Mice deficient in Ext2 lack heparan sulfate and develop exostoses

Heparan sulfate proteoglycans (HSPG) are present in the extracellular matrix and on the surface of virtually every cell type. They consist of a core protein to which heparan sulfate is covalently attached. In vitro, HS interacts with a multitude of molecules, including growth factors and extracellular matrix proteins. Much of our knowledge about the in vivo function of HS in extracellular signaling cascades stems from studies in Drosophila (Lin,2004). However, in recent years, analyses of targeted mutations in animals and human disorders caused by mutations in HS production have vastly increased our understanding of the in vivo role of HS in development and disease (for reviews see Esko and Selleck,2002; Nakato and Kimata,2002).

Hereditary multiple exostoses (HME) is a skeletal disorder characterized by the presence of multiple bony protuberances called exostoses (or osteochondromas), usually arising in the epiphyseal growth plate of bones formed by endochondral ossification. Although exostoses can be found on almost every long bone, they appear more frequently at the distal end of the femur,tibia, fibula, humerus and ribs. Often, the formation of an exostosis interferes with normal bone development, resulting in deformities such as shortening and bowing of the bones(Hennekam, 1991; Solomon, 1963). HME is a multigenic disease and follows an autosomal dominant mode of inheritance. Approximately 80% of affected individuals have a positive family history,although exostoses can also appear sporadically. Genetic linkage studies and mutation analyses have identified two main loci as being associated with the disease: EXT1, located on chromosome 8q24.1; and EXT2,located on chromosome 11p11 (Ahn et al.,1995; Stickens et al.,1996).

HS is a glycosaminoglycan (GAG), a linear polysaccharide composed of alternating D-glucuronic acid (GlcA) and N-acetyl-D-glucosamine(GlcNAc) subunits. The biosynthesis of HS is a complex process that involves a whole host of enzymes (for review see Esko and Selleck, 2002). The EXT1 and EXT2 genes encode the exostosins, glycosyltransferases involved in the synthesis of HS(Lind et al., 1998; McCormick et al., 1998; Wei et al., 2000). EXT1 and EXT2 are type II transmembrane proteins involved in HS chain elongation,catalyzing the alternating transfer of GlcA and GlcNAc residues. In vivo, EXT1 and EXT2 form a hetero-oligomeric complex in the Golgi apparatus and this complex has a higher glycosyltransferase activity than either protein alone(McCormick et al., 2000; Senay et al., 2000). The majority of cases of HME are caused by frameshift or missense mutations in EXT1 or EXT2, creating truncated forms of the proteins they encode(Ahn et al., 1995; Stickens et al., 1996; Wuyts and Van Hul, 2000; Zak et al., 2002). The clinical syndromes of HME caused by mutations at the EXT1 or the EXT2 loci are identical, and no phenotype-genotype correlations exist(Cook et al., 1993; Le Merrer et al., 1994; Wu et al., 1994).

The initial identification of EXT genes as glycosyltransferases did not yield an immediate understanding of the biological processes they control. If HS is present on almost every cell type, then why do mutations in EXT genes lead primarily to a bone phenotype? Further insight into the function of EXT genes was obtained through the identification of the Drosophila Ext genes tout-velu (homolog of EXT1), sister of tout-velu(homolog of EXT2) and brother of tout-velu (homolog of EXTL3)(Bellaiche et al., 1998; Han et al., 2004; Takei et al., 2004). Mutations in these genes result in the accumulation of Hedgehog, Wingless and Decapentaplegic proteins in front of the mutant cells, suggesting that HS is required for generating morphogen gradients. This was an important finding, as mutations in EXT genes affect bone formation, and all three morphogens[Hedgehog, Wnt proteins and bone morphogenic proteins (BMPs)] have a role in skeletal development (Karsenty and Wagner,2002).

Targeted deletion of Ext1 in mice results in early embryonic lethality and Ext1-null embryos die around the time of gastrulation. Inactivation of Ext1 abolishes the production of HS, illustrating the importance of Ext genes in HS synthesis(Lin et al., 2000). Ext1 heterozygous mice show somewhat increased chondrocyte proliferation and delayed hypertrophic differentiation owing to increased Indian hedgehog [Ihh (Hilton et al.,2005)], but are not reported to show any gross skeletal phenotype. Other evidence of a direct relationship between Ihh, Ext genes and bone development has come from the analysis of mice with a hypomorphic Ext1 allele (Ext1^gt/gt) that was generated by a gene-trap method (Koziel et al.,2004). In contrast to Ext1-null mice, Ext1^gt/gt mice survive until E14.5 and some embryos can be recovered at E16.5. Interestingly, in the growth plates of Ext1^gt/gt mice, Ihh has an expanded range of signaling. These data suggest that Ext1 might play a role in endochondral ossification by modulating the distribution of Ihh. As EXT1 and EXT2 form a complex and HME is caused by mutations in either gene, we generated mice deficient for Ext2 and analyzed the homozygous and heterozygous phenotypes to better understand the etiology of this disease.

