Bud specific N-sulfation of heparan sulfate regulates Shp2-dependent FGF signaling during lacrimal gland induction

Lacrimal gland development follows the classic branching morphogenesis paradigm, which requires elaborate epithelial-mesenchymal interactions to induce selective epithelial budding. Previous studies have demonstrated that lacrimal gland morphogenesis is dependent on Fgf10 expressed in the periocular mesenchyme. In mice, both homozygous and heterozygous deletions of Fgf10 abolish lacrimal gland development(Entesarian et al., 2005; Makarenkova et al., 2000). Similarly in humans, autosomal dominant aplasia of lacrimal and salivary glands, or ALSG syndrome, is associated with FGF10 mutations(Entesarian et al., 2005). Conversely, exogenous Fgf10 can induce ectopic lacrimal gland formation in eye explant cultures, and transgenic expression of Fgf10 leads to lacrimal gland like structures in the cornea(Govindarajan et al., 2000; Makarenkova et al., 2000). Fgf7, a close homolog of Fgf10, is also expressed in the periocular mesenchyme. Although a lacrimal gland phenotype has not been observed in the Fgf7 deletion mutant, Fgf7 can also induce ectopic lacrimal glands in transgenic animals or explant cultures(Lovicu et al., 1999; Makarenkova et al., 2000). Biochemical studies have shown that both Fgf7 and Fgf10 strongly interact with Fgfr2b. Indeed, antisense DNA against Fgfr2b in ex vivo cultures can disrupt the budding outgrowth of the lacrimal gland epithelium,and the human lacrimo-auriculo-dento-digital (LADD) syndrome, a symptom of which is lacrimal duct aplasia, is caused by mutations in FGFR2 and FGFR3 genes (Makarenkova et al.,2000; Rohmann et al.,2006). Therefore, FGF signaling is essential for lacrimal gland development.

The FGF-FGFR interaction is known to require heparan sulfate proteoglycans at the cell surface. Heparan sulfates are linear glycosaminoglycan molecules modified by a series of sulfotransferase enzymes, which generate significant structural diversity among different tissues(Esko and Selleck, 2002). However, within each tissue, the composition of heparan sulfate is remarkably consistent, suggesting that the biosynthesis of heparan sulfate is tightly regulated during development (Ledin et al., 2004; Maccarana et al.,1996; van Kuppevelt et al.,1998). The sulfation of heparan sulfate is important for its interaction with FGF and FGFR, as evident in the studies of branching morphogenesis in the Drosophila trachea and vertebrate lung development. The Drosophila heparan sulfate N-deacetylase/N-sulfotransferase (Ndst) gene sulfateless is required for the FGFR-dependent MAPK activation during tracheal development, but not for EGFR-dependent MAPK signaling in the amnioserosa. Moreover, the trachea branching morphogenesis defect in the sulfateless mutant can be partially rescued by overexpression of FGF ligand (Lin et al., 1999). This is consistent with the idea that heparan sulfate is involved in stabilizing FGF-FGFR interactions. Drosophila Hs2st or Hs6stmutations abolish 2-O and 6-O-sulfation of heparan sulfate respectively;however, a compensatory elevation of sulfation levels exists at other residues. Probably because the overall sulfation level is maintained, Hs2st or Hs6st single mutants display only partial embryonic lethality with few tracheal defects(Kamimura et al., 2006). In Hs2st; Hs6st double mutants, however, the trachea branching morphogenesis is severely disrupted because of the loss of FGF signaling. In the vertebrate lung, it is observed that the branching epithelium tubules express sulfated heparan sulfate at high levels, and that chemical inhibition of heparan sulfate sulfation prevents FGF10-induced lung budding(Izvolsky et al., 2003). In addition, the Hs6st1 mutant mouse exhibits impaired lung alveolarization among other defects(Habuchi et al., 2007). Therefore, the role of heparan sulfate in FGF dependent branching morphogenesis is conserved in both fly and mouse.

