Evc is a positive mediator of Ihh-regulated bone growth that localises at the base of chondrocyte cilia

The human disorder EvC is a chondroectodermal dysplasia comprising skeletal and craniofacial abnormalities in conjunction with dysplastic teeth and nails. The skeletal features are short ribs, short limbs, a defect of the lateral aspect of the proximal end of the tibia leading to genu valgum deformity,fusion of the hamate and capitate bones of the wrist and postaxial polydactyly. There is progressive shortening along the proximodistal axis of the limbs with zeugopod (radius-ulna/tibia-fibula) shortening being more marked than the shortening of the stylopod (humerus/femur), and in the autopod shortening of the distal phalanges being more marked than that of the proximal phalanges. Craniofacial abnormalities include multiple labiogingival frenulae,premature eruption of teeth including the presence of teeth at birth, small conical teeth and missing primary or permanent teeth. Approximately two thirds of affected individuals also have a cardiovascular malformation. Thirty of the 52 cases in Victor McKusick's description of the disorder in the Amish died within six months of birth, owing to the respiratory problems resulting from the small thorax and cardiovascular malformations(McKusick et al., 1964). Although medical management of both respiratory and cardiovascular problems has improved dramatically, there is still a significant mortality associated with this disorder.

By positional cloning we identified two genes, EVC and EVC2, which when mutated give rise to this condition(Ruiz-Perez et al., 2000; Ruiz-Perez et al., 2003). These genes are in close proximity, with a divergent orientation; the transcription starts sites are separated by 2,624 bp in human and by only 1,647 bp in mouse. EVC and EVC2 encode novel proteins with putative transmembrane domains and regions of coiled-coiled structure, but on database interrogation show no similarity with any other proteins or motifs that give clues to their function. To determine which molecular pathways and developmental processes are perturbed in EvC, we developed anti-Evc antibodies to study subcellular localisation and generated a mouse model. As the majority of mutations identified in EVC(Tompson et al., 2007) are predicted to cause loss of function, we ablated gene function in the mouse.

The major features of EvC are shortening of the long bones and ribs. Although many signalling molecules and pathways are involved in skeletal development, Indian hedgehog (Ihh) is the master regulator(Kronenberg, 2003). In the growth plates of long bones, Ihh is secreted by the prehypertrophic chondrocytes, generating a gradient of signal that coordinates chondrocyte differentiation, chondrocyte proliferation and perichondrial development(Kronenberg, 2003). Previous work involving chimeras and knockout mice has shown that Ihh regulates chondrocyte differentiation by stimulating the synthesis of parathyroid hormone-like peptide (Pthrp; also known as Pthlh - Mouse Genome Informatics)in the periarticular region (Chung et al.,1998; St-Jacques et al.,1999; Vortkamp et al.,1996). Pthrp, which is also a paracrine signalling molecule,diffuses back towards the centre of the developing bones and acts on the parathyroid hormone (Pth)/Pthrp receptor (PPR; also known as Pthr1-Mouse Genome Informatics) to prevent proliferative columnar chondrocytes differentiating into postmitotic hypertrophic cells(Karaplis et al., 1998; Lanske et al., 1996; Weir et al., 1996). By doing this, Pthrp suppresses Ihh expression in early hypertrophic cells and an Ihh-Pthrp negative feedback loop is established to regulate the distance from the joint at which chondrocytes leave the proliferative pool and undergo hypertrophy (Chung et al.,1998; Vortkamp et al.,1996). Accordingly, Ihh-depleted mice have abnormal growth plates with undetectable levels of Pthrp expression in which chondrocytes undergo hypertrophy closer to the ends of bones and lack the characteristic stacked columns of proliferating chondrocytes. In addition, Ihh mutants display Pthrp-independent defects, including dramatic reduction in chondrocyte proliferation and failure of cortical bone formation(Razzaque et al., 2005; St-Jacques et al., 1999).

In vertebrates, there are three hedgehog (Hh) signalling molecules [sonic hedgehog (Shh), Ihh and desert hedgehog (Dhh)] whose effects are mediated by three different Gli transcription factors (Gli1-3). The essence of Hh signalling is that binding of Hh molecules to their receptor, patched (Ptch),releases inhibition of smoothened (Smo), the activator of the Hh pathway. Smo prevents phosphorylation and cleavage of full-length Gli3 (Gli3-190) to the transcriptional repressor Gli3R (Gli3-83) and promotes Gli transcriptional activation (Hooper and Scott,2005).

