receptor (pthrpr) genes leads to early cessation of chondrocyte proliferation and subsequent acceleration of chondrocyte differentiation (Karaplis et al., 1994; Lanske et al., 1996; Kobayashi et al., 2002). One way in which PTHrP prevents chondrocyte maturation is through inhibition of Runx2 expression in proliferating and pre-hypertrophic chondrocytes (Guo et al., 2006). In addition to its role in Runx2 mRNA expression, we have found that PTHrP down-regulates Runx2 and Runx3 protein levels in a cyclin D1-dependent manner (Zhang et al., 2008) (Fig. 3). The inhibitory effect of PTHrP on Runx2 and Runx3 protein levels may, therefore, be mediated by cyclin D1-induced phosphorylation and ubiquitination of Runx2 and Runx3 (Fig. 3). The specific E3 ligase that cooperates with cyclin D1 in ubiquitination and proteasome-dependent degradation of Runx2 and Runx3 remains to be identified.
Acetylation of lysine residues within histone protein tails reduces their affinity for DNA, thereby relaxing the chromatin structure and increasing the accessibility of DNA to transcription factors. This is accomplished through the activity of histone acetyl-transferases, or HATs. Histone deacetylases (HDACs) are capable of reversing this process by removing the acetyl moiety and, therefore, allowing histones to bind tightly to DNA again. Both HATs and HDACs are reported to interact with Runx2 and modulate Runx2 stability and transcriptional activity. It was even reported that Runx2 itself is regulated through acetylation. For example, it was found that BMP-2 could induce Runx2 acetylation, and that Smad1 and Smad5 could facilitate an interaction between Runx2 and p300, a protein possessing HAT activity (Jeon et al., 2006). Acetylation inhibited Smurf1-mediated degradation of Runx2 and also stimulated Runx2 transcriptional activity. Furthermore, lysine-to-arginine substitutions of Runx2 residues 225/230/350/351 (corresponding to residues 219/224/344/345 of human type I Runx2) abolished Runx2 acetylation and Smurf1-dependent degradation (Fig. 2). Interestingly, the TGF-β signaling pathway was shown to inhibit Smurf-dependent Runx3 ubiquitination and degradation, also through acetylation by p300 (Jin et al., 2004). Acetylation may protect Runx proteins from degradation by masking the lysine residue to which the ubiquitin moiety is normally bound.
Independently, it has been shown that the activation domain of Runx2 physically interacts with p300 in ROS 17/2.8 osteoblastic cells (Sierra et al., 2003). Functionally, p300 potentiates Runx2-dependent activation of the osteocalcin gene promoter. The authors found that this effect, however, is independent of intrinsic p300 HAT activity, since a HAT-deficient p300 mutant enhances Runx2-dependent osteocalcin promoter activity to the same extent as wild-type p300. It is possible; therefore, that p300 may facilitate acetylation of Runx2 through another protein with HAT capabilities. For example, Runx2 is also reported to associate with P/CAF (p300/CBP-associated factor) (Sierra et al., 2003; Jeon et al., 2006). Furthermore, P/CAF was able to cooperate with p300 in activation of the osteocalcin gene promoter (Sierra et al., 2003). MOZ (monocytic leukemia zinc finger protein) and MORF (MOZ-related factor) are two other HAT proteins found to interact physically with Runx2 and enhance Runx2 transcriptional activity, though they have not been shown to induce Runx2 acetylation (Pelletier et al., 2002).
Histone deacetylase inhibitors were shown to increase Runx2 transcriptional activity as well as positively regulate osteoblast differentiation and new bone formation (Schroeder and Westendorf, 2005; Jeon et al., 2006). Furthermore, several HDACs are independently reported to interact with and modulate the function of Runx2. For example, HDAC3 interacts with the amino terminus of Runx2 and was shown to repress Runx2 transactivation of the Osteocalcin promoter (Schroeder et al., 2004). Suppression of HDAC3 through RNA interference in MC3T3 cells resulted in the increased expression of Runx2 target genes. In a separate study, HDAC3 was identified as a co-repressor of Runx2 in regulation of bone-sialoprotein gene expression (Lamour et al., 2007). Both HDAC4 and HDAC5 were shown to be required in TGF-β/Smad3-mediated inhibition of Runx2 activity (Kang et al., 2005). Runx2, Smad3, and HDAC4 or HDAC5 form a ternary complex at Runx2 DNA-binding elements that represses Runx2 activity. HDAC6 interacts with the carboxy-terminus of Runx2 and acts as a co-repressor of Runx2 for p21 gene expression in pre-osteoblasts (Westendorf et al., 2002). Finally, HDAC7 also interacts with the carboxy-terminus of Runx2 and suppresses Runx2 activity through a deacetylase-independent mechanism (Jensen et al., 2008). Furthermore, suppression of HDAC7 by RNA interference accelerated osteoblast differentiation of C2C12 cells in the presence of BMP-2.