A 12 kb genomic clone was obtained by screening a 129Sv mouse genomic library (Stratagene) with a probe derived from the 5′ end of the mouse Ext2 cDNA. A 2.1 kb XhoI-BstEII fragment from intron 1 was cloned upstream of a neomycin-resistance (Neo) gene and downstream of a thymidine kinase (TK) gene to form the 5′ vector arm. The neo and TK genes are under control of the phosphoglycerokinase (PGK)promoter and the herpes simplex virus (HSV) promoter, respectively. The 3′ vector arm was created by cloning a 3.5 kb NotI-HindIII fragment downstream of the neo gene. The targeting vector was linearized by digestion with NotI and electroporated into ES cells. After 10 days of treatment with G418 (Geneticin,180 μg/ml Invitrogen) and FIAU[1(–2-deoxy-2-fluoro-b-D-arabinofuranosyl)-S-iodouracil] for positive-negative selection, respectively, ES colonies were picked and tested for homologous recombination. Selected clones were karyotyped to test for chromosomal abnormalities. Three independent heterozygous ES clones were injected into C57BL/6 blastocysts to generate chimeric mice. A total of seven chimeric animals were produced, two of which transmitted the mutation through the germline. Mice heterozygous for the Ext2 mutant allele gave rise to litters of normal size when bred to wild-type animals.

ES cell DNA and mouse tail DNA were screened for the Ext2 targeted mutation by digestion with SpeI or HindIII, followed by southern blot hybridization with probe A or probe B, respectively (see Fig. S1 in the supplementary material). Replacement of exon 2 of the Ext2gene with a neo cassette introduced SpeI and HindIII sites that could be used to distinguish the wild-type allele from the targeted allele. Upon digestion of genomic DNA with SpeI and hybridization with probe A, wild-type and targeted alleles produced fragments of 9 kb and 5.3 kb, respectively. Following digestion with HindIII and hybridization with probe B, wild-type and targeted alleles produced fragments of 6 kb and 3.5 kb, respectively. Probe A is a 1.8 kb SpeI-HindIII fragment external to the region of vector homology. Probe B is a 300 bp PCR amplification product (forward primer,5′ACATTATGATCACATATTGC; reverse primer,5′GGAGTACAATGAACTGCTGACG3′), 3′ from exon2 and internal to the region of vector homology. PCR genotyping was accomplished using a four-primer PCR amplification (wild-type allele: Ext2-GT-150F,GGTCTGGACGATAGGTGTCAGG; Ext2-GT-640F, GTGACGTAGTAGATTCGGTGC; the amplification product is 190 bp) (targeted allele: Ext2-GT-810R, GTTGAACAATCCAATCCACGC;Ext2-GT-NEO-R, CATGCTCCAGACTGCCTTGG; the amplification product is 260 bp).

TSC were isolated as previously described(Tanaka et al., 1998) with some modifications: 3.5 dpc mouse blastocysts were isolated from Ext2^+/– C57BL/6 mice crossings. They were individually plated into four-well plates and cultured on mouse embryonic fibroblasts (MEF) in MEF medium + Fgf4 (25 ng/ml) + heparin (1 μg/ml). The medium was changed on day 3 and the blastocyst outgrowths were trypsinized on day 4 or 5. Flat colonies appeared between day 6 and day 10 and were split once they had reached 50% confluence. The medium was changed every 2 days and the cells were passaged at 80 to 90% confluence at 1 part in 10 or 20. When the blastocyst outgrowths or the cell lines were trypsinized, the doses of Fgf4 and heparin used for the cultures were 1.5 × the normal dose. TSC were weaned from the MEFs after two passages.

Tissues were fixed in 4% paraformaldehyde in phosphate-buffered saline,decalcified in EDTA, paraffin embedded, sectioned at 5 μm and stained with von Kossa's stain, Safranin O/Fast Green or Masson Trichrome. Briefly, for Safranin O/Fast Green staining, deparaffinized rehydrated sections were stained in Weigert's Iron Hematoxylin (Sigma,), 0.02% aqueous Fast Green(Sigma) followed by a rinse in 1% acetic acid and 0.1% aqueous Safranin O(Sigma). For Masson Trichrome staining, a kit was used according to the instructions of the manufacturer (Sigma).

For immunohistochemistry, tissues were fixed, embedded and sectioned as described above. For HS immunostaining, sections were deparaffinized, washed in PBS and any endogenous peroxidase activity was blocked using 0.3% hydrogen peroxide (Sigma) in methanol for 30 minutes at ambient temperature. Sections were rinsed in PBS and incubated with M.O.M. blocking solution (M.O.M. kit,Vector Laboratories) for 1 hour at room temperature. After rinsing in PBS,sections were incubated overnight at 4°C with the mouse monoclonal antibody 10E4 (Seikagaku) at a dilution of 1:100. Unbound antibody was removed by washing in PBS. Bound antibody was detected using an HRP-conjugated anti-mouse IgM secondary antibody (Jackson ImmunoResearch Laboratories) for 1 hour at ambient temperature. After washing in PBS, the peroxidase was detected using a DAB detection kit (Zymed). Sections were washed in tap water and counterstained with Methyl Green (Sigma). Control sections were overlaid with 2.5 mU of heparitinase I (Sigma) in 0.5 ml of buffer (100 mM NaCl, 1 mM CaCl₂, 50 mM sodium HEPES, pH 7.0, containing 25 mg of BSA) for 2 hours at 37°C. After rinsing in PBS, sections were incubated with the HS antibody 10E4 as described above.

