Making no bones about it: Transcription factors in vertebrate skeletogenesis and disease

Abstract

Skeletogenesis is a complex multi-step process, which involves many genes and pathways. The tightly regulated interplay between these genes in these pathways ensures a correct and timely organogenesis and it is imperative that we have a fair understanding of the major genes and gene families involved in the process. This review aims to give a deeper insight into the roles of 3 major transcription factor families involved in skeleton formation: Sox, Runx and Pax and to look at the human skeleotogenic phenotypes associated with mutations in these genes.

INTRODUCTION

The vertebrate skeleton is a highly complex organ, which helps in maintaining balance and giving mechanical support and protection to vital internal organs. The incidence of skeletogenic disorders in humans is about in 1 in 4000 with many being lethal at an early age [1]. To have a better understanding of the process it is imperative that we understand the genes and the gene families involved and the molecular processes behind them. This will lead to a better understanding and management of the various diseases associated with the skeletal system.

Chondrocytes are the first skeleton-specific cell type to appear during development and defects in chondrogenesis lead to chondrodysplasias and osteoarthritis [2, 3]. Early chondrocyte differentiation and subsequent maturation are controlled by Sox9 and its family members, Sox5 and Sox6 [4, 5] and these continue to play an important role all the way through to late skeletogenesis [6].

Runx2 is important for the initial commitment of perichondrial cells to the osteoblast lineage [7, 8], whereas Runx3 has no independent role in skeletogenesis, double knockout mice of Runx2 and Runx3 show complete lack of hypertrophic chondrocytes [9, 10].

Pax1 and Pax9 are two other important transcription factors (TFs) involved in skeletogenesis. In Pax1 null mice the entire axial skeleton is defective with reduced or loss of ossification centers [11], whereas the Pax9 null mice display no vertebral column defects but show preaxial polydactyly, cleft secondary palate and lack the derivatives of pharyngeal pouches [12].

This review will attempt to look in a greater detail at these genes and how they affect skeletogeneis and known human diseases associated with them.

Sox5, Sox6 and Sox9 during skeletogenesis

Sox stands for Sry-related high mobility group (HMG) box as Sry was the first member in the family to be discovered. Sox9 belongs to the SoxE subgroup whereas Sox5 and Sox6 are part of the SoxD subgroup. These proteins have the most important role to play in the initiation and progression of chondrogeneis and skeletogeneis. The first skeletal specific cell type to appear during chondrogenesis is the chondrocyte. Precartilaginous condensation marks the first step of chondrogenesis. At this stage, these skeletal precursor cells stop expressing type I collagen and hyaluronan. Instead, they start expressing adhesion proteins like N-cadherin and tenascin-C which allow the cells to aggregate tightly [13]. The transcription factor Sox9 has been shown to be required for these precartilaginous cell condensations and their survival but the mechanism behind it remains elusive [14, 15]. These skeletal precursor cells which are bipotential at this stage have the ability to become chondrocytes or osteoblasts. They express both Sox9 as well as Runx2 which is a master regulator in osteoblastogenesis. The commitment to chondrogenesis is largely determined by the key chondrogenic transcription factor, Sox9, whose expression is absolutely necessary [16]. It inhibits Runx2 expression through another transcription factor, Nkx3.2 (also known as Bapx1), and interacts directly with Runx2 to repress its activity pivoting it towards the chondrogenic fate [17, 18]. Other transcription factors have been implicated in establishing chondrocyte commitment but the in vivo relevance has only been shown for Pax1, Pax9, Nkx3.1 and Nkx3.2 [17, 19–21].

During the next stage of chondrogenesis, the prechondrocyte cells in the centre of the precartilaginous condensation undergo differentiation to form the early chondrocytes. Studies have shown that Sox9 has the ability to directly activate a 48-bp enhancer sequence in the intron of Col2a1 which is highly expressed during this phase and this activity was potentiated by two other proteins, Sox5 and Sox6 [22, 23]. Sox9 was found to bind as a homodimer to a pair of consensus sequence in cartilage genes like Col2a1 and Matn1 [24]. Early chondrocyte differentiation and subsequent maturation are governed by Sox9, Sox5 and Sox6, also known as the Sox chondrogenic trio which activate cartilage-specific genes [4]. Overexpression of the Sox trio in cell culture was shown to be sufficient to induce chondrocyte differentiation of mesenchymal cells and nonchondrocyte cell lines, establishing the importance of the Sox trio in directing chondrogenesis [25].

