Spatial and temporal changes in the distribution of proteoglycans during avian neural crest development

Neural crest cells migrate long distances along pathways containing an intricate extracellular matrix (ECM). As a consequence, the ECM is thought to play a central role in several aspects of neural crest development. In vitro, neural crest cells migrate avidly on numerous ECM molecules including fibronectin, laminin and collagens (Newgreen and Erickson, 1986; Perris and Johansson, 1987; Perris et al. 1989; Perris et al. 1990a), suggesting that individual matrix components may serve as permissive migratory substrates. Ultrastructural studies performed in situ reveal that neural crest cells form specialized contacts with this fibrillar matrix network encountered during migration (Löfberg et al. 1980; Newgreen and Erickson, 1986; Perris et al. 1990b). In ovo injections of antibodies against indi-vidual matrix molecules or their cell surface receptors results in abnormal neural crest development in vivo (Bronner-Fraser, 1985, 1986a; Bronner-Fraser and Lallier, 1988; Bronner-Fraser, 1988). Moreover, trans-plantations of regionally and temporally defined matrices adsorbed onto membrane microcarriers have provided evidence that the ECM can prematurely promote the onset of neural crest cell movement in vivo (Löfberg et al. 1985, 1988).

A logical first step in establishing the role of the ECM in neural crest cell migration is to determine its structural and molecular composition at various phases of neural crest development. Ultrastructural and immunohistochemical studies have revealed that the interstitial matrix along trunk neural crest migratory pathways consists of a fibrillar collagenous network, which contains abundant amounts of fibronectin, tenascin/cytotactin and glycosaminoglycans (Newgreen and Erickson, 1986; Perris and Bronner-Fraser, 1989; Perris et al. 1990a; Newgreen et al. 1986, 1990). Basement membrane matrices, enriched in laminin and collagen type IV, also line some neural crest migratory routes (Newgreen and Erickson, 1986; Perris and Bronner-Fraser, 1989; Perris et al. 1990b).

Although a great deal of information has been compiled regarding the distribution and possible func-tion of cell adhesion glycoproteins such as fibronectin, laminin, cytotactin/tenascin and various collagens during neural crest cell migration, far less is known about the role of proteoglycans in this process. Proteoglycans represent a heterogeneous population of molecules that contribute to the compositional diversity of the ECM. Indirect evidence that proteoglycans might be well-represented during neural crest development has emerged from a series of histochemical studies using cationic dyes, in situ metabolic labelling and differential enzymatic degradation (Kvist and Finne-gan, 1970; Pintar, 1978; Löfberg et al. 1980; Newgreen et al. 1982, 1986; Perris et al. 1990b). From these observations it was concluded that several distinct families of matrix and cell-associated proteoglycans were expressed at various phases of neural crest development. However, the nature and spatiotemporal distribution of specific populations of proteoglycans present at these early stages of development, as well as their relationship to their potential ligands, such as collagens and hyaluronan, has not been determined.

In this study, we have examined the distribution of various proteoglycan subclasses in situ and have determined their spatial and temporal organization relative to the development of the trunk neural crest in the chick embryo. For this purpose, a panel of monoclonal antibodies that specifically detect native carbohydrate structures of chick proteoglycans were used in combination with differential enzymatic degra-dation and antibodies generated against chondroiti-nase-digested murine proteoglycans. In addition, a specific probe for hyaluronan (Knudsen and Toole, 1985; Ripellino et al. 1985) was used to map the distribution of this potential matrix and cell surface-associated ligand for proteoglycans. Our observations indicate that a time-related and region-specific organiz-ation of diverse types of proteoglycans occurs during neural crest cell migration and gangliogenesis, and the onset of vertebral chondrogenesis.

Native CS/KS-PG monomers were extracted from bone marrow of femurs and tibias of 17 day old chick embryos using 4 M guanidine–HCl in the presence of protease inhibitors as previously described (Oegema et al. 1975; Sorrell et al. 1988). Solid CsCl was added to the extract to give the final density of 1.5 g ml⁻¹, and the solution was centrifuged at 100000g for 48 h. The bottom one third of the gradient, with the density of 1.6 g ml⁻¹, was collected, dialyzed against 0.15 M NaCl, then water and lyophilized as previously described (Sorrell et al. 1988). The purified, high buoyant density PGs (DI fraction), which have been shown to be substituted with both CS and KS chains, were then used to immunize mice. The immunization protocol and the protocol for the production of hybridoma were as previously published (Caterson et al. 1987; Sorrell et al. 1990). Clones resulting from the fusion were tested in ELISA for the presence of antibodies that recognized the marrow PGs as well as PGs from other sources. Antibodies produced by nine such clones, 7D4, 6C3, 4C3, 4D3, 3D2, 1B4, 4D1, 2D3 and 8C2, were identified and classified as IgM kappa. Each clone was then subcloned and used to produce ascites fluid. The ascites fluid was further screened for the ability of the antibody to identify epitopes on a number of different purified CS/KS-PG monomers, in their intact form or after alternative digestion with chondroitinase ABC and keratanase (see below).

MAbs 2B6, 1B5, 3B3 and 5D4 were obtained from Dr Bruce Caterson, Division of Orthopaedics, Department of Surgery, University of North Carolina, Chapel Hill, NC. The mAb 33–2, kindly provided by Dr Douglas Fambrough (The John Hopkins University, Baltimore, MA), reacts with a core protein-associated epitope located on a large molecular mass HS-PG (MW>300×10³Mr) of embryonic chick skeletal muscle (Bayne et al. 1984). The HNK-1 antibody was purified from ascites fluid (American Tissue Culture, Inc) by column chromatography on protamine sulfate followed by ammonium sulfate precipitation. In some cases, the purified HNK-1 antibody was biotinylated using biotin-N-hydroxysuccinamide ester (Calbiochem) according to a previously published procedure (Perris and Johansson, 1987; Perris et al. 1989).

