Transcriptional Regulation of Osteoblasts

Endochondral bone formation is controlled by a hierarchy of transcription factors that are expressed in a defined temporal sequence (fig. 1). RUNX2 is expressed very early in skeletal development, first appearing with the formation of mesenchymal condensations in areas destined to become bone and persisting through subsequent stages of bone formation [Ducy et al., 1997]. Several other transcription factors function together with RUNX2 to move cells down chondrocyte or osteoblast lineages. For osteoblasts, this is accomplished by Osterix (OSX), which commits osteochondroprogenitor cells to the osteoblast lineage [Nakashima et al., 2002]. Subsequently, another factor, ATF4, controls the transcriptional activity of mature osteoblasts [Yang et al., 2004]. RUNX2 also participates in the chondrogenic lineage. However, at early stages, it is likely suppressed by the chondrocyte-specific factors Sox 8/9 [Zhou et al., 2006]. Evidence for this model largely comes from genetic studies showing bone phenotypes of decreasing severity as Runx2, Osx or Atf4 are knocked out. Thus, skeletal development in Runx2-deficient embryos fails to progress beyond the cartilage anlage stage [Komori et al., 1997; Otto et al., 1997]. These embryos exhibit no sign of cartilage hypertrophy, bony collar development, vascular invasion or marrow formation. In contrast, Osx^–/– embryos have hypertrophic cartilage, early collar formation and normal levels of RUNX2, suggesting that this factor is downstream of RUNX2 [Nakashima et al., 2002]. ATF4^–/– mice form bone, but at reduced rates, consistent with this factor playing a regulatory role in bone formation [Yang et al., 2004].

RUNX2 mRNA is present throughout bone development beginning at embryonic day 9.5 (E9.5) in the mouse and persisting in active osteoblasts and osteocytes throughout adult life [Ducy et al., 1997]. There is good evidence that Runx2 regulates separate events at different times during skeletal development. As noted above, it has early roles in cartilage hypertrophy, bony collar formation and vascularization, but is also required for terminal differentiation of osteoblasts and osteocytes via induction of genes involved in extracellular matrix (ECM) formation and mineralization. These findings suggest that other factors and signals must be able to modify Runx2 activity according to stage of development without necessarily altering levels of Runx2 protein. Figure 2 illustrates how RUNX2 activity can be regulated without altering Runx2 protein levels. In this study, differentiation of MC3T3-E1 preosteoblast was induced by growth in ascorbic acid-containing medium, thereby allowing cells to secrete a collagenous ECM. As ECM accumulates, there is a dramatic induction of osteoblast marker genes such as bone sialoprotein (Bsp) and Ocn, but, surprisingly, RUNX2 mRNA and protein levels remain relatively constant.

There are at least 3 possible ways RUNX2 transcriptional activity could be regulated in the absence of changes in RUNX2 protein levels. First, covalent modification could alter RUNX2 transcriptional activity. Second, levels of cofactors or ability of cofactors to interact with RUNX2 could be regulated. Lastly, Runx2 protein levels could be controlled by proteosome-mediated degradation. All 3 types of regulation have been previously described, although we will focus on the first 2.

The first example involves a pathway incorporating integrin-mediated activation of the ERK/MAP kinase pathway that results in phosphorylation and stimulation of RUNX2 transcriptional activity. Integrins provide a direct link between cells and ECM to convey information about mechanical loads and ECM stiffness experienced by bone to control differentiation [You et al., 2001; Engler et al., 2006]. Integrins are also important signal transduction molecules in their own right that activate Ras-ERK and p38 MAP kinase, calcium channels and mechanosensors [Hynes, 2002].

Osteoblast differentiation requires synthesis of a collagen-containing ECM. As is shown in figure 2, bone sialoprotein and osteocalcin are only expressed in cells grown under conditions where they can synthesize a collagenous matrix [Franceschi, 1992]. Using a combination of chemical inhibitors and genetic approaches, we previously established that the response of osteoblasts to ECM is mediated by α2β1 integrins and the ERK/MAP kinase pathway [Xiao et al., 1997, 1998, 2000].

