Heparan sulfate biosynthetic gene Ndst1 is required for FGF signaling in early lens development

Multiple signaling pathways are involved in early-lens morphogenesis,including bone morphogenic proteins (BMPs), fibroblast growth factors (FGFs),Wnts and hedgehog (Hh). Genetic inactivation of Bmp4 or Bmp7in mice abolished Sox2 expression in the head ectoderm and disrupted lens-placode formation (Furuta and Hogan,1998; Wawersik et al.,1999). Similarly, suppression of FGF signaling via a pharmacological inhibitor or dominant-negative FGF receptor 1 (FGFR1) resulted in the downregulation of Pax6, Sox2 and Foxe3 expression,and in defects in lens formation (Faber et al., 2001). In further support of a role of FGF signaling in early lens development, a mutation in the FGFR signaling mediator Frs2α led to anophthalmia or microphthalmia(Gotoh et al., 2004). Hh signaling is necessary for early eye-field specification and lens regeneration(Macdonald et al., 1995; Tsonis et al., 2004). However,hyperactive Hh signaling from the embryonic midline may also result in lens degeneration, as demonstrated in cavefish development(Yamamoto et al., 2004). Wnt signaling is also known to be involved in the development of the lens. Misexpression of a Wnt receptor, frizzled 3, led to ectopic Pax6 expression and eye formation in Xenopus(Rasmussen et al., 2001). By contrast, a mouse mutant deficient for the Wnt signaling co-receptor Lrp6 exhibited abnormal cell death in lens epithelium, whereas the conditional knockout of β-catenin in periocular ectoderm resulted in ectopic lentoid formation (Smith et al.,2005; Stump et al.,2003).

Many of the signaling pathways described above are known to be dependent on the presence of heparan sulfate proteoglycans (HSPGs) on the cell surface(reviewed in Lin, 2004). HSPGs are glycoproteins containing covalently linked heparan sulfate glycosaminoglycan chains. These linear polysaccharides exhibit enormous structural heterogeneity because of variable N-deacetylation of N-acetylglucosamine residues, N- and O-sulfation, and epimerization of uronic acid residues (Esko and Selleck,2002). Previous studies have demonstrated that cell-surface proteoglycans can serve as co-receptors for FGF(Rapraeger et al., 1991; Yayon et al., 1991). This is supported by crystallographic structures of heparan sulfate associated with an FGF-FGFR complex (Pellegrini et al.,2000; Schlessinger et al.,2000). Recently, the role of HSPGs in morphogen diffusion was illuminated by genetic studies of Drosophila proteoglycan core proteins and glycosaminoglycan biosynthetic enzymes. It was demonstrated that loss of HSPGs prevented the transport of Dpp, wingless (Wnt) and Hh molecules,resulting in the disruption of morphogen gradients(Belenkaya et al., 2004; Han et al., 2004; Kirkpatrick et al., 2004). In vertebrate development, the proteoglycan core-protein gene glypican 3(Gpc3) genetically interacts with Bmp4 during limb development, and loss of Wnt signaling is also observed in mouse Gpc3mutants (Paine-Saunders et al.,2000; Song et al.,2005). By contrast, a mutation in the glycosaminoglycan biosynthetic gene UDP-glucose dehydrogenase (Ugdh) inhibited the signaling of FGF, but not of Nodal or Wnt3, in mesoderm- and endoderm-migration during gastrulation(Garcia-Garcia and Anderson,2003). Another glycosaminoglycan biosynthetic gene, Ext1,is required for FGF8 signaling in CNS development, whereas a hypomorphic mutation in Ext1 expanded the range of Indian hedgehog (Ihh)signaling during chondrocyte maturation(Inatani et al., 2003; Koziel et al., 2004). Interestingly, recent studies have shown that zebrafish ext2 and extl3 regulate Fgf10, but not Fgf4, signaling during limb development(Norton et al., 2005). These findings demonstrate the potential of HSPGs in regulating specific signaling pathways in a contextdependent manner.

The enzyme N-acetylglucosamine N-deacetylase-N-sulfotransferase (Ndst) catalyzes the first sulfation step during the synthesis of heparan sulfate. Consistent with its crucial role in HSPG modification, a Drosophila Ndst mutant, sulfateless, exhibited a segment-polarity phenotype as a result of impaired Wnt signaling (Lin and Perrimon,1999). Furthermore, FGFR-dependent MAPK activity was also reduced in the sulfateless mutants during mesoderm and trachea development,and genetic interactions were demonstrated between sulfateless and the FGF-receptor gene (Lin et al.,1999). Finally, mosaic analysis showed that the loss of sulfateless prevented the diffusion of Dpp- and Hh-molecules in wing imaginal discs (Belenkaya et al.,2004; Han et al.,2004). These results established that Ndst genes are essential for the transport of morphogenic molecules and for their subsequent signaling.