Inside the cell, FGF signaling is transmitted by multiple pathways,prominent among which is the Raf/MAP-kinase kinase (MEK)/extracellular signal-regulated kinase (ERK) pathway. Here, Shp2, a ubiquitously expressed protein tyrosine phosphatase, positively regulates MAPK activity by modulating many of its components (Lai et al.,2004; Neel et al.,2003). In support of this model, Shp2-null mouse embryos die at the implantation stage because of a defective Src-Ras-MAPK pathway downstream of Fgf4 signaling (Yang et al.,2006). During mouse eye development, both lens and retina formation depend on FGF signaling mediated by Frs2α(Gotoh et al., 2004). Targeted mutations in the Shp2-binding site of Frs2α strongly reduce ERK signaling, thus disrupting lens induction. Recent studies have also suggested the role of Shp2 in branching morphogenesis. During embryonic lung development, Shp2 transcripts are predominantly detected at the epithelial bud cells, and expression of dominant negative Shp2 inhibits epithelial branching and Fgf10-induced endodermal budding(Tefft et al., 2005). Further studies show that the level of phospho-ERK is reduced and that the epithelial cell proliferation, but not differentiation, is impaired. Interestingly, the conditional knockout of Shp2 in the mammary gland does not affect normal branching and terminal budding in virgin females, but pregnancy-induced lobulo-alveolar outgrowth is disrupted due to a Stat signaling defect(Ke et al., 2006). Therefore,further studies are required in order to understand the function of Shp2 in branching morphogenesis.

We have recently shown that Ndst1 is required for FGF signaling during lens induction (Pan et al.,2006). In the present study, we further explore the extracellular and intracellular regulation of FGF signaling, and demonstrate that the lacrimal gland bud-specific modification of heparan sulfate by Ndstcontrols the selective Fgf10/Fgfr2 interaction at the cell surface. This leads to Shp2-dependent activation of ERK signaling inside the cell and preferential outgrowth of the lacrimal gland bud. Therefore, lacrimal gland budding is controlled by the Ndst-Fgfr2-Shp2 genetic pathway.

Ndst1, Ndst1^flox and Shp2^flox mice have been previously described (Grobe et al., 2005; Zhang et al.,2004). Fgfr2^flox mice were kindly provided by Dr David Ornitz (Washington University Medical School, St Louis, MO)(Yu et al., 2003). Ndst2 mice were kindly provided by Dr Lena Kjellén (University of Uppsala, Uppsala, Sweden) (Forsberg et al., 1999). Le-Cre mice were kindly provided by Drs Richard Lang (Children's Hospital Research Foundation, Cincinnati, OH), Peter Gruss (MPI for Biophysical Chemistry, Göttingen, Germany) and Ruth Ashery-Padan (Tel Aviv University, Tel Aviv, Israel)(Ashery-Padan et al., 2000). R26R mice were obtained from the Jackson Laboratory (Bar Harbor, ME)(Soriano, 1999). All experiments were performed in accordance with institutional guidelines.

Lacrimal glands in embryos carrying the Le-Cre transgene were visualized by their GFP expressions under a Leica MZ16F fluorescent dissecting microscope. Alternatively, lacrimal glands in newborn animals were revealed by aceto-carmine stain. Briefly, decapitated heads were dissected to reveal the lacrimal glands located in the facial subcutaneous tissue between the skin and cranial bones, then fixed in 4% PFA at 4°C overnight. After dehydration in 70% ethanol overnight, the heads were incubated in 0.5% aceto-carmine [0.5 g carmine stain (C-1022, Sigma, St Louis, MO) dissolved in 100 ml boiling 45%acetic acid] for 5-10 minutes, and destained in 70% ethanol for 3 minutes, 1%acid alcohol (1% HCl in 70% ethanol) for 2 minutes and 5% acid alcohol (5% HCl in 70% ethanol) for 1 minute. Lacrimal glands were examined under a Leica MZ16F dissecting microscope and photographed with a Leica DFC320 camera.

BrdU, TUNEL, X-gal staining and heparan sulfate immunohistochemistry were performed as previously described (Pan et al., 2006). For regular immunohistochemistry, paraffin sections were cleared in xylene and rehydrated through a series of decreasing ethanol solutions. Antigen retrieval was performed by microwave heating for 10 minutes at sub-boiling condition in citrate buffer (10 mM sodium citrate, pH 6.0). After cooling, endogenous peroxidase activity was quenched with 3%H₂O₂ in 10% methanol/PBS solution for 30 minutes, and non-specific interaction was blocked by 5% goat serum in PBS at room temperature for 1 hour. The sections were next incubated with primary antibody at 4°C overnight in a humid chamber, followed by sequential treatment with a biotin-conjugated secondary antibody and ABC reagent (Vectastain ABC Kit,Vector Laboratories, Burlingame, CA). Finally, the sections were incubated with DAB solution for color reaction.