Proteoglycan monomers from shark cranial cartilage (Al), embryonic chick epiphysial cartilage (DI), bovine articular cartilage (A1D1) bovine nasal cartilage (A1D1) and Swarm rat chondrosarcoma (A1D1) were purified according to previously published procedures (Oegema et al. 1975; Sorrell et al. 1988, 1990). Wells of microtiter plates (Costar half area plates) were coated with PG monomers 25–50 μg ml⁻¹ (dry weight) in PBS overnight at 4 °C. Proteoglycan-free areas of the wells were then blocked with 1 % BSA (v/w) in PBS for 1–2 h at 37°C. In some experiments, wells with immobilized PGs were digested for 5 min at room temperature with 10μi.u.ml⁻¹− 100mi.u. ml⁻¹ chondroitinase ABC (Sigma), or 70mi.u.ml⁻¹Pseudomonas endo-β-galactosidase (keratan-ase; ICN Biochemicals Inc.), dissolved in Tris-HCl buffer, pH 7.6, containing protease inhibitors and 1% BSA. Digested and untreated proteoglycan-coated wells were extensively rinsed with cold PBS and incubated with the various mAbs (1:100–1:700000) diluted in PBS for 1 h at 37 °C, followed by sequential incubation with horseradish peroxidase-conjugated goat anti-mouse Ig (Southern Biotech-nology) diluted 1:500 in 0.1 M Tris-0.9 % NaCl pH 7.6 for 1 h at 37°C, and o-phenylenediamine (0.4 mg ml⁻¹)/H₂O₂ (0.0015%). The enzyme reaction was stopped by addition of 2 M H₂SO₄ and the adsorbance readings were performed at 492 nm in a Bio-Rad microplate reader. Competitive ELISA was carried out by coating microwell plates with 50μgml⁻¹ monomers of shark cranial cartilage proteoglycan as indicated above. Serial 50% dilutions of shark cranial cartilage, chick bone marrow, chick epiphysial cartilage, bovine articular cartilage or bovine nasal cartilage were prepared in PBS with 1 % BSA. Equal aliquots of diluted antigen, or buffer alone, were added to a constant volume of antibody (6C3 diluted 1:25000; 7D4 diluted 1:150000; 4C3 diluted 1:6000) in the same buffer, incubated at 37°C for 1 h, and then added to the coated wells. Standard ELISA was performed as above. The % inhibition was determined for three assays, averaged and plotted versus μgml⁻¹ of competing antigen. The amount of competing antigen required to give 50 % inhibition was then obtained by extrapolation from the inhibition curves.

A 65×10³ M_r proteolytic fragment corresponding to the hyaluronan-binding region of the bovine nasal cartilage proteoglycan was generously provided by Dr Torward Laurent (Department of Medical and Physiological Chemis-try, Biomedical Center, Uppsala, Sweden). The fragment was generated by affinity chromatography as previously described (Tengblad, 1979; Laurent and Tengblad, 1980). The PG fragment, with attached low molecular weight hyaluronan (16–30 oligosaccharides) to preserve the hyaluronan-binding activity, was biotinylated according to the same procedure as indicated for the HNK-1 antibody. Purified hyaluronan (human umbilical cord; M_r 1000000; Sigma) was bound to AH-Sepharose 4B (Pharmacia) as described by Tengblad (1979). Before use, the biotinylated hyaluronan-binding fragment was then dissociated from the hyaluronan oligosac-charides by transfer to a 4M guanidine–HCl–50mM sodium acetate buffer, pH 5.8, containing 10 mM EDTA, 100 mM 6-aminohexanoic acid, 5mM benzamidine hydrochloride, and 2mM phenylmethylsulfonyl fluoride (PSMF; Sigma).

Chick embryos at developmental stages 14–23 were rapidly frozen in liquid nitrogen-cooled isopentane and immediately transferred to methanol precooled to −80°C and stored at this temperature for 3 days. They were then successively transferred to −35°C for 1 day, −20°C for 2 days and 4°C for 2 days. Following cryofixation, embryos were embedded in paraplast after one change of methanokhistosol (7:3), one change of methanokhistosol (3:7), two changes of histosol, one change of histosol:paraplast (1:1), and three changes of paraplast at 20 min for each step. A number of other procedures, including direct fixation in modified Kamovsky’s fixative, 4% paraformaldehyde, Sainte-Marie, periodate– lysine–paraformaldehyde and Zenker’s fixatives followed by paraplast embedding or cryosectioning were also tested and were found to be significantly less efficient in retaining reactivity for the various anti-proteoglycan mAbs. One exception was the mAb 33–2 (anti-HS-PG antibody) for which optimal preservation of immunoreactivity was obtained after methanol fixation and cryosectioning. The latter was accomplished according to previously published procedures (Krotoski et al. 1986). Serial transverse, horizontal and parasagittal sections through the mid-trunk level of the embryo were cut at 10–20 μm and mounted on gelatinized or albuminized slides.