Early chondrocytes further enlarge to form prehypertrophic chondrocytes. These cells eventually stop proliferating and become hypertrophic chondrocytes. This process requires the down-regulation of the Sox trio which negatively regulates hypertrophy to prevent the premature differentiation of prehypertrophic chondrocytes [26]. At the same time, the transcription factors Runx2 and Runx3 through the interactions with other factors positively regulate hypertrophy [27]. Other factors like Msx2, Mef2c, Mef2d and Fra2 have also been associated with the positive regulation of hypertrophy though their mode of action is not well understood [28]. The Wnt/β-catenin pathway plays an important role in supporting osteoblastogenesis. It is hypothesized to down-regulate Sox9 expression and upregulate Runx2 expression, hence favoring the osteoblast differentiation [29–31]. The transactivation domain of Sox9 was also shown to physically interact with β-catenin to enhance chondrocyte differentiation and with parts of the transcriptional machinery such as transcriptional co-activators CBP/p300 [24].

The control of the expression of the Sox trio themselves in chondrocytes remains unclear. Through in vitro studies, Sox9 was proposed to self-regulate its expression via a positive feedback loop [32]. miR-145 was reported to negatively regulate chondrocyte differentiation by targeting Sox9 in mesenchymal stem cells (MSCs) [33]. A recent study performed using rat chondrosarcoma cells found Sox9 binding sites located within its introns which may support the proposition [34]. From the same study, Sox9 binding sites were found in the Sox5 promoter and its introns suggesting that Sox9 may regulate Sox5 expression directly. In another recent study, miR-194 was found to regulate chondrogenesis by targeting Sox5 in human adipose-derived stem cells [35].

Relevance to human diseases and phenotype of knockout mice

Heterozygous mutations in and around Sox9 were shown to cause a semi-lethal syndrome known as campomelic dysplasia, characterized by a severe form of human chondrodysplasia often accompanied by male sex reversal and defects in other non-skeletal organs, thus identifying Sox9 as a critical player in chondrogenesis [35, 36]. Heterozygous Sox9 mutant mice showed the same skeletal malformations as humans with campomelic dysplasia and die soon after birth [15]. A delay in chondrogenic mesenchymal condensation and premature mineralization was observed indicating that Sox9 is needed for initiating condensation and the inhibition of hypertrophy in proliferating chondrocytes. This was supported by another observation that Sox9-null cells were excluded from wildtype cells during mesenchymal condensation and that these mutant cells do not express chondrogenic genes like Col2a1, suggesting that Sox9 is required for a chondrogenic cell fate [16]. When Sox9 was inactivated prior to the onset of mesenchymal condensation, mesenchymal condensations were completely absent, and Runx2 expression was not detected, indicating that Sox9 expression is required for the formation of osteochondroprogenitor cells [14]. Conditional Sox9 inactivation after mesenchymal condensation using Col2a1-Cre was observed to cause condensed mesenchymal cells to stop differentiation and impaired those differentiated chondrocytes in their proliferation and maturation process.

The ectopic expression of Sox9 in limb buds of transgenic mice showed ectopic formation of cartilage with the induction of Sox5 and Sox6 expression, while lack of Sox9 abolishes Sox5 and Sox6 expression in chondrocytes, indicating that Sox9 is required for the downstream expression of Sox5 and Sox6 [14, 36]. Sox5 and Sox6 are coexpressed with Sox9 from the prechondrocyte stage onwards during chondrogenesis and are required for the overt chondrocyte differentiation [23, 37]. Sox5-null mice die at birth from respiratory distress and were observed to have a smaller ribcage and a cleft secondary palate whereas Sox6-null mice die at birth or soon after with a short sternum as compared to the wildtype [38]. In general, single gene knockouts for Sox5 and Sox6 demonstrated mild skeletal defects. However, when both genes are inactivated, the mice die three days before birth with severe defects in cartilage formation, demonstrating the functional redundancy between Sox5 and Sox6 in chondrogenesis. This severe chondrodysplasia phenotype observed is comparable to the phenotype of the conditional knockout of Sox9 after mesenchymal condensation in the mice.

