Possible effects of EXT2 on mesenchymal differentiation - lessons from the zebrafish

Bone formation and homeostasis are complex processes in which many cell types and various signalling pathways are involved. Chondrocytes and osteoblasts originate from the same precursors - mesenchymal stem cells, which can also differentiate towards adipocytes, fibroblasts, myoblasts and epithelial cells. The osteoblast lineage is under strict control of RUNT-RELATED TRANSCRIPTION FACTOR 2 (RUNX2) and its downstream target - SP7/OSTERIX, which can work in a Runx- dependent and/or an independent manner. Runx2 and Osterix mRNA are expressed in the immature chondro/osteoprogenitor cells and osteoblasts only [1] and alterations in the expression of either of the two genes affect bone development. Runx- and Osterix-null mice have normal cartilage but their reduced or absent expression of type I collagen, bone sialoprotein, osteonectin, osteopontin, and osteocalcin indicate that bone development is blocked at the step of pre-osteoblast to osteoblast differentiation [2]. Mammalian cells treated with an Osterix inhibitor, dexamethasone, enter an adipogenic- instead of osteoblastic lineage [3]. This imbalance between bone and fat is a known phenomenon. For example, it has been shown that knockout mice, which are heterozygous for Peroxisome proliferator-activated receptor γ (PPARγ have impaired adipogenesis, coinciding with an increased osteoblast number [4]. Other signalling molecules such as wingless (Wnt), bone morphogenic protein (BMP), and hedgehog were also shown to trigger the switch between different lineages including a bone-to-fat change. Remarkably, in all of these pathways, receptor-ligand binding and gradient formation is dependent on heparan sulphates (HS).

Heparan sulphate (HS) are glycosaminoglycans, heavily sulphated linear polysaccharides, that are present in all type of cells. Once they become attached to a core protein they form proteoglycans. The biosynthesis of HS take place in the Golgi apparatus and endoplasmic reticulum, where the elongation of glycosaminoglycan chains is maintained by type II glycosyltransferases encoded by the EXOSTOSINs genes, EXT1 and EXT2[5]. Several genes are involved in the biosynthesis and degradation of HS, and mutations affecting the HS production have serious consequences. Abnormal accumulation of HS, due to its impaired degradation, causes mucopolysaccharidosis, a progressive disorder affecting mental and physical abilities, causing damage to various organs and leading to premature death. Patients with mucopolysaccharidosis often display skeletal abnormalities such as short stature or abnormal bone density [6, 7]. Decreased levels of HS due to mutations in EXT1 or EXT2 also lead to a skeletal abnormality resulting in one of the most common benign bone tumours in young adults – osteochondroma [8]. The hereditary form of osteochondroma, multiple osteochondromas (MO; previously named multiple hereditary exostosis, MHE or hereditary multiple exostosis, HME), is a syndrome that is characterized by the development of multiple tumours (osteochondromas) at different sites of the endochondral skeleton [9]. MO is also associated with various other skeletal and non-skeletal phenotypes such as short stature, bone bowing (Figure 1), impingement of tendons, muscles or nerves as well as low bone density, lipid deposition within osteochondromas, pain and scarring [9‐13].

Several mice models have been developed to study the role of EXT1 or EXT2 in bone and osteochondroma formation [14]. Zebrafish (Danio rerio) have also been shown to be a powerful animal model with morphological and developmental pathways comparable to those seen in humans [15]. We use zebrafish dackel (dak) mutants that carry a nonsense mutation in the ext2, gene which is 84.7% identical (at protein level) with human EXT2[16]. The ext2 ^-/- fish have been used as a model for MO. They have shown to mimic the cartilage phenotype (organization and behaviour) that is common to all models and the dental phenotype present in a number of patients but never described in mice [17‐20].

In this study we show that bone development in the zebrafish ext2 ^-/- mutant is affected at two levels/stages in osteogenesis. Firstly development of the scaffold that is needed for osteoblasts to generate the bone is delayed/absent because chondrocytes fail to undergo terminal differentiation. Secondly, bone formation fails to progress from pre-osteoblasts towards osteoblasts and this change coexists with abnormal lipid depositions and premature adipocyte differentiation. Compounds stimulating fat-to-bone shift, GW9662 and purmorphamine, stimulate bone development in WT and ext2 heterozygote but do not rescue the ext2-null bones. Reduced expression of xbp1, the master regulator of osterix, suggests that unfolded protein responses might play an important role in MO pathogenesis. Beside the “low bone-high fat phenotype”, the ext2-null fish also have a muscle phenotype, i.e. muscles are shorter and thicker, and therefore might have different mechanical properties. Bone bowing, weak muscles and muscle fatigue are often observed in MO patients. Based on our findings in the fish model we speculate that bone bowing may occur as a result of weaker “fat bones” being distorted by muscles (with different mechanical properties). In support of this concept misshaped clavicles and bowed Meckel’s are a very frequent phenomenon in the ext2-null mutant fish (data not shown).

Impaired bone development in the ext2 ^-/- fish has been described previously [17‐19]. In order to identify at what step bone formation is affected and what mechanisms underlie changes in the ext2 mutant, we examined the expression of various bone molecular markers in the ext2 ^-/- fish and compared it with its siblings (Figure 2).

