Heparan sulphate proteoglycans have been implicated in the binding and presentation of several growth factors to their receptors, thereby regulating cellular growth and differentiation. To investigate the role of heparan sulphate proteoglycans in mouse spinal neurulation, we administered chlorate, a competitive inhibitor of glycosaminoglycan sulphation, to cultured E8.5 embryos. Treated embryos exhibit accelerated posterior neuropore closure, accompanied by suppression of neuroepithelial bending at the median hinge point and accentuated bending at the paired dorsolateral hinge points of the posterior neuropore. These effects appear specific, as they can be prevented by addition of heparan sulphate to the culture medium, whereas heparitinase-treated heparan sulphate and chondroitin sulphate are ineffective. Both N- and O-sulphate groups appear to be necessary for the action of heparan sulphate. In situ hybridisation analysis demonstrates a normal distribution of sonic hedgehog mRNA in chlorate-treated embryos. By contrast, patched 1 transcripts are abnormally abundant in the notochord, and diminished in the overlying neuroepithelium, suggesting that sonic hedgehog signalling from the notochord may be perturbed by inhibition of heparan sulphation. Together, these results demonstrate a regulatory role for heparan sulphate in mouse spinal neurulation.
Neurulation is a fundamental morphogenetic event that culminates in the elevation, apposition and fusion of the neural folds in the dorsal midline (Copp et al., 1990; Smith and Schoenwolf, 1997). Failure of neural tube closure leads to the clinically important birth defects spina bifida and anencephaly (Copp and Bernfield, 1994). Although the molecular mechanisms that control neural tube closure are poorly understood, a role for extracellular signalling mechanisms is emerging. For example, mice that lack the function of fibroblast growth factor (FGF) receptor 1, which transduces the signalling of several FGFs, exhibit defective closure of the caudal neural folds (Deng et al., 1997). A similar defect arises in mice homozygous for the curly tail (ct) mutation, in which reduced expression of the Wnt5a gene has been identified (Gofflot et al., 1998). Moreover, recent studies demonstrate a critical role for sonic hedgehog (Shh) signalling in the regulation of neural plate bending in the spinal region of the mouse embryo (Ybot-Gonzalez et al., 2002). Hence, three distinct signalling pathways have been implicated in mouse spinal neurulation. A common feature of these signalling pathways is the role played by heparan sulphate proteoglycans (HSPGs) in the presentation of ligands to responding cells. The present study examines the role of HSPGs in mouse spinal neurulation.
Heparan sulphate is a glycosaminoglycan (GAG) of repeating disaccharide subunits, comprising glucosamine and glucuronic/iduronic acid, covalently linked to a core protein backbone (Bernfield et al., 1999). A family of HSPGs exists, some of which are localised to the cell surface through a transmembrane core protein (syndecans) or by a glycosylphosphatidylinositol linkage (glypicans), while others form part of the extracellular matrix (perlecan and agrin). HSPGs bind to and regulate the activity of many important signalling molecules. For example, analysis of the Drosophila tout-velu mutant and the Ext1-deficient mouse, both of which lack heparan sulphate co-polymerase, shows that HSPGs are essential for the function of several members of the hedgehog family of signalling proteins (Bellaiche et al., 1998; Lin et al., 2000; The et al., 1999). Similarly, examination of the Drosophila sugarless and sulfateless mutants provides evidence for a role of HSPGs in fibroblast growth factor (FGF) and Wnt signalling (Häcker et al., 1997; Lin et al., 1999). Furthermore, tissue-specific developmental defects are seen after modification of the number and position of the sulphate groups on heparan sulphate. For example, mice homozygous for the Hs2st gene trap mutation are unable to add 2-O sulphate groups to heparan sulphate and exhibit renal, eye and skeletal defects (Bullock et al., 1998), while targeted disruption of the Ndst1 gene, which encodes N-deacetylase N-sulphotransferase 1, leads to pulmonary hypoplasia, atelectasis and respiratory distress syndrome (Ringvall et al., 2000).