It has been shown that vertebrate Hh signalling is mediated through primary cilia. Cilia are essential for Shh-mediated patterning in early limb bud and neural tube development (Huangfu et al.,2003; Liu et al.,2005). Important components of the pathway, suppressor of fused(Sufu) and the Gli proteins, localise to primary cilia and Smo translocates to cilia upon pathway activation (Corbit et al., 2005; Haycraft et al.,2005). The recent generation of a Prx1Cre (also known as Prrx1-cre) conditional allele of Ift88 has enabled study of cilia-dependent processes in mesenchyme-derived tissues of the limb and confirms, as anticipated, that cilia are also required for Ihh signalling(Haycraft et al., 2007). In mutants with disrupted anterograde or retrograde intraflagellar transport,Gli3 processing is abnormal with an increase in full-length Gli3 and decreased levels of Gli3R (Haycraft et al.,2005; Huangfu and Anderson,2005; Liu et al.,2005; May et al.,2005). Despite the increase in full-length Gli3 in intraflagellar transport (IFT) mutants, all have low expression levels of the readouts of the pathway, Gli1 and Ptch1(Huangfu et al., 2003; Liu et al., 2005), proving that inhibition of Gli processing is not sufficient to produce transcriptional activation and that additional IFT-dependent events are required to promote Gli activator functioning.

Here we report that Evc localises at the base of primary cilia and demonstrate that defective transduction of Ihh signalling underlies the skeletal phenotype associated with Evc mutation. Our findings show that Evc acts as a positive mediator of Hh signalling downstream of Smo, but that it is not essential for Gli processing.

The Evc targeting vector was engineered to simultaneously create an in-frame fusion of β-galactosidase (β-gal) to the first amino acid of Evc and ablate gene function by deleting the downstream remaining Evc exon 1 sequence (Fig. 1A). The 5′ and 3′ DNA arms of homologous sequence were isolated from a 129/SvJ BAC clone (Genome Systems) and the lacZcassette from plasmid pSDKLacZpA. The linearised targeting vector was electroporated into AB1 ES cells and homologous recombinant clones injected into C57BL/6 blastocysts. Male chimeras were bred to C57BL/6 female mice and germline transmission confirmed by PCR and Southern blot(Fig. 1B). Mice were PCR genotyped by simultaneous amplification of Evc exon 1 (170 bp) and lacZ (475 bp) (Fig. 1C). Absence of Evc transcript in Evc^-/- mice was confirmed by RT-PCR(Fig. 1D). Given the close proximity of Evc2 to Evc, we compared Evc2expression in chondrocytes from two E17.5 mutant embryos with two wild-type littermates by quantitative PCR and found similar expression (data not shown). As Evc 3′ untranslated exons overlap with Crmp13′ exons, we checked that removal of the Evc transcript did not affect transcription of Crmp1(Fig. 1D). Mice were maintained by crossing Evc^+/- males to F1 females from C57Bl/6×129 crosses.

Amino acids 459-999 of the mouse Evc protein (GenBank CAB76567) were expressed with a 6×His tag in E. coli and purified by Ni²⁺ chelation chromatography (Novagen). The protein (150 μg)was used to immunise sheep (Diagnostic Scotland). Total IgGs were prepared from final serum (Protein G HiTrap, Amersham) and anti-Evc antibodies purified by affinity to the antigen conjugated to AminoLink Coupling Gel (Pierce).

Tissue was frozen in OCT and cryosections (10 μm) dried onto charged slides. mIMCD3 cells were cultured using standard conditions. Tissue sections and cells were fixed in 4% (w/v) paraformaldehyde (PFA) in PBS at 4°C. Primary antibodies were: sheep polyclonal anti-Evc (5 μg/ml); mouse monoclonal anti-γ-tubulin (1:800, GTU-88, Sigma) and anti-acetylated tubulin (1:2000, 6-11B-1, Sigma). Secondary antibodies were: donkey anti-sheep Alexa Fluor 594 (Molecular Probes) and goat anti-mouse FITC (Sigma). Samples were mounted in Vectashield with DAPI (Vector) and images captured on an Axioplan 2 fluorescence microscope (Zeiss).