Abnormal lipid deposition coinciding with impaired bone formation is not common to all types of proteoglycan deficiencies (see Additional file 2). b3gat3- and uxs1-homozygote mutants, that are upstream of ext2 in the biosynthesis pathway and lack heparan and chondroitin sulphates, have a very mild bone phenotype and do not show increased lipid deposition (this work and data not shown). Interestingly, the fam20b and xylt1 mutants downstream of uxs1 and upstream of b3gat3 and ext2 were shown to have enhanced bone ossification [32]. Unfortunately nothing is known about fam20b and xylt1 lipid metabolism. The ext2 ^-/- and gpc4 ^-/- , two mutants with reduced HS-levels only, have high lipid content; but only the ext2 mutants have severely reduced bone formation, while the gpc4-null fish have very mild bone impairment. The slc35b2 homozygote mutant, which has diminished levels of all sulphated proteoglycans, has an even more severe bone phenotype than the ext2 ^-/- fish and show only very mild enhancement of lipid deposition. Why different proteoglycan deficiencies have such different effects on bone and lipid metabolisms is not clear. Holmborn and coauthors [33] showed that, in the ext2 homozygote mutant, the remaining HS are over-sulphated which changes their properties (i.e. increase occurrence of protein-interacting domains). Although, heparin, a highly sulphated glycosaminoglycan and a potent anticoagulant, which is often used in clinical practice, negatively affects bone density and is known to increase lipid deposition in sera, the role of over-sulphation of (proteo-)glycans would need to be confirmed.

Craniofacial skeletal development in zebrafish is of mixed origin being derived from cranial neural crest and/or mesoderm [34]. The presence of one functional copy of the ext2 gene is sufficient for the maintenance of normal differentiation of chondrocytes, osteoblasts and other mesenchyme-derived cells. Reduction in HS-levels in the ext2 ^-/- larvae clearly affects skeletal development. Loss of bones cannot be linked specifically to one type of precursor cell as both neural crest- and mesoderm-derived structures are affected. Despite their origin, two populations of osteoblasts with different sensitivity to hedgehog signalling have been described in zebrafish [35]. As no defects in the hedgehog signalling were found in the craniofacial skeleton of the ext2 ^-/- fish, it is unlikely that bone defects could be linked to a specific type of hedgehog-sensitive osteoblast. However, it is possible that there are multiple types of osteoblasts existing in fish, differing in their sensitivity for HS.

Bone homeostasis depends on the balance between osteoblastic and osteoclastic activity. Lipids are known to attract osteoclasts while suppressing osteoblastogenesis (for review see [36]). Unfortunately, we were not able to test this in zebrafish as the first osteoclasts develop by 16 dpf, beyond the time of premature death of the ext2 ^-/- fish. Nevertheless, observations from patient material suggest that indeed both osteoblasts and osteoclasts are affected by HS-deficiencies [10, 26] or by HS abnormal accumulation [7] and, in both cases, bone mineral density is altered. Osteoblasts and adipocytes might not be the only lineages affected by imbalanced HS. EXT1-null embryonic stem cells also appear to have impaired differentiation hematopoietic lineages [37], while osteochondromas exhibit impaired vascularisation [38].

Fatty acids, when not stored in adipocytes, accumulate into the circulation [39]. Although premature adipocyte-like cells were detected in the ext2 ^-/- fish it is unlikely that they would be able to store all the lipids as cytoplasmic droplets. Therefore, Oil red O stain in vasculature could reflect only a surplus of fatty acids/lipids. However, it is also possible that mutation in the ext2 gene leads to an abnormal intravascular accumulation of lipids. The changes in bones and fat that we have described in fish were a characteristic of an organism homozygous for a mutation in the ext2 gene in all cells. Since MO patients are mostly heterozygous for a mutation in EXT they should have very mild (if any) systemic phenotype. However, if findings from this fish model are true for humans, strong focal changes should be expected at the site where loss of heterozygosity/haplo-insufficiency occured. Not much is known about lipid metabolism in patients with MO. Lemos and co-authors [10] reported lower bone mineral density of femoral neck and lumbar spine in MO patients near osteochondromas. In addition, single reports described deposition of fat within the cartilaginous cap of osteochondromas [11] and development of lipoma, a benign bone tumour, or fat-pads in association with osteochondromas [40, 41]. These finding might have been coincidental in MO but increased lipid levels often remain asymptomatic. In light of our findings in the fish model on the bone-fat imbalance the status of lipids in human MO seems worth investigating.

Humans, mice and fish with MO are often short in stature and have bowed bones. Recently, Jones and co-authors [42] demonstrated that osteochondroma growing on account of deranged bone growth is apparent only in some individuals and other mechanisms must contribute to the short bone phenotype. Also bone bowing does not always require osteochondroma formation to generate the observed anatomical changes (K. Jones, University of Utah School of Medicine, personal communication). The presence of muscle phenotype needs to be confirmed in non-fish MO. Further work will show how (if) muscles with different mechanical properties contribute to the formation of shorter and bowed bones in patients.

Our data indicated that HS have multiple functions during endochondral bone development. First of all, HS are required for terminal differentiation of the cartilaginous template and consecutive formation of a scaffold that is needed for further bone development. Secondly, normal expression of runx2 and impaired expression of osterix in the ext2 ^-/- fish indicated that HS are required by osteoblast precursors for their further differentiation within the osteoblastic lineage. Furthermore, the increased lipid deposition in the ext2 ^-/- fish suggest that HS are involved in determining the cell lineage when mesenchymal precursor cell differentiates into bones and/or fat. PCR analyses confirm the increase in the expression of lipid markers and down-regulation of early skeletal markers. It still remains to be established how HS are involved in this shift, but lower expression of xbp1, a master regulator of osterix, suggests that HS affect the unfolded protein response, a pathway which is known to control bone formation and lipid metabolism.