HSPGs are found in the basement membrane of the neuroepithelium of the closing neural tube, as well as in the adjacent tissues of the posterior neuropore region (Copp and Bernfield, 1988; O’Shea, 1987). Moreover, degradation of heparan sulphate by heparitinase disrupts cranial neurulation in cultured rat embryos, leading to the development of exencephaly (Tuckett and Morriss-Kay, 1989). The purpose of the present study was to determine whether HSPGs, and specifically the sulphation pattern of heparan sulphate, could also regulate spinal neurulation. We cultured E8.5 mouse embryos in the presence of an inhibitor of sulphation, chlorate, to examine the importance of the sulphate group in HSPG function. The sulphate donor for GAG sulphation is 3′-phosphoadenosine 5′-phosphosulphate (PAPS), which is synthesised by PAPS synthetase. Chlorate acts as a sulphate analogue and competes with sulphate in PAPS synthesis, resulting in inhibition of GAG sulphation (Conrad, 1998; Greve et al., 1988). We find that chlorate acts to hasten the closure of the spinal neural tube, and we present evidence suggesting that this effect may be mediated, in part, via inhibition of Shh signalling.
MATERIALS AND METHODS
Culture of intact embryos
Random-bred CD1 mice were mated overnight and the day of finding a copulation plug was designated embryonic day (E) 0.5. Embryos at E8.5 were explanted into Dulbecco’s modified Eagle’s medium (DMEM) containing 10% foetal calf serum and cultured in undiluted rat serum, in a roller incubator maintained at 38°C (Copp et al., 2000). Cultures were stabilised over a 3 hour period, then chlorate, β-D-xyloside (p-nitrophenyl-β-D-galactopyranoside), sulphate, bovine kidney heparan sulphate, shark cartilage chondroitin-4-sulphate or chondroitin-6-sulphate (all from Sigma), or chemically modified bovine kidney heparan sulphate (Seikagaku) were added in various combinations and culture was continued for 14 hours. Heparan sulphate was degraded by incubation with 0.1 U/ml heparitinase (Seikagaku) at 37°C overnight, and the enzyme was then heat inactivated.
After culture, embryos were selected for further analysis only if they exhibited vigorous blood flow in the yolk sac circulation, and if the heart beat was regular and above 100 per minute. They were dissected from the yolk sac and amnion, inspected for closure of the cranial and spinal neural tube, then scored using the Morphological Scoring System of Brown and Fabro (Brown and Fabro, 1981). Crown-rump length, head length and posterior neuropore length (the distance between the rostral end of the posterior neuropore and the tip of the tail bud) were measured using an eyepiece graticule attached to a Zeiss SV6 stereomicroscope. Some embryos were sonicated and used to determine total protein content using the BCA Protein Assay Kit (Pierce and Warriner), whereas others were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) before embedding in paraffin wax and sectioning at 6 μm, transversely through the posterior neuropore region. Sections were stained with Haematoxylin and Eosin or used for immunolocalisation of chondroitin sulphate and heparan sulphate. Other embryos were processed for whole-mount mRNA in situ hybridisation for Shh and patched 1 (Ptch).
Mouse monoclonal antibodies 10E4 (Seikagaku) and CS-56 (Sigma) were used to detect heparan sulphate and chondroitin sulphate respectively (Avnur and Geiger, 1984; David et al., 1992). A monoclonal mouse IgM antibody (DAKO) raised against Aspergillus niger glucose oxidase (an enzyme neither present nor inducible in mammalian tissues) was used as a negative control. Rehydrated tissue sections were incubated with 3% hydrogen peroxide and the blocking solutions from the HistoMouse-SP kit (Zymed) to reduce nonspecific background staining, then incubated with 10 μg/ml primary antibody for an hour at room temperature. The signal was amplified using a biotinylated anti-mouse secondary antibody and a streptavidin-horseradish peroxidase conjugate and detected using 3,3′-diaminobenzidine tetrahydrochloride.
Labelling and analysis of GAGs
GAGs were labelled using carrier-free [35S]sulphate and analysed by anion exchange chromatography (Solursh and Morriss, 1977; Copp and Bernfield, 1988). Embryos were stabilised in culture for 3 hours, then carrier-free [35S]sulphate (Amersham Pharmacia) was added to the culture medium to a final concentration of 100 μCi/ml (final concentration of sulphate: 0.75 μM) and culture was continued for a further 5 hours. After removal of yolk sac and amnion, embryos were washed using ice-cold DMEM and PBS in order to remove unincorporated label. Embryos were stored in TE (50 mM Tris, pH 7.5; 2 mM EDTA) at –70°C.
Labelled GAGs were extracted by sonicating embryos in TE on ice. An aliquot was removed for scintillation counting and protein quantification, then 50 μg each of carrier hyaluronan and chondroitin-6-sulphate (Sigma) were added and GAGs were precipitated overnight at –20°C using three volumes of 1.3% potassium acetate in 95% ethanol. GAGs were pelleted by centrifugation at 14,000 g for 15 minutes, then re-suspended in de-ionised water. Pronase (Protease XIV, Sigma) was added (2 mg) and samples were incubated at 55°C overnight to degrade proteins. The enzyme was heat-inactivated, the GAGs were re-precipitated using potassium acetate and ethanol, and the pellet was dissolved in de-ionised water.
