SULF1 and SULF2 regulate heparan sulfate-mediated GDNF signaling for esophageal innervation

Heparan sulfate (HS) regulates a number of extracellular signaling pathways that are essential for embryonic development(Lin, 2004). HS is a linear polysaccharide chain covalently attached to a protein core of heparan sulfate proteoglycans (HSPGs) and consists of repeating disaccharide units of uronic acid linked to glucosamine (Esko and Lindahl, 2001). During HS biosynthesis in the Golgi, disaccharide units can be sulfated at the 2-O position of uronic acid and N-, 3-O and 6-O positions of glucosamine by specific sulfotransferases. Sulfation along the HS chain is incomplete, thus creating highly sulfated domains separated by partially sulfated and non-sulfated domains. The highly sulfated domain interacts with a variety of extracellular signaling ligands and receptors and is therefore considered the major signaling domain of HS. However, owing to a lack of biochemical tools and genetic models through which specific sulfate groups of HS can be modified, it remains controversial whether such interactions are determined non-selectively by the overall negative charge, or selectively through establishment of fine sulfated HS sequences - an `HS code'-within the highly sulfated domain(Kreuger et al., 2006).

Sulfs are newly discovered extracellular heparan sulfate 6-O-endosulfatases that have unique structural features, enzymatic activities and signaling functions (Ai et al., 2005). Sulfs contain the essential enzymatic sequences conserved among all sulfatases, as well as distinct hydrophilic sequences that are required both to dock Sulfs on the cell surface and for their enzymatic activities(Dhoot et al., 2001; Ai et al., 2006). SULF1 and SULF2 in vertebrates have similar substrate specificity towards a selective subset of 6-O-sulfate groups within the highly sulfated domain of HS chains,implicating these enzymes as regulators of the `HS code'(Morimoto-Tomita et al., 2002; Ai et al., 2003). Sulfs remodel HS 6-O-sulfation pattern on the cell surface to regulate HS binding to signal ligands and receptors in a diversity of signaling pathways, including Wnt, FGF and HGF, BMP and SHH (Ai et al.,2005). Sulfs are dynamically expressed in the embryonic tissues(Dhoot et al., 2001; Danesin et al., 2006; Lum et al., 2007). Avian SULF1 controls Wnt-dependent myogenic specification and is implicated in SHH-regulated oligodendroglial specification(Dhoot et al., 2001; Danesin et al., 2006).

The functions of Sulfs in mammals are unknown. Recent studies report that Sulf single- and double-mutant mice appear normal at birth, and Sulf2^-/- and Sulf1^-/-;Sulf2^-/- mice have reduced body weight and double-mutant mice die soon after birth(Lum et al., 2007; Lamanna et al., 2006),although the cellular basis for this growth phenotype was not identified. To investigate the developmental signaling functions of Sulfs, we established independent lines of Sulf mutant mice using gene targeting. We identify a primary neuronal innervation defect of Sulf1^-/-;Sulf2^-/- esophagi, providing an explanation for the severe growth defects in these mice. Furthermore, we show that this esophageal defect in Sulf1^-/-;Sulf2^-/- mice is due to aberrant GDNF signaling.

The neuronal innervation of muscles along the gastrointestinal tract is controlled by GDNF, a HS-dependent neurotrophic factor derived from the target muscles (Baloh, et al., 2000; Barnett et al., 2002; Rickard et al., 2003). In the embryonic esophagi, muscle progenitors express GDNF beginning at embryonic day(E) 10, peaking between E11 and E16 and diminishing at E18(Golden et al., 1999). GDNF not only promotes the proliferation of the enteric neural crest precursors and support their neuronal and glial differentiation(Heuckeroth et al., 1998), but also acts as a target-derived chemoattractant for directed neurite outgrowth of both `intrinsic' enteric neurons and `extrinsic' neurons whose cell bodies are located in ganglia outside of the esophagus(Young et al., 2001; Yan et al., 2004). Neurons reach the target muscle during the initiation of GDNF expression in the embryonic esophagus (Durbec et al.,1996; Sang and Young,1997), and the formation of functional innervations and enteric glial cells initiates in the embryo and proceeds postnatally until completion around two weeks after birth (Sang and Young, 1997; Breuer et al.,2004). The esophageal muscle in the muscularis externa (ME)matures from smooth muscle to skeletal muscle along with the neuronal innervation (Rishniw et al.,2003). In the adult esophagi, the striated muscle in the ME is innervated by both intrinsic and extrinsic neurons(Sang and Young, 1998; Neuhuber et al., 2006),whereas the smooth muscle in the muscularis mucosae is directly innervated mostly by intrinsic neurons, which elicit the smooth muscle contractility in response to extrinsic nerves, as suggested by previous electrophysiological studies (Kamikawa and Shimo,1979; Storr et al.,2001; Worl et al.,2002). Defects in neural innervation can lead to neonatal death and a variety of esophageal disorders, such as achalasia and a motility disorder, congenital idiopathic megaesophagus(Longstretch and Walker, 1994; Neuhuber et al., 2006).