There are four known Ndst enzymes in mammals, and biochemical experiments suggest that they might have different substrate specificities(Aikawa et al., 2001). Targeted deletion of Ndst1 in mice resulted in embryonic lethality as a result of lung defects, whereas brain and ocular defects had also been noted(Fan et al., 2000; Grobe et al., 2005; Grobe et al., 2002; Ringvall et al., 2000). Ndst2 mutants had impaired mast-cell development(Forsberg et al., 1999; Humphries et al., 1999),whereas the Ndst3 mutant did not exhibit an obvious phenotype(Grobe et al., 2002). Importantly, Ndst1 and Ndst2 double-homozygous mutants exhibited early embryonic lethality, similar to that observed in Ext1- and Ext2-null mutants(Forsberg et al., 1999; Grobe et al., 2002; Lin et al., 2000; Stickens et al., 2005). These results demonstrate both the functional specificity and redundancy among the Ndst-family enzymes.

We have previously characterized cranial facial-developmental defects in Ndst1 mutants and showed that Ndst1 genetically interacted with Shh. In addition, we found that fibroblast cells derived from Ndst1-mutant embryos failed to respond to FGF stimulation in vitro,suggesting a role of Ndst1 in FGF signaling(Grobe et al., 2005). In this study, we further examined lens development in Ndst1 mutants, and demonstrated that loss of Ndst1 function disrupted lens-vesicle invagination and lens cell differentiation. Importantly, we showed that BMP-and Wnt-signaling were not affected in Ndst1-mutant lenses. Instead, Ndst1 loss of function led to a reduced binding of FGF ligand or FGF-FGFR complex on the cell surface. Consistent with this, MAPK signaling was downregulated during lens development. Therefore, Ndst1 was important for lens-specific FGF signaling during development.

Ndst1 mice were maintained in a C57BL/6 background(Grobe et al., 2005). Bmp4 mice were kindly provided by Simon Conway (Indiana University School of Medicine, Indianapolis, IN, USA) and Bridget Hogan(Lawson et al., 1999). TOPGAL mice are obtained from Jackson Laboratory(DasGupta and Fuchs,1999).

Lens tissue was dissected in ice-cold PBS and immediately placed in liquid nitrogen. RNA was isolated from tissue extracts using a RNA-isolation kit(Qiagen, Valencia, CA), and reverse transcription was carried out according to the manufacturer's instructions (Invitrogen, Carlsbad, CA). The primers used for PCR were: Ndst1 (forward:5′-ACCACAGCCAGCCAGAACGCTTGTG-3′; reverse:5′-AGCTGCGCTCTTCCCCTTTACTGTC-3′), Ndst2 (forward:5′-CCTTGCAGAACCGTTGTC-3′; reverse:5′-CAGCCATTCCAATCCTG-3′), Ndst3 (forward:5′-TGTGTTTCCTGTGAGTCCAGATGTGTG-3′; reverse:5′-ATTGTCCTCCTCACTTCCATCAGCCTG-3′) and Ndst4 (forward:5′-AACAGGAAATGACACTTATTGAAACG-3′; reverse:5′-ACTTTGGGGCCTTTGGTAATATG-3′).

Pregnant mice were injected 2 hours prior to dissection with BrdU dissolved in PBS at 0.1 mg BrdU per 1 g body weight. The embryos were fixed in 4% PFA at 4°C overnight, incubated in 30% sucrose/PBS at 4°C overnight and embedded in OCT compound. Antigen retrieval was performed on 10 μm cryosections by microwave heating for 10 minutes at sub-boiling condition in citrate buffer at pH 6.0, and treated with 1 N HCl for 90 minutes at room temperature. Next, the sections were blocked with 10% normal goat serum in PBS at room temperature for 2 hours prior to the addition of an anti-BrdU antibody(Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA,USA). After overnight incubation at 4°C, the sections were treated with secondary antibody and with the nuclear stain Hoechst for 2 hours at room temperature, and then examined under a Leica DM500 fluorescent microscope. The cell-proliferation rate was calculated as the ratio of BrdU-positive cells versus Hoechst-positives cells, and analyzed by the Student's t-test.

TUNEL staining was performed with an in situ cell-death detection kit(Roche, Indianapolis, IN, USA). Briefly, cryosections were processed for antigen retrieval as described above, incubated with blocking buffer (0.1 M Tris-HCl, pH 7.5, 3% BSA, 20% serum) for 30 minutes at room temperature and then with TUNEL reaction mixture for 2 hours at 37°C. After rinsing with PBS, the sections were blocked again with 0.05% blocking reagent (supplied in the TSA Indirect Tyramide Signal Amplification Kit, Perkin Elmer Life Science,Boston, MA, USA) for 30 minutes and then incubated with TUNEL-POD for 30 minutes at 37°C. Finally, the signal was developed with DAB substrate and detected under a Leica DM500 microscope.

