ABSTRACT
Chondroitin sulfate proteoglycans have been implicated in the regulation of cell migration and pattern formation in the developing peripheral nervous system. To identify whether the large aggregating proteoglycan versican might be mediating these processes, we prepared monospecific antibodies against a recombinant core protein fragment of chick versican. The purified antibodies recognize the pre-dominant versican splice-variants V0 and V1. Using these antibodies, we revealed a close correlation between the spacio-temporal expression of versican and the formation of molecular boundaries flanking or transiently blocking the migration pathways of neural crest cells or motor and sensory axons. Versican is present in the caudal sclerotome, the early dorsolateral tissue underneath the ectoderm, the pelvic girdle precursor and to a certain extent in the perinotochordal mesenchyme. Versican is completely absent from tissues invaded by neural crest cells and extending axons. Upon completion of neural crest cell migration and axon outgrowth, versican expression is shifted to pre-chondrogenic areas. Since versican inhibits cellular interactions with fibronectin, laminin and collagen I in vitro, the selective expression of versican within barrier tissues may be linked to a functional role of versican in the guidance of migratory neural crest cells and outgrowing axons.
INTRODUCTION
Neural crest cells migrate along two major routes in the trunk region of the developing chick embryo (for reviews see Le Douarin, 1982; Bronner-Fraser, 1994). Those that follow the ventral pathway through the somites give rise to sensory (dorsal root ganglia) and sympathetic (autonomic) ganglia as well as adrenomedullary cells. Those migrating along the dorsolateral pathway about 24 hours later form melanocytes. After this initial phase of neural crest cell migration, motor axons leave the ventral neural tube. At the level of the developing hindlimb, they join the sensory axons extending from the dorsal root ganglia (DRG) and grow in the direction of the future plexus region (Tosney and Landmesser, 1985).
Although the trunk neural crest cells emigrate from the neural tube in an unsegmented fashion, they move on their ventral route selectively through the rostral, but not the caudal halves of the somites (Bronner-Fraser and Stern, 1991). Similarly, outgrowing axons of motor and sensory neurons avoid the caudal tissue and extend through the rostral sclero-tomes (Keynes and Stern, 1984). This regional-specific migration-behaviour has been attributed to properties of the invaded somite tissue, which may either promote cell locomotion or provide barrier functions (reviewed by Tosney, 1991; Bronner-Fraser, 1994). Similar combinations of migration pathways and flanking barrier tissues seem to guide moving neural crest cells and nerve axons in other embryonic tissues. Additional regions with neural crest cell and/or axon migration-inhibiting activity include: the perinotochordal mesenchyme (Newgreen et al., 1986; Pettway et al., 1990; Tosney and Oakley, 1990), the pelvic girdle precursor (Tosney and Landmesser, 1984) and transiently, the tissue between the somites and the ectoderm (Erickson et al., 1992).
Several extracellular matrix components, which promote neural crest cell migration and axon outgrowth, are expressed along the migration routes (reviewed by Bronner-Fraser, 1994). In particular the glycoproteins fibronectin and laminin have been identified as potent migration substrates for neural crest cells, in vitro. These cell-matrix interactions seem to be mediated by β1-integrin receptors on the surface of neural crest cells ( Lallier andBronner-Fraser, 1993; Lallier et al., 1994).
Barrier tissues are rich in chondroitin-6-sulfate (CS-6) and carbohydrates that are recognized by peanut agglutinin (PNA) (Oakley and Tosney, 1991; Oakley et al., 1994). Both glyco-conjugates are absent from the migration pathways during the period of active cell and axon locomotion. In the extracellu-lar matrix, chondroitin-6-sulfate chains are covalently attached to proteoglycan core proteins (reviewed by Kjellén and Lindahl, 1991), which eventually also carry carbohydrates bearing the PNA-binding disaccharide Gal-β(1-3)-GalNAc (Shinomura et al., 1990). Cell adhesion-inhibiting functions have been attributed to several extracellular chondroitin sulfate proteoglycans (reviewed by Ruoslahti, 1989; Jackson et al., 1991; Toole, 1991; Margolis and Margolis, 1993). It seems likely therefore that they play active roles in the formation of tissues avoided by neural crest cells and extending axons. Most of the proteoglycans expressed in the caudal halves of the somites, the perinotochordal region, the pelvic girdle tissue and in early phases between somite and ectoderm, have to date only been characterized with regard to their glycosaminoglycan moiety. To date the best described proteoglycan, which is localized in the posterior sclerotome during motor axon outgrowth (Tan et al., 1987) has been named CTB-proteoglycan (CTB: cytotactin-binding). The core protein of the CTB-proteoglycan carries chondroitin sulfate side chains and oligosaccharides recognized by PNA. Its relationship to other large chondroitin sulfate proteogly-cans is unknown.