Whole skeletal preparations of 2-week-old mice were prepared and stained with Alizarin Red and Alcian Blue as previously described(McLeod, 1980). For the measurements of bone length and localization of exostoses, skeletons were prepared by Dermestid beetles as described(Hefti et al., 1980).

A 10 mg/ml stock of bromodeoxyuridine (BrdU; Sigma) was injected intraperitoneally into one-week-old mice at a dose of 100 μg BrdU/g. Mice were sacrificed 1 hour after injection, bones were harvested and processed as described in the section `Histological analyses and immunohistochemistry'. Staining for BrdU was carried out on paraffin sections using a kit according to manufacturer's directions (Zymed).

Poly (A) RNA was isolated (Micro-FastTrack, Invitrogen) from TSC according to manufacturer instructions. cDNA was synthesized using a kit from Invitrogen(Two-step RT-PCR system) and used as template for PCR reactions using the following primers: Ext2-exon2 forward, 5′GTGGATGATGCCGGTGTTCC 3′;Ext2-exon3 reverse, 5′ CAGGCAACATATTGAACAGC 3′; Ext2-exon8 forward, 5′ CCTACAGATCATCAATGACAGG 3′; Ext2-exon9 reverse,5′ AGCAGCTTGGACAGACTGG 3′. Primer sets represent coding regions that span introns. E-cadherin primers were used as a control for RNA integrity. PCR amplification was performed using 30 cycles and products were analyzed on 2% agarose gels.

Paraffin sections were placed on acid-etched, TESPA-treated slides and prepared for in situ hybridization as described(Albrecht et al., 1997). Plasmids were linearized with the appropriate restriction enzymes to transcribe either sense or antisense ³⁵S-labeled riboprobes. Probes were as follows: collagen type 2 (Col2a1), collagen type 10(col10a), Indian hedgehog (Ihh) and patched 1(Ptch1) (Albrecht et al.,1997; Ferguson et al.,1999); parathyroid hormone/parathyroid hormone-like peptide receptor (Pthr1) (Kobayashi et al., 2002); Fgf8(Crossley and Martin, 1995);brachyury (Wilkinson et al.,1990); Apoe(Basheeruddin et al., 1987); H19 (Poirier et al.,1991); Lim1 (Shawlot and Behringer, 1995); Snai1(Nieto et al., 1992); Otx2 (Simeone et al.,1993); and Hesx1(Thomas and Beddington, 1996). Slides were washed at a final stringency of 65°C in 23 ×SSC, dipped in emulsion and exposed for 1-2 weeks. Slides were counterstained with Hoechst 33342 or Hematoxylin.

Cell culture supernatants (with or without serum) were collected on ice in the presence of protease inhibitors [pepstatin (5 μg/ml), leupeptin (1μg/ml), aprotinin (10 μg/ml), benzamidine (10 μg/ml) (Sigma)],centrifuged at 150 g for five minutes at 4°C to remove non-adherent cells and cell debris, aliquoted and stored at –80°C. Whole-cell lysates were prepared by scraping cultured cells into RIPA buffer(20 mM Tris pH 7.2, 10 mM EDTA, 0.3 M NaCl, 0.1% SDS, 1% Triton X100, 0.05%Tween 20) containing the same protease inhibitors. Cell debris was removed by centrifugation at 20,000 g for 20 minutes at 4°C and the protein aliquots were stored at –80°C until used. The BCA protein assay kit (Pierce, Rockford, IL) was used to determine the protein concentration of the supernatants and of the cell lysates. The polyclonal antibody against Ext2 was a generous gift from Dr Takahiko Shimizu (Tokyo Metropolitan Institute of Gerontology)(Morimoto et al., 2002).

Trophoblast stem cells or ES cells were cultured for 72 hours in a medium composed of 70% MEF-conditioned medium and 30% sulfate-depleted RPMI medium or for 48 hours with 50 μCi/ml of ³⁵SO₄ (Dupont NEN) in sulfate-reduced F12 medium, respectively. [³⁵S]GAG chains were isolated by DEAE chromatography and a sample was treated with chondroitinase ABC at 37°C overnight (Bame and Esko,1989). Treated and untreated chains were separated by anion-exchange HPLC. Glycosaminoglycans were eluted with a linear gradient of NaCl (0.2-1 M) using a flow rate of 1 ml/minute and increasing the NaCl concentration by 10 mM/minute. The effluent from the column was monitored for radioactivity with an in-line radioactivity detector (Radiomatic floOne/beta,Packard Instruments) with sampling rates every 6 seconds and data averaged over 1-minute intervals.