GDNF signaling requires 2-O- and 6-O-sulfate groups of HS or heparin, a highly sulfated HS derivative, for binding to HS and for GDNF signaling(Barnett et al., 2002; Rickard et al., 2003). Whether Sulfs regulate the GDNF signaling pathway has not been investigated. In this study, we establish that Sulfs are required for GDNF signaling during esophageal innervation. We show that Sulf mutant HS has a highly selective increase of the trisulfated disaccharides without affecting other structural properties, establishing Sulfs as major regulators of cellular HS 6-O-desulfation. Furthermore, SULF1 and SULF2 are dynamically and differentially expressed by the GDNF-expressing muscle progenitors and neuronal progenitors to promote GDNF-mediated neurite sprouting. By further demonstrating that neurite outgrowth of the esophageal explants requires GDNF,but not neurotrophins or other previously known Sulf-regulated signaling ligands, we establish functional specificity of Sulfs in the GDNF pathway during esophageal innervation. Our findings provide the first evidence that Sulf enzymes are in vivo regulators of an `HS code' that controls HS fine sulfated structures to coordinate the responses of Sulf-expressing cells to developmental signals.

Sulf mutant mice were generated as described (see Figs S1, S2 in the supplementary material) (Tompers and Labosky, 2004). Genotyping was performed by DNA extraction from tail followed by PCR. The Sulf single-mutant mice, maintained as heterozygous,were backcrossed into C57BL6 background (sixth generation). The Sulf1^-/- and Sulf2^-/- mice were mated to generate the double-heterozygous mutants. Sulf1^-/-;Sulf2^-/- mice or embryos were generated by crossing Sulf1^+/-;Sulf2^+/- mice with Sulf1^-/-;Sulf2^+/- mice or by mating between Sulf1^-/-;Sulf2^+/- mice. The Sulf1^-/-;Sulf2^-/- mice were in a mixed 129SvEv/C57BL/6 genetic background.

The lower-half thoracic segment of the esophagus was dissected from control and Sulf1^-/-;Sulf2^-/- mice, rinsed gently in ice-cold PBS and fixed in periodate-lysine-paraformaldehyde (PLP) for 2 hours at 4°C. Tissues were then embedded in OCT and 10 μm serial cryosections were collected. Sections were stained with Hematoxylin and Eosin (H&E). Adjacent, unstained sections were used for immunohistochemical studies (see below).

Immunohistochemistry on tissue sections was performed as described previously (Ai et al., 2003),except for an additional 1-hour blocking with MOM blocking reagent (Vector)for mouse monoclonal antibodies. Antigen-antibody complexes were detected either by fluorescence or by chromogenic substrate. For explant cultures or whole-mount staining, the washes were extended to 30 minutes each. Primary antibodies included: (1) rabbit anti-glial fibrillary acidic protein (GFAP)(Dako; 1:100); (2) rabbit anti-MS1HD (1:100); (3) rabbit anti-MS2HD (1:100);(4) mouse anti-skeletal fast myosin (clone MY-32) alkaline phosphatase conjugate (Sigma; 1:300); (5) goat anti-GDNF (R&D Systems; 2 μg/ml);(6) mouse anti-neuron-specific class III β-tubulin antibody (clone TuJ1)(R&D Systems; 1:500); (7) rabbit anti-smooth muscle-specific SM22 (a gift from Dr Mario Gimona, Austrian Academy of Sciences; 1:1000); (8) rabbit anti-p75 (Upstate; 1:200); (9) goat anti-GFRα1 (R&D systems; 1μg/ml); (10) rabbit anti-RET (Santa Cruz Biotechnology; 1:50); (11) rabbit anti-phosphorylated p44/42 MAPK (Cell Signaling; 1:500).

The ∼400 μm esophagus was dissected from E11.5 mouse embryos and cultured in 500 μl growth medium (DMEM plus 10% fetal bovine serum;Mediatech) on presolidified collagen gel in a 24-well plate. GDNF, NGF, BDNF,NT3 or NT4 (R&D Systems) were added immediately to the collagen/DMEM mixture at various concentrations. Esophagus explants were cultured on collagen gel for 4-5 days before fixation with 4% paraformaldehyde in PBS followed by immunohistochemistry as described above. After antigen-antibody complexes were detected by diaminobenzine (DAB) substrate, the explants were examined using a Nikon TS100 microscope and photographed using a Nikon E4300 camera to show the whole explant with extended neuritis. To quantify the number of neurons in the explant, immunostained explants were examined using a Leica DMR microscope and photographed using a Leica DC300F camera.