The role of RUNX2 in MAPK responsiveness was established in studies with the Ocn gene. MAPK activation via transfection of cells with constitutively active MEK1 (MEK-SP) could induce Ocn expression, while dominant-negative MEK1 (MEK-DN) was inhibitory [Xiao et al., 2000]. Mutational analysis of the Ocn promoter for MAPK-responsive sequence elements established the involvement of 2 RUNX2-binding sites called OSE2a and b. Mutation of either site reduced MAPK responsiveness [Frendo et al., 1998].

As might be expected, activation of MAPK, either by ECM synthesis or transfection with MEK-SP, results in increased RUNX2 phosphorylation [Xiao et al., 2000]. Examination of different truncations of the RUNX2 protein showed that the C-terminal proline/serine/threonine region of RUNX2 was required for both MAPK responsiveness and phosphorylation [Xiao et al., 2000]. More detailed deletional analysis further localized a minimal region for MAPK responsiveness between amino acids 254 and 320 (data not shown). The specific identification and functional significance of ERK/MAPK phosphorylation sites in RUNX2 will be reported elsewhere. A number of other stimuli that act through the ERK/MAPK pathway, including FGFR2 activation by FGF2 and mechanical loading of osteoblasts, also stimulate RUNX2 phosphorylation and transcriptional activity [Wang et al., 2002; Xiao et al., 2002; Ziros et al., 2002; Kim et al., 2003; Kanno et al., 2007].

We took a transgenic approach to examine the importance of the ERK/MAPK pathway and RUNX2 phosphorylation in osteoblast function in vivo [Ge et al., 2007]. Mice were generated using a 0.6-kb mOG2 promoter to specifically drive MEK-SP and MEK-DN expression in osteoblasts. Transgene expression was detected only in bones and had a developmental time course of expression that parallels expression of the endogenous Ocn gene. ERK1/2 phosphorylation in calvarial osteoblasts derived from these animals was shown to increase by 50% in cells from Mek-sp mice and to decrease by 50% in Mek-dn cells [Ge et al., 2007]. Thus, transgene expression leads to subtle changes in MAPK activity that resemble fluctuations normally induced by physiological stimuli. Skeletal whole mounts revealed that Mek-dn decreased skeletal size and calvarial mineralization, while these parameters were increased in Mek-sp mice (fig. 3a–f). Histology of long bones revealed an additional interesting difference between wild-type and transgenic mice (fig. 3g). At E15.5, long bones are normally undergoing endochondral ossification in diaphyseal regions. However, in Mek-dn mice, this process is drastically delayed with only early bony collar formation being visible. In contrast, in Mek-sp mice, endochondral bone formation is accelerated.

Transgenic modification of osteoblast MAPK activity was also found to alter RUNX2 phosphorylation and transcriptional activity. RUNX2 phosphorylation was increased in Mek-sp cells as was luciferase activity of an Ocn reporter gene. Also, as expected, Mek-sp stimulated in vitro osteoblast differentiation as measured by induction of osteoblast marker mRNAs or mineralization, while differentiation was inhibited in Mek-dn cells [Ge et al., 2007].

To provide evidence that MAPK effects on skeletal development are mediated by RUNX2, we used a genetic approach. Runx2^+/– mice are known to have a characteristic phenotype (hypoplastic clavicles, patent fontanelles) that resembles the human genetic disease cleidocranial dysplasia [Otto et al., 1997]. We reasoned that if MAPK acts by altering RUNX2 activity, calvaria and clavicles should be selectively sensitive to the Mek transgene when RUNX2 is limiting (that is, in Runx2^+/– mice). To test this, Runx2^+/– mice were crossed with Mek-sp or Mek-dn transgenic lines and the resulting skeletal phenotypes were examined at E19 (fig. 4). Note the reduced calvarial mineralization and tiny clavicles in the Runx2^+/– embryos. The presence of the Mek-sp transgene led to a clear rescue of the cleidocranial dysplasia phenotype with increased clavicle size and calvarial mineralization. In the presence of Mek-dn, effects on Runx2^+/– mice were even more dramatic. In this case, the Mek-dn transgene exacerbated effects of Runx2 haploinsufficiency with a further reduction in calvarial mineralization and near disappearance of clavicles. Notably, Runx2^+/– Mek-sp embryos did not survive the birth process due to severe skeletal defects. These experiments provide strong evidence that the ERK/MAPK pathway, via actions on RUNX2 transcriptional activity, is important for normal osteoblast differentiation and bone formation in vivo.