In cases where antibody application was preceded by enzymatic digestion, sections were incubated with the various enzymes for 1–3 h at 37 °C. Streptomyces hyaluronidase (Sigma) was used at 115i.u.ml⁻¹ in 0.1M sodium acetate buffer, pH5.2, containing 250μgml⁻¹ ovomucoid (trypsin inhibitor O-IV), 1 mM PMSF, 1 mM EDTA, 1 mM iodoacetida-mine and 200 Ki.u. ml⁻¹ aprotonin (all Sigma). Chondroiti-nases ABC and AC II (Sigma) were used at 0.2i.u. ml⁻¹ in 0.1M bicine buffer (Calbiochem), pH8.0, containing 0.1% BSA. Keratanase (Pseudomonas sp. IFO 13309) and endo-β-galactosidase (Escherichia freundii; ICN Biochemicals Inc.) were applied at 0.0li.u.ml⁻¹ in 50 mM Tris-HCl buffer, pH 7.2, containing 80 mM NaCl, 1 mM EDTA, 1 mM iodoaceta-mide, 1 mM PSMF, and 5 μgml⁻¹ pepstatin A (Sigma).

Incubation with primary antibodies (1:50–1:150 in 0.01M phosphate buffer, pH 7.2, containing 0.1 % BSA) was carried out overnight at 4°C, except for the HNK-1 labelling that was performed for 2h at room temperature. Antibody–antigen binding was visualized by indirect immunofluorescence using lg-class and species specific secondary antibodies, directly conjugated to fluorescein, or rhodamine (Zymed Labora-tories). Double labelling was performed using biotinylated HNK-1 antibody, visualized by incubation of the sections with streptavidin–Texas Red complexes (1:200; Amersham). Labelled sections were mounted in glycerol-Tris-HCl buffer, pH8.0, containing 2 mg ml⁻¹ l,4-diazabicyclo(2,2,2,)octane. In some cases, comparative stainings with anti-PG antibodies and the HNK-1 antibody were carried out on adjacent sections to avoid possible steric hindrance effects of IgM antibodies.

Stage 14–23 embryos were fixed in phosphate-buffered 4% paraformaldehyde, pH7.2, containing 0.5% cetylpyrinidium chloride and 0.1% polyvinylpyrrolidone (M_r 40000; Sigma) for 12–18 h at 4 (Derby and Pintar, 1978). Fixed embryos were rinsed, embedded in gelatin and sectioned as described above. Tissue sections were incubated with 30-100pg biotinylated hyaluronan probe in 0.1M phosphate-buffered saline, pH7.2, at 4°C overnight. Sections were then extensively rinsed in buffer containing 0.01 % Nonidet P-40, incubated with streptavidin-fluorescein and finally mounted as described for antibody-labelled sections. Control sections were predigested with Streptomyces hyaluronidase (Sigma) in the presence of protease inhibitors (see above), or incubated with inactivated probe. For the latter purpose, the biotinyl-ated hyaluronan-binding fragment was preincubated with a molar excess (100 μgml⁻¹) of hyaluronan oligosaccharides (16–30) for 1 h at 37°C, dialyzed to remove excess nonbound oligosaccharides, and then applied to the sections. In some cases, sections were double-labelled with the hyaluronan probe and the HNK-1 antibody.

In a number of cases, stained sections were analyzed by confocal laser microscopy in conjunction with computerized image analysis. For this purpose, we used a Bio-Rad MRC 500 laser unit attached to an epifluorescent microscope (Nikon, Labphot), a videoprinter VP 3500 (Seikosha), a 3D6 personal computer, and Bio-Rad software for image analysis. Relative staining intensities were analyzed according to an arbitrary value scale of 0-255 in a series of 74 transversal sections cut between somite 15 and 22 of stage 17 embryos (29–31 somites) and stained with mAb 4C3; and a series of 16 horizontal sections spanning >20 somites of similar age embryos stained with mAb 7D4. Transverse sections were analyzed in the medial–lateral plane by single spot measurements across their dorsal portion, starting and ending in the subepidermal space. A single spot measurement was also taken within cells of the neural tube where no immunoreactivity was observed and was used to correct for the background nonspecific fluorescence. Analysis in the dorsal–ventral dimension was carried out by consecutive single spot measurements along an arbitrary vertical line drawn across the section and passing through the subepidermal space, the lateral sclerotome, the perichordal area and finally the periaortic area. Horizontal sections containing the mid-portion of the neural tube were analyzed in the rostral–caudal dimension by recurring spot measure-ments at the borders and center of each somite, and within the intersomitic clefts, starting from the most recently formed somite and progressing rostrally. In this case, control values for correction of the nonspecific background fluorescence were obtained from the dermamyotome, which consistently lacked proteoglycan immunoreactivity.

The characteristics and specificities of the various mAbs are summarized in Table 1. Immunization of mice with high buoyant density PGs extracted from bone marrow of 17 days old chick embryos resulted in nine clones producing mAbs specific for CS and KS glycosaminoglycans. MAbs 7D4, 6C3, 4D3 and 4C3 all reacted with epitopes expressed on native CS moieties of cartilage PGs from a variety of sources including chick epiphysis, bovine arteries and nose, chick bone marrow and shark cranium (Fig. 1). Although all five mAbs detected epitopes on CS-PGs, direct and competitive ELISA using five distinct PG types as model antigens clearly demonstrated that each antibody detected a different structural element (Fig. 1A). ELISA also showed that, while the antigenic profiles of the chick CS-PGs were related, the antigenic profiles of the other CS-PGs were characteristic for each type. Combined use of ELISA and chondroitinase ABC digestion proved that the immunological diversity resided in the CS chains. Specific épitopes were differentially removed by various concentrations of enzyme (Fig. IB), consistent with the idea that each of these antibodies recognized a unique structural unit. For example, the epitope recognized by mAb 6C3 was ten-fold more sensitive to digestion than those recognized by 7D4 or 4C3. Recent data indicate that mAbs 7D4, 6C3 and 4C3 primarily recognize CS chains with prevalent 6-sulfated groups, whereas mAb 4D3 seems to react primarily with 4-sulfated chondroitins (Sorrell, unpublished).