RNA whole-mount in situ hybridization was performed as previously described(Zhang et al., 2002). RNA in situ hybridization on sections was carried out according to a standard protocol (Dakubo et al., 2003). The following probes were used: BF2, ERM (from Bridget Hogan, Duke University Medical Center, Durham, NC, USA), Hes1 (from Naoki Takahashi, Nara Institute of Science and Technology, Nara, Japan), Math5 (also known as Atoh7 - Mouse Genome Informatics; from Tom Glaser, University of Michigan, Ann Arbor, MI, USA), Ndst1, Pax6,Six3 (from Guillermo Oliver, St Jude Children's Research Hospital,Memphis, TN, USA) and Sox2. At least three embryos of each genotype were analyzed for each probe.

X-gal staining, in situ hybridization and regular immunohistochemistry were performed as previously described (Zhang et al., 2003). Immunohistochemistry of phospho-Smad(Smad-P) and phospho-Erk (Erk-P) was carried out according to published procedures (Ahn et al.,2001). Briefly, mutant and control embryos were matched by somite numbers, and processed for coronal section on a Leica cryostat. For antigen retrieval, the sample slides were incubated in citrate buffer (10 mM sodium citrate, pH 9.0) at 80°C for 30 minutes, followed by treatment with 2%H₂O₂ to quench the endogenous peroxidase activity. After 1 hour of blocking at room temperature with 5% goat serum in PBS, the slides were incubated with primary antibody diluted in the blocking solution overnight at 4°C. Next, the slides were blocked for 30 minutes with 0.05%blocking reagent (TSA Indirect Tyramide Signal Amplification Kit, Perkin Elmer Life Science, Boston, MA, USA) and sequentially incubated with a biotin conjugated anti-rabbit antibody and ABC reagent (Vectastain ABC Kit, Vector Labs, Burlingame, CA, USA). To amplify the immunoperoxidase signal, the specimens were incubated with biotinyl tyramide diluted 1:50 in tyramide diluent for 10 minutes and then in 1:250 streptavidin-HRP for 30 minutes. Finally, the sections were incubated with DAB solution for color reaction.

As a control, we also performed phospho-Erk1/2 immunostaining on embryos treated with the FGFR1 inhibitor PD-173074 (a gift from Pfizer, New Jersey,NJ, USA) or the MAPK kinase inhibitor U0126 (Cell Signaling Technology,Beverly, MA). Prior to immunohistochemistry, the control embryos were incubated in RPMI containing 1% BSA; 50 μM U0126 or 40 μM PD-173074 at 37°C; and 5% CO₂ for 30 minutes. This effectively abolished the phospho-Erk1/2 expression in the embryos, thus validating the specificity of the phospho-Erk1/2 staining in our experiment(Corson et al., 2003).

The antibodies we used were: anti-phospho-Erk1/2, anti-phospho-Smad1/5/8,anti-phospho-Smad2 (all from Cell Signaling Technology, Beverly, MA, USA),anti-phospho-Smad1 (PS1) antibody [kindly provided by Peter ten Dijke (Leiden University Medical Center, Leiden, The Netherlands) and Carl-Henrik Heldin(Ludwig Institute for Cancer Research, Uppsala, Sweden)], anti-Pax6 (the Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA),anti-AP2α (Santa Cruz biotechnology, Santa Cruz, CA, USA), anti-Pax2,anti-Prox1 (both from Covance, Berkeley, CA, USA), anti-αA crystallin(kindly provided by Samuel Zigler, National Institute of Health, Bethesda,Maryland, USA), anti-Six3 (kindly provided by Guillermo Oliver, St Jude Children's Research Hospital), and 10E4, HepSS-1 and 3G10 (all from Seikagaku,Tokyo, Japan).

Mutant embryos and their matched littermates were harvested and sectioned as above. Prior to the assay, the frozen sections were incubated in 0.5 mg/ml NaBH₄ for 10 minutes and then in 0.1 M glycine for 30 minutes. For analysis of FGF2 binding to heparan sulfate, the embryo sections were next quenched with 2% H₂O₂ and blocked with 0.05% TSA blocking reagent. Biotinylated FGF2 was produced as previously described(Bai et al., 1999) and incubated with the sections at 4°C overnight. The bound FGF2 was detected using a Vectastain ABC kit (Vector Labs, Burlingame, CA, USA) and stained in DAB solution (Sigma, St Louis, MO, USA). As a negative control, heparan sulfate on sections was degraded with 10 units of heparitinase I (Seikagaku,Tokyo, Japan) in 50 mM Tris-HCl pH 7.4, 50 mM sodium acetate, 1 mM calcium chloride and 1 mg/ml BSA at 37°C for 2 hours prior to the assay.