Evc ^+/-, Evc ^-/- and wild-type control embryos were stained for β-gal activity with X-Gal as previously described (Hogan et al., 1994; Schatz et al., 2005). E15.5 embryo heads and newborn hindlimbs were stained with X-Gal and embedded in a gelatine/albumin mix for vibratome sectioning. Sections (100 μm) were mounted in 50% (v/v) glycerol in PBS for photographing. Skeletal preparations of P6 mice were conducted as previously described(Kessel and Gruss, 1991). Digital radiographs of euthanised P1 and P18 mice were collected at X-ray exposures of 30 kV for 5 seconds and 45 kV for 10 seconds, respectively. Bones measurements were compared by ANOVA.

Embryo limbs were fixed in 4% PFA in PBS and paraffin embedded. P16 limbs were decalcified in 4% formaldehyde:nitric acid:H₂O (1:1:8) for a week prior to embedding. Sections (7 μm) were stained with Haematoxylin-Eosin or by trichromic staining (0.13% Light Green SF Yellowish,0.16% Orange G, 0.2% fuchsin acid). Standard von-Kossa method was used for detection of mineralisation. E17.5 pregnant females were sacrificed 5 hours after receiving an intraperitoneal injection of BrdU cell proliferation labelling reagent (GE Healthcare). Proliferating cells were detected on 7μm paraffin sections by immunocytochemistry (BrdU Staining Kit, Zymed). BrdU-positive and BrdU-negative nuclei were counted in columnar chondrocytes of Evc^-/- and wild-type mice (12 and seven sections,respectively). The percentage BrdU-positive cells was calculated for each section and data compared by ANOVA.

Hindlimbs from E16.5 littermates were fixed overnight in 4% PFA in PBS,decalcified in 14% (w/v) EDTA in PBS and paraffin embedded. Sections (10μm) were hybridised using 2×10⁷ cpm/ml ³⁵S-labelled antisense riboprobes, except for Col2a1 and Col10a1, which were digoxigenin labelled (Roche). Hybridisation steps were as in Lescher et al. (Lescher et al.,1998). After stringency washes, dehydrated sections were coated with nuclear emulsion (K5, Ilford) and exposed for 2-4 weeks at 4°C. Signal was developed in Kodak D-19 developer and fixer and sections counterstained with nuclear Fast Red. Hybridisation probes were obtained by request, except for Gli1 (AB025922) and mouse PPR(NM_011199), which were RT-PCR amplified. Bright-field images were obtained on a Nikon DS-5L1 digital camera and dark-fields produced in Photoshop(Adobe).

Mouse embryonic fibroblasts (MEFs) were established from Evc^-/- and control E14.5 embryos(Todaro and Green, 1963). Chondrocytes were isolated from tibial epiphysial cartilage dissected from two E17.5 Evc^-/- and two Evc^+/+littermates as described (Shingleton et al., 2000) with the following modifications to enzyme incubation times: hyaluronidase, 5 minutes; trypsin, 10 minutes; collagenase, 5 hours. Chondrocytes were incubated in DMEM containing 10% fetal calf serum at 2×10⁶ cells in 6-well plates for a maximum of 7 days following dissection. For Hh pathway induction experiments,1×10⁵ cells were grown overnight in 12-well plates and treated for 72 hours with either 2 μM purmorphamine (10 mM stock in DMSO,Calbiochem) or DMSO alone. Simultaneous RT-PCR amplification of Ptch1(nt 1944-2303; NM_008957) and Hprt (nt 108-294; NM_013556) in MEF cDNA was performed for 22 cycles under standard PCR conditions. Assays were carried out twice on cultures derived from three Evc^-/-,two Evc^+/+ and one Evc^+/- embryo. Ratios of Ptch1 to Hprt band intensity were determined for each culture before and after treatment and compared by ANOVA. For the quantitative PCR experiments, chondrocyte cDNA samples were loaded onto Taqman Low Density Array (TLDA) microfluidic cards (ABI, Forster City, CA). The following primers and probes were used: Gli1 (ABI assay ID. Mm00494645_m1), Ptch1 (ABI assay ID. Mm00436026_m1), Evc(ABI assay ID. Mm00469587_m1), Evc2 (ABI assay ID. Mm00507589_m1) andβ-actin (Actb; ABI assay ID. Mm00607939_s1). Relative quantification of genes was performed using the ABI Prism 7900HT Sequence Detection System and expressed as arbitrary units as determined by 2^{-(Ct gene-Ct β-actin)}.