In vivo experiments reveal that HDAC4 plays a critical role in the regulation of chondrocyte hypertrophy (Vega et al., 2004). Mice nullizygous for HDAC4 exhibit an early onset of chondrocyte hypertrophy, leading to premature ossification of developing bones. This phenotype is identical to that of mice over-expressing Runx2 in proliferating chondrocytes (Takeda et al., 2001; Ueta et al., 2001). Over-expression of HDAC4 in proliferating chondrocytes, however, results in a phenotype similar to Runx2 loss-of-function, where chondrocyte hypertrophy and endochondral bone formation are inhibited. This suggests that HDAC4 normally inhibits Runx2 function in pre-hypertrophic or proliferating chondrocytes. In vitro experiments show that HDAC4 binds the Runt-homologous DNA-binding domain of Runx2 and inhibits Runx2 transactivation activity and DNA-binding capabilities (Vega et al., 2004). Based on the mechanistic studies of Jeon et al. summarized above, it is also possible that, in vivo, HDAC4 may promote Runx2 degradation through de-acetylation.
Runx2 is an essential transcription factor in the development and maintenance of mammalian bone and teeth. In humans, two functional copies of Runx2 are required for proper skeletal and dental development, since haploinsufficiency of Runx2 results in the autosomal-dominant disorder cleidocranial dysplasia. The phenotypes of this disease vary in severity, depending upon the nature of the Runx2 mutation. There have even been cases reported where no Runx2 mutation can be found, suggesting that mutations in Runx2 regulatory elements or proteins may be to blame. Further research aimed at identifying these factors will no doubt aid in our understanding of Runx2 regulation, but may also add to our list of the molecules important to bone and tooth development.
Through mouse models and cell-culture studies, it is now well-established that Runx2 is necessary for osteoblast differentiation and chondrocyte hypertrophy, two processes essential for intramembranous and endochondral ossification. Additionally, Runx2 is essential for proper tooth morphogenesis. Runx2 gene expression is observed in early mesenchymal condensations during mouse embryonic development. This, however, precedes the expression of the Runx2 target genes that are involved in chondrocyte hypertrophy and osteoblast differentiation, suggesting that Runx2 protein levels and activity are suppressed prior to the initiation of these developmental events.
At the molecular level, Runx2 serves as a scaffold on which multiple proteins assemble to form regulatory complexes that either activate or repress transcription. Regulation of Runx2 occurs on multiple levels through multiple signaling pathways initiated by various extracellular stimuli. Mounting evidence shows that Runx2 protein levels and activity are regulated independently of mRNA expression and processing through post-translational modifications, including phosphorylation, ubiquitination, and acetylation. These modifications likely affect the protein-protein interactions between Runx2 and other members of Runx2 transcriptional regulatory complexes, ultimately determining whether these complexes will activate or repress transcription. For example, phosphorylation or acetylation of Runx2 may provide a ‘docking site’ for a transcriptional co-activator or co-repressor unable to bind in the absence of modification. Alternatively, phosphorylation of a specific residue may provide a recognition sequence for an ubiquitin E3 ligase that can target Runx2 for proteasomal degradation. Acetylation, however, may block the availability of a lysine residue to an ubiquitin moiety, resulting in increased protein stability. These are just a few examples of how post-translational modifications may affect Runx2 activity and stability, and much work is yet to be done in discerning exactly how the modifications described in this review ultimately affect Runx2 protein interactions and activity.
Of the known Runx2 interacting proteins, only a few were discussed in this review, and more are certain to be discovered in the future. In summary, many pieces of information have been uncovered with regard to the regulation of Runx2 expression and activity. These include the identification of extracellular signals and corresponding signaling pathways that control Runx2 mRNA and protein expression, as well as the identification of proteins that interact with Runx2 in the activation or repression of specific genes involved in bone formation. Future research will no doubt aid in better understanding of how these pieces fit together in the coordinated regulation of this protein that is so important in human bone and tooth development.
This work was supported by Grants R01-AR051189, K02-AR052411, and R01-AR054465 from NIAMS (to D.C.); R01-DK072230 from NIDDK and a Department of Defense Grant W81XWH-07-1-0160 (to G.X.); and R21-AR53586 from NIAMS (to L.X.). JHJ is sponsored by NIAMS T32-AR053459 training grant.