To better understand the function of the Ext2 gene, we inactivated it through homologous recombination in ES cells. A targeting vector was constructed in which a neomycin-resistance cassette replaced exon 2, which contains the start of translation (see Fig. S1 in the supplementary material). Three independent Ext2 heterozygous ES clones were injected into C57BL/6 blastocysts and chimeric mice were mated with wild-type C57BL/6 to generate germline mutant mice.

We expected that inactivation of Ext2 would also abolish HS synthesis. To test this hypothesis, we isolated wild type, Ext2^+/–and Ext2^–/– TSC, and collected RNA for RT-PCR analyses (Fig. 1A). As expected, no amplification product was obtained in an RT-PCR reaction using primers for exon 2. However, an amplification product of the expected size was obtained when primers for exon 8 were used, indicating that a truncated form of the Ext2 RNA transcript was being made. Cell lysates from wild-type,heterozygous and Ext2^–/– TSC were analyzed by western blot (Fig. 1B). Incubation with a polyclonal antibody against EXT2 showed an expected band of 82 kDa in cell lysates from wild-type and Ext2^+/–TSC. However, heterozygotes showed a second band of lower molecular weight and of varying intensity, suggesting that a truncated form of the protein was being produced. Immunoblots of lysates from Ext2^–/–cells revealed that these cells express only the truncated form of the protein.

ES cells derived from wild-type, heterozygous and null embryos were grown in culture with ³⁵SO₄ to label sulfated glycosaminoglycans, which in wild-type cells consist of a mixture of HS and chondroitin sulfate (CS) (Lin et al.,2000). A sample of the ³⁵S-GAG chains was analyzed by anion exchange chromatography directly and another sample was first treated with chondroitinase ABC, which depolymerizes chondroitin sulfate. As shown in Fig. 1C, wild-type and heterozygous Ext2 ES cells both produced two peaks of labeled material. The first peak was resistant to chondroitinase and signifies HS, whereas the second peak was sensitive to chondroitinase and signifies CS. Ext2^–/– ES cells only produced one peak, which co-eluted with CS and was sensitive to chondroitinase, and thus no HS was being produced.

Next, we verified that HS was not produced by staining sections of E7.5 wild-type and Ext2^–/– embryos with an antibody that recognizes HS (10E4) (Fig. 1D). In wild-type embryos, a strong signal was detected in the basement membranes of the ectodermal, endodermal and mesodermal cell layers. By contrast,specific staining for HS was not detected in Ext2^–/– mutants. Therefore, although an immunoreactive fragment of the Ext2 protein is produced, it is inactive in HS synthesis.

Mice heterozygous for the Ext2 mutant allele appeared phenotypically normal and gave rise to litters of normal size when crossed with wild-type animals. We did not detect any Ext2^–/– neonates from nearly 300 offspring of Ext2 heterozygous intercross matings, indicating that mutant mice died during embryogenesis (Table 1). To characterize the embryonic lethality, timed Ext2^+/– matings were set up and embryos were genotyped and morphologies examined starting at E6.0. At E6.0, Ext2 mutants were indistinguishable from wild-type embryos (data not shown). However, at E6.5 the first visible and reproducible differences with wild-type embryos became apparent (Fig. 2A,B). Mutant embryos appeared to have formed a normal ectoplacental cone, but the extra-embryonic regions were underdeveloped and the egg cylinder did not elongate. At E7.5, all wild-type embryos had initiated gastrulation and had formed a primitive streak. Although all of the mutants were smaller than wild-type embryos, there was heterogeneity with respect to the degree of egg cylinder elongation and formation of extra-embryonic structures connecting the embryonic structures to the ectoplacental cone(Fig. 2C,D). Despite the morphological variation of Ext2^–/– embryos,none of the mutants showed any visible evidence of primitive streak formation(Fig. 2E-H). By E8.5, the Ext2^–/– embryos had degenerated and lacked recognizable structures (data not shown).

We examined the mutants in greater detail in order to determine the precise stages of development impacted by the lack of HS. At E7.5, control embryos had developed a primitive streak and mesodermal cells were visible between the ectoderm and visceral endoderm. At this stage, we observed three different classes of morphologically distinct Ext2^–/–mutants (Fig. 2F-H). In all three types of mutants, parietal endoderm and Reichert's membrane had elongated to almost the same size as the wild-type embryo. However, no histologically recognizable primitive streak or mesoderm (embryonic or extra-embryonic) formed in any of the mutants.