Protein lysate from E14.5 limbs was prepared in RIPA buffer supplemented with one Mini Complete Protease Inhibitor tablet (Roche) and 0.1M PMSF. Total protein was electrophoresed on 7.5% SDS-PAGE gels and transferred to Hybond-C membrane (Amersham). Gli3 was detected using rabbit anti-Gli3 (1:200, a gift of B. Wang, Cornell University, NY) followed by goat anti-rabbit peroxidase(1:5000, Jackson ImmunoResearch) and SuperSignal West Dura Extended Duration Substrate (Pierce). Protein loading was assessed using rabbit anti-β-catenin (BD Biosciences).

Using the Evc antibodies and γ-tubulin and acetylated tubulin as markers of centrioles and cilia axoneme, respectively, we found that Evc localises at the distal end of the maternal centriole (the basal body), at the base of the axoneme (Fig. 2A,B), suggestive of localisation to the transition zone. Specificity of the Evc immunostaining was demonstrated by absence of signal in chondrocytes from Evc^-/- mice(Fig. 2C). Cilia were present in Evc^-/- chondrocytes, demonstrating that Evc is not essential for ciliogenesis (Fig. 2C).

During construction of the Evc targeting vector, we inserted a lacZ reporter cassette directly under control of the Evcpromoter to determine the Evc spatiotemporal expression pattern during development. By whole-mount X-Gal staining of Evc^+/- mouse embryos, we first observed lacZexpression at E11.5 in the orofacial region in the lateral nasal process,maxillary and mandibular processes (Fig. 3A-D), followed at E12.5 by expression in the mesenchymal condensations of the skeletal system at the time they initiate chondrocyte differentiation (Fig. 3C). At E15.5, vibratome sectioning of the orofacial region demonstrated β-gal activity in the cartilages of the nasal septum and nasal capsule and those enclosing the Jacobson organ (Fig. 3E). Strong β-gal activity was also evident in the upper and lower lip mesenchyme and in the mesenchyme outlining the growing bones of the maxilla and mandible (Fig. 3H). By E15.5, lacZ expression was general to all the cartilaginous components of the skeleton, including the chondrocranium, the vertebrae, the rib cage and the axial skeleton (Fig. 3F,G). In the developmentally more advanced elements of the axial skeleton, lacZ expression was narrowed to the chondrocytes of the epiphysis and the perichondrium and absent in the central bone metaphyses. However, in the phalanges in which chondrocytes have not yet undergone hypertrophy, the entire element was lacZ-positive. In order to determine which type of chondrocytes express Evc, we studied newborn skeletal growth plates in which chondrocyte layers are more easily distinguishable. Vibratome sectioning of stained newborn skeletal growth plates showed β-gal activity in the perichondrium and in resting and proliferating chondrocytes (Fig. 3M,N), but no activity in the prehypertrophic and hypertrophic chondrocytes. Finally, we observed β-gal activity in the cranial sutures,suggesting a role in both endochondral and intramembranous ossification. We detected lacZ expression in nails(Fig. 3L), the dermal papilla of the vibrissae (Fig. 3J) and in the mesenchyme surrounding the developing tooth buds(Fig. 3I), in keeping with the ectodermal phenotype seen in EvC patients.

Evc^+/- mice have no discernable defects. Offspring from heterozygous Evc^+/- crosses were genotyped at E18.5 just prior to birth and there was no deviation from mendelian ratios byχ ² tests and thus no prenatal loss of Evc^-/- embryos was detected. However, approximately half of the Evc^-/- offspring were missing 2 days after birth. These neonatal losses are unlikely to be due to cardiovascular malformation because no overt cardiovascular malformations were observed on histological examination of 14 Evc^-/- mice (eight at E14.5, one at E15.5, four at E18 and one newborn) (results not shown). Surviving Evc^-/- mutants were unable to feed on a normal diet, but were able to survive to adulthood when supplied with soft, well-hydrated food;they did not breed. The phenotype of Evc^-/- mice recapitulates the human disorder with short ribs, short limbs and dental anomalies, although they do not have polydactyly(Fig. 4A,B). Preliminary examination of the teeth revealed variability in the abnormalities of the incisors, including absence of the upper incisors and a single upper incisor. We frequently observed a smaller first molar, in comparison with wild-type dentition (Fig. 4C).