Labelled GAGs were separated by anion exchange chromatography on DEAE cellulose (DE52; Whatman) using a linearly increasing elution gradient of sodium chloride solution (up to 0.7 M), at a flow rate of 15 ml/hour. Total elution volume was 25 ml and fractions were collected at 2-minute intervals. Radioactivity in each fraction was measured using a Wallac 1410 liquid scintillation counter. Recovery from the anion exchange column was 99.3±0.4%. The efficiency of scintillation counting was 87.1%.
In situ hybridisation
Whole-mount in situ hybridisation using digoxigenin-labelled riboprobes was performed as described (Tautz and Pfeiffle, 1989), using plasmids for Shh (Echelard et al., 1993) and Ptch (Goodrich et al., 1996). Briefly, embryos fixed in 4% paraformaldehyde and stored in 100% methanol were rehydrated, incubated in 6% hydrogen peroxide for 1 hour, and then digested with Proteinase K (10 μg/ml) for 2 minutes. The embryos were re-fixed in 0.2% glutaraldehyde and 4% paraformaldehyde, and hybridised overnight at 70°C. After high stringency washes, the riboprobes were localised using a sheep anti-digoxigenin antibody and detected by incubation in nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate. Transverse paraffin wax-embedded sections (10 μm) were prepared through the posterior neuropore region.
Embryonic parameters were compared between treatment groups by Student’s t-test or one-way analysis of variance, with subsequent pairwise comparisons against the PBS-treated control group by Dunnett’s test. Statistical significance level was P<0.05.
Titration of chlorate concentration
Chlorate concentrations of up to 30 mM have been used in both cell and organ cultures to inhibit sulphation, with no significant effect on GAG or protein synthesis or on cell viability (Conrad, 1998; Davies et al., 1995; Greve et al., 1988; Miao et al., 1996). Initially, therefore, we performed a titration series in cultures of whole mouse embryos in order to identify a chlorate concentration that might affect spinal neural tube closure without generalised toxic effects on embryonic growth and development. Embryos were explanted from the uterus between the 10- and 14-somite stages, stabilised in culture for 3 hours, then cultured in the presence of different concentrations of chlorate for 14 hours.
All embryos exposed to chlorate concentrations of 30 mM or less had a vigorous yolk sac circulation and regular heart beat. Moreover, comparison of crown-rump length, head length, Brown and Fabro morphological score and protein content showed no statistically significant differences between embryos treated with chlorate and those in the control group, cultured in the absence of chlorate (Fig. 1). All embryos had between 19 and 23 pairs of somites with no significant difference in somite number between treatment groups (Fig. 2A). No obvious difference in gross morphology was found between the two groups, apart from the posterior neuropore length, which differed significantly as described in the following sections.
Inhibition of GAG sulphation results in accelerated posterior neuropore closure
The posterior neuropore is a region of unclosed neural folds at the caudal end of the E8.5-E10 embryo (Van Straaten et al., 1992). To determine whether inhibition of GAG sulphation has an effect on spinal neurulation, posterior neuropore length was compared among embryos cultured in the presence of different chlorate concentrations (Fig. 2B). Those exposed to 30 mM chlorate showed a statistically significant reduction of 34.2% in posterior neuropore length compared with embryos that were not exposed to chlorate, whereas no statistically significant difference in posterior neuropore length occurred in embryos cultured in the presence of up to 20 mM chlorate. Hence, closure of the posterior neuropore appears to be accelerated by chlorate treatment, an effect that is specific for spinal neurulation and not secondary to either developmental retardation (as somite number is not affected, Fig. 2A) or growth retardation (as all growth parameters are normal, Fig. 1). All subsequent experiments were carried out at a chlorate concentration of 30 mM.
In contrast to chlorate, there was no significant effect of 1 mM β-D-xyloside on posterior neuropore closure (not shown), whereas cranial neurulation was frequently disturbed, as described previously (Morriss-Kay and Crutch, 1982). Moreover, caudal somitogenesis was disturbed in β-D-xyloside-treated embryos, but not in chlorate-treated embryos. Hence, chlorate and β-D-xyloside have differing effects on neurulation in the spinal and cranial regions.