The antibodies used were: anti-BrdU (G3G4, Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), anti-Cre (#69050-3, Novagen,Madison, WI), anti-phospho-ERK1/2 and anti-phospho-Shp2 (#9101 and #3751, Cell Signaling Technology, Beverley, MA), anti-GFP (a gift from Dr Pamela Silver,Harvard Medical School, Boston, MA), anti-Pax6 (PRB-278P, Covance, Berkeley,CA), anti-Shp2 (Sc-280, Santa Cruz Biotechnology, Santa Cruz, CA), and 10E4,3G10 (Seikagaku, Tokyo, Japan).

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). Briefly, mouse embryos were fixed in 4% PFA, cryoprotected in 30% sucrose buffer, and embedded in OCT compound. After sectioning on a Leica CM1900 cryostat, the samples were collected on glass slides and dried. RNA in situ probes dissolved in hybridization buffer (50% deionized formamide, 10% dextran sulfate, 1 mg/ml rRNA, 1× Denhardt's solution, 5× SSC, 5 mM EDTA)were heat denatured at 70°C for 10 minutes and applied to the sections. The slides were covered with a cover slip to reduce evaporation. After overnight incubation at 65°C in a humid chamber, the slides were rinsed three times in wash buffer (1× SSC, 50% formamide, 0.1% Tween-20) at 65°C, and twice in MABT buffer (100 mM maleic acid, 150 mM NaCl, pH 7.5,0.1% Tween-20) at room temperature. The sections were incubated with blocking solution (20% heat-inactivated sheep serum/2% blocking reagent (Roche,Indianapolis, IN) in MABT buffer) for 1 hour at room temperature. To detect the signal, anti-DIG antibody (Roche, Indianapolis, IN) was diluted 1:1500 in blocking buffer and added to the slides. After overnight incubation at 4°C, the slides were washed in MABT buffer and stained with BM purple AP staining solution (Roche, Indianapolis, IN).

The following probes were used: Erm and Pea3 (both from Dr Bridget Hogan, Duke University Medical Center, Durham, NC), Fgf10and Fgfr2b (both from Dr Suzanne Mansour, University of Utah, Salt Lake City, UT), and Ndst1 (Pan et al., 2006). At least three embryos of each genotype were analyzed for each probe.

In situ binding of the FGF-FGFR complex with heparan sulfate was carried out using the LACE assay as previously described(Allen and Rapraeger, 2003; Pan et al., 2006). Briefly,embryos were harvested and fixed in 4% PFA at 4°C overnight prior to paraffin embedding. Deparaffinized and rehydrated 5 μm sections were incubated in 0.5 mg/ml NaBH₄ for 10 minutes_, in 0.1 M glycine for 30 minutes and 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 15% fetal bovine serum in RPMI-1640 at 4°C overnight, followed by 2 hours incubation at room temperature with cy3-labeled anti-human Fc IgG secondary antibody. After washing in PBS, the slides were mounted with n-propyl gallate (NPG) antifading reagent, and examined using a Leica DM500 fluorescent microscope.

Lacrimal gland explant cultures were performed according to an established protocol (Makarenkova et al.,2000). After washing, 80-120 μm diameter heparin acrylic beads(Sigma, St Louis, MO) were incubated with 250 μg/ml recombinant FGF(R&D Systems, Minneapolis, MN) or BSA dissolved in PBS at 4°C overnight. E13.5-E14.5 embryos carrying the Le-Cre transgene identified by GFP expression were dissected in PBS to remove the ocular tissue, which includes the eye and the adjacent mesenchyme and ectoderm. The FGF- or BSA-soaked beads selected under microscope were nudged into the periocular mesenchyme using forceps. When placed on a Millipore filter (0.8μm pore), the eye rudiment floated on top of the culture media [CMRL1066 supplemented with 10% FBS, 4 mM L-Glutamine, 0.1 mM non-essential amino acids and antibiotics (Gibco, Carlsbad, CA)]. After 2 days' incubation at 37°C with 5% CO₂, the culture was examined for GFP-expressing lacrimal gland buds.

To perform genetic analysis of lacrimal gland development, we first characterized the Le-Cre transgene, which carries a Cre-IRES-GFP cassette driven by a Pax6 ectodermal enhancer active in the lacrimal gland (Ashery-Padan et al., 2000). In the Le-Cre transgenic mouse, GFP expression was clearly detected in the lens and corneal epithelium at E12.5 prior to lacrimal gland development (data not shown). At E14.5, visible at the temporal side of the eye was a small protrusion of lacrimal gland primordium,which continued to extend until E16.5 to form an elongated duct with a rounded tip at its distal end (Fig. 1A,B, arrowheads). At the time of extensive branching morphogenesis at P0, the Le-Cre GFP reporter expression illuminated the fine branches within the exorbital (arrow) and the intraorbital (arrow head) lobes of the lacrimal gland (Fig. 1C). The GFP expression of the Le-Cre transgene continued to be present in adult animals, faithfully following the development of the lacrimal gland (data not shown).