We focussed on analysis of the skeletal abnormalities and commenced by studying radiographs and skeleton preparations. At birth, there was no difference between the head-to-tail length of wild-type and mutant mice, but the rib cage was narrow and the radius, ulna, femur and tibia were shorter in the mutant mice (see Table S1 in the supplementary material). At P18,radiographs revealed that the mutant mice are smaller than littermates, that the rib cage is small and that shortening of the radius, ulna and tibia is more pronounced than for the humerus and femur, indicating the same pattern of bone shortening in the mouse mutant as in patients with EvC(Fig. 4A). These observations were confirmed by measurements taken from the radiographs, with decreased radius/humerus (P=0.00065), tibia/femur (P=0.0045) and rib cage/body length (P=0.00853) ratios at P18 (see Table S1 in the supplementary material). Alcian Blue-Alizarin Red skeletal staining also demonstrated this pattern of limb shortening and in addition showed an irregular margin between bone and cartilage in the basiocciput and costochondral junctions (Fig. 4B), as well as premature mineralisation of some of the pedicles between the vertebral bodies and vertebral laminae(Fig. 4B, arrows).

Histological analysis of embryonic growth plates revealed shorter proliferative and hypertrophic chondrocyte layers in the epiphyses of the long bones with hypertrophy of chondrocytes occurring closer to the articular region than in the wild-type controls. Chondrocytes in the proliferative zone in Evc^-/- mice had the characteristic flattened shape(Fig. 5A). We consistently observed an abnormality in the shape of the upper end of the tibia(Fig. 5A,C), comparable to that seen in EvC patients. We studied mineralisation by von-Kossa staining and detected delayed formation of the periosteum adjacent to the prehypertrophic and hypertrophic chondrocytes (Fig. 5B, red arrows). Similarly, trichromic staining in older mice demonstrated delayed formation of the secondary ossification centres(Fig. 5C).

The phenotypic observations of epiphyseal shortening due to chondrocytes hypertrophying nearer to the articular region, defective perichondrium to periosteum induction and mineralisation of synchondroses are all factors compatible with impaired Ihh signalling in mice(Hilton et al., 2005; Koziel et al., 2005; Razzaque et al., 2005; St-Jacques et al., 1999). In addition, Evc localises to the base of cilia, structures that mediate Hh signalling (Huangfu and Anderson,2005). We therefore decided to determine whether defective Hh signalling underlies the Evc^-/- skeletal phenotype and assessed expression of Ihh signalling molecules in the growth plates of Evc^-/- mice by radioactive in situ hybridisation. We found Ihh expression in prehypertrophic chondrocytes of Evc^-/- mice to be normal. However, expression of the Ihh downstream targets Ptch1 and Gli1 was drastically reduced in the adjacent perichondrium and proliferating chondrocytes, providing evidence of defective Ihh signalling (Fig. 6A). To corroborate this, we derived mouse embryonic fibroblasts(MEFs) from Evc^-/- and control littermates and assessed induction of Ptch1 expression under exogenous stimulation of the pathway with purmorphamine (Sinha and Chen, 2006; Wu et al.,2004). RT-PCR results from MEF cultures showed that the Ptch1 response of purmorphamine-treated Evc^-/-MEFs was diminished compared with wild-type MEFs(Fig. 6B). The mean ratio of Ptch1:Hprt expression was significantly different between Evc^-/- and wild-type fibroblast cultures grown in the presence of purmorphamine (ANOVA, P=0.003). In addition, we assessed Ptch1 and Gli1 transcript levels following purmorphamine treatment by quantitative RT-PCR in chondrocytes derived from Evc^-/- and wild-type littermates. The Ptch1(Fig. 6C) and Gli1(Fig. 6D) response of Evc^-/- purmorphamine-treated chondrocytes was greatly diminished compared with control cells. Since purmorphamine is a Hh agonist that targets Smo, these studies confirm abnormal Hh signalling in Evcmutants and demonstrate an intracellular defect downstream of Smo.

The Evc expression pattern we observed in the growth plate(Fig. 3M,N) is similar to that of Ihh target genes. To investigate whether Evc could be under Ihh regulation, we assessed Evc expression in wild-type chondrocytes by quantitative PCR. Exogenous stimulation of the pathway with purmorphamine did not alter Evc transcript levels (data not shown).