Cells were cultured in growth medium plus 1% antibiotics (Gibco). To activate signaling, cells (1×10⁵ per well of a 24-well plate)were serum-starved in DMEM for 6 hours before adding GDNF or NGF at various concentrations. After stimulation, cells were lysed for western blot or for co-immunoprecipitation.

Immunoblot analysis was as described previously(Dhoot et al., 2001). The intensity of the signal was quantified by Multi-analysis software (Bio-Rad). The primary antibodies for immunoblots included: rabbit anti-MAPK (1:4000;Sigma), mouse anti-phosphorylated MAPK (1:1000; Sigma), rabbit anti-phosphorylated AKT (1:1000; Cell Signaling), rabbit anti-GDNF (0.5μg/ml); mouse anti-phosphorylated Tyr (1:1000; Calbiochem); rabbit anti-RET(1:500).

Esophagus was dissected from P12 or adult mice. To measure contractions of skeletal muscles, the 5-mm longitudinal esophagus slices were cut out, tied with silk threads at both ends and hooked up to tungsten needle tips. One of the needles was connected to a force transducer (AM801, SensoNor, Horten,Norway) and the other to a micromanipulator to stretch the muscle length to 1.1 times the slack. Platinum electrode wires were placed near the tubes on both sides. The esophagus slices were stimulated by 1-millisecond duration square pulses (TSS20, Intracel, UK), one single pulse for twitch and 30 Hz for 1 minute for tetanus. To measure contractions of smooth muscle, the smooth muscle cross-rings were dissected from 1-mm esophagus tubes and then hooked up to two needle tips. The rings were stretched to 1.2-1.3 times the slack. The muscle preparations were then immersed into the normal external solution [150 mM NaCl, 4 mM KCl, 2 mM Ca-methanesulphonate (Ca-Ms), 2 mM Mg-Ms, 5.6 mM glucose, 5 mM N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid, pH 7.4] in a well on a bubble plate to allow for rapid solution exchange. Force records were amplified with a bridge-amplifier, which was connected to a data-acquisition system (PowerLab, ADInstruments, Colorado Spring, CO). Depolarizing external solution had K-Ms substituted equally for NaCl with other chemicals in the same concentrations. Compounds used in the recoding were applied at the following final concentrations: 10 μM atropine, 2μg/ml alpha-BTx, 100 μM ATP, 124 mM K⁺, 30 μM histamine and 30 μM carbachol. All experiments were carried out at 30°C.

SULF1 digestion of heparin or heparin-conjugated agarose beads (Sigma) was performed as described previously (Ai et al., 2006). To allow GDNF binding to heparin, 20 μl of digested heparin-conjugated agarose beads were incubated with GDNF in PBS (a total volume of 100 μl) at room temperature for 1 hour. The amount of GDNF bound to the heparin-conjugated agarose beads was assayed by immunoblot analysis. To allow GDNF binding to GFRα1, 10 ng GDNF, 1 μg GFRα1-Fc and heparin were mixed in 50 μl PBS for 30 minutes at room temperature. The GDNF-heparin-GFRα1-Fc complex was purified with 10 μl protein A-agarose beads. The amount of GDNF bound to GFRα1 was assayed by immunoblot analysis.

Mouse embryonic fibroblasts (MEFs) were isolated from the skin of E14.5 mouse embryos after dissociation with dispase II (Boehringer Mannheim; 2 mg/ml). MEFs were cultured in DMEM with 10% fetal bovine serum. The radiolabeling, preparation and structural analysis of the HS were performed as described previously (Ai et al.,2003).

To investigate Sulf regulation of developmental signaling, we generated mice with targeted deletions of the second coding exon (exon 2) of Sulf1 and Sulf2 using Cre-loxP technology (see Figs S1, S2 in the supplementary material). Exon 2 encodes essential amino acids in the enzymatic domains of SULF1 and SULF2 (Dhoot et al., 2001). Although Sulf mutant embryos express partial Sulf transcripts and polypeptides recognizable by anti-Sulf antibodies (see Fig. S1E and Fig. S2D in the supplementary material; data not shown), removal of exon 2 eliminates the full-length Sulf transcripts (see Fig. S1E and Fig. S2D in the supplementary material) and disrupts Sulf protein function, as shown by the disaccharide analyses of ³⁵S-radiolabeled glycosaminoglycans(GAGs) (Fig. 1A). GAGs were isolated from cultured primary mouse embryonic fibroblasts (MEFs) of the E14.5 skin that express both SULF1 and SULF2 (see Fig. S1E and Fig. S2D in the supplementary material). Loss of Sulf1 selectively increased the abundance of the trisulfated disaccharide IdoA2S-GlcNS6S (ISMS), the known substrate of Sulf enzymes, by ∼30% (Ai et al., 2003), and proportionately decreased the disulfated IdoA2S-GlcNS (ISM) disaccharide, whereas it had no effect on the other two disaccharides, GMS and IMS (Fig. 1A, Table 1). By contrast, Sulf2^-/- and Sulf1^+/-;Sulf2^+/- MEFs had unchanged levels of ISMS disaccharide, suggesting redundant Sulf protein activity in MEF cells. The magnitude of the increase in ISMS in various Sulf mutant MEFs was generally proportional to the number of mutant Sulf alleles(Table 1). Most significantly,HS isolated from Sulf1^-/-;Sulf2^-/- MEFs had double the ISMS content (63%), a level comparable to that of heparin, a highly sulfated HS derivative. Sulf mutant MEFs of all genotypes had normal HS chain length, anionic properties, N-sulfation and chondroitin sulfate contents (data not shown). These results establish that Sulf mutants have loss-of-function mutations, and that Sulf enzymes are the major regulators of HS 6-O-sulfation in vivo.