For in situ binding of the FGF-FGFR complex with heparan sulfate, we carried out the ligand and carbohydrate engagement (LACE) assay, as described(Allen and Rapraeger, 2003). Briefly, the frozen sections were incubated in 0.5 mg/ml NaBH₄ for 10 minutes, in 0.1 M glycine for 30 minutes and then blocked with 2% BSA. Next, the slides were incubated with 20 μM FGF, 20 μM human FGFR-Fc chimera (both from R&D Systems, Minneapolis, MN) and 10% fetal-calf serum in RPMI-1640 at 4°C overnight. After washing in PBS, the bound FGFR-Fc was detected using a cy3-labeled anti-human Fc IgG secondary antibody, and the fluorescence signal was examined using a Leica DM500 fluorescent microscope.

To study the role of the Ndst1 gene in eye development, we first examined its expression pattern by RNA in situ hybridization. Strong expression was detected in both the lens placode and in the optic vesicle in E9.5 mouse embryos (Fig. 1A). As a control, we also detected Pax6 expression in the eye primordium,whereas no signal was observed with the Ndst1 sense probe. This ubiquitous expression of Ndst1 was also found in the later stages of eye development. At E12.5, Ndst1 expression was present in the lens epithelia and in the retina. The specificity of RNA in situ hybridization was again demonstrated by the lack of signal using the Ndst1 sense probe and the restricted staining pattern observed with the Pax6 probe.

Next, we sought to confirm the in situ hybridization results by RT-PCR analysis. Total RNA was isolated from the lens and retina of E12.5 mice, and subjected to RT-PCR with specific primers for Ndst1-4 and for the mitochondria ribosomal subunit L19(Fig. 1B). In the lenses, we detected the expression of Ndst1 and, to a lesser extent, Ndst2 only. By contrast, transcripts of all four Ndst genes were present in the retinae. As a control, similar levels of ribosomal subunit L19 signals were observed in the retina and in the lens. Furthermore,no signal was detected in a RT-PCR reaction without reverse transcriptase. Previous studies have shown that Ndst2-null mice are normal, with the exception of defects in connective-tissuetype mast cells(Forsberg et al., 1999; Humphries et al., 1999). Together, these results suggest that, of the Ndst genes, Ndst1 most probably plays the dominant role during lens development.

We next examined homozygous Ndst1-null mice from E12.5 to E17.5 and observed ocular phenotypes in all embryos collected (n=18, Fig. 2A). Among them, four out of 18 embryos (22%) exhibited microphthalmia with reduced retinae and lenses,eight out of 18 embryos (44%) retained the retina but lacked lens, and the remaining six out of 18 embryos (34%) had no retina or lens. The size of embryos without lenses was indistinguishable from that of wild-type litter mates, whereas the embryos without eye structures were sometimes smaller and showed additional brain defects.

To investigate the origin of the ocular defects, we next analyzed homozygous Ndst1 mutants at E13.5. The range of lens phenotypes was apparent in the Ndst1-mutant embryos and, even in the least-affected embryos, lenses were smaller and pinched at the anterior region(Fig. 2D,E). Unlike the wild-type littermates, the lens lumen of which was mostly filled with lens fibers, much of the anterior-lens vesicle was still empty in mutant lenses. To uncover the molecular changes in Ndst1 mutants, we also performed RNA in situ hybridization on frozen sections to stain for Pax6 gene expression (Fig. 2B-I). Pax6 expression should be restricted to the epithelial cells of the lens at this stage; however, in mutants, Pax6 transcripts were detected throughout the lens (Fig. 2E). In more-severely affected mutants, the entire lens was reduced to a small cluster of cells connected to the surface ectoderm with a residual lens stalk (Fig. 2G,arrow) and the retinae were mis-shaped. Finally, some Ndst1 mutants lacked any apparent retina or lens structure(Fig. 2H). There were sometimes pigmented cells left at the presumptive eye region(Fig. 2I, arrowhead). Interestingly, Pax6 expression could still be detected in the surface ectoderm (Fig. 2I). The severity of the lens defects observed in these E13.5 Ndst1 embryos suggested that failure of lens development probably originated even earlier during development.

We next asked whether the Ndst1-mutant eye phenotype was restricted to the lens (Fig. 2J-S). Sox2 is a major early neural marker during development; Six3, Hes1 and Math5 are transcription factors that define retinal progenitor cells; and, finally, BF2 marks the posterior region of the retina in topographic axon mapping(Brown et al., 2001; Darnell et al., 1999; Furukawa et al., 2000; Wang et al., 2001; Yuasa et al., 1996; Zhu et al., 2002). RNA in situ hybridization showed that transcripts of these genes were present in both wild-type and Ndst1-mutant retinae. Therefore, retinal patterning and differentiation appeared to have at least been initiated in the absence of the lens in Ndst1 mutants.