MAbs 4D1, 3D2, 1B4, 8C2 and 5D4 also recognized epitopes on CS-PG monomers, bdt unlike the five mAbs described above, these epitopes were not removed by chondroitinase predigestion of the antigens. In fact, anti-genicity of PGs for the latter mAbs sometimes increases following chondroitinase treatment (data not shown). In contrast, digestion of immobilized embryonic chick cartilage CS-PG abolished virtually all immunoreactivity for these mAbs (Fig. 1C). The moderate reduction in immunoreac-tivity for bovine nasal cartilage CS-PG is likely to be due to differences in the structure of the KS chains of these two species-specific PGs.

The overall staining patterns obtained with mAbs 7D4, 6C3 and 4C3 were similar and therefore are described collectively. However, the extent and intensity of the labelling obtained with these mAbs decreased in the order 4C3>7D4>6C3, even though mAb 7D4 reacted 15 times more strongly with isolated chick proteoglycans than mAb 4C3 (Fig. 1A). Minor differences in the immunoreactive patterns obtained with these mAbs were also observed within the sclerotome and in the perinotochordal area.

Analysis of the distribution of proteoglycans (PGs) and hyaluronan was carried out in the mid-trunk region at two developmental stages: during neural crest cell migration (stages 14–19 at level of somites 12–27; Fig. 2A) and following gangliogenesis (stages 21–23 between wing and limb buds). Four generalized patterns of immunoreactivity were distinguishable and those will be referred to as patterns I–IV (Table 2).

Proteoglycans detected by the mAbs 7D4/6C3/4C3 were distributed widely along neural crest cell mi-gratory pathways. The immunostaining patterns pro-duced by antibodies 7D4/6C3/4C3 (Fig. 2B) were similar to that of mAb 3B3, which recognizes chondroi-tin-6-sulfate (Ch6S) moieties following chondroitinase ABC digestion. Thus, it is possible that the native epitopes recognized by these antibodies were expressed on CS chains rich in 6-sulfated groups. This staining pattern will be refered to as pattern I (Table 2). At initial stages of neural crest cell migration, pattern I CS-PGs were detected within the subepidermal basement membrane, surrounding the neural tube and noto-chord, and in the ventral portion of the sclerotome (Fig. 2B). Fibrillar staining was also observed in the intersomitic clefts and in the perinotochordal region. However, these PGs were generally absent from regions through which neural crest cells initially will migrate (Fig. 3C,D; Bronner-Fraser, 1986b; Newgreen et al. 1986).

During the course of neural crest cell migration, pattern I CS-PGs increased considerably in extent and intensity within the sclerotome (Fig. 3A,B). Their deposition appeared to follow a rostral-to-caudal and ventral-to-dorsal sequence, first appearing in the ventral sclerotome concomitant with sclerotome formation and later appearing in the dorsal sclerotome (4–5 somites rastrad). Twelve to fourteen somites rostral to the most recently formed somite, pattern I CS-PGs were observed throughout the sclerotome. The relative levels of immunoreactivity within the different portions of the sclerotome were quantitated by confocal laser microscopy in conjunction with computerized image analysis. The dorsal sclerotome, containing migrating neural crest cells (Fig. 2F,G) exhibited a 10-fold lower staining intensity for 7D4-immunoreactive CS-PGs than did the corresponding ventral sclerotome (Fig. 4B,D). The subepidermal space and neural tube basement membranes exhibited higher levels of immu-noreactivity than either the dorsal or ventral sclerotome (Fig. 4A,C). The differences between dorsal and ventral immunoreactivty appeared segmental, being greater at the level of the rostral sclerotome (containing neural crest cells) than at the caudal sclerotome (Fig. 4D,E).

At advanced stages of neural crest cell migration, there was a marked rearrangement of pattern I CS-PG immunoreactivity. Proteoglycan fibrils were observed primarily in the caudal sclerotome, but were absent from the rostral sclerotome in the vicinity of migrating neural crest cells (Figs 3E,F and 4). At final stages of neural crest cell migration and the beginning of their aggregation into peripheral ganglia, pattern I CS-PG fibrils surrounded the ganglia, but were not found within them (Fig. 3G,H).

CS-PGs rich in unsulfated chondroitin (ChO) chains (recognized by mAb 1B5; pattern II) were detected for the first time during active neural crest cell migration, where they appeared confined to the notochord and its adjacent fibrillar matrix. Similarly, CS-PGs recognized by mAb 4D3 and CS-PGs rich in Ch4S chains (mAb 2B6; pattern III) were observed around the notochord and ventrolateral subepidermal space, but were absent from all other regions of the embryo.