We next examined whether the Le-Cre transgene could also specifically target Cre expression in the lacrimal gland. For this,we performed immunohistochemistry on E14.5 embryo sections and observed co-expression of nuclear Pax6, cytoplasmic GFP and nuclear Cre recombinase in the lacrimal gland bud cells (Fig. 1D-F). Furthermore, we crossed the Le-Cre with a Cre reporter mouse, R26R, which expresses lacZ only after a stop cassette is removed by Cre-mediated recombination(Soriano, 1999). This experiment also allowed a genetic fate mapping of the Le-Cre-positive cells, because even transient expression of Cre under the direction of Le-Cre permanently marked all the progeny cells with constitutive lacZ expression. In the Le-Cre;R26R mice, X-gal staining showed that the lacZ reporter was expressed throughout lacrimal gland development, closely resembling the GFP expression pattern described above (data not shown). At birth, both intraorbital and exorbital lobes of the lacrimal gland were labeled with strong X-gal staining, which upon sectioning, was found to be present only in the Pax6-positive epithelial cells but not in any mesenchymal cells(Fig. 1G-I). This result demonstrates that the Le-Cre transgene can efficiently delete loxP flanked DNA in lacrimal gland cells. Furthermore, consistent with the idea that the lacrimal gland mesenchymal cells are derived from neural crest cells, our fate-mapping results confirm that the developing lacrimal bud is restricted to contribute only to the epithelial cells of the lacrimal gland (Johnston et al.,1979).

To investigate the role of heparan sulfate in lacrimal gland morphogenesis,we examined the distribution of both modified and unmodified heparan sulfate using the 3G10 antibody, which detects the heparan sulfate stub region exposed by Heparitinase I digestion (David et al.,1992). At E15.5, we observed ubiquitous 3G10 staining in many regions of the eye, including the Pax6-positive conjunctival epithelium and the protruding lacrimal bud (Fig. 2A, arrow). By contrast, the 10E4 antibody, which recognizes only the sulfated heparan sulfate, exhibited a similar albeit more restricted staining pattern (Leteux et al.,2001; van den Born et al.,2005). Although 10E4 still labeled both the lacrimal gland bud and its surrounding mesenchyme, the staining was much weaker in the posterior stalk region of the lacrimal gland when compared with the very tip of the lacrimal bud, with the strongest staining in the basolateral membranes of the tip epithelial cells (Fig. 2D,F, arrow). As a control, no 10E4 staining was observed in Heparitinase I-treated sections (data not shown). Therefore, the distal tip of the lacrimal bud expresses the specifically modified heparan sulfate.

This restricted 10E4 staining pattern indicates that heparan sulfate sulfation is differentially regulated during lacrimal gland development. We thus performed RNA in situ hybridization to investigate the expression of the heparan sulfate N-deacetylation and N-sulfation gene Ndst1, which is essential for generating the N-sulfated 10E4 epitope during embryonic development (Pallerla et al.,2007; Pan et al.,2006). Using an Ndst1 antisense probe, we indeed detected Ndst1 transcripts in the lacrimal gland buds with strongest expression in the Pax6-positive tip cells(Fig. 2G-I, arrowheads). As a control, the Ndst1 sense probe did not detect any signal (data not shown). Ndst1 expression thus correlates with the N-sulfation pattern of heparan sulfate in lacrimal budding.

The specific Ndst1 expression in the lacrimal gland bud raises the interesting issue of whether the Ndst1 gene and its sulfation function may play a regulatory role in lacrimal gland development. For this,we analyzed the systemic knockout of Ndst1(Grobe et al., 2005; Pan et al., 2006). At birth,carmine staining revealed extensively branched lacrimal glands in wild-type control pups (Fig. 2J, arrow). Ndst1^-/- animals, however, exhibited significant ocular defects with no detectable lacrimal gland(Fig. 2K, arrow). To confirm the absence of the lacrimal gland, we also crossed the Ndst1 null allele with the Le-Cre transgene. In Ndst1^-/-; Le-Cre pups at P0, again no GFP-positive lacrimal gland was observed(Fig. 2L, arrow). These results demonstrate that the heparan sulfate Ndst1 gene is essential for lacrimal gland development.