In search of the mechanism for lens defects, we next studied lens induction in Ndst1 mutants. At the 24-somite stage, the mutant lens placode was morphologically indistinguishable from wild-type controls(Fig. 3A). However, at the 30-somite stage, wild-type embryos had formed the lens pit, whereas Ndst1-mutant embryos exhibited less-advanced indentation in the lens placode (Fig. 3A). At the 35-somite stage (E11.0), lens vesicles were either entirely absent (data not shown) or reduced in size in Ndst1 mutants(Fig. 3A). Throughout lens-vesicle development, the rates of BrdU incorporation in mutant embryos were consistently reduced in comparison to wild-type controls(Fig. 3B). By contrast, few apoptotic cells were observed in either wild-type or Ndst1-mutant lens vesicles (Fig. 3C, arrow),even though there was a significant increase in TUNEL staining in periocular mesenchyme in Ndst1-mutants (Fig. 3C, arrowhead). These results suggest that Ndst1 mutants were defective in lens cell proliferation.

We next studied the molecular defects underlying lens induction in Ndst1 mutants. Pax6 and Six3 are crucial transcription factors for lens induction and morphogenesis that are expressed at increasing levels as the lens placode invaginates to form the lens vesicle. At the 26- and 28-somite stages (E10.5), mildly affected Ndst1-mutant embryos also expressed the Pax6 and Six3 proteins at high levels(Fig. 4B,E, arrows); however,the expressions of these proteins were less elevated in severely affected mutants (Fig. 4C,F,arrowheads). Of note, we did not observe a reduction in Pax6expression after lens-placode invagination, suggesting that Ndst1 is specifically required for Pax6 expression during lens induction(Fig. 4G-O). Although the disruption of Pax6 and Six3 expressions could be secondary to lens-development failure, it may also directly contributes to lens-induction defects in Ndst1 mutants.

AP2α, a transcription factor required for lens development, was found to be expressed in both wild-type and Ndst1-mutant lens primordia at the 32-somite stage (data not shown). In 35-somite embryos, however, some of the Ndst1 mutants expressed AP2α in overlying head ectoderm only, and not in lens vesicles (Fig. 4I, five out of eight embryos). Therefore, AP2α expression was specifically lost between the 32- and 35-somite stages. Normally, αA crystallin is expressed in the lens pit at E10.5 and Prox1 expression initiates in the lens placode at E9.5(Robinson and Overbeek, 1996; Wigle et al., 1999). None of these molecules were observed in the more severely affected homozygous Ndst1-null mutants (Fig. 4L, four out of six mutants for Prox1; Fig. 4O, two out of three mutants for αA crystallin). Notice that some Ndst1 mutants exhibited relatively mild lens-vesicle defects and that these embryos also preserved Pax6, AP2α, Prox1 and αA crystallin expressions. This is consistent with the observation that Ndst1 mutants displayed a range of phenotypes, including some that developed both lens and retina. Taken together, these molecular defects show that Ndst1 inactivation results in a delay or even failure of lens-vesicle development.

BMP/TGFβ signaling are known to play important roles in lens development. We thus examined the intracellular mediators of BMP/TGFβsignaling - the phospho-Smad1 (Smad1-P) and phospho-Smad2(Smad2-P) proteins - in Ndst1-mutant lenses. To validate the Smad1-P antibody, we first compared its staining pattern in wild-type and Bmp4-mutant embryos. Consistent with previous reports, we observed specific Smad1-P expression in the first branchial arch(Fig. 5A, arrowhead) and olfactory placode (Fig. 5A,arrow) at E9.5 (Ahn et al.,2001; Faber et al.,2002). Not surprisingly, this coincides with strong Bmp4expression in these locations (Dudley and Robertson, 1997). In Bmp4-knockout embryos, however, this Smad1-P staining pattern was abolished. Therefore, Smad1-Pimmunohistochemistry reliably detected active Bmp4 signaling during embryonic development.

In wild-type embryos, Smad1-P was present in the presumptive lens ectoderm and optic vesicle as early as the 20-somite stage (E9.5), forming an anterior-posterior gradient (Fig. 5B). This expression pattern persisted in 30 somite-stage embryos(E10.5) as lens placodes invaginated. In homozygous Ndst1-knockout embryos (KO/KO), similar Smad1-P staining was detected, even as the lens-vesicle formation was disrupted (Fig. 5B, arrowhead). Furthermore, both wild-type and Ndst1-mutant lens vesicles strongly expressed the Smad2-Pprotein at E11.0 (Fig. 5B,arrow). Therefore, the BMP/TGFβ signaling mediated by phospho-Smad proteins was unaffected in Ndst1 mutants. To further test the genetic interaction between Ndst1 and Bmp4, we crossed Ndst1 mice with a Bmp4-mutant strain carrying a LacZ knock-in allele (Bmp4^LacZ)(Lawson et al., 1999). As shown in Fig. 5C, the loss of one copy of the Bmp4 gene did not enhance the lens phenotype in either heterozygous or homozygous Ndst1-mutant embryos (n=24 for Ndst1^KO/+ Bmp4^LacZ/+, n=10 for Ndst1^KO/KO Bmp4^LacZ/+), and the Bmp4expression reported by the β-galatosidase activity was also unchanged in the Ndst1-mutant background. Taken together, these results suggest that BMP/TGFβ signaling was not affected by Ndst1 inactivation in the lens.