To confirm that the effect of chlorate is due to competitive inhibition of GAG sulphation, embryos were cultured in the presence of PBS, chlorate, or chlorate plus sulphate. Exogenous sulphate in the culture medium is predicted to compete out the effect of chlorate in inhibiting GAG sulphation. As in the dose-response experiment (Fig. 2B), we found that 30 mM chlorate leads to a reproducible shortening of the posterior neuropore (Fig. 3A). Moreover, exogenous sulphate (10 mM) was able to block the effect of chlorate in inducing premature neuropore closure (P<0.01). There was no difference in posterior neuropore length between embryos cultured in the presence of chlorate plus sulphate and those exposed to PBS (P>0.05). Hence, acceleration of posterior neuropore closure by chlorate is due to competitive inhibition of GAG sulphation.
Specific requirement for heparan sulphate in posterior neuropore closure
Heparan sulphate and chondroitin sulphate are the principal sulphated GAGs synthesised by the mouse embryo during neurulation (Copp and Bernfield, 1988; Solursh and Morriss, 1977). To assess their relative requirements for posterior neuropore closure, embryos were cultured in the presence of chlorate together with either exogenous heparan sulphate or chondroitin sulphate. Heparan sulphate effectively blocked the action of chlorate on posterior neuropore closure (Fig. 3B), so that posterior neuropore length did not differ between embryos cultured in the presence of both chlorate and heparan sulphate, and those exposed to PBS (P>0.05). However, heparan sulphate that had been degraded by overnight treatment with heparitinase before addition to cultures was unable to block the effect of chlorate in accelerating posterior neuropore closure (Fig. 3C). Exogenous chondroitin-4-sulphate (not shown) and chondroitin-6-sulphate (Fig. 3D) also failed to prevent the chlorate-induced effect on posterior neuropore closure. These findings suggest that an intact heparan sulphate chain is required to counteract the effect of chlorate on spinal neurulation.
Effect of chlorate on GAG sulphation
We studied the effect of chlorate on GAG sulphation by culturing embryos in the presence of [35S]sulphate followed by extraction of 35S-labelled GAGs and analysis by anion exchange chromatography. Because the molar concentration of sulphate in these cultures was 1.3×104 times lower than in the blocking experiments (Fig. 3A), the addition of [35S]sulphate was not expected to block the effect of chlorate.
Embryos cultured in the absence of chlorate gave an elution profile (red line, Fig. 4A) in which heparan sulphate eluted at a sodium chloride concentration of 0.33±0.01 M (mean±s.e.m.) and was sensitive to heparitinase digestion (not shown), while chondroitin sulphate eluted at a sodium chloride concentration of 0.43±0.01 M and was sensitive to digestion by chondroitinase ABC. By contrast, in cultures of chlorate-treated embryos (green line, Fig. 4A), sulphation of chondroitin sulphate was completely abolished, while the degree of sulphation of heparan sulphate was reduced by 46.0%, giving rise to Peak HS′. This incomplete abolition of heparan sulphation probably results from chlorate (30 mM) inhibiting O-sulphation more effectively than N-sulphation (Safaiyan et al., 1999).
Importantly, addition of exogenous heparan sulphate (100 ng/ml) to the culture medium did not block the effect of chlorate in inhibiting GAG sulphation (blue line, Fig. 4A). This finding demonstrates that the action of exogenous heparan sulphate in abrogating the chlorate effect on neuropore closure (Fig. 3B) is not via a diminution of the inhibitory action of chlorate on sulphation of endogenously synthesised heparan sulphate. Rather, exogenous heparan sulphate seems to have a more direct effect, making good the deficiency of sulphated endogenous heparan sulphate in the embryonic tissues.
We treated samples of material from Peak HS′ with chondroitinase, heparitinase or PBS before analysis using anion exchange chromatography (Fig. 4B). Chondroitinase had no effect on Peak HS′, whereas heparitinase reduced its height to give Peak HS′′, thus confirming that Peak HS′ consists of under-sulphated heparan sulphate.
Requirement for both N- and O-sulphate groups in heparan sulphate
Heparan sulphate contains both N- and O-linked sulphate groups. To determine which is needed for regulating spinal neurulation, embryos were cultured in the presence of chlorate plus either de-N- or de-O-sulphated heparan sulphate. Neither heparan sulphate species was able to prevent the neuropore closure effect of chlorate (Fig. 5), suggesting that both N- and O-sulphate groups are needed for heparan sulphate to regulate spinal neurulation.