Cilia and IFT proteins have been shown to be required for proteolytic processing of Gli3. Gli3R is dramatically decreased and full-length Gli3 dramatically increased in IFT mutant embryos at E9.5-11.5(Haycraft et al., 2005; Liu et al., 2005). Western blot analysis of protein extracted from E10.5 Evc^-/- and control embryos revealed normal levels of both full-length Gli3 and Gli3R(results not shown). However, as there is no detectable β-gal activity in Evc^+/- or Evc^-/- embryos at this point in development, we tested whether Evc is required for Gli3 processing in limb extracts from E14.5 embryos. Levels of Gli3, particularly full-length Gli3,have previously been shown to be very low in the E14.5 limb(Hilton et al., 2005). We did not observe a change in the levels of either full-length Gli3 or Gli3R in E14.5 Evc^-/- limb extracts when compared with littermate controls (Fig. 6E).

One of the features of Ihh-knockout mice is a striking proliferation deficiency known to be caused by the increase in Gli3 repression(Hilton et al., 2005; Koziel et al., 2005). To test whether proliferation is affected in Evc mutants, we undertook in vivo BrdU labelling experiments followed by immunohistochemistry. No significant difference was observed in the percentage of BrdU-positive chondrocytes between Evc^-/- and wild-type mice in the proximal tibia at E17.5 (Fig. 7A). As proliferation and apoptosis (data not shown) do not seem to be affected, we hypothesised that the growth plate shortening in Evc^-/- mice is likely to be due to premature chondrocyte differentiation.

To assess differentiation, we first analyzed expression of Col2a1and Col10a1 and observed that Evc^-/- chondrocytes undergo normal endochondral ossification from Col2a1-positive resting-proliferating chondrocytes to Col10a1-expressing hypertrophic chondrocytes (Fig. 7B). Since Pthrp is the principal molecule controlling chondrocyte maturation(Kronenberg, 2003), we went on to study Pthrp signalling and found that Pthrp expression in articular chondrocytes, which is normally maintained by Ihh signalling, was decreased (Fig. 7B). We observed normal expression of the Pth/Pthrp receptor, PPR, in the prehypertrophic zone, but expression in the osteoprogenitor cells of the perichondrium (Naski et al.,1998) (Fig. 7B,arrows) was reduced in the Evc^-/- growth plate(Fig. 7B), in keeping with the mineralisation defect in the perichondrium evidenced by the von-Kossa staining.

To further clarify whether Evc^-/- growth plate shortening is due to premature differentiation from distal to columnar chondrocytes, or from columnar to hypertrophic chondrocytes, or both, we supplemented the histological analysis by studying Fgfr1 expression as a marker of distal and hypertrophic chondrocytes and Fgfr3expression as a marker of columnar and early hypertrophic cells. The hybridisation results showed that there was no shortening of the distal Fgfr1 expression domain in Evc^-/- mice, but marked shortening of the zone expressing Fgfr3(Fig. 7C), from which we conclude that differentiation from distal to columnar chondrocyte, which is influenced by Gli3R (Koziel et al.,2005), is proceeding normally and hypertrophic differentiation is occurring prematurely.

In this study, we demonstrate that defective Ihh signalling underlies the skeletal features in mice lacking Evc. By in situ hybridisation, we have shown that there is decreased expression of the Ihh targets Gli1, Ptch1 and Pthrp in the growth plates of Evc^-/- mice,although expression of Ihh itself is normal. Reduced target gene expression could result from defective production or diffusion of the signal or from defective cell response upon Ihh arrival. lacZ expression analysis of the growth plate helped to distinguish between these possibilities as X-Gal predominantly stained proliferating and distal chondrocytes and the perichondrium, cells that respond to Ihh signal, but left the region of Ihh-secreting prehypertrophic cells unstained. From this we deduced that Evc was likely to be involved in transduction of the response to Ihh signal. Confirmation came from the in vitro experiments in which we supplemented MEFs and chondrocytes with the Smo agonist purmorphamine. These experiments verified, by an independent method, that Evc^-/- cells exhibit diminished upregulation of Ptch1 and Gli1 in response to signal, and demonstrated that the defect lies downstream of Smo. Smo blocks proteolytic processing of Gli3 and also plays a role in conferring activator status to full-length Gli3. Western blot analysis and indirect evidence (normal proliferation and normal differentiation from distal to columnar chondrocytes) demonstrate that Gli3R is not increased, indicating that Gli3 processing occurs normally in Evc-depleted mice.