Canonical Wnt signaling results in the inhibition of GSK-3β kinase and in the accumulation of β-catenin in the nucleus, which allows TCF-family transcription factors to activate downstream target genes. Using a transgenic mouse line carrying a LacZ reporter driven by multimerized TCF after activation by β-catenin (TOPGAL)(DasGupta and Fuchs, 1999), we also assayed the canonical Wnt signaling activity in Ndst1-mutant lenses (Fig. 5D). In both wild-type and Ndst1-mutant embryos, similar TOPGAL transgene expressions were observed in periocular tissue, whereas no β-galatosidase activity was detected in the lens. Therefore, the lens defects in Ndst1 mutants are unlikely to be caused by abnormal canonical Wnt signaling.

We next analyzed the expression pattern of heparan sulfate during lens development. The monoclonal antibody 10E4 recognizes an epitope unique to heparan sulfate, whereas the HepSS-1 antibody binds to N-sulfated heparan sulfate domains (Leteux et al.,2001; van den Born et al.,2005). In wild-type embryos, both antibodies stained the basal membranes of the optic vesicle and the lens vesicle(Fig. 6). We further demonstrated that this staining pattern was specific to heparan sulfate because sections treated with heparitinase I completely lost the staining(Fig. 6). Heparitinase I digestion also generated a heparan sulfate `stub' motif, which was the epitope of the 3G10 antibody (David et al.,1992). In heparitinase I-treated sections, we observed specific 3G10 staining in the developing eye (Fig. 6). Therefore, heparan sulfate was abundantly expressed during lens formation.

Ndst1 catalyzes the N-deacetylation and N-sulfation of heparan sulfate. Interestingly, we observed a complete loss of 10E4 and HepSS-1 staining in KO/KO embryos, but 3G10 staining after the Heparitinase I treatment remained intact (Fig. 6). Because the 3G10 antibody detects heparan sulfate stubs that remain after heparitinase digestion, these findings indicate that heparan sulfate chains were still being made in the Ndst1-mutant embryos, but that these were undersulfated.

The sulfation pattern of heparan sulfate is important for its interaction with FGF ligands and receptors. We thus performed LACE assays and asked whether Ndst1 mutants defective in heparan sulfate modification also exhibited reduced FGF binding (Allen and Rapraeger, 2003). Embryo sections were incubated with FGF2 tagged with biotin and the binding of FGF2 to eye tissue was detected by biotin histochemistry. In wild-type eyes, biotinylated FGF2 was specifically localized at the basal membrane of lens- and retinal-cells, and the FGF2-binding pattern closely resembled the distribution of endogenous heparan sulfate during eye development (Fig. 7A). As a control, no staining was observed in the absence of biotinylated FGF2 (data not shown). More importantly, prior treatment of embryo sections with heparitinase I completely abolished the staining(Fig. 7A). This demonstrates that the binding of biotinylated FGF2 on these tissue sections crucially depends on intact heparan sulfate. In Ndst1-mutant embryos,incubation with the same concentration of biotinylated FGF2 resulted in much-weaker staining as compared with the wild-type controls, and significant binding of FGF2 to lens cells was observed only after a 10-fold increase in FGF2-ligand concentration (Fig. 7A). Therefore, the Ndst1 mutation resulted in a reduced affinity of FGF2 to the lens cell basement membrane.

We next tested whether the assembly of the FGF-FGFR complexes was also affected in Ndst1 mutants. For this experiment, we assayed FGF8,which has been shown to be important for early eye development, and FGF19, the mouse homolog of which (FGF15) is strongly expressed in the optic vesicle(Lovicu and Overbeek, 1998; McWhirter et al., 1997; Vogel-Hopker et al., 2000). Eye sections from E10.5 embryos were incubated with purified FGF and with FGFR fused with the human IgG Fc domain (FGFR-Fc), and bound FGFR-Fc was probed with an anti-IgG antibody. In wild-type-embryo sections, we observed specific binding of FGFR2c and FGFR3c to lens- and retina-cells in the presence of FGF8b (Fig. 7B). As a control,no signal was detected without FGF8b or after the treatment of tissue sections with heparitinase I (data not shown). This demonstrated that the observed FGFR binding in situ was mediated by FGF and cell-surface heparan sulfate. In E10.5 Ndst1-mutant sections, binding of FGFR3c-FGF8b was reduced throughout the eye region (Fig. 7B). In comparison, FGFR2c-FGF8b binding was weaker in the retina(Fig. 7B, arrow) and became almost undetectable in the Ndst1-mutant lens(Fig. 7B, arrowhead). The more-pronounced loss of FGFR2c-FGF8b binding to the lens was especially interesting considering the significant lens phenotype in Ndst1mutants. Finally, FGFR4-FGF19 exhibited weak but unchanged binding to both wild-type and Ndst1-mutant eyes, suggesting that Ndst1inactivation disrupted the assembly of some, but not all, FGF-FGFR pairs in the eye tissues.