Effect of chlorate on neuroepithelial bending
Next, we studied the morphology of the posterior neuropore in embryos exposed to chlorate in culture (Fig. 6). Gross observation of chlorate-treated embryos revealed that, in addition to their reduced neuropore length, a marked convexity is invariably present in the neural plate of the neuropore region (Fig. 6B). This contrasts with the normal posterior neuropore morphology observed in PBS-treated embryos (Fig. 6A) and in embryos treated with chlorate plus exogenous heparan sulphate (Fig. 6C).
Transverse histological sections (Fig. 6D) confirm that the E9.5 posterior neuropore exhibits a median hinge point overlying the notochord, together with paired dorsolateral hinge points on the neural folds (Fig. 6E). These hinge points help to elevate the neural folds and bring the fold apices into apposition in the dorsal midline. The non-bending regions of the neural plate lying between the median and dorsolateral hinge points appear straight and rigid (Shum and Copp, 1996; Ybot-Gonzalez and Copp, 1999) (Ybot-Gonzalez et al., 2002). Strikingly, embryos cultured in the presence of 30 mM chlorate do not have a median hinge point (Fig. 6F). Instead, the midline neural plate is flat, or even convex, whereas bending at the paired dorsolateral hinge points is accentuated, and this appears to compensate for the lack of a median hinge point in bringing the neural fold apices towards the dorsal midline. The region of the neural plate between the paired dorsolateral hinge points appears to have lost its rigidity and buckles, yielding the convex appearance observed in whole embryos (Fig. 6B). Despite this abnormal neural plate morphology, immunohistochemical staining for activated caspase 3 does not provide evidence for increased programmed cell death in chlorate-treated embryos (not shown).
Formation of the chlorate-induced neuroepithelial morphology could be blocked by addition of exogenous sulphate to the culture medium (Fig. 6G), yielding a posterior neuropore morphology similar to that seen in embryos exposed only to PBS. This result correlates with the finding that exogenous sulphate blocks chlorate-induced premature posterior neuropore closure, and is consistent with competitive inhibition of GAG sulphation by chlorate.
Supplementation of the culture medium with exogenous heparan sulphate also prevents chlorate-induced changes in posterior neuropore morphology (Fig. 6C,H). This suggests that the exogenous heparan sulphate in the culture medium is an effective substitute for ‘normally sulphated’ heparan sulphate synthesised endogenously by the embryo, and correlates with the finding that exogenous heparan sulphate blocks chlorate-induced premature posterior neuropore closure. By contrast, addition of exogenous chondroitin-6-sulphate to the culture medium does not prevent the chlorate-induced change in neuroepithelial morphology (Fig. 6I). In these embryos, the median hinge point is absent, the paired dorsolateral hinge points exhibit accentuated bending and neuroepithelial rigidity is lost in the region between the dorsolateral hinge points, as in embryos exposed to chlorate alone (Fig. 6F).
Chlorate diminishes immunostained GAGs in the mouse embryo
We assessed the presence of sulphated GAGs after chlorate treatment by immunostaining transverse sections of the posterior neuropore region with anti-heparan sulphate (10E4) and anti-chondroitin sulphate (CS-56) antibodies. Embryos cultured in the presence of PBS show strong staining for both heparan sulphate (Fig. 7A) and chondroitin sulphate (Fig. 7D) in the neuroepithelial basement membrane, extending from the median hinge point to the neural fold apices. Staining is also seen in the basement membrane of the surface ectoderm and gut endoderm, and in the extracellular matrix of the paraxial mesoderm. By contrast, staining for both sulphated GAGs is greatly reduced in chlorate-treated embryos. Staining for heparan sulphate is very weak throughout the section (Fig. 7B), whereas staining for chondroitin sulphate, although very much reduced in the neuroepithelium and neuroepithelial basement membrane, shows a less marked reduction than heparan sulphate in the underlying mesoderm (Fig. 7E). Importantly, chlorate-treated embryos exposed to exogenous heparan sulphate (Fig. 7C) exhibit an intensity of heparan sulphate staining that is intermediate between PBS-treated (Fig. 7A) and chlorate-treated embryos (Fig. 7B), with strongest staining in the surface ectodermal basement membrane and in the neural folds. This finding demonstrates that exogenous heparan sulphate, which is able to block the effect of chlorate on posterior neuropore closure, gains access to the cells of the posterior neuropore.