Runx2 and Runx3 in skeletogenesis

The Runx family of genes encode for transcription factors that contain the characteristic DNA-binding runt domain which derived its name from the Drosophila pair-rule gene, runt, owing to the high degree of homology between the two sequences [39]. This highly conserved 128-amino-acid runt motif found proximal to the N-terminus has functions in (1) DNA binding, recognizing a canonical DNA motif TGPyGGTPy (where Py refers to pyrimidine) [40], (2) proteinprotein interactions [41] and (3) nuclear import that is in addition to the conserved nuclear matrix-targeting signal (NMTS) in the C-terminus [42].

Runx2 is a crucial factor for the initial commitment of perichondrial cells and condensed mesenchymal anlagen of the intramembranous bones to osteoblast lineage cells [7, 8]. In endochondral ossification, the link between chondrocyte maturation and osteoblast differentiation hinges on Ihh signalling. While Runx2 regulates Ihh in the prehypertrophic chondrocytes, Ihh induces Runx2 expression in the adjacent perichondrium [43]. Runx2 expression, however, is not sufficient for osteoblast differentiation as reflected by the ectopic maturation of chondrocytes without any defects in osteoblast differentiation in transgenic mice constitutively expressing Runx2 [44]. Further commitment of the Runx2-expressing osteoblast progenitor cells to fully committed osteoblasts in both endochondral and intramembranous bones requires a Krüppel-like zinc finger domain-containing transcription factor Sp7 (Osterix) [45]. The activity of Osterix is enhanced through interaction with nuclear factor of activated T cells (Nfatc1) transcription factor [46].

The Runx2^+/− mice appeared normal but on closer examination revealed a defect in intramembranous ossification characterized by hypoplastic clavicles and delayed fusion of the cranial fontanelles. These abnormalities reflected some of the symptoms in the human skeletal disorder, cleidocranial dysplasia [8, 47, 48]. Runx2^−/− mice died from respiratory failure shortly after birth owing to the inability to respire due to a non-ossified rib cage. The mutant mice were clearly smaller with shorter limbs and snout and were devoid of an ossified skeleton. Analysis of all bones showed the absence of osteoblasts while chondrocytes were still present. This demonstrated that Runx2 is essential for osteoblast differentiation and has no positive regulatory functions in chondrocyte differentiation and proliferation. Although the deletion of the Runx2 gene has an impact on both intramembranous and endochondral ossification, the former appears more sensitive to Runx2 deficiency.

Runx3, on the other hand, has no apparent role in skeletogenesis as the Runx3^−/− mice either present a severe limb ataxia phenotype [49, 50] or die of starvation shortly after birth owing to excessive growth of gastric endothelial cells [51] with no overt skeletal defects. However, Runx3 was noted to cooperate with Runx2 in chondrocyte maturation evident from the lack of hypertrophic chondrocytes or the expression of the hypertrophic chondrocyte marker, Col10a1, in the skeleton of a Runx2^−/−Runx3^−/− mouse embryos [9, 10]. These observations suggest that Runx2 and Runx3 play compensatory roles in chondrocyte maturation during endochondral ossification. However, Runx2 dominates in advancing chondrocyte maturation over Runx3 as chondrocyte maturation was more impeded in Runx2^−/− mice than in Runx3^−/− mice [9].