The phenotype in Ihh^-/- mice is far more severe than that in mice lacking Evc. This is because the absence of Ihh is associated not only with removal of Gli activator functions, but also with increased Gli3 repression, with some aspects of the Ihh^-/- phenotype resulting from decreased Gli activation and some being due to increased Gli3R repression. Analysis of Ihh^-/-,Gli3^-/- double knockout mice, in which Gli3 repression is abolished in an Ihh^-/- background, has shown which features of the Ihh^-/- phenotype result from elevated Gli3R(Hilton et al., 2005; Koziel et al., 2005). The proliferation deficiency in the growth plate of Ihh^-/-mice is fully rescued in the double knockouts and is therefore attributable to increased Gli3 repression. Gli3R levels are normal in Evc^-/- mice and, consistent with this, chondrocyte proliferation in Evc^-/- mice is normal. By contrast, in the growth plates of the Ihh^-/-,Gli3^-/- mice,there is no Gli1 expression and there is only partial recovery of the reduced levels of Pthrp and Ptch1 expression. Similarly, the bone collar defect is not completely rescued. Lack of function of Gli activators is the likely cause for these features of Ihh^-/- mice, which are not restored in the double knockouts (Hilton et al.,2005; Koziel et al.,2005). In Evc^-/- mice, there is low expression of Gli1, Ptch1 and Pthrp in the growth plate as observed in the double knockouts. Likewise, both Evc^-/- mice and the double knockouts have defective bone collar formation as manifest by decreased PPR expression in the perichondrium and in the von-Kossa staining. We postulate that Evc^-/- mice are likely to be deficient in the same Ihh-dependent Gli activator functions that are absent in the Ihh^-/-,Gli3^-/- mice.

Upregulation of Ptch1 and Gli1 expression has previously been observed in cells outlining the cartilage growth plate when Ptch1 expression is specifically diminished in chondrocytes(Long et al., 2004). The extended range of signalling has been attributed to the absence of Ptch-mediated ligand sequestration (Chen and Struhl, 1996; Lewis et al., 2001). We observed similar expression of Ptch1 and Gli1 in the weakly Evc-expressing cells surrounding the articular surface in Evc^-/- mice(Fig. 6A). This indicates that in mice lacking Evc, Hh signalling is most markedly perturbed in those cells that would normally have the highest levels of Evc expression.

Growth plate, and hence bone, shortening could result from one or more of the following: decreased proliferation, increased apoptosis, premature differentiation either of distal to columnar chondrocyte or from proliferating to hypertrophic chondrocytes. In Evc^-/- mice,proliferation appears normal, consistent with the normal Gli3R levels observed in western blotting, and apoptosis also appears normal. Differentiation from distal to columnar chondrocyte, which is regulated by Gli3R levels(Koziel et al., 2005), is also normal in Evc^-/- mice. As shown by the considerable reduction in size of the Fgfr3 expression domain, the Evc^-/- growth plate shortening is explained by the shorter region of proliferating columns of chondrocytes, which in turn is due to the premature onset of hypertrophic differentiation. Hypertrophic differentiation is regulated by Pthrp, itself a target of Gli activation(Koziel et al., 2005). Thus,we can conclude that premature hypertrophic differentiation in Evc^-/- chondrocytes is due to decreased Pthrpexpression secondary to defective Hh signalling.

Defects in IFT cause embryonic lethality and it is only with the advent of mice carrying a conditional Ift88 allele and the Prx1Cretransgene that it has been possible to study endochondral bone formation in IFT mutants (Haycraft et al.,2007). Evc^-/- and the Prx1Cre Ift88conditional mice both have decreased activation of Ihh targets Ptch1and Gli1 in the growth plates of long bones and delayed bone collar formation, features that are not fully rescued by Gli3R derepression in the Ihh^-/-,Gli3^-/- mice and that are expected to be associated with defective Gli activation. Evc mutants differ from the IFT conditionals with respect to Gli3R, which is decreased in the IFT nulls. Chondrogenic clumps of cells surrounding the perichondrium are described in Prx1Cre Ift88 conditional mice, a feature that we have not observed in the Evc mutant and not reported in the Ihh^-/-,Gli3^-/- mice. An additional difference between IFT and Evc mutants relates to Ihh expression, which appears normal in Evc^-/- mice but is reduced in the Prx1Cre conditional allele of Ift88.