After neural crest cells had aggregated to form peripheral ganglia, pattern I CS-PGs displayed a metameric arrangement related to the position of the peripheral ganglia. Between adjacent ganglia, these CS-PGs were distributed uniformly throughout the lateral sclerotome, including the perinotochordal region. In contrast, the ganglia and the ventral roots, containing axons and neural crest-derived Schwann cells, were devoid of pattern I CS-PGs (Fig. 5A). CS-PGs carrying predominantly Ch6S chains (mAb 3B3) persisted in their colocalization with CS-PGs detected by mAbs 7D4/6C3/4C3 (Fig. 5B), with the exception of the lateral subepidermal space and the sclerotome adjacent to the neural tube where the latter PGs were absent. Ch6S-PGs were distributed metamerically in the ventral portion of the embryo, within perinotochordal matrix fibrils extending well into the sclerotome between adjacent ganglia at sites of prospective vertebral arch formation (Fig. 5B). No immunoreactivity was observed in the sympathetic ganglia or in the regions of the presumptive adrenal medulla, splanchnic sympath-etic plexus and mesenteric ganglia.

Pattern III CS-PGs recognized by mAb 4D3 were concentrated in the subepidermal space, the sclerotome surrounding the neural tube (including its basement membrane) and the perinotochordal region. Between adjacent ganglia, pattern III CS-PGs formed a dense fibrillar network, continuous with that surrounding the notochord and neural tube, and extending ventrally between the notochord and dorsal aorta in prospective regions of vertebral arch formation.

The distribution of CS-PGs bearing predominantly Ch4S (pattern III; mAb 2B6 in conjunction with chondroitinase AC II digestion) was metameric along the rostrocaudal extent of the sclerotome (Fig. 5C,D), with immunoreactivity enriched between adjacent ganglia in the region of the prospective vertebral arches. Ch0-PGs (pattern II; recognized by mAb 1B5) were solely observed around the notochord. Both Ch0 and Ch4S-bearing CS-PGs were absent from the most dorsal regions of the embryo and the sclerotome ventral to the notochord.

Chondroitinase ABC/AC II digestion of the sections prior to application of the mAbs 7D4, 4C3, 6C3 and 4D3 entirely abolished staining, demonstrating that the antibodies reacted exclusively with native CS chains of tissue proteoglycans. In contrast, if the chondroitinase ABC/AC II digestion was omitted prior to staining with antibodies 3B3 (Ch6S), 1B5 (Ch0) and 2B6 (Ch4S), no immunoreactivity was observed, which confirmed the specificity of the staining with these antibodies for epitopes generated by such enzymatic treatment (Couchman et al. 1984).

The distribution of DS-bearing proteoglycans was defined by comparing the staining patterns observed for the mAb 2B6 following either chondroitinase ABC or chondroitinase AC II pre-digestion. DS-PGs were not immunohistochemically detectable at stages of initial neural crest cell migration. During advanced stages of neural crest cell migration, immunoreactivity for DS-PGs was confined to the notochordal basement mem-brane and its associated fibrillar matrix (pattern III). At postganglionic stages, DS-PG immunoreactivity became more heavily concentrated in the perinotochor-dal region, and also appeared in the ventral sclerotome, perioaortic region, and subepidermal space. DS-PGs appeared in a metameric pattern similar to that of Ch4S-PGs in the ventral sclerotome at sites of presumptive vertebral arch formation.

At initial stages of migration, immunoreactivity with the 5D4 antibody (pattern I KS-PGs; Table 2) was similar to that of Ch6S, but was considerably less widespread and weaker in interstitial matrices. In contrast, immunolabelling with the 4D1, 1B4, 2D3, 3D2 (pattern II KS-PGs; Fig. 2C) and 8C2 antibodies (pattern IV KS-PGs) was observed primarily around the notochord, with some discontinuous staining in the subepidermal space for pattern IV KS-PGs.

With progressive neural crest cell migration, the intensity and distribution of pattern I, II, and IV KS-PGs increased in a region-specific manner. A punctate staining for pattern I KS-PGs appeared at advanced phases of neural crest cell migration and accompanied the distribution of pattern I CS-PGs. Distribution of pattern I KS-PGs within the sclerotome initially was uniform, but became progressively excluded from the rostral portion during the course of neural crest cell migration (Fig. 2E).

Pattern IV KS-PGs extended throughout the sclero-tome, the intersomitic space, the perinotochordal matrix and were sparsely distributed around the neural tube surface (Fig. 5E,F). Deposition of pattern IV KS-PGs preceded and was more abundant than that of pattern I KS-PGs. In contrast to pattern I CS-PGs (Fig. 3A,B) which increased in distribution in a ventral-to-dorsal sequence, Pattern IV KS-PGs were aug-mented in a dorsal-to-ventral sequence (Fig. 5E,F) during the course of neural crest development.

Pattern II KS-PGs were heavily concentrated in the perinotochordal and supra-aortic region, and the subectodermal basement membrane (Fig. 2D). In contrast, immunoreactivity for pattern IV KS-PGs was persistently confined to the perinotochordal area.

Immunoreactivity for pattern I KS-PGs increased and largely coincided with that of pattern I CS-PGs in the dorsal regions of the embryo and in the lateral sclerotome (Fig. 5G) at interganglionic levels. How-ever, some differences were noted; for example, these KS-PGs filled the entire subepidermal space and were sparsely distributed in the perinotochordal region and ventral sclerotome at axial levels containing the peripheral ganglia (Fig. 5G). No staining was observed in the sclerotome surrounding the neural tube at any axial level, though immunoreactivity was observed within the neurogenic region and roofplate of the neural tube.