We next investigated the tissue specific requirement of Ndst1using a floxed allele of Ndst1(Grobe et al., 2005; Wang et al., 2005). For the mesenchymal-specific knockout of Ndst1, we employed the Wnt1-Cre transgene, which is specifically activated in the neural-crest derived mesenchymal cells(Danielian et al., 1998). This complemented the Le-Cre transgene mediated Ndst1 knockout in the lacrimal epithelial cells. In wild-type E14.5 embryos, both the distal lacrimal gland bud and its surrounding mesenchyme stained for 10E4 antibody(Fig. 3A). By contrast, 10E4 staining was restricted to the lacrimal gland epithelial cells in the Ndst1^flox/flox; Wnt1-Cre embryos and mesenchymal cells in the Ndst1^flox/flox; Le-Cre embryos,respectively (Fig. 3B, arrow; Fig. 3C, arrowhead). In the Ndst1^flox/flox; Wnt1-Cre; Le-Cre and the Ndst1^-/- embryos, 10E4 staining was completely abolished in the lacrimal gland region, whereas 3G10 staining remained unchanged(Fig. 3D, arrow; data not shown). Therefore, complementary abrogations of heparan sulfate sulfation were achieved with Wnt1-Cre or Le-Cre-mediated Ndst1knockouts.

To analyze the phenotypes of these tissue-specific Ndst1 deletion mutants, we next performed carmine staining. At birth, well-formed lacrimal glands were detected in both control and in the Ndst1^flox/flox; Wnt1-Cre mice without any overt ocular phenotype (Fig. 3E,F,arrowheads). By contrast, the Ndst1^flox/flox; Le-Cre pups were frequently micropthalmic without apparent lacrimal glands (Fig. 3G, arrowhead). To further characterize the phenotype, we also visualized the lacrimal gland by GFP expression from the Le-Cre transgene. In contrast to the extensively branched lacrimal glands in the control Le-Cre embryos,50% of Ndst1^flox/flox; Le-Cre newborn mice showed much reduced lacrimal glands (Fig. 3H,I, arrows, n=24). In another 46% of the Ndst1^flox/flox; Le-Cre animals, no lacrimal gland was observed, even though GFP expression was still detectable in the cornea(Fig. 3J, arrow, n=22). We suspected that the incompletely penetrant lacrimal gland defects in Ndst1^flox/flox; Le-Cre mice might be due to genetic redundancy of the Ndst2 gene(Holmborn et al., 2004). Indeed, by crossing the Ndst1^flox/flox; Le-Crewith Ndst2 animals, we showed that the double mutants, Ndst1^flox/flox; Ndst2^-/-; Le-Cre,completely abolished lacrimal gland development(Fig. 3K, arrow, n=18). Taken together, our results demonstrate that lacrimal gland development depends on Ndst-mediated heparan sulfate modification in the epithelial cells.

To determine the specific defect of lacrimal gland development in Ndst1 mutant embryos, we next examined the lacrimal gland budding during embryonic development. At E14.5, a small lacrimal gland bud identified by its GFP fluorescence had grown into the surrounding periocular mesenchyme in the Le-Cre transgenic embryo, but no such outgrowth was observed in the Ndst1^flox/flox; Le-Cre embryos under a fluorescent dissecting microscope (Fig. 4A,B, arrow and arrowhead). Fgf10-Fgfr2b and Bmp7 signaling are known to be important for lacrimal gland development, but the expression of Fgf10 and Bmp7 in the periocular mesenchyme, as well as of Fgfr2b in the conjunctival epithelium remained unchanged (Fig. 4C-F, arrows, and data not shown). Similarly, the 3G10 staining of total heparan sulfate in the conjunctival epithelium, which comprises the precursors of the lacrimal gland bud, was indistinguishable between the E12.5 control and Ndst1^flox/flox; Le-Cre embryos(Fig. 4G,H, arrows). In the fornix, or the deepest rim of the conjunctival epithelium, however, the Ndst1 mutant cells lacked 10E4 staining in both basal and lateral membranes (Fig. 4I,J, arrows,epithelium outlined with broken lines), although the epithelial cells further away from the eye retained 10E4 staining(Fig. 4I,J, arrowheads). Consistent with this early molecular defect, histological analysis of E14.5 Ndst1 mutant embryos showed either no budding or only very limited outgrowth of the lacrimal gland, probably reflecting the variable severities of the Ndst1 mutant phenotype(Fig. 4L, Fig. 5D-F). The presumptive lacrimal gland budding sites in the E14.5 Ndst1 mutants were also defective in cell proliferation as shown by the lack of BrdU staining,although no abnormal cell apoptosis was observed by TUNEL staining(Fig. 4L, arrow, and data not shown). By contrast, extensive BrdU incorporation was observed in the nuclei of control lacrimal glands (Fig. 4K, arrow). Therefore, the epithelial specific knockout of Ndst1 disrupted heparan sulfate modification and cell proliferation at the lacrimal gland primordium.