Based on previous reports of FGF expression patterns, we extended our analysis to FGF1, 2, 3, 5, 8, 9 and 15 to assay their pairwise interactions with FGFR1b, 1c, 2b, 2c, 3b, 3c and 4 (Chow and Lang, 2001; de Iongh and McAvoy, 1993; Kitaoka et al.,1994; Lovicu and Overbeek,1998; Martinez-Morales et al.,2005; McWhirter et al.,1997; Nguyen and Arnheiter,2000; Vogel-Hopker et al.,2000; Zhao et al.,2001). This includes all the FGF and FGFR variants known to be present during early eye development. It also represents all six subfamilies of FGFs, except the FGF11-14 subfamily, which does not interact with FGF receptors. In wild-type embryos, the binding of each FGF-FGFR pair largely confirmed previous results obtained in cell culture or in biochemical studies(Fig. 7C). In Ndst1mutants, however, FGF-FGFR interactions were altered. Except for FGF19, the interaction of each FGF with at least one FGFR variant was disrupted by Ndst1 deletion. Different FGF-FGFR pairs were affected differentially by Ndst1 inactivation, and each pair exhibited distinct binding activities in different tissues (Fig. 7C and data not shown). Overall, 18 FGF-FGFR pairs required Ndst1 function in the lens, suggesting that Ndst1-modified heparan sulfate potentially regulates a large number of FGF-FGFR interactions during lens development.

The significant loss of in situ FGF-FGFR binding to Ndst1-mutant tissue raised the possibility that FGF signaling was compromised during eye development. To test this idea, we first examined the expression of the phospho-Erk1/2 (Erk-P) proteins - the downstream effectors of the FGF-MAPK pathway. Using a phospho-specific antibody against Erk1/2, we detected Erk-P expression in the developing optic cup and lens vesicle in E10.5 embryos (Fig. 8A). The specificity of the immunohistochemistry assay was demonstrated in embryos cultured in the presence of the MEK inhibitor U0126,which acts upstream of Erk1/2 (Favata et al., 1998). After treatment, Erk-P expression was completely lost throughout the embryos, including the eye tissues(Fig. 8A). Furthermore, we cultured wild-type embryos with PD173074, a potent FGFR inhibitor(Skaper et al., 2000). This treatment also abolished Erk-P staining in the developing optic cup and lens vesicle (Fig. 8A). Together, these results confirm previous reports that Erk-Pexpression directly correlates with FGFR-MAPK signaling activity during eye development (Corson et al.,2003; Govindarajan and Overbeek, 2001; Lovicu and McAvoy, 2001).

In wild-type embryos, strong Erk-P immunostaining was observed in lens placodes at the 27-somite stage, whereas little Erk phosphorylation was detected in the Ndst1-mutant lens ectoderm(Fig. 8B, arrowhead). Similarly, wild-type embryos exhibited strong Erk-P expression in the invaginating lens vesicle at the 30- and 32-somite stages. In mutant embryos,where lens development failed to progress beyond initial lens-placode invagination, the lens tissues expressed Pax6 but not the Erk-Pproteins (Fig. 8B, arrow). Interestingly, strong expression of Erk-P remained in the mutant optic vesicle throughout development. These results show that the MAPK pathway was specifically disrupted in the Ndst1-mutant lens.

Downstream to FGF-MAPK signaling, ETS-domain transcription factors are both transcriptional effectors and direct targets of the pathway(Tsang and Dawid, 2004). In particular, previous experiments have demonstrated that the expression of the Pea3 (also known as Etv4 - Mouse Genome Informatics) group of ETS-domain transcription factors [ER81 (also known as Etv1 - Mouse Genome Informatics),ERM (also known as Etv5 - Mouse Genome Informatics) and Pea3] closely mimic FGF-signaling activities (Munchberg and Steinbeisser, 1999; Raible and Brand, 2001; Roehl and Nusslein-Volhard, 2001). Using RNA in situ hybridization, we observed expression of ERM in developing wild-type lens vesicles at E10.5. By contrast, ERM expression was significantly downregulated in Ndst1-mutant lenses. Consistent with lens-specific loss of Erk-P, the reduction of ERM was also confined to the developing lens, because midbrain- and branchial-arches still exhibited strong ERM expression. Together, these results show that FGF-MAPK signaling was disrupted in the Ndst1-mutant lens.