RUNX2 directly or indirectly interacts with a large number of nuclear factors. For example, CBF-β forms heterodimers with all members of the Runx family [Ito, 2004], while ATF4-RUNX2 interactions may mediate some of the responses of osteoblasts to parathyroid hormone [Jiang et al., 2004; Xiao et al., 2005]. SMAD proteins, mediators of BMP/TGF-β actions, can also stimulate RUNX2 activity [Afzal et al., 2005]. In addition, there are a number of inhibitory factors. SOX9, mentioned above, suppresses RUNX2-dependent chondrocyte hypertrophy [Zhou et al., 2006]. TWIST may prevent RUNX2 stimulation of mineralization in developing cranial bones to prevent craniosynostosis [Bialek et al., 2004]. Lastly, histone deacetylases are known to bind RUNX2 and keep chromatin in a deacetylated, inactive state [Imai et al., 2004].

Studies on Bsp revealed an important regulatory function of interactions between RUNX2 and DLX5 [Roca et al., 2005]. We previously characterized a 2.4-kb promoter fragment of this gene and showed that it contains sufficient information to drive osteoblast-selective expression in vivo [Benson et al., 2000; Gopalakrishnan et al., 2003]. Detailed characterization of the proximal Bsp promoter identified 2 RUNX2-binding sites (R1, R2) and an adjacent homeodomain protein-binding site (C site) that binds DLX5 (fig. 5a). Interestingly, levels of RUNX2 associated with chromatin do not appreciably change during osteoblast differentiation. In contrast, DLX5 only becomes chromatin associated in differentiated cells (fig. 5b). However, total DLX5 protein was equivalent in control and differentiated cells indicating that the affinity of Dlx5 for chromatin (and possibly Runx2) increases with differentiation. Subsequent analysis revealed that RUNX2 must be associated with both sites for full activity of the Bsp promoter. Furthermore, the R2 site and the C site functionally cooperate to regulate promoter activity (fig. 5c). Consistent with this observation, RUNX2 and DLX5 physically interact as measured by pull-down assays (fig. 5d).

As one of the earliest determinants of the bone cell lineage, RUNX2 first appears soon after the formation of mesenchymal condensations. Although present throughout subsequent stages of osteoblast differentiation, RUNX2 activity is highly regulated. As was discussed, this regulation can be the result of RUNX2 phosphorylation or interaction with partner proteins such as DLX5, in both cases leading to activation of RUNX2 transcriptional activity. A number of outstanding questions remain regarding RUNX2 function. First, it is not known if RUNX2 becomes associated with all target genes it will eventually regulate soon after it is first expressed or if it associates with different genes at different stages of differentiation. If the former is true, then RUNX2 may be largely reliant on the posttranslation controls and accessory factor binding described here to become active at the appropriate time. If the latter is the case, RUNX2 synthesis and degradation may be more dynamic with transcriptional activity being dependent on RUNX2 protein levels as well as levels of accessory factors. A second area relates to the degree to which posttranslational modification of RUNX2 can control its activity. As noted above, there are now several examples of RUNX2 being positively or negatively regulated by phosphorylation. However, other types of modification such as acetylation have also been reported [Jeon et al., 2006]. It will be important to understand how these different modifications are coordinated to control RUNX2 activity. Lastly, the full breadth of factors associating with RUNX2 is presently unknown, but current evidence suggests that RUNX2 likely exists in large macromolecular complexes that have a punctate distribution in the nucleus [Zaidi et al., 2002]. A major challenge will be to isolate and characterize these complexes and determine how their composition is regulated during skeletal development.