As FGF signaling promotes cell proliferation during branching morphogenesis, we next examined whether the lack of heparan sulfate modification in Ndst1 mutants disrupted FGF signaling. In a ligand and carbohydrate engagement assay (LACE), Fgf10 and Fgfr2b form a tight binding complex that is detectable by an antibody against the IgG Fc domain tag fused with Fgfr2b on E14.5 wild-type embryo sections(Fig. 5A)(Allen and Rapraeger, 2003; Pan et al., 2006). This Fgfr2b-binding pattern was completely lost in the absence of Fgf10 or on sections treated with Heparitinase I enzyme, demonstrating an obligatory complex of Fgf10, Fgfr2b and heparan sulfate in situ (data not shown). As expected, the immunofluorescence pattern closely resembled the heparan sulfate 10E4 staining pattern, with strongest staining in the cells at the tip of the lacrimal gland bud (Fig. 5A,arrow). In the Ndst1^flox/flox; Le-Cre embryos,however, the Fgf10/Fgfr2b staining was abolished in the epithelial cells of lacrimal gland but not in the surrounding mesenchymal cells(Fig. 5D, arrow). Similar bud-specific and Ndst1-dependent staining was also observed for the Fgf7/Fgfr2b combination, but no binding signal was observed for Fgf10/Fgfr2c,Fgf7/Fgfr2c or Fgf2/Fgfr2b (Fig. 5B,E, arrow and data not shown). Finally, we have previously shown that Fgf1/Fgfr2b binding is insensitive to the Ndst1 mutation during early lens induction. Indeed, we also detected in the lacrimal gland epithelial and mesenchymal cells almost ubiquitous Fgf1/Fgfr2b binding, which was unaffected by the Ndst1 mutation(Fig, 5C,F, arrow). Therefore,the Ndst1-dependent heparan sulfate modification specifically regulates Fgf10/Fgfr2b and Fgf7/Fgfr2b binding at the tip of the lacrimal gland bud.

We next performed functional tests of the LACE assay results in explant culture. Eye rudiments were dissected from E13.5 or E14.5 Le-Cremouse embryos and cultured for 2 days. Under fluorescent microscope,GFP-expressing lacrimal gland buds were observed extending from the edge of the cornea, sometimes with multiple branches at the distal tip(Fig. 5G,J, asterisk). Consistent with previous studies that Fgf10 is the primary inducer of the lacrimal gland, heparin acrylic beads soaked with Fgf10 induced ectopic lacrimal gland buds, while beads soaked in BSA did not induce bud growth(Fig. 5J, arrowheads). In the Ndst1 mutants, however, the budding rate of endogenous lacrimal gland was reduced from 63% to 18%, confirming the in vivo finding that Ndst1 was required for lacrimal gland development(Fig. 5H,K, P<0.001, Fisher's exact test). Interestingly, the budding of ectopic lacrimal gland was not significantly reduced in the Ndst1mutants (P=0.246, Fisher's exact test), suggesting that the concentrated Fgf10 protein released from beads may partially compensate for Ndst1 inactivation in the culture. This is reminiscent of the rescue of tracheal branching morphogenesis in Drosophila sulfateless mutants by FGF overexpression (Lin et al.,1999). Finally, we further reduced the extent of heparan sulfate sulfation by compounding Ndst1 and Ndst2 mutations. In the Ndst1^flox/flox;Ndst2^-/-; Le-Cre double mutants, both endogenous and ectopic lacrimal glands were completely abolished in the explant culture experiments, demonstrating the essential role of Ndst genes in Fgf10-induced lacrimal gland budding (P<0.001 for endogenous budding, P<0.02 for ectopic budding, Fisher's exact test). Therefore, Fgf10 depends on Ndst function to promote lacrimal gland development.

We next sought to provide genetic evidence that Fgfr2, the strong binding partner of Fgf10 in LACE assay, is also required for lacrimal gland budding. The Le-Cre transgene was crossed with an Fgfr2^flox allele, which carries two loxP sites flanking exons 7-10 of Fgfr2 gene(Yu et al., 2003). As expected, although the Le-Cre control pups at birth clearly formed the lacrimal gland as marked by abundant GFP expression, none of the Fgfr2^flox/flox; Le-Cre animals developed lacrimal glands (Fig. 6A,B). The mutant phenotype can be traced to lacrimal gland induction, as budding of the lacrimal gland was observed in Le-Cre controls but not in the Fgfr2^flox/flox; Le-Cre embryos(Fig. 6C,D). By contrast, no lacrimal gland defect was observed in the mesenchymal specific Fgfr2knockout, Fgfr2^flox/flox; Wnt1-Cre(Fig. 6E,F). These results demonstrate the essential role of Fgfr2 in lacrimal gland epithelium cells for budding morphogenesis.