The Sulf single-mutant mice are viable, fertile and have a normal life span of over 2 years (Fig. 1B, Table 1); however, Sulf2^-/- mice weighed ∼10% less than wild-type littermates. The Sulf1^-/-;Sulf2^-/- embryos and neonates appeared normal at birth (data not shown) and fully viable based on observed mendelian frequencies at E11.5 (14/107, expected 1/8), at E14.5(14/69, expected 1/4) and at birth (34/311, expected 1/8). We identified a primary developmental defect in esophageal innervation of Sulf^-/-;Sulf2^-/- embryos that is likely to cause the severe postnatal growth defect of a majority of Sulf1^-/-;Sulf2^-/- pups (41/76) that was evident as early as postnatal day 3 (P3) (Fig. 1B), with death at ∼P14. Sulf1^-/-;Sulf2^-/- pups that survived into adulthood (35/76) were runted and had greatly reduced fertility(Fig. 1B, Table 1). Approximately 60% of adult Sulf1^-/-;Sulf2^-/- mice developed megaesophagus phenotypes as early as 2 months of age, characterized by food accumulation in the esophagus, coughing, labored breathing and lung infection(Fig. 1C). The esophagus of Sulf1^-/-;Sulf2^-/- mice had an enlarged lumen with a normal arrangement of the muscle layers and esophageal epithelium by H&E staining and by immunostaining with specific markers for each cell type, although these tissue layers appeared much thinner owing to the dilation of the esophagus (Fig. 1C; data not shown). We did not observe inflammatory cells in the Sulf1^-/-;Sulf2^-/- esophagi at P12 or in surviving adults (data not shown), ruling out the possibility of immune cell-mediated tissue injury. In addition, we did not detect compensatory expression of the intact Sulf gene in single-mutant esophagi (data not shown).

As the megaesophagus phenotype was prevalent only in Sulf1^-/-;Sulf2^-/- mice, and not in Sulf1^-/- (0%) or Sulf2^-/- mice(∼1-2%), this suggested that SULF1 and SULF2 have redundant functions that contribute to the megaesophagus phenotype in Sulf1^-/-;Sulf2^-/- mice. To investigate this possibility, Sulf1 and Sulf2 expression in embryos was assayed by in situ hybridization and by immunostaining using specific antibodies generated against the hydrophilic domains of SULF1 (MS1HD) and SULF2 (MS2HD) (see Fig. S3A,B in the supplementary material). The tissue distribution of Sulf mRNA completely overlaps with protein expression,establishing the specificity of the Sulf antibodies(Fig. 1 and see Fig. S3C in the supplementary material) and consistent with Sulfs being membrane-docking and lacking free secretion (Dhoot et al.,2001). The expression levels of Sulfs were relatively low before E9.5 (data not shown). At E14.5 and E16.5, SULF1 and SULF2 were expressed at higher levels in partially overlapping patterns in a variety of embryonic tissues, including the floor plate of the neural tube, bone, cartilage,skeletal muscle and lung (see Fig. S3C in the supplementary material)(Lum et al., 2007).