The persistent staining of chondroitin sulphate in chlorate-treated embryos contrasts with our finding of virtually no 35S-labelled chondroitin sulphate detectable by ion-exchange chromatography after chlorate treatment (Fig. 4A). It seems likely that the anti-chondroitin sulphate antibody is detecting ‘normally sulphated’ chondroitin sulphate synthesised prior to the addition of chlorate to the culture medium, which must exhibit slower turnover than heparan sulphate. The overall reduction of 10E4 and CS-56 staining in the sections of chlorate-treated embryos suggests that sulphate groups comprise an important component of the epitopes recognised by these antibodies.
Effect of chlorate on sonic hedgehog signalling
Shh has been suggested to participate in the regulation of spinal neurulation by contributing to the formation of the median hinge point, while inhibiting bending at paired dorsolateral hinge points (Ybot-Gonzalez et al., 2002). To determine whether absence of the median hinge point and the accentuated bending at dorsolateral hinge points in chlorate-treated embryos may be related to disruption of Shh signalling, we compared the expression patterns of Shh and Ptch by in situ hybridisation. Shh is a ligand of Ptch and Shh signalling up-regulates Ptch expression (Goodrich et al., 1996; Marigo et al., 1996). Hence, an abnormal pattern of Ptch expression may indicate disruption of Shh signalling.
The expression patterns of Shh and Ptch transcripts in embryos cultured in the presence of PBS are essentially as described (Echelard et al., 1993; Goodrich et al., 1996). In the posterior neuropore region, Shh transcripts are detected in the notochord and in the ventral part of the hindgut (Fig. 8A), while Ptch transcripts are present in the neuroepithelium, particularly at the median hinge point, with staining intensity decreasing progressively towards the paired dorsolateral hinge points (Fig. 8C). The notochord is only weakly positive for Ptch, which is also seen in the hindgut (both dorsal and ventral parts) and weakly in mesoderm immediately lateral to gut and notochord.
Embryos cultured in the presence of chlorate exhibit a Shh expression pattern closely similar to that in PBS controls (compare Fig. 8A with 8B). By contrast, the distribution of Ptch transcripts differs in several respects in chlorate-treated embryos (Fig. 8D,E). Notably, the notochord is more strongly positive for Ptch mRNA in chlorate treated embryos than PBS controls, whereas the overlying neuroepithelium exhibits weaker signal and the ventrodorsal gradient of expression is less pronounced. Furthermore, Ptch expression is less restricted to axial tissues: variable amounts of punctate staining are detected even in the most lateral paraxial mesoderm.
We have examined the role of heparan sulphate in spinal neurulation, using chlorate to competitively inhibit GAG sulphation. Chlorate specifically blocks median hinge point formation in the bending neural plate and accelerates closure of the posterior neuropore, the latter being mediated through accentuated bending at the paired dorsolateral hinge points. These chlorate-induced effects can be prevented by supplementation of the culture medium with ‘normally sulphated’ heparan sulphate, but not by addition of chondroitin sulphate, heparitinase-treated heparan sulphate or heparan sulphate lacking either the N- or O-linked sulphate groups. These findings are summarised in Table 1. Below, we consider the possible roles of HSPGs in neurulation through their influence on the actin cytoskeleton, the cell cycle and the propagation of molecular signals, including Shh.
Differing actions of chlorate and β-D-xyloside
The accelerating effect of chlorate on spinal neurulation contrasts with the lack of effect of β-D-xyloside on posterior neuropore closure. The reason for this difference probably lies in the different biochemical actions of the two agents. Chlorate inhibits sulphation of GAG chains, but does not block their addition to the protein backbone, whereas β-D-xyloside acts as a substitute for xylose in the link between sulphated GAGs and their protein backbones, yielding ‘naked’ proteoglycan backbones and free, normally sulphated GAG chains. Our findings suggest that correct sulphation of heparan sulphate, but not its attachment to the protein backbone, is required for normal spinal neurulation, whereas intact proteoglycans appear to be necessary for cranial neurulation, which is inhibited by both β-D-xyloside and heparitinase (Morriss-Kay and Crutch, 1982; Tuckett and Morriss-Kay, 1989). Although the identity of the HSPGs that participate in neurulation is unclear, the presence of perlecan during mouse neurulation has been demonstrated (O’Shea, 1987; Trasler and Morriss-Kay, 1991).