To further investigate the lacrimal budding defect in Fgfr2^flox/flox; Le-Cre mutants, we performed the BrdU incorporation assay to examine the cell proliferation. At E14.5 when the lacrimal bud had just emerged from the conjunctival fornix, the cells within the tip of the bud were rapidly dividing as shown by incorporation of BrdU in their nuclei (Fig. 6G,arrowhead). In the Fgfr2^flox/flox; Le-Cremutants, however, the presumptive lacrimal gland precursors remained as a thin layer of epithelial cells with very little BrdU incorporation(Fig. 6H). Similar to Ndst1 mutants, TUNEL assay did not detect excessive cell death in the Fgfr2^flox/flox; Le-Cre embryos (data not shown). Finally, we hypothesized that, as a primary inducer of lacrimal gland, FGF signaling may also be responsible for the upregulation of heparan sulfate modification in the lacrimal gland bud tip cells. This cannot be directly tested in the Fgfr2^flox/flox; Le-Cre embryos at E14.5 as these mutants never develop lacrimal buds. At E12.5, however, the conditional knockout of Fgfr2 by Le-Cre abolished FGF signaling in the fornix of the conjunctival epithelium but not in the ectodermal cells further away from the eye. Indeed, although the far side of ectodermal epithelium still stained for 10E4 antibody in the basal-lateral membranes, this heparan sulfate staining was lost in the Fgfr2 mutant conjunctival fornix cells (Fig. 6I,J, arrow and arrowheads, also outlined). Thus, FGF signaling promotes heparan sulfate modification in lacrimal gland precursor cells.

Activated FGF receptors can tyrosyl phosphorylate Shp2 protein, which further recruits downstream molecules required for MAPK signaling(Hadari et al., 1998). We thus examined the activity of Shp2 protein using a phospho-specific antibody. Not surprisingly, phospho-Shp2 expression was detected in the control lacrimal gland bud, but not in the Fgfr2 mutant conjunctival epithelium, while the retinal phospho-Shp2 remained unaffected(Fig. 6K,L). This suggested that Fgfr2 inactivation may result in downregulation of MAPK signaling during lacrimal gland induction.

To further explore the FGF downstream signaling, we deleted Shp2 in the lacrimal gland by crossing the Le-Cre transgene with a Shp2^flox allele that, upon Cre mediated recombination,creates a Shp2-null allele (Zhang et al., 2004). Elimination of Shp2 was demonstrated by immunohistochemistry using an antibody specific to the C-terminal domain of Shp2. Compared with the ubiquitous expression in control embryos, Shp2 in the Shp2^flox/flox; Le-Cre embryos was specifically lost in the conjunctival epithelium, as expected with the epithelial-specific deletion of Shp2 (Fig. 7A,B, broken lines). Similar to the Fgfr2 and Ndst1 knockouts, the Shp2 mutants never exhibited much cell proliferation in the fornix of the conjunctival epithelium, as shown by the BrdU incorporation assay or abnormal cell apoptosis as revealed by TUNEL assay results (Fig. 7C,D; data not shown). As a consequence, there was no detectable lacrimal gland budding at E14.5 or any lacrimal gland structure at birth(Fig. 7F,H versus control Fig. 7E,G). Importantly, the 10E4 staining of heparan sulfate was disrupted in the fornix of the conjunctival epithelium in the Shp2 mutants(Fig. 7I,J, arrows). Thus, Shp2 function is also essential for heparan sulfate modification during lacrimal gland development.