The first group (class 1) represented about 25% of all mutants and was characterized by a round-shaped egg cylinder that appeared to be directly connected to the ectoplacental cone (Fig. 2F). No extra-embryonic ectoderm, endoderm or mesoderm was present. The egg cylinder consisted of only two layers, ectoderm and visceral endoderm, with no primitive streak or mesoderm. The cells of the ectodermal layer were pycnotic and TUNEL assays confirmed a significant increase in the number of apoptotic cells in this layer(Fig. 2J). We excluded class 1 embryos from further analyses as they were no longer viable.

A second class of mutants (class 2), representing about 50% of all mutants,developed some extra-embryonic structures, but they were composed of immature extra-embryonic ectoderm and endoderm, when compared with wild type, and lacked extra-embryonic mesoderm entirely(Fig. 2G). The egg cylinder had elongated to about one-third the size of the wild-type embryos, but only consisted of embryonic ectoderm and visceral endoderm. TUNEL assays showed only a few apoptotic cells (Fig. 2K).

A third group of Ext2 mutants (class 3), showed further elongation of the embryonic and extra-embryonic regions, but still no embryonic or extra-embryonic mesoderm was present (Fig. 2H). Interestingly, these mutants had initiated the formation of a head fold and histological analyses showed the presence of neuroectoderm. This region of the embryo contained a high number of TUNEL-positive cells(Fig. 2L). Further histological analyses indicated that cavitation was taking place in the region of TUNEL-positive cells (data not shown).

H19, a marker for extra-embryonic tissues(Poirier et al., 1991), was expressed in Ext2^–/– embryos and confirmed that extra-embryonic ectoderm and endoderm, and ectoplacental cone and trophoblast giant cells were present and not significantly affected(Fig. 3A,B). Apolipoprotein E(Apoe), a marker for visceral endoderm(Basheeruddin et al., 1987),was expressed in Ext2^–/– embryos, indicating that Ext2 is not required for ectoderm or endoderm formation(Fig. 3C,D). Interestingly,expression of Fgf8, one of the earliest markers for primitive streak formation(Crossley and Martin, 1995; Maruoka et al., 1998), was detected in wild-type embryos as well as in Ext2 mutants, albeit at reduced levels (Fig. 3E,F). However, as Fgf8 is normally expressed just prior to streak formation in a patch of epiblast cells on the proximal prospective posterior side, ands in the visceral endoderm (Crossley and Martin, 1995), this expression does not necessarily indicate that Ext2 mutants are capable of forming a primitive streak. By contrast, another marker of the primitive streak and mesodermal cells, brachyury (T)(Wilkinson et al., 1990), was present in wild-type embryos, but not in Ext2 mutants at E6.5(Fig. 3G,H).

Both class 2 (Fig. 3I,J) and class 3 (Fig. 3K,L) mutants expressed T at E7.5. At the midstreak stage, Lim1 mRNA is expressed at low levels in the primitive streak and at higher levels in mesodermal cells migrating away from the streak. By E7.5, its expression is restricted to the node (Barnes et al.,1994). In Ext2^–/– class 3 mutants,we found several cells expressing Lim1 in the region where the primitive streak would normally form (Fig. 3O,P). However, Snai1, which is expressed in cell populations that will become migratory, including primitive streak and nascent mesoderm (Nieto et al., 1992; Smith et al., 1992), was not expressed in Ext2^–/– embryos(Fig. 3Q,R). These results suggest that Ext2^–/– embryos initiate the formation of a primitive streak, but fail to properly form mesoderm.

Our histological analyses indicated that some E7.5 Ext2^–/– embryos can initiate the formation of a head fold and accumulate neuroectoderm. Otx2 is normally detected in anterior neuroectoderm at late streak stage, in the region encompassing the prospective forebrain and midbrain (Ang et al., 1994; Simeone et al.,1993). Hesx1 is initially expressed in anterior visceral endoderm (AVE) and subsequently in definitive endoderm. Hesx1expression then intensifies and spreads laterally into neuroectoderm, but remains restricted to the most anterior region of the brain(Hermesz et al., 1996; Thomas and Beddington, 1996). Therefore, we performed in situ hybridization with Otx2 and Hesx1 probes, and were able to confirm that the anterior cell mass observed in the Ext2^–/– mutants is indeed neuroectoderm (Fig. 3S,T).

Humans with HME are heterozygous for either EXT1 or EXT2. Close examination of the skeletons of Ext2^+/– mice revealed the presence of one or more exostoses on the ribs of 7 out of 25 (28%) Ext2 heterozygotes, whereas exostoses were never detected in wild-type animals(Fig. 4A,B). In humans, ribs are a common location for exostoses formation, although nearly every bone,with the exception of membranous bones, is involved. However, we did not find exostoses on the long bones (tibia, femur, ulna and radius) of the Ext2^+/– mice. Despite this difference, the exostoses we examined in the mice were very similar to those seen in humans and varied in shape and size from a small widening of the metaphyseal region to a large bulky outgrowth. Most mice only had a single exostosis (range 1-3). The distribution of lesions was uniform throughout the rib cage and there was no preferential orientation of the projection (ventrally, dorsally or laterally).