Pattern IV KS-PG immunoreactivity increased at postmigratory stages, and was apparent throughout the lateral sclerotome at all axial levels. These KS-PGs were widespread within the ganglia and ventral roots, and were largely absent from the medial sclerotome and around the neural tube (Fig. 5H). To determine whether pattern IV KS-PG immunoreactivity was present on neural crest-derived cells, sections were double-labelled with mAb 8C2 and the HNK-1 anti-body. Although 8C2 staining was present within the ganglia and ventral roots, it appeared distinct from that produced by the HNK-1 antibody. In the ventrolateral sclerotome, pattern IV KS-PGs were concentrated in the area occupied by neural crest cells forming the sympathetic ganglia. Although outlining the dorsal aorta, pattern IV immunoreactivity largely was absent from the peri-aortic region through which adrenome-dullary ceils and the splanchnic sympathetic plexus precursors move. Pattern IV KS-PGs were pronounced at the sclerotomal-myotomal interface and the noto-chordal basement membrane.

Pattern II KS-PGs remained concentrated in the perinotochordal space and along the ventral neural tube following ganglion formation. Pattern II KS-PGs also prominently stained the mesonephric tubules and fibrils within the adjacent periaortic sclerotome.

Keratanase (Pseudomonas sp.) and endo-β-galactosi-dase (Escherichia freundii) digestion of sections reduced immunoreactivity for mAbs 5D4, 4D1, 1B4, 2D3, 8C2 and 3D2, indicating that the epitopes recognized by these antibodies were on native KS chains. At stages of neural crest cell migration, keratanase removed all immunoreactivity for mAbs 5D4, 1B4 and 4D1, whereas some immunoreactivity remained around the notochord after digestion with endo-β-galactosidase. At postganglionic stages, kera-tanase and endo-β-galactosidase treatment completely abolished immunoreactivity for mAb 4D1, but not 5D4. Similarly, some mAb 8C2 immunoreactivity persisted subepidermally and around the dorsal neural tube after endo-β-galactosidase, whereas keratanase digestion eliminated all staining. The differential susceptibility to enzymatic degradation indicates that the populations of KS-PGs detected during neural crest cell migration were heterogeneous, carrying unbranched KS chains with both monosulfated and non-sulfated N-acetyllac-tosamine repeating units (Ito et al. 1986). Interestingly, the differential susceptibility to enzymatic digestion suggested that pattern IV KS-PGs associated with migrating and post-migratory neural crest cells (mAb 8C2) were abundant in undersulfated KS chains, whereas those largely absent from the migratory pathways (mAbs 5D4, 4D1 and 3D2) were more heavily sulfated.

At initial neural crest cell migration, immunoreactivity for the mAb 33–2 (pattern IV HS-PG; Table 2) was evident in the all basement membranes surrounding the neural tube and notochord, along the basal side of the epidermis, and on the dorsolateral dermamyotome. In contrast to CS/DS/KS-PGs, no HS-PG staining could be observed in the interstitial matrix. During the course of neural crest cell migration, pattern IV HS-PG immunoreactivity persisted along basement membranes and also became evident within the somitic sclerotome (Fig. 6A). In this region, it occurred in a uniformly arranged fibrillar pattern that accompanied the ex-pression of CS/KS-PGs (Fig. 6B). With continued neural crest cell migration through the rostral portion of the somites and subsequent coalescence of the pregang-lionic cells, HS-PG-containing matrix fibrils became restricted to the ganglia and little or no immunoreac-tivity could be detected in the surrounding sclerotome.

Following gangliogenesis, HS-PG immunoreactivity was primarily localized in basement membranes, but also occurred in a punctate pattern within the newly formed dorsal root ganglia and surrounding Schwann cells of the ventral roots/rami communicans (Fig. 6C). No staining could be seen within the sympathetic ganglia, the sclerotome surrounding the dorsal root ganglia and notochord, or the periaortic region. The pattern of HS-PG immunoreactivity coincided with that of its potential matrix ligands collagen type IV (Perris et al. 1990a) and laminin (Krotoski et al. 1986). The one exception is that collagen type IV and laminin remain detectable within the sclerotome after gangliogenesis, whereas HS-PG does not remain detectable (Fig. 6C).

The distribution of hyaluronan was determined by using a specific probe (Knudsen and Toole, 1984; Rippellino et al. 1985) consisting of a biotinylated fragment from the hyaluronan-binding region of the bovine nasal cartilage proteoglycan. At stages of early neural crest cell migration, hyaluronan was found widely distributed along neural crest migratory pathways. This glycos-aminoglycan did not appear to be associated with basement membranes, but rather occurred in fibrillar structures apposed to basement membranes. A con-siderable amount of hyaluronan was seen within the epithelial somites and around the notochord. High levels of staining were detected above the dorsal neural tube in the vicinity of early migrating neural crest cells, particularly along the subepidermal region through which neural crest cells migrate to form melanocytes (Fig. 7A).

During the period of early and advanced neural crest cell migration, abundant amounts of hyaluronan were detected in the ventral regions of the sclerotome. The hyaluronan density, as evaluated by staining intensity, was greatest adjacent to the notochord and decreased progressively outward into the distal sclerotome. No hyaluronan was detected in the dermamyotome or neural tube (Fig. 7B). Along the rostrocaudal extent of the sclerotome, hyaluronan-rich fibrils were uniformly distributed and extended between, but not across the intersomitic clefts. As neural crest cell migration ceases and gangliogenesis commences, the levels of hyaluro-nan appeared reduced in the immediate vicinity of neural crest cells. Expression of hyaluronan was weaker in the peripheral ganglia when compared with the strong and uniform interstitial expression observed throughout the sclerotome and subepidermal space (Fig. 7C).

Streptomyces hyaluronidase digestion of the sections prior to application of the probe completely eliminated reactivity. Similarly, inactivation of the probe by oligosaccharide-saturation of its binding site signifi-cantly depleted its ability to detect hyaluronan in tissue sections.