We next assayed MAPK signaling activities in Ndst1, Fgfr2 and Shp2 mutants. As a downstream effector of MAPK signaling, phospho-ERK was expressed in the control E12.5 conjunctival fornix, which contained the thickening precursor cells of the lacrimal gland(Fig. 7K, arrow). In all three mutants, no induction of phospho-ERK expression was detected in the same region (Fig. 7L-N, arrowheads). In the control E14.5 lacrimal gland, the strongest phospho-ERK expression was at the very tip of the bud, which correlated with the cell proliferation pattern shown in the BrdU assay as well as heparan sulfate pattern by 10E4 staining (Fig. 7O, arrow). By contrast, the Le-Cre-mediated conditional knockout of Ndst1,Fgfr2 and Shp2 abolished phospho-ERK expression at the conjunctival epithelium, even though retinal expression of phospho-ERK remained unaffected (Fig. 7P-R). The PEA3 family transcription factors, including Pea3 and Erm are well characterized FGF signaling downstream genes. By RNA in situ hybridization, we also observed a complete loss of Pea3 and Erm expressions in the tip of lacrimal gland primordium in all three mutants (Fig. 7S-V, and data not shown). Taken together, our results support that Ndst1, Fgfr2 and Shp2 potentiate FGF-ERK signaling in lacrimal gland development.

The spatial relationships between heparan sulfate, ligands and receptors are still not well understood. A previous study has suggested that heparan sulfate acts in trans to promote VEGF signaling in endothelial cells, and,similarly, a cell-nonautonomous function of heparan sulfate has also been proposed for FGF signaling (Jakobsson et al., 2006; Richard et al.,1995; Sherman et al.,1998). By genetic manipulation, however, we have recently shown that heparan sulfate probably functions cell-autonomously in tumor vasculogenesis (Fuster et al.,2007). In this study, tissue-specific knockout experiments further show that both heparan sulfate and Fgfr2 are required in the lacrimal gland epithelium, and that mesenchymal knockouts of Ndst1 and Fgfr2 do not perturb lacrimal gland development. Therefore, at least for lacrimal gland FGF signaling, heparan sulfate appears to function in cis with Fgfr2 in the lacrimal gland epithelium.

We have also provided evidence that the heparan sulfates in the lacrimal gland epithelium are differentially modified. Despite the ubiquitous heparan sulfate expression throughout the length of the lacrimal gland, only the distal tip of the lacrimal gland bud contains uniquely sulfated heparan sulfate as identified by the 10E4 antibody. Notably, this heparan sulfate modification pattern correlates exactly with the in situ binding of Fgf10/Fgfr2b and Fgf7/Fgfr2b at the lacrimal gland bud, shown by LACE assay,and localized activation of FGF signaling, as shown by phospho-ERK staining. By contrast, Fgf1 and Fgf2 exhibit either ubiquitous binding or very weak binding respectively, with Fgfr2b in the developing lacrimal gland. Finally,genetic ablation of Ndst1 and Ndst2 genes disrupted endogenous lacrimal gland development in vivo and Fgf10 induced ectopic lacrimal gland budding in explant culture, demonstrating the essential role of heparan sulfate N-sulfation in Fgf10/Fgfr2b signaling. By its developmentally dynamic expression pattern, heparan sulfate could thus not only potentiate but also restrict FGF signaling within the lacrimal gland tip,providing a mechanism to promote directional outgrowth of the lacrimal gland bud.

How then is the tissue specific heparan sulfate modification generated? It is known that heparan sulfate biosynthetic genes are differentially expressed during development, and at least in the case of Ndst genes,post-transcriptional regulation also plays a role in their expression(Grobe and Esko, 2002). However, it remains to be determined what upstream signaling regulates these heparan sulfate modification genes. In this study, our data show that heparan sulfate modification itself depends on functional Fgfr2 and Shp2 in lacrimal gland primordium, suggesting reciprocal regulation between heparan sulfate and FGF signaling. This feedback model is attractive because it may help to stabilize the Fgf10 responsive zone to the tip of lacrimal bud with specific heparan sulfate modification.

In summary, we have provided genetic evidence that specifically modified heparan sulfate at the cell surface plays a functionally important role in lacrimal gland development by spatially regulating Fgf10/Fgfr2b signaling(Fig. 8). This leads to activation of intracellular ERK signaling mediated by Shp2 phosphatase and preferential cell proliferation at the lacrimal gland bud. As restrictive heparan sulfation modification is widely observed in development, the genetic cascade of Ndst-Fgfr2-Shp2 may thus be a general mechanism in the branching morphogenesis process.

The authors thank Drs Bridget Hogan, Lena Kjellén, Richard Lang,Peter Gruss, Suzanne Mansour, David Ornitz, Ruth Ashley-Padan and Pamela Silver for mice and reagents; Dr Lixing Reneker for help with immunohistochemistry; and Dr Rebecca Chan for discussion. The work was supported by grants from March of Dimes Birth Defects Foundation (#5-FY05-49 to X.Z.) and from NIH (EY017061 to X.Z., R37GM33063 and HL57345 to J.D.E., and CA78606 to G.S.F.).