SULF1 and SULF2 were found to be dynamically and differentially expressed in the embryonic esophagus (Fig. 2). They were first detectable around E11.5, peaked around E14.5,decreased dramatically by E18.5 and were undetectable 2 weeks after birth and in the adult (Fig. 2; data not shown). SULF1 expression did not co-localize with neuronal β-tubulin(TUBB3 - Mouse Genome Informatics), as labeled by the TuJ1 antibody(Fig. 2F). Instead, Sulf1 mRNA and protein were detected at the esophageal ME at E14.5,when the smooth muscle forms prior to maturation into the skeletal muscle(Fig. 2B-E). To confirm muscle expression of SULF1, double labeling was performed with antibodies against SULF1 and the smooth muscle marker SM22 (TAGLN - Mouse Genome Informatics). Although both antibodies were raised in rabbit, we distinguished membrane-bound SULF1 from intracellular SM22 by their differential subcellular localization (Fig. 2C-E). We found that SULF1 outlined the membrane of SM22-expressing cells at the outer layer of E14.5 esophagus, and was localized at both outer and inner muscle layers of the esophagus at E16.5 (Fig. 2H and see Fig. S3D in the supplementary material). Esophageal muscle progenitors also express GDNF(Golden et al., 1999). To test whether SULF1 and GDNF co-localize, we characterized GDNF expression in the esophagus at E14.5 and E16.5 for comparison with SULF1 expression. GDNF was found on the cell surface and appeared diffusely across the esophageal ME at E14.5 (Fig. 2G), including the SULF1-expressing outer layer, and overlapped with SULF1 at E16.5(Fig. 2H-J). By contrast, SULF2 was detected within the esophageal muscle layers and was tightly associated with the neuronal marker TuJ1 (Fig. 2K-N and see Fig. S3E in the supplementary material). As SULF2 and neuronal β-tubulin have differential subcellular localization, the observed close association between these two proteins indicates that SULF2 is expressed by innervating neuronal progenitors. The differential expression of SULF1 and SULF2 by muscle progenitors and by innervating neural progenitors suggests Sulf regulation of neuron-muscle interaction in the embryonic esophagi.

The esophageal defects in Sulf1^-/-;Sulf2^-/-mice may result from impaired muscle contractility owing to defects in muscle maturation or in neuronal innervation. To distinguish these possibilities, we characterized the formation of the esophageal muscle(Fig. 2). The striated muscles of ME expressed the myogenic genes - including those encoding myosin heavy chain [MHC (MYHS)], MYF5, MYOD (MYOD1 - Mouse Genome Informatics) and myogenin- normally at P14, when Sulf1^-/-;Sulf2^-/- pups start to die (Fig. 3A; data not shown). Also, Sulf1^-/-;Sulf2^-/- esophagi complete skeletal muscle maturation normally at P15(Rishniw et al., 2003), as shown by the expression of fast skeletal myosin in the esophageal abdominal segment connecting to the stomach (Fig. 3A). In addition, Sulf1^-/-;Sulf2^-/-mice that survived into adulthood expressed fast skeletal muscle myosin and nicotinic acetylcholine receptor clusters normally in the esophageal striated muscle (Fig. 3A; data not shown). Furthermore, isolated thoracic longitudinal segments of the adult Sulf1^-/-;Sulf2^-/- esophagi produced a comparable contractile force in response to twitch and tetanus electrical stimuli with those of controls (Fig. 3B) (Worl et al.,2002). To distinguish forces generated by nerve stimulation from forces generated from direct muscle stimulation, we applied alpha-bungarotoxin(alpha-Btx) to block electrical stimuli-induced nerve input by inhibiting postsynaptic nicotinic acetylcholine receptors. Alpha-Btx significantly inhibited both twitch and tetanus forces in both Sulf1^-/-;Sulf2^-/- and control esophagi in a similar manner, demonstrating that the skeletal muscle innervation is normal in Sulf1^-/-;Sulf2^-/- esophagi. After the treatment with alpha-BTx, the remaining forces were resistant to atropine, an inhibitor of muscarinic receptors on smooth muscle(Fig. 3B), but were completely inhibited by further addition of tetrodotoxin in Sulf1^-/-;Sulf2^-/- and control esophagi (data not shown). These results establish that the electrical stimuli-induced forces are produced largely by skeletal muscle, not by smooth muscle, in the esophagus. The electrical stimuli elicited relatively low levels of smooth muscle contractility, as evidenced by the small shoulders after the tetanus stimuli and by their sensitivity to inhibition by atropine(Fig. 3B). We observed no significant difference in skeletal muscle contractility between the double-mutant and control esophagi. Therefore, the development, maturation and function of esophageal skeletal muscle is unaffected in Sulf1^-/-;Sulf2^-/- mice.

By contrast, smooth muscle contractility was found to be impaired in Sulf1^-/-;Sulf2^-/- esophagi. The contractile forces of the isolated smooth muscle ring at the thoracic segments of the esophagi were elicited by various stimuli such as electrical, high K⁺, carbachol, ATP and histamine(Storr et al., 2001; Worl et al., 2002). The smooth muscle contractility elicited by electrical stimuli (the small shoulders after tetanus stimuli, Fig. 3B) was comparable in control and Sulf1^-/-;Sulf2^-/-mutant esophagi. In addition, compared with Sulf1^+/-;Sulf2^+/- controls, the Sulf1^-/-;Sulf2^-/- esophageal smooth muscles showed a greatly diminished response to carbachol, an agonist of the muscarinic receptor, but only partially reduced response to other stimuli(Fig. 3C,D). A similar decrease in carbachol-induced smooth muscle contractility was observed in Sulf1^-/-;Sulf2^-/- esophagi at P15 (data not shown). No smooth muscle contractility defects were observed in the lower sphincter muscle of the Sulf1^-/-;Sulf2^-/-esophagi (data not shown). The esophageal defects of Sulf1^-/-;Sulf2^-/- mice are different from symptoms of achalasia in which the lower sphincter muscle contractility is deregulated owing to a loss of inhibitory nitrinergic neurons(Longstretch and Walker, 1994; Holland et al., 2002; Neuhuber et al., 2006), but are similar to those observed in congenital idiopathic megaesophagus. Carbachol-induced smooth muscle contractility is likely to correlate with the level of neuronal innervation in the esophagus, as shown in studies of sprouty 2 mutant mice (Taketomi et al.,2005). The selective impairment in carbachol-induced smooth muscle contractility of the Sulf1^-/-;Sulf2^-/- esophagi therefore suggests specific neuronal innervation defects rather than a general disruption of the smooth muscle structure in the muscularis mucosae.