In the present study, the distribution of proteoglycans was related spatially and temporally with the develop-ment of avian neural crest cells, by using a panel of monoclonal antibodies produced against proteoglycan moieties. Four patterns of immunoreactivity were distinguished by this comparative study: (1) CS/KS-PGs rich in Ch6S chains and recognized by mAb 7D4, 6C3, 4C3 and 5D4 were abundantly distributed throughout the embryo during all phases of neural crest development, but were largely absent from the areas occupied by migrating neural crest cells and their derivatives (pattern I); (2) CS/KS-PGs bearing primar-ily unsulfated chains and recognized by mAbs 4D1, 1B4, 3D2 and 2D3 consistently were restricted to the perinotochordal region (pattern II); (3) CS/DS-PGs rich in Ch4S chains and recognized by mAbs 4D3 and 2B6 were present in the ventral subepidermal space and notochordal region during neural crest cell migration and appeared in areas of presumptive chondrogenesis after the time of gangliogenesis (pattern III); and (4) KS-PGs and an HS-PG, recognized by 8C2 and 33–2 antibodies, respectively, that were widely distributed along the migratory pathways during the course of neural crest cell movement, and which later occurred within the structures formed by the neural crest. Hyaluronan was often colocalized with the first set of proteoglycans, but was also prevalent around migrating neural crest cells which seemed to have reduced levels of hyaluronan at postmigratory stages.

The production of a number of monoclonal anti-bodies with the unique ability to recognize specific moieties of PGs enabled the identification of various classes of PGs in situ. When tested in ELISA, all mAbs were found to react with chick cartilage CS/KS-PGs, as well as with a variety of other species-specific proteogly-cans. Further characterization of the antibodies by competitive ELISA and differential enzymatic diges-tion demonstrated that mAbs 7D4, 6C3, 4C3 and 4D3 specifically recognized discrete epitopes on native CS chains, whereas mAbs 4D1, 1B4, 2D3, 8C2 and 3D2 reacted with KS chains of proteoglycans. The specificity of the epitopes recognized by both anti-CS and anti-KS antibodies was shown by their distinguishable affinities for different purified proteoglycans and by their divergent immunochemical properties (Sorrell et al. 1988, 1990; Sorrell and Caterson, 1990). The identity of the epitopes recognized by the anti-KS antibodies was further demonstrated by their differential affinity for keratanase-digested proteoglycan antigens.

Staining of chick embryos during the course of neural crest development with mAbs 7D4, 6C3 and 4C3 yielded strikingly similar patterns. However, this does not necessarily imply that the mAbs recognized the same proteoglycan molecule. The possibility remains that the mAbs may identify distinct proteoglycan classes that colocalize and share common glycosamino-glycan epitopes. Alternatively, the glycosaminoglycan epitopes may reside on different proteoglycan core proteins. Changes in the distribution patterns observed at various phases of development may have reflected developmentally related substitutions of the glycos-aminoglycan determinants of the same proteoglycans molecules.

At stages of initial neural crest cell movement, the interstitial matrix around the dorsolateral neural tube and sclerotome contained abundant amounts of hyalur-onan, but little or no proteoglycan. Other areas that are rich in a variety of CS/KS-PGs (including the subepi-dermal space, the area around the ventral neural tube, the notochord and the intersomitic clefts) either entirely lack neural crest cells at early stages of migration (Bronner-Fraser, 19866; Newgreen et al. 1986) or, in the case of the intersomitic clefts, contain only a minority of neural crest cells (Lallier and Bronner-Fraser, 1988). With progressive neural crest cell migration within the somites and concomittant dispersion of the sclerotomal cells, there was a concomitant increase in the amount of CS/KS-PGs deposited in the interstitial matrix of the sclerotome. This accumulation of PGs appeared following the entrance of neural crest cells, but extended rapidly throughout the sclerotome; thus neural crest cells moved toward embryonic areas dense in proteoglycans.

With progressive migration and the condensation of neural crest cells into peripheral ganglia, the rostral portion of the sclerotome became virtually devoid of CS/KS-PGs (recognized by the mAbs 5D4, 4D1 and 3D2) immunoreactivity. However, some hyaluronan, HS-PG, and KS-PGs, identified by the 8C2mAb, were observed. It is unclear why HS-PG and some forms of KS-PGs are expressed within the peripheral ganglia and ventral roots, but these molecules may have been retained on the cell surface or synthesized by the neural crest-derived neurons and/or glia. The absence of certain CS/KS-PGs from areas occupied by migrating and post-migratory neural crest cells is intriguing in the light of previous reports suggesting that these molecules may inhibit neural crest cell migration. It has been shown that the CS/KS-PGs from bovine nasal cartilage impede neural crest cell locomotion on a number of motility-promoting matrix molecules in vitro (Perris and Johansson, 1987; Perris, 1990; Perris et al. 1990b). The Swarm rat chondrosarcoma proteoglycan and a chick brain proteoglycan also have been found to be nonpermissive for neural crest cell attachment and migration (Newgreen, 1982; Tan et al. 1987). These findings in conjunction with previous in vivo obser-vations in both avians and amphibians (Newgreen et al. 1982, 1986; Tucker, 1986; Tan et al. 1987; Perris et al. 19896; Perris, 1990) suggest that the local assembly of CS/KS-PGs in the matrix may participate in the restriction and cessation of neural crest cell movement. The proteoglycans found to be inhibitory for neural crest cell migration in vitro, such as the bovine nasal cartilage CS/KS-PGs and the chondrosarcoma CS-PGs, have a high content of low sulfated chondroitin/ high sulfated keratan and prevalence of Ch4S chains (Heinegård and Paulsson, 1984). Interestingly, we observed that proteoglycans with such characteristics were invariably secluded from the primary migratory pathways of the neural crest. In contrast, there is evidence that PGs recognized by mAbs 6C3, 7D4, and 4C3 are oversulfated (Sorrell, unpublished).