MicroRNA control of Runx2 and human disease

Recently a host of miRNAs have been discovered as a form of intermediate regulatory mechanism employed by the Runx2 transcription factor. The miR23a~27a~24-2 cluster was found to bind to the 3’UTR of Satb2 to inhibit its activity. Runx2 directly represses the transcription of the miR23a~27a~24-2 cluster thus releases the direct inhibition of Satb2, a Runx2 repressor, to retard osteogenesis. There is a feedforward mechanism whereby miR23a binds directly to the 3’UTR of Runx2 to induce Runx2 transcription which in turn represses the miR23a~27a~24-2 cluster.

Constitutive Runx2 expression through the final stages of osteoblast differentiation results in osteopenia in mice. The increase in these miRNAs during the end stages of osteoblastogenesis is thus believed to be one mechanism to interrupt sustained bone formation to prevent osteopenia. [52].

Another recent study, in MC3T3E1 and ATDC5 cells has established that at least 10 miRNAs (miR23a, miR-30c, miR-34c, miR-133a, miR-135a, miR204, miR205, miR217, miR-218, miR338) directly target the 3’UTR of the Runx2 mRNA and through that significantly inhibit osteogenic differentiation [53].

A new study has found evidence that miR-3960 directly targets Hoxa2 which is a repressor of Runx2 expression and miR-2861 directly targets Hdac5 to release the inihibition on Runx2 resulting in an increase in Runx2 protein production. Runx2 was also found to bind to the promoter of the miR-3960/miR-2861 cluster to increase its transcriptional activity. Hence, an autoregulatory relationship was described between Runx2 and the miR-3960/miR-2861 cluster, found clustered at the same loci and transcribed from the same miRNA polycistron [54].

Currently, miRNAs targeting Runx3 or regulated by Runx3 in the context of bone formation are yet to be discovered.

Pax1 and Pax9 in skeletogenesis

The Pax gene family constitutes a group of genes encoding transcription factors with a highly conserved DNA-binding domain, the paired-box. Genes within the family are further divided into subfamilies based on the presence of a combination of domains: paired-domain containing two Helix-turn-helix motifs [55], paired-type homeodomain and octapeptide motif (HSVSNILG) [56]; their sequence similarity; and overlapping domains of expression. Identified initially through similarity to the paired-box in the Drosophila gene paired, Pax1 and Pax9 are two of the Pax genes in the same subfamily, essential for the early stages of axial skeleton formation [57].

Of all the nine members of the Pax gene family in the mouse, Pax1 and Pax9 are the only Pax genes that are expressed in sclerotomal cells. They contain only the paired-domain and the octapeptide motif, and share a high protein sequence similarity of 79%, diverging only at their C-terminal ends. Moreover, they share similar expression domains (but not identical), especially in the sclerotome and later in the intervertebral disc anlagen [58].

The Sonic Hedgehog (Shh) morphogen emanating from the notochord and floor plate of the neural tube induces the expression of Pax1 transcripts at E8.5 in the ventro-medial deepithelializing somites to specify their sclerotomal fates [59]. Pax9 transcripts are expressed slightly later (E9.0) and restricted to the caudal half of the sclerotome, unlike Pax1 which is expressed in the rostral half as well. Subsequently Pax1 and Pax9 become restricted to the intervertebral disc anlagen by E12.5 [58, 60].

The importance of Pax1 in the development of vertebral column, scapula and sternum was initially identified through several spontaneous mouse mutants: undulated (un) [61], Undulated short-tail (Un^s) [62], undulated-extensive (un^ex) [63] and undulated intermediate (un-i) [64] which consist of either point mutations or the deletion of the entire Pax1 locus [11, 56, 60]. Subsequent targeted disruption of Pax1 in mice confirmed its role in the proper formation of these skeletal structures. Pax1 heterozygotes were externally normal like wild-type mice, but displayed abnormalities of some skeletal elements such as the first two cervical vertebrae, lumbar vertebrae and sternum with an overall penetrance of 88%. Pax1^null mice were smaller than wild-type mice and had a charactersitic shortened, kinked-tail phenotype. The entire axial skeleton encompassing the vertebral column, scapula, sternum and tail were all defective with reduced or lost ossification centers, fusion of pedicles, loss of acromion process and inappropriate ossification of some of the intersternebra. Deformation in the lumbar region was more severe, with split vertebrae, lack of intervertebral discs and formation of ventral rod-like cartilage structures [11].