Mouse exostoses resembled those in humans. The lesions were located near the costochondral junction and were composed of cortical and medullary bone with an overlying hyaline cartilage cap(Fig. 4C-E). Characteristic for exostoses, the bone marrow cavity was continuous with that of the underlying bone. The cartilage cap was covered with a thin layer of fibrous tissue that was continuous with the perichondrium. Within the cartilage cap we found random patches of calcification. Some chondrocytes appeared to be lined up,resembling the organization of chondrocytes in a normal growth plate. Histological sections of the exostosis cartilage cap stained positive for HS with the HS antibody 10E4 (Fig. 4F). These results demonstrate that chondrocytes of the exostosis produce HS, and are therefore not null for Ext2.

Frequently, the formation of exostoses is accompanied by short stature(Hennekam, 1991). It has been previously reported that Ext1^+/– mice have a∼10% reduction in bone length (Lin et al., 2000). We examined bone length in Ext2^+/– mice on the C57BL/6 background. Skeletons at 12 weeks of age were cleaned by Dermestid beetles and found no significant difference in either total length [wild type, 95.2±0.534 mm(n=27); Ext2^+/–, 96.0±0.637 mm(n=25); P=0.351, two-tailed t-test], measured from tip of the snout to base of the tail, or in the ratio of total length divided by length of tibia plus femur [wild type, 2.90±0.012 mm(n=25); Ext2^+/–, 2.91±0.014 mm(n=25); P=0.887, two-tailed t-test). Our reevaluation of Ext1^+/– mice also showed no significant differences in bone length (B.M.Z., D.S., M. Hilton, G. Evans,Z.W., D. Wells and J.D.E., unpublished).

We also observed a 100% penetrant phenotype in postnatal Ext2 heterozygotes: nodules or single misplaced chondrocytes were detected on every rib and appeared throughout the entire length of the costochondral cartilage(Fig. 5A). These displaced chondrocytes were present as single cells(Fig. 5B) or as clusters(Fig. 5C) near the perichondrium. On rare occasions, a nodule was large enough to be detected on a whole-mount skeletal stain (Fig. 5D). The abnormal chondrocytes were always found with overlying perichondrium. In many cases, this appeared to make the perichondrium bulge out.

Interestingly, these chondrocyte abnormalities seen in the Ext2^+/– mice, between 1 and 2 weeks of age, occur at the same time the rib cartilage undergoes hypertrophic differentiation. Before 1 week of age, chondrocytes in the ribs of wild-type mice have a uniform undifferentiated appearance throughout the costochondral cartilage(Fig. 6A,B; also see Fig. 10). Around 2 weeks of age, however, chondrocytes in the center of the skeletal element mature and become hypertrophic, while cells in the periphery remain small and undifferentiated. In situ hybridization confirmed that the morphological changes correspond to changes in expression of differentiation-specific genes. Up to 1 week of age, all chondrocytes express Col2, an early marker of chondrocyte differentiation (Fig. 6C,D). At 2 weeks of age, expression of Col2 is restricted to cells at the periphery (Fig. 6E,F). Chondrocytes in the center now express Col10, a marker of hypertrophic chondrocytes (Fig. 6G).

This raises the issue of whether the formation of the nodules was the result of a defect in proliferation and/or differentiation. Older animals tended to have nodules that were bigger, suggesting that nodules could result from clonal expansion of one or more displaced chondrocytes (compare Fig. 5B,C). When we injected BrdU into 2-week-old Ext2^+/– mice and collected ribs 1 hour later, we found BrdU-positive cells in the proliferative zone of the growth plate of the ribs, as expected (Fig. 7A-C). There were no BrdU-positive cells in any other region of the costochondral cartilage where the nodules form or in the nodules themselves of the Ext2^+/– ribs. The nodules,therefore, did not appear to be highly proliferative.

As the displaced cartilage had the appearance of prematurely differentiated chondrocytes, we performed in situ hybridization analysis with a Col10 antisense probe to check for evidence of cartilage hypertrophy(Fig. 7D). At 2 weeks of age,hypertrophic cells in the center of the rib cartilage expressed Col10, while cells in the periphery did not. The nodules of displaced chondrocytes expressed Col10, but were separated from the centrally located hypertrophic cells by a layer of chondrocytes that did not express Col10. This thin layer of cells did express Col2 (data not shown). These data indicate that cells in the nodule had differentiated prematurely. Staining with the HS antibody 10E4 further showed that the chondrocytes in the nodules still produced HS, and thus were not Ext2^–/– cells(Fig. 7E).

We next asked what mechanism underlies these defects in cartilage formation. The Ihh/Pthr1 pathway was a good candidate, as altered activation of this pathway can potentiate chondrocyte differentiation. We observed that,although Ihh is expressed in the growth plate of rib cartilage at the time the defects appear, it is absent from the rest of the rib cartilage and from areas where premature differentiation was observed (data not shown). To determine hedgehog activity more sensitively, we crossed the Ext2^+/– mice with mice carrying lacZ under control of the hedgehog receptor patched 1 (Ptch1). At 1-2 weeks of age, when the premature hypertrophic differentiation was observed, no lacZstaining could be found in the rib cartilage other than in the growth plate of the rib (data not shown; see Fig. 10 for schematic representation of expression domains). This result is in agreement with our in situ analysis and indicates that there is no aberrant hedgehog activity in the Ext2^+/–chondrocytes.