Several explanations could account for the low levels of CS/KS-PG immunoreactivity around migrating neural crest cells and their derivatives. Migrating neural crest cells could suppress the secretion and matrix assembly of CS/KS-PGs by the surrounding tissues. Alternatively, neural crest cells and/or adjacent sclero-tomal cells could specifically degrade the CS/KS-PGs. A third possibility is that a high rate of hyaluronan shedding from migrating neural crest cells could exclude proteoglycans from the cell surface. High levels of hyaluronan were observed in association with migrating neural crest cells within the dorsolateral sclerotome. It has been suggested that high levels of hyaluronan may be required for the emigration of neural crest cells from neural tubes cultured in the presence of serum CS-PGs (Luckenbill-Edds and Carrington, 1988). Moreover, enhanced hyaluronan synthesis elevates the motility of fibroblasts within collagen gels that contain serum CS-PGs (Chen et al. 1989). We have recently demonstrated that inhibition of neural crest cell migration by CS/KS-PGs requires the close apposition of the proteoglycans with the cell surface, via binding to cell surface hyaluronan (Perris and Johansson, 1990). Thus, in areas sparsely popu-lated with proteoglycans, continuous shedding of hyaluronan from migrating cells may displace proteo-glycans from the cell surface, providing a proteoglycan-free microenvironment. In contrast, in regions rich in proteoglycans, such displacement of PGs could be reduced by overwhelming levels of hyaluronan, produc-ing a proteoglycan-rich environment around the cells. Consistent with previous observations (Derby, 1978; Pintar, 1978), we noted that the level of hyaluronan around post-migratory neural crest cells was consider-ably lower than that around migrating cells. Thus, the balance between extracellular proteoglycans and intra-cellular synthesis of hyaluronan may participate in and/or be a consequence of the regulation of neural crest cell migration.

Our findings indicate that a distinct segregation of proteoglycan populations occurs during development, in a fashion that relates to both neural crest cell migration and formation of the axial vertebrae. In the perinotochordal region, CS/KS-PGs recognized by the mAbs 4D3, 4D1, 2B6, 3D2 and rich in Ch4S/ChO chains accumulated around the notochord and its surrounding fibrillar network, whereas Ch6S-bearing proteoglycans, CS/KS-PGs reactive with mAbs 7D4/ 6C3/4C3, 8C2 and 5D4 and HS-PG were considerably less abundant or entirely absent from this region. However, these latter proteoglycans together with DS-PGs were enriched in the dorsal and lateral sclerotome, and in the subectodermal space. The differential occurrence of characteristic cartilage proteoglycans detected by the former antibodies produced a meta-meric pattern that inversely correlated with ganglion formation and ‘prematurely’ defined the presumptive chondrogenic sites of vertebral arch formation. In fact, these proteoglycans were restricted to the perinoto-chordal space and ventral sclerotome at axial levels containing the ganglia, but extended throughout the lateral and dorsal sclerotome at interganglionic levels. The fibrillar network of cartilage-type proteoglycans emanating from the notochord extended to the region of the ventral motor roots (Bronner-Fraser, 1986b).

In conclusion, our comprehensive analysis of the spatial and temporal distribution of proteoglycans indicates that these molecules have differential distri-bution patterns at times of neural crest migration and that these patterns are modified during development. In several in vivo (Newgreen et al. 1982, 1986; Tucker, 1986; Tan et al. 1987; Perris, 1990; Perris et al. 1990b) and in vitro (Newgreen, 1982; Newgreen et al. 1982; Tan et al. 1987; Perris, 1990; Perris and Johansson, 1987, 1990) systems, localized accumulation of proteoglycans correlates with the restriction of neural crest cell migration. Consistent with this, the dorsal subepider-mal space is extremely rich in proteoglycans at stages when neural crest cells begin their ventral migration, although neural crest cells do not migrate into the subepidermal region until later stages (Serbedzija et al. 1989). The area around the ventrolateral neural tube, the intersomitic clefts and the perinotochordal region are also particularly densely populated with proteogly-cans and these regions also lack migrating neural crest cells. During neural crest cell migration, the caudal sclerotome retains a proteoglycan-rich network, whereas the rostral sclerotome containing neural crest cells loses its proteoglycan-immunoreactivity. Finally, proteoglycans of prevalent cartilage-type accumulate metamerically in subnotochordal and periaortic re-gions, which are inversely correlated with sites of sympathetic ganglion and aortic plexus formation. Future efforts will be directed toward isolation of region- and age-specific classes of proteoglycans and disturbance of their deposition and/or assembly in situ in order to clarify their roles in neural crest cell migration, gangliogenesis and formation of the axial vertebrae. The use of our immunological probes will facilitate these studies, since these identify specific populations of PGs that have been extracted from tissue.

We thank Drs Bruce Caterson and Douglas Fambrough for providing various antibodies, and Dr Forward Laurent for donation of the purified proteoglycan fragment. Kristie Bruk, Mary Flowers, and Michael Artinger are thanked for their excellent technical assistance. This study was supported by NIH grant USPHS HD-15527 and the Muscular Dystrophy Association. MB-F is a Sloan Research Fellow.