To test whether esophageal innervation is defective in Sulf1^-/-;Sulf2^-/- mice, we identified innervating nerve fibers of both intrinsic and extrinsic neurons using the TuJ1 antibody. At E14.5, before the neurites extend across the ME, the Sulf1^+/-;Sulf2^+/- control and Sulf1^-/-;Sulf2^-/- esophagi exhibited comparable staining with TuJ1, GDNF, GDNF receptor α 1 (GFRα1), RET(previously known as c-RET), or the low-affinity neurotrophin receptor p75(NGFR - Mouse Genome Informatics) (Fig. 4A; data not shown). However, at E18.5, Sulf1^-/-;Sulf2^-/- esophagi had much reduced levels of neurite sprouting and innervation density at the smooth muscle of the muscularis mucosae, although the number of intrinsic neurons and the circumference of the esophagi were unchanged(Fig. 4B). In addition, we found no difference in p75 expression on the cell body or in the number of p75-expressing intrinsic neurons between control and Sulf1^-/-;Sulf2^-/- esophagi(Fig. 4B). However, p75 expression on neurites innervating the smooth muscle was decreased in Sulf1^-/-;Sulf2^-/- esophagi, providing additional evidence that the innervation of the muscularis mucosae is defective. Reduced smooth muscle innervation in Sulf1^-/-;Sulf2^-/- esophagi persisted postnatally and in the adult (Fig. 4C and see Fig. S4 in the supplementary material), which directly affected smooth muscle contractility (Fig. 3C,D) and led to gradual enlargement of the esophageal lumen from a ∼20% increase in the circumference of smooth muscle at P15(Fig. 4C), to almost a doubling in the circumference in adults (Fig. 1C and see Fig. S4 in the supplementary material).

To test whether enteric glial cells might also be affected in mutant esophagi, we labeled the esophagus using antibodies against the glial cell markers, glial fibrillary acidic protein (GFAP) and S100 (S100a1 - Mouse Genome Informatics) at P15 and in the adult when enteric glial cells form and mature (Fig. 4C and see Fig. S4 in the supplementary material; data not shown). We found that the number of GFAP-expressing cells along the thoracic segments of the Sulf1^-/-;Sulf2^-/- esophagi was reduced to one-third and to half of that in the control esophagi at P15 and in adult,respectively (Fig. 4C and see Fig. S4 in the supplementary material). Since esophageal innervation is controlled by GDNF (Yan et al.,2004) and GDNF supports enteric glial differentiation from neural crest progenitors (Heuckeroth et al.,1998), the observed esophageal defects in Sulf1^-/-;Sulf2^-/- mice suggest that Sulf regulation of the GDNF signaling pathway establishes the esophageal innervation that is required for postnatal feeding and growth.

To investigate whether SULF1 and SULF2 are required for GDNF-induced neurite outgrowth of the endogenous neurons in the embryonic esophagus,esophageal explants from E11.5 embryos were cultured on collagen gels(Yan et al., 2004). After 4 days, the explants adhered to the collagen gel and exhibited a GDNF dose-dependent neurite sprouting, as shown by TuJ1 immunoreactivity extending from the explants in GDNF-induced cultures(Fig. 5A). The sprouting was not promoted by neurotrophins NGF (NGFβ - Mouse Genome Informatics),BDNF, NT3 (NTF3), NT4 (NTF5) or any other previously known Sulf-regulated signaling ligand, including FGF2, HGF, VEGF165 (NRP1), Wnts, BMP2 and sonic hedgehog (Fig. 5A-C and see Fig. S6 in the supplementary material) (Ai et al., 2005). Furthermore, heparin, but not dermatan sulfate,blocked the GDNF-induced neurite outgrowth (see Fig. S6A,B in the supplementary material), establishing the HS-dependent GDNF activity. The neurite outgrowth from control esophageal explants could be induced by GDNF at 10 ng/ml, with maximal induction at 50 ng/ml GDNF(Fig. 5A,C). By contrast, Sulf1^-/-;Sulf2^-/- explants failed to extend neurites at 10 ng/ml GDNF, and the neurite outgrowth was less than half of the control levels at 20 ng/ml. The defect in the neurite outgrowth of Sulf1^-/-;Sulf2^-/- esophageal explants was fully rescued by 100 ng/ml GDNF (Fig. 5A,C). A lack of neurite outgrowth by Sulf1^-/-;Sulf2^-/- esophageal explants at 10 ng/ml GDNF could be due to disrupted GDNF signaling in intrinsic neurons, or to a reduction in the number of intrinsic neurons. To distinguish between these two possibilities, we quantified the total number of TuJ1-immunoreactive intrinsic neurons within the explants after 4 days in culture. We found that GDNF is essential for neuron number, as control explants cultured without any neurotrophic factors or in the presence of NGF had only one-third of the neurons as explants treated with 10 ng/ml GDNF(Fig. 5E). Sulf1^-/-;Sulf2^-/- esophageal explants, although they showed no significant neurite outgrowth at 10 ng/ml GDNF, had the same number of intrinsic neurons as the controls(Fig. 5E), consistent with the in vivo phenotype. In addition, a few neurons within GDNF-treated control explants migrated out of the explants (data not shown), whereas neurons of the Sulf1^-/-;Sulf2^-/- explants never migrated out of the explants (Fig. 5D),suggesting a defect in GDNF-induced neural migration.