An interesting finding in this study was that Ndst1 inactivation disrupted signaling of FGF, but not of BMP or Wnt, during lens development. This is surprising considering that the Drosophila Ndst gene sulfateless is essential for all three signaling pathways. It is possible that specific heparan sulfate modifications generated by Ndst1 are only required for FGF signaling, but not for BMP/Wnt signaling. However,another explanation may be that FGF signaling is more sensitive to defective heparan sulfate than the other two signaling pathways. In BMP/Dpp- and Wnt-signaling, heparan sulfate may be required primarily for morphogen movement in the developing field (Lin,2004). Thus, in Drosophila tissue where Dpp is abundantly expressed, heparan sulfate mutation has no obvious effect on Dpp-controlled patterning (Haerry et al.,1997; Lin and Perrimon,1999). By contrast, cell-surface heparan sulfate acts as a co-receptor for FGF signaling, forming a trimeric complex with FGF and the FGF receptor. Therefore, all mutant cells deficient for heparan sulfate fail to respond to FGF signaling. In vertebrate lens development, the lens vesicle develops from a single layer of placodal cells, which is directly exposed to strongly expressed inductive signals, such as BMP4, from the optic vesicle. It is therefore expected that the Ndst1-mutant lens vesicle will maintain its BMP response. By contrast, heparan sulfate is required for FGF-signal binding on the cell surface, and Ndst1 inactivation will thus abrogate the FGF-signaling activity during lens-vesicle development.

Ndst1 also exhibited remarkable specificity in regulating FGF-FGFR interactions. Previous studies have shown that N-sulfation of heparan sulfate is essential for its binding to FGF2(Turnbull et al., 1992). Consistent with this model, we observed reduced FGF2 binding to lens cells in Ndst1 mutants. In addition, Ndst1 inactivation significantly reduced binding of lens cell heparan sulfate to FGF8b-FGFR2c and FGF8b-FGFR3c complexes. Surprisingly, the binding affinity of FGF19-FGFR4 to lens cell heparan sulfate remained unchanged in Ndst1 mutants. This suggested that Ndst1-mediated structural remodeling of heparan sulfate was necessary to allow for the assembly of some, but not all, FGF-FGFR complexes on the cell surface. Such distinct requirement for heparan sulfate motifs in FGF signaling has also been observed in earlier studies(Ford-Perriss et al., 2002). In particular, a mutation in the heparan sulfate 2-O-sulfotransferase(HS2ST) gene was shown to disrupt binding of FGF8b-FGFR2c, but not FGF8b-FGFR3c, on E10.5 heart sections(Allen and Rapraeger, 2003). In zebrafish limb development, ext2 and extl3 mutants specifically affected FGF10, but not FGF4, signaling(Norton et al., 2005). Taken together, these results suggest that heparan sulfate-synthesizing enzymes can play an active role in regulating different FGF-signaling pathways.

In this study, we have systematically analyzed in situ FGF-FGFR binding on the cell surface of the lens. Our results are mostly consistent with mitogenesis studies performed in cell culture and with binding studies by surface plasmon resonance, although a few differences were noted(Mohammadi et al., 2005; Ornitz et al., 1996; Zhang et al., 2006). These differences probably result from the fact that our assay involved endogenous heparan sulfate on developing lenses, whereas the other systems depend on exogenous heparin. Nevertheless, our data confirm that many of the FGF-FGFR interactions require N-sulfated glucosamine residues in heparan sulfate, suggesting that Ndst1 inactivation could potentially disrupt multiple FGF-FGFR-signaling pathways during eye development. Recent studies demonstrating modest. or even no, lens defects in FGFR1, FGFR2 and FGFR3 single mutants support this idea (Garcia et al., 2005; Huang et al.,2003; Zhao et al.,2006). Therefore, our study of the Ndst1 gene provides an attractive model to unravel the complexity of FGF signaling in eye development.

The authors thank S. Conway, T. Glaser, C. Heldin, B. Hogan, G. Oliver, N. Takahashi, P. ten Dijke and Pfizer for mouse, antibodies, probes and reagents;E. Creshaw for help with immunohistochemistry; and M. Rauchman for helpful discussion. The work was supported by grants from March of Dimes Birth Defects Foundation (#5-FY05-49 to X.Z.); NIH (RO1 EY017061 to X.Z. and R37GM33063 and HL57345 to J.D.E.); and DFG (German Research Council) (GR1748/1-1 and GR1748/1-2 to K.G.).