Heparan sulphate and the role of actin microfilaments in neurulation
Actin microfilament bundles are found in neuroepithelial cells (Baker and Schroeder, 1967; Karfunkel, 1974; Morriss-Kay and Tuckett, 1985; Morriss and New, 1979; Nagele and Lee, 1980; Sadler et al., 1982; Ybot-Gonzalez and Copp, 1999) and their contraction has been postulated to cause neuroepithelial cell wedging, leading to elevation and bending of the neural folds (Karfunkel, 1974). Moreover, HSPGs are known to co-localise with, and bind to, actin and other components of the cytoskeleton in epithelial cells, thus participating in organisation of the cytoskeleton (Bernfield et al., 1999; Carey et al., 1994; Carey et al., 1996; Fernandez-Borja et al., 1995). Indeed, apical microfilament bundles in the cranial neuroepithelium were found to be poorly organised when cranial neural tube closure was inhibited by β-D-xyloside treatment of cultured rat embryos, which blocks the attachment of GAGs to their proteoglycan core proteins (Morriss-Kay and Crutch, 1982). The breadth of the apical region of neuroepithelial cells was widened in these embryos, suggesting that the microfilament bundles were not under tension. Moreover, exposure of cultured mouse embryos undergoing spinal neurulation to cytochalasin D also resulted in disassembly of the apical microfilaments in the neuroepithelium (Ybot-Gonzalez and Copp, 1999). This led to loss of rigidity of the neural plate, resembling the morphology seen in chlorate-treated embryos. This suggests that microfilament assembly and contraction in the neuroepithelium require heparan sulphate, and that this heparan sulphate needs to be ‘normally’ sulphated. Importantly, however, formation of the median and dorsolateral hinge points during mouse spinal neurulation was not inhibited by cytochalasin D (Ybot-Gonzalez and Copp, 1999), indicating that additional factors must also act to prevent median hinge point formation in the chlorate-treated embryos.
Heparan sulphate and the cell cycle
Besides apical contraction, neuroepithelial cells may adopt a wedge shape by broadening of the cell base. This appears to be an important mechanism for hinge point formation during mouse spinal neurulation, in the absence of a requirement for actin microfilaments (Ybot-Gonzalez and Copp, 1999). Interkinetic nuclear migration results in the cell nucleus occupying a basal position during S- and early G2-phases of the cell cycle, and this is accompanied by cell wedging. Indeed, more cells are in G2-phase in the median hinge point than in non-bending regions of the neural plate of the chick embryo, and these cells have a prolonged cell cycle (Smith and Schoenwolf, 1987; Smith and Schoenwolf, 1988). Similarly, cells in the median hinge point have a higher S-phase labelling index but lower mitotic index than elsewhere in the spinal neuroepithelium of the mouse embryo (Gerrelli and Copp, 1997).
Nuclear heparan sulphate is known to influence the cell cycle (Fedarko et al., 1989; Fedarko and Conrad, 1986; Ishihara and Conrad, 1989). Heparan sulphate has been localised in the nuclei of hepatoma cells in culture, and these heparan sulphate molecules contain highly sulphated residues (Fedarko and Conrad, 1986). As the hepatoma cells become confluent and stop dividing, the amount of nuclear heparan sulphate increases up to threefold (Ishihara and Conrad, 1989). Heparan sulphate, extracted from confluent cell cultures and added to growing cells, is taken up and transported to the cell nuclei (Fedarko et al., 1989), leading to inhibition of cell division. By contrast, heparan sulphate obtained from cells in the logarithmic phase of growth is less effectively taken up and does not affect cell division. Thus, prolongation of the cell cycle in median hinge point cells during normal neurulation could require the presence of nuclear heparan sulphate. Accordingly, perturbation of sulphation of heparan sulphate by chlorate could cause cells in the median hinge point to progress through the cell cycle, resulting in loss of basal localisation of the cell nuclei and disruption of cell wedging.
Heparan sulphate on the cell surface and in the extracellular matrix is also able to regulate cell cycling through its interaction with growth factors (Conrad, 1998). Binding of FGF to high-affinity FGF receptors leads to dimerisation and mutual tyrosine phosphorylation of these receptors, resulting in biological effects such as cell proliferation (Schlessinger et al., 1995). The growth factor-receptor interaction is facilitated by heparan sulphate. However, heparan sulphate that inhibits FGF-stimulated cell proliferation has also been described. For example, inhibitory heparan sulphate prevents the human breast cancer cell line MDA-MB-231 from responding to FGF2 (Delehedde et al., 1996). MDA-MB-231 cells do not normally have a mitogenic response to FGF2, but blocking sulphation of heparan sulphate by chlorate enables these cells to respond to FGF2 and proliferate. Heparan sulphate that inhibits a mitogenic response to FGF1 and FGF7 has also been described (Bonneh-Barkay et al., 1997; Pye et al., 2000), and this inhibitory activity has been correlated with the sulphation pattern of heparan sulphate. Thus, cells in the median hinge point could have a prolonged cell cycle because of their inability to respond to growth factors (such as FGFs), owing to the inhibitory action of heparan sulphate. Blocking sulphation of heparan sulphate by chlorate treatment releases the cells from this inhibition, enabling their increased proliferation, and abolishing median hinge point formation.