To investigate mechanisms of Sulf regulation of GDNF signaling, we utilized biochemical and tissue culture assays. SULF1 and GDNF are co-expressed by the esophageal muscle progenitors and their expression levels are co-regulated to peak between E14 and E16 (Golden et al.,1999), suggesting that they have related functions. To test whether SULF1 regulates GDNF signaling, we first examined whether SULF1 modulates GDNF binding to HS by comparing GDNF binding to heparin-agarose beads that were predigested with either purified SULF1 or enzymatically inactive QSULF1(C-A) (Fig. 6A)(Ai et al., 2003). SULF1 activity significantly decreased GDNF binding to heparin by up to 6-fold at sub-saturating amounts of GDNF, and by ∼30% under conditions of excess GDNF (Fig. 6A). Second, we tested whether SULF1 activity affects HS-regulated GDNF binding to GFRα1. GDNF and the extracellular domain of GFRα1 (GFRα1-Fc)were incubated with various amounts of heparin predigested by SULF1 or enzymatically inactive QSULF1(C-A) to allow the formation of a GDNF-heparin-GFRα1-Fc ternary complex. As reported previously, heparin enhanced GDNF binding to GFRα1 by 2-fold(Fig. 6B)(Rickard et al., 2003). However, SULF1 had no effect in heparin-regulated GDNF binding to GFRα1(Fig. 6B), indicating that SULF1-regulated HS 6-O-desulfation does not control GDNF binding to GFRα1. These findings suggest that SULF1 functions in the embryonic esophagus to control the affinity of GDNF for HS in the extracellular matrix to facilitate GDNF binding to RET receptor on GDNF-responding innervating neurons.

SULF2 is expressed by esophageal muscle-innervating neurons and thus may regulate the response of these neurons to GDNF. To test whether SULF2 regulates signaling activity of GDNF in neurons, we established stably transfected neuroblastoma-glioma NG108-15 cell lines that expressed comparable levels of SULF2 or inactive QSULF1(C-A) (data not shown), or an empty expression vector. NG108-15 cells do not express detectable levels of endogenous Sulfs, as judged by immunostaining (data not shown). In response to GDNF, NG108-15 cells showed dose-dependent activation of the signaling pathway by phosphorylating the RET receptor and the downstream kinases AKT (also known as PKB and AKT1 - Mouse Genome Informatics) and Erks (MAPK1) 3- to 10-fold above uninduced levels (Fig. 6C,D and see Fig. S5 in the supplementary material). Expression of SULF2, but not inactive QSULF1(C-A) or empty vector, further increased the phosphorylation levels of RET and AKT by 2- to 3-fold and Erks by up to 2-fold in response to GDNF (Fig. 6C,Dand see Fig. S5A in the supplementary material). Additionally,SULF2-expressing NG108-15 cells showed sustained activation of downstream AKT and Erks over 30 minutes after GDNF stimulation, compared with controls in which GDNF activity rapidly declined 15 minutes after stimulation(Fig. 6E and see Fig. S5B in the supplementary material). By contrast, SULF2 had no effect on NGF-induced phosphorylation of AKT or Erks in PC12 cells(Fig. 6F), demonstrating the functional specificity of SULF2 in the GDNF signaling pathway. This result is also consistent with previous findings that NGF signaling is independent of HS sulfation (Barnett et al.,2002). These studies of neural cell lines provide evidence that SULF2 enhances GDNF signaling in neural progenitors during esophageal innervation.