What is the mechanism of cartilage hypertrophy?Pthlp^–/– mice show hypertrophic differentiation that is reversed by introducing a constitutively active Pthr1 under the control of the Col2 promoter(Soegiarto et al., 2001). To test whether the abnormal chondrocyte hypertrophy seen in the nodules of the Ext2^+/– ribs was due to reduced Pthlp signaling, we crossed the constitutively active Col2-Pthr1 transgene into the Ext2^+/– mice(Fig. 8). However, this was unable to prevent the premature hypertrophic differentiation of the rib cartilage of Ext2^+/– mice. Taken together, these data suggest that the rib cartilage abnormalities seen in the Ext2^+/– mice are not the direct result of alterations in Ihh/Pthr1 signaling.

The phenotype seen in the ribs prompted us to look for phenotypes in the growth plates of long bones. There were no overt morphological differences in growth plates in histological sections of ulna, radius and femur from E18.5 wild-type and Ext2^+/– mice(Fig. 9A,B). Mineralization of the cartilage matrix, formation of the bony collar surrounding the hypertrophic chondrocytes and formation of trabecular bone all appeared to be normal (Fig. 9E,F).

To assess chondrocyte proliferation, we injected pregnant females with bromodeoxyuridine (BrdU) and embryos collected at E18.5. We found no significance in the overall number of BrdU-positive cells (wild type 33.1±2.5, n=12; Ext2^+/–34.4±2.0, n=11; P=0.70, unpaired t-test) in the proliferating zone of the growth plate(Fig. 9G,H).

However, closer examination of the chondrocytes in the proliferative zone showed a loss of the characteristic columnar organization normally seen in wild-type growth plates (Fig. 9A-D). Although the organizational defect in Ext2^+/– long bones was subtle, it was consistent for all (8/8) bones examined. Similar chondrocyte disorganization was observed in embryonic mice homozygous for a gene trap insertion in the Ext1 gene(Ext1^gt/gt)(Koziel et al., 2004).

In wild-type mice, Pthr1 and Ihh are expressed in prehypertrophic chondrocytes in domains that were partially overlapping(Fig. 9K,L and Fig. 9M,N,respectively). Ptch1 was expressed in the perichondrium, adjacent to the expression domain of Ihh and in proliferating chondrocytes(Fig. 9I,J). Ext2^+/– mice expressed Ptch1, Ihh and Pthr1 in the appropriate places, but there was a small decrease in the expression domains of Pthr1 and Ihh, while the Ptch1 expression domain in Ext2^+/– seemed comparable with that in wild type. We then analyzed the distribution of Ihh protein by immunohistochemistry using Ab80, a sonic hedgehog antibody that crossreacts with Ihh. We did not see any obvious difference in the localization of the hedgehog protein between wild type and Ext2^+/– (data not shown). These data show that inactivation of one allele of Ext2 interferes with normal chondrocyte development. However, as we could not detect any differences in total length of bones, this differentiation defect does not appear to influence the endochondral ossification process in any significant way.

Ext1 and Ext2 are glycosyltransferases involved in the chain elongation of heparan sulfate (Lind et al.,1998; McCormick et al.,2000; Wei et al.,2000). Prior studies of Ext1 in mice demonstrated that this gene is essential for HS synthesis and early development(Lin et al., 2000). We have shown for the first time that the expression of Ext2 is essential for HS synthesis in mice, confirming studies in Drosophila(Bornemann et al., 2004; Takei et al., 2004) and zebrafish (Lee et al.,2004).

Inactivation of Ext2 results in early embryonic lethality around the time of gastrulation. The morphogenic movements and differentiation of cells that accompany gastrulation are in large part directed by signaling molecules, such as members of the transforming growth factor β (Tgfβ) superfamily(Zhao, 2003), Wnt proteins(Kelly et al., 2004) and Fgf proteins (Ciruna and Rossant,2001). Many of these signaling molecules can bind HS, at least in vitro (for a review, see Lin,2004). Our data show that inactivation of Ext2 abolishes production of HS, possibly resulting in an effect on one or more of these signaling pathways. One candidate is Ihh, as mice with a gene trap mutation in the Ext1 gene have altered Indian hedgehog signaling(Koziel et al., 2004). However, Ihh-null mutants die around mid-gestation, between 10.5 and 12.5 dpc (St-Jacques et al.,1999), much later than the Ext2-null mice.