Heparan sulphate and sonic hedgehog signalling
Heparan sulphate is required for localisation and propagation of the hedgehog signal (Bellaiche et al., 1998; Lin et al., 2000; The et al., 1999) through its interaction with the transmembrane protein Dispatched. This promotes the release of hedgehog from producing cells and helps to propagate the signal through the tissues to the receiving cells (Burke et al., 1999). In addition, a specific requirement for sulphate groups on heparan sulphate is indicated by the disruption of hedgehog signalling where sulphation of heparan sulphate is perturbed, as in the Drosophila sulfateless mutant (The et al., 1999). Thus, it could be postulated that the absence of a median hinge point and the increased bending at the paired dorsolateral hinge points in the chlorate-treated embryos are caused by disruption of Shh signalling. Indeed, dorsolateral bending of the neural plate in the posterior neuropore has been shown to be negatively regulated by Shh (Ybot-Gonzalez et al., 2002). However, induction of midline bending by the notochord (Smith and Schoenwolf, 1989; Van Straaten et al., 1985) does not appear to depend primarily on Shh action. For example, median hinge point formation is not abolished in embryos lacking Shh function (Ybot-Gonzalez et al., 2002). We conclude that modulation of Shh function may be responsible for some, but not all, of the abnormalities of posterior neuropore closure observed in chlorate-treated embryos.
We have shown that chlorate alters the expression pattern of Ptch mRNA, whereas Shh expression appears unaffected. We cannot exclude the possibility that Ptch expression is affected directly by chlorate, or indirectly affected by altered neural plate bending. Nevertheless, our results are consistent with a reduced propagation of the Shh signal in the presence of chlorate. It is well established that the influence of Shh from the notochord and floor plate induces a ventrodorsal gradient of Ptch expression in the neuroepithelium (Marigo and Tabin, 1996). We suggest that during normal development, Shh signalling from the notochord is propagated towards the overlying neuroepithelium through the action of heparan sulphate in the intervening extracellular matrix. This suggestion is supported by the immunohistochemical studies of Martí et al. (Martí et al., 1995), which show that Shh peptide is ‘cleared’ from notochordal cells (despite continuing Shh transcription in these cells) and accumulates in the adjacent floor plate region of E9.5 mouse embryos [see Figure 5F,G by Martí et al. (Martí et al., 1995)]. We suggest, moreover, that inhibition of heparan sulphation by chlorate treatment abolishes this propagation of the Shh signal. Shh peptide is now able to induce strong expression of Ptch within the notochord, at the site of Shh production, as well as in adjacent paraxial mesodermal cells that normally receive only low concentrations of Shh peptide, whereas Ptch is induced to a lesser degree in the overlying neuroepithelium, and the dorsoventral gradient of neuroepithelial Ptch expression is diminished.
The situation we observe in chlorate-treated mouse embryos differs from that in the Drosophila tout-velu mutant, where Ptc protein is absent, rather than ectopically expressed (Bellaiche et al., 1998). This difference could be explained by a quantitative and qualitative difference in heparan sulphate. In the tout-velu mutant, heparan sulphate is almost undetectable, owing to a lack of heparan sulphate co-polymerase, which is required for chain elongation (Toyoda et al., 2000). By contrast, chlorate treatment merely inhibits O-sulphation and, to a lesser extent N-sulphation, and does not have a significant effect on chain elongation or other structural modifications of heparan sulphate (Conrad, 1998; Greve et al., 1988; Safaiyan et al., 1999).
In conclusion, we have demonstrated a role for HSPGs in regulating mouse spinal neurulation. This role seems likely to be mediated via a combination of the effects of heparan sulphate on the cytoskeleton, cell cycle and the propagation of extracellular signals.
We are grateful to Andrew McMahon and Matthew Scott for kind gifts of the Shh and Ptch plasmids. This work was supported by the Wellcome Trust (A. J. C.) and by an Overseas Graduate Scholarship from the National University of Singapore (G. W. Y.).