To further demonstrate that Sulfs enhance GDNF signaling in vivo, we assayed the endogenous level of MAPK activation in E14.5 esophagi by immunohistochemistry. Compared with robust MAPK phosphorylation in cell bodies located within the control esophageal muscles, Sulf1^-/-;Sulf2^-/- enteric neurons showed a much reduced level of MAPK phosphorylation (Fig. 6G). By contrast, Sulf1^-/-;Sulf2^-/-epithelial cells surrounding the esophageal lumen exhibited normal MAPK phosphorylation (Fig. 6G). Although MAPK phosphorylation can be triggered by several signals, the GDNF pathway is one of the major signaling pathways in esophageal neuronal progenitors at E14.5. Therefore, this observation is consistent with our finding that Sulfs promote GDNF signaling activity, as shown by esophageal explant and cell signaling assays.

In this study, we have utilized gene targeting in combination with physiological, biochemical and cell biological approaches to investigate functions of Sulfs and HS 6-O-sulfated sequences in developmental signaling. Our results reveal a novel regulatory function of Sulfs in GDNF signaling during esophageal innervation, and thus provide the first evidence that Sulfs are essential in vivo regulators of HS-mediated developmental signaling. Mechanistically, we show that Sulf activity decreases GDNF binding to HS,which promotes signaling. This result suggests that Sulfs have dual functions:to control the binding of GDNF to cell surface and matrix and to regulate neuronal reception of the GDNF signal at the receptor level. Sulf activity is not sufficient to release GDNF from the extracellular matrix, as E14.5 Sulf double-mutant esophagi have normal GDNF immunoreactivity(Fig. 4A). One attractive model is that SULF1 would mobilize GDNF in the matrix by reducing the non-specific GDNF-HS interaction, thereby promoting GDNF transmission from esophageal muscle to innervating neurons and facilitating high-affinity receptor interaction and signaling response in GDNF-receiving neurons.

Our studies also show that Sulfs are enhancers of GDNF signaling rather than obligatory components in the GDNF pathway. Neuroblastoma-glioma cells,which do not express Sulfs, are capable of transmitting the GDNF signal. In addition, high concentrations of GDNF rescue the neurite outgrowth defect of Sulf mutant esophagi in our explant assays, demonstrating that Sulf-deficient neurons are competent to receive GDNF signal. The regulatory roles of Sulfs in GDNF signaling contrast with the obligatory roles of GDNF or GDNF receptors,as demonstrated by the observed differences in the number of enteric neurons between Sulf mutant mice and mice deficient in GDNF or GDNF receptors(Baloh et al., 2000). In addition, we did not observe any significant innervation defects in the Sulf double-mutant colon by immunolabeling (see Fig. S7 in the supplementary material). These findings suggest that Sulf activity is not crucial for enteric neural crest progenitor migration or myenteric plexus formation during development, consistent with the relatively late onset of Sulf expression. However, Sulfs are essential for the developmental transmission of GDNF signaling during esophageal innervation in the embryo. The esophageal smooth muscle is largely innervated by intrinsic neurons. Although extrinsic innervation might also be affected, the observed ∼50% reduction of smooth muscle innervation in Sulf double-mutant esophagi indicates Sulf function in intrinsic neurons, which is consistent with the defects in GDNF-dependent neurite outgrowth of the Sulf mutant esophageal explants.

As Sulf-deficiency leads to specific changes in HS 6-O-sulfation without disrupting the overall charge or structure of HS chains, our findings establish that Sulf desulfation generates HS with a distinct 6-O-sulfated fine structure - an `HS code' - to control the specificity of extracellular signaling responses to HS-dependent ligands and receptors. HS 6-O-sulfation is known to be strictly controlled by developmental stage, tissue type and in tumors (Esko and Lindahl,2001; Kreuger et al.,2006). Here, we show that Sulfs have dynamic expression and are enzymatically active in embryos, indicating their dynamic regulatory roles. Sulf mutant mice not only provide new tools for studies to decipher the HS-sulfation code for regulation of ligand-receptor interactions, but also open new opportunities to investigate the genetic and developmental signaling mechanisms underlying the motility disorder, congenital idiopathic megaesophagus. Additionally, Sulf mutant mice provide a new therapeutic model for developmental investigations of GDNF-mediated neuroprotection in Parkinson's disease and brain injury(Airaksinen and Saarma, 2002)and in other diseases involving Sulf-mediated HS-regulated signaling.

We thank Dr Mario Gimona for providing SM22 antibody, Drs Sarah Wilcox-Adelman, Jennifer Chen, Cathy Soula and Willington Cardoso for helpful discussion, Monika Kimanova for generating mouse chimeras and Aliete Langsdorf for maintaining mouse colonies. This work was supported by a grant from the National Institute of Child Health and Human Development to C.P.E., a National Institute of Health postdoctoral fellowship to X.A. and a grant (#13401) from the Swedish Research Council to M.K.-G. The authors declare that they have no conflicting financial interests.