Targeted inactivation of Pax9 surprisingly does not give rise to any vertebral column defects. While Pax9 heterozygotes are perfectly normal, Pax9^null mutants display several defects [12].

Pax9^null mice display preaxial polydactyly, cleft secondary palate and lack the derivatives of pharyngeal pouches (parathyroid glands, thymus and ultimobrancial bodies) and all teeth. This phenotype, distinct from that of Pax1^null, corroborates with the Pax9 expression sites in the neural crest-derived cells of the craniofacial and tooth mesenchyme [12].

Despite the complete lack of vertebral column defects, and the possession of a distinct set of phenotypic changes in the Pax9^null mice, it was postulated that Pax9 and Pax1 may have a genetic interaction due to their high sequence similarity and overlapping expression domains in the sclerotome. It was hypothesized that they may have redundant roles in their site of co-expression - the sclerotome. This prompted the generation of the Pax1/Pax9 double null mice [20].

A study of Pax1^nullPax9^null (double null) mice revealed that these two closely related TFs indeed have redundant roles in vertebral column development. While Pax1 can fully compensate for the loss of Pax9 in the vertebral column, absence of Pax1 can only be partially compensated for by Pax9. In accordance to their redundant roles, there is a gene-dosage effect observed in the compound mutants and the disruption of both Pax genes leads to an overt phenotype in the vertebral column where there are no vertebral bodies or intervertebral discs (IVD) and proximal parts of the ribs are also defective. The vertebral column defects in Pax1^nullPax9^null double mutants were more severe than that in Pax1^null single mutants [20].

The development of the axial skeleton itself is a multi-step process beginning with somitogenesis, followed by de-epithelialization of somites, proliferation of the sclerotomal cells which then migrate and condense around the notochord, which subsequently undergoes endochondral ossification [65]. While Pax1 and Pax9 are not required for the formation of the sclerotome per se, it is hypothesized that they are needed to maintain the proliferative capacity of sclerotomal cells, sufficient to attain a critical density of cells for mesenchymal condensation to form, upon which chondrogenesis takes place [20]. Indeed, the essential role of Pax1 in regulating cell proliferation is evident through its genetic interaction with another TF, the mesenchyme forkhead-1 (Mfh1). Mfh1 is also expressed in the sclerotome and has been shown to synergize with Pax1 to control the mitotic activity of sclerotomal cells [66].

Furthermore, Pax1 and Pax9 have been shown in an in vitro study to directly bind to the promoter and trans-activate Bapx1, another key TF known to be essential for the proper differentiation of prechondroblast into chondrocytes in axial skeletal formation [19, 67]. This lends support that both the Paxes are involved in the early stages of axial skeleton formation and are critical for its development.

Therefore, identification of the target genes of Pax1 and Pax9 will help illuminate the early events of regulation involved in the commitment of MSCs towards the osteo-chondrogenic lineage.

Human disease associated with the Pax gene family

The diseases of Pax1 and Pax9 are not limited to that of the mice. In fact, similar phenotypes of malformed vertebral column have been observed in human fetuses suffering from the Jarcho-Levine syndrome, whereby PAX1 and PAX9 protein expression was significantly reduced [68]. Similarly, PAX1 mutations have been associated with certain forms of Klippel-Feil syndrome [69]. The conserved roles of Pax1 and Pax9 in mouse and humans indicates the suitability of mouse as a model system to study such developmental disorders.

CONCLUSION

It is becoming increasingly clear that a broader systems biology approach is required to understand the complex developmental systems of vertebrates and in turn get a clearer picture of developmental diseases. Efforts are underway to elucidate comprehensive gene regulatory networks that will eventually lead to better comprehension of developmental disorders and how to manage them. The mouse will continue to be an indispensible ally in this effort and mapping multiple mutations in important genes will allow us to decipher human phenotypes better and hopefully lead to the development of remedies or cures.