Evc-depleted mice phenocopy most of the EvC features and therefore represent a good model for this syndrome. They reproduce the skeletal and dental anomalies of the condition and the perinatal lethality. Similar neonatal losses in Ihh^-/- and Pthrp^-/-mice have been attributed to respiratory failure secondary to the small rib cage (Karaplis et al., 1998; St-Jacques et al., 1999). The lack of polydactyly in Evc^-/- mice came as a surprise as EvC individuals invariably have postaxial polydactyly of the hands. Neither Gli1 nor Gli2 mutant mice manifest limb patterning defects(Park et al., 2000). However,the spontaneous and targeted Gli3 mutants have polydactyly, and polydactyly also occurs in human disorders caused by Gli3 mutations such Pallister-Hall syndrome and Greig cephalopolysyndactyly syndrome(Ehlen et al., 2006). In many of the mutants with polydactyly, including the IFT mutants, both Gli3 activator function and Gli3 repression are perturbed. It is apparent that the balance between Gli3 activation and Gli3 repression across the limb bud is crucial in specifying the number of digits(Ahn and Joyner, 2004; Wang et al., 2007). The fact that we did not detect lacZ expression in the limb buds of Evc^-/- mice when patterning is being established indicates that Evc expression in the limb buds is low compared with that in bone anlagen, and this might explain the absence of polydactyly in Evc^-/- mice. Regardless, as anteroposterior patterning in the limb bud is regulated by Shh rather than Ihh, the polydactyly seen in EvC individuals suggests that EVC may also play a crucial role in transduction of Shh signalling during human limb bud morphogenesis. The occasional observation of a single central incisor in Evc^-/- mice is also suggestive of defective Shh signalling in the mouse given that this is a feature seen in association with holoprosencephaly due to SHH mutations. Similarly, temperospatial expression of lacZ in tissues where Shh is expressed, such as the orofacial region and vibrissae (Bitgood and McMahon, 1995), is consistent with Evc participation in the transduction of the Shh signal.

We have shown that Evc localises at the distal end of the basal body and the base of the axoneme and that it is integral to Ihh signalling in developing bones. Intracellular transduction of Hh signal is not yet fully understood. There have been surprises - knockdowns of two of the components of the pathway first identified in Drosophila, Sufu and Fu, indicate that their roles differ between Drosophila and vertebrates(Svard et al., 2006; Varjosalo et al., 2006). Evc adds to the differences between Drosophila and vertebrates, as there are no recognisable Evc homologues in organisms other than vertebrates. In the last few years, new players have been identified acting downstream of Smo in mice [Rab23 (Eggenschwiler et al.,2006) and tectonic (Reiter and Skarnes, 2006)], in zebrafish [Iguana (also known as Dzip1 - ZFIN)(Sekimizu et al., 2004; Wolff et al., 2004)] and in chick [Talpid (Davey et al.,2006)], the precise role of which remain to be elucidated. Here we show that Evc is a novel basal body component of Hh signalling indispensable for normal endochondral growth, acting downstream of Smo to facilitate transcription of the Ihh-regulated genes.

We thank Paul Cairns and Weiping Li for help with blastocyst injections and embryo transfers; Dr Ralf Kist for the vectors used in the construction of the targeting vector; Dr Erik E. Turner for plasmid pSDKpA; Dr Andreas Schedl for the AB1 ES cells; Dr Fanxin Long, Dr Matthew Hilton and Dr Andrew McMahon for the Ihh, Ptch1 and Pthrp probes; Dr Bjorn Olsen for Col2a1 and Col10a1 probes; Dr Andrea Vortkamp for Fgfr1, Fgfr3 probes; and Dr Baolin Wang for the Gli3 antibody. This work was supported by the MRC, Spanish Ministerio de Educacion y Ciencia, BBSRC, BHF and the Nuffield Foundation's Oliver Bird Rheumatism Programme. V.L.R.-P. acknowledges Dr Santiago Rodriguez de Córdoba for support.