The lack of Ext2 leads to a gastrulation defect and abnormalities in the formation of extra-embryonic structures. As proteoglycans are required for gastrulation and specifically to promote Fgf signaling(Garcia-Garcia and Anderson,2003), it is plausible that FGF signaling is affected in Ext mutants. However, Fgf4 and Fgfr2 mutant mice die shortly after implantation (Arman et al.,1998; Feldman et al.,1995), and thus have an earlier phenotype than Ext-null mice. Ext mutants share some similarities with Fgfr1 and Fgf8 mutants,although they differ greatly with respect to formation of extra-embryonic structures.

The lack of properly organized extra-embryonic structures and lack of mesoderm are more reminiscent of defects caused by mutations in members of the TGFβ-superfamily. Bmp4 is required for formation of the extra-embryonic mesoderm and epiblast cell proliferation(Winnier et al., 1995), while Nodal plays a role in primitive streak formation(Conlon et al., 1994). Although inactivation of Wnt3a affects gastrulation(Yoshikawa et al., 1997), Wnt3a-null embryos have a defect that is less severe than that of Ext1 or Ext2 mutants. It seems more likely that the defects observed in the Ext mutants are the result of alterations in multiple signaling pathways.

HME is characterized by the presence of multiple bony outgrowths or lumps on the metaphyses of long bones. In this study, we show that exostoses formed in Ext2 heterozygote mice with all of the features found in human exostoses. To determine why exostoses in mice only form on the ribs is in principle no different from asking why individuals with HME do not have exostoses on all of their bones. Is there a phenotype that is common to all heterozygotes that could predispose them to exostoses formation? Indeed, we found a subtle disorganization of late proliferative chondrocytes in the growth plates of all bones of the Ext2 heterozygotes and a small but reproducible decrease in the expression zone of Ihh and Pthr1. However, Ihh expression was greatly reduced postnatally and, thus, it is possible that there was a difference in diffusion of the Ihh protein, too subtle to be detected with our methods. Alternatively, the disorganization of chondrocytes could be the result of changes in the ECM caused by the reduction of HS.

Between 1 and 2 weeks of age, cells in the center of the rib cartilage become hypertrophic and express Col10, while cells in the periphery remain less differentiated and continue to express Col2. The mechanisms that control this differentiation process have not been fully elucidated; however, it has been demonstrated that in the long bones, Ihh and Pthlh signaling are required for proper chondrocyte differentiation. The premature hypertrophic differentiation seen in the Ext2^+/– mice suggests that the Ihh/Pthr1 pathway might be involved; however, our observations do not support this idea.

The EXT1 and EXT2 genes have been suggested to function as tumor suppressor genes on the basis of studies showing loss of heterozygosity for markers around the EXT1 and EXT2 loci in both sporadic and exostosis-derived chondrosarcomas(Hecht et al., 1995). Our analyses show a reduction in the amount of HS in the chondrocyte zones of exostoses, but never a total loss, lending further support to the model that exostoses formation is the result of haploinsufficiency.

The dependence of FGF signaling pathways on HS has been well documented. In the long bones, FGF signaling shortens proliferative columns both by decreasing chondrocyte proliferation directly and by suppressing Ihh expression (Minina et al.,2002; Naski and Ornitz,1998; Ornitz and Marie,2002). Bmps antagonize the effects of Fgf signaling(Minina et al., 2002; Minina et al., 2001). Mutations in Ext1 and Ext2 decrease HS synthesis, which might result in reduced Fgf signaling, leading to defects in chondrocyte differentiation. Our preliminary data support the hypothesis that altered Fgf signaling could cause the Ext2 phenotype (B.M.Z., D.S., M. Hilton, G. Evans,Z.W., D. Wells and J.D.E., unpublished).

An intriguing issue is whether the premature differentiation of chondrocytes leads to the formation of exostoses or whether distinct processes cause the two phenomena. It has been suggested that the formation of the exostoses is the result of a defect in the bony collar that surrounds the growth plate. Alternatively, the passage of the growth plate might result in vascularization of the tissue, causing the chondrocytes in the nodule, which have an opposite orientation of differentiation, to initiate the outgrowth of an exostosis at a 90° angle from the parent bone(Fig. 10). During rib growth,the ratio of cartilage to bone in the ribs is essentially constant. The probability of the growth plate passing a nodule is relatively small, as the nodules appear only between 1 and 2 weeks, and are rare compared with the overall number of chondrocytes. Thus, the random distribution of nodules relative to the growth plate could explain the low penetrance and variable distribution of exostoses.

We thank Doris Brown for all the help in creating the Ext2knockout mice and Ernestina Schipani for providing the constitutively active Pthr1 mice. We are grateful to Helen Capili for her outstanding patience and skill in sectioning the many skeletal elements. We thankTakahiko Shimizu for his generous donation of the EXT2 polyclonal antibody and Denise Piscopo for reading and discussing the manuscript. This work was supported by funds from the National Institutes of Health (AR46238 and AG23218 to Z.W. and GM33063 to J.D.E.). D.S. was a Wyeth fellow of the Life Sciences Research Foundation. B.M.Z. was supported by a minority supplement to grant GM33063.