Recently, the primary structures of a number of proteoglycans associated with the nervous system have been determined, based on their cDNA sequences (reviewed by Lander, 1993; Margolis and Margolis, 1993). The proteoglycans expressed in the extracellular space of embryonic and adult brain tissues mainly belong to the family of large aggregating proteoglycans. This family currently consists of aggrecan (Doege et al., 1987), versican (Zimmermann and Ruoslahti, 1989), neurocan (Rauch et al., 1992) and brevican (Yamada et al., 1994). All four members are substituted with chondroitin sulfate side chains. Their core proteins are composed of distinct gly-cosaminoglycan attachment domains separating homologous globular domains at both ends of the core protein. These amino- and carboxyl-terminal structures bear a hyaluronan-binding domain and a set of EGF-, lectin- and complement regulatory-elements, respectively.
Recently, we have demonstrated that at least three isoforms of human versican (also known as PG-M) exist as a result of alternative splicing processes involving exon 7 and 8 of the versican gene (Dours-Zimmermann and Zimmermann, 1994; Naso et al., 1994). The three splice-variants differ in length and composition of their chondroitin sulfate attachment domains and have been named V0, V1 and V2 (Dours-Zimmermann and Zimmermann, 1994). Versican has been localized in brain tissues (Perides et al., 1992), in skin (Zimmermann et al., 1994) and in aorta (Yao et al., 1994). Furthermore, versican/PG-M is transiently expressed in the pre-chondrogenic mesenchyme of developing chick limb buds (Kimata et al., 1986; Shinomura et al., 1990).
A number of observations made in vitro provide evidence for a cell adhesion destabilizing function of versican/PG-M: (i) immobilized versican interferes with the substrate attachment of primary fibroblasts to collagen I, fibronectin, vitronectin and laminin (Yamagata et al., 1989); (ii) versican, present in the pericellular matrix of cultured fibroblasts, is excluded from focal contacts (Yamagata et al., 1993a) and (iii) inhibition of versican synthesis represses the malignant cell-adhesion phenotype of MG63 osteosarcoma cells (Yamagata and Kimata, 1994).
Although the structure and cell adhesion impairing proper-ties of versican match well with the profile of candidate inhibitors of neural crest cell and motor axon locomotion, no detailed study of the versican distribution in early development of the trunk region has been done so far. We therefore prepared monospecific antibodies against chicken versican and used them in immunohistochemical experiments. We demonstrate here the close correlation between versican expression and the formation of tissues that inhibit the migration of neural crest cells and the outgrowth of motor and sensory axons.
MATERIALS AND METHODS
Preparation of the bacterial fusion protein
A bacterial fusion protein containing amino acids 2887 to 3246 of chick versican was prepared as previously described (Zimmermann et al., 1994). A cDNA fragment covering bases 8803 to 9882 (Shinomura et al., 1993) was amplified from chicken genomic DNA by polymerase chain reaction using a PCR amplification kit (Perkin-Elmer, Norwalk, CT). The upper primer included at the 5′ end an SfiI restriction site and the lower primer a NotI restriction site, in addition to the versican-specific sequences. 300 ng of genomic DNA was used as template in a 100 μl reaction. 28 cycles consisting of 1 minute denaturing at 94°C, 2 minutes primer annealing at 60°C and 2 minutes extension at 72°C were run on a thermocycler (Perkin-Elmer, Norwalk, CT). The final extension was prolonged to 9 minutes. The PCR product was restricted with SfiI and NotI and ligated into the pDS9-cassette expression vector (Zimmermann et al., 1994). The fusion protein encoding portion of the resulting construct was verified by DNA sequencing.
M15[pREP4] cells (Stüber et al., 1990) were used as bacterial hosts for the expression of the recombinant protein. The expression of fusion protein was induced by addition of IPTG to a final concentration of 2 mM. After 5 hours, the cells were centrifuged at 6,000 g for 10 minutes. The pellet was resuspended in 0.15 M NaCl / 0.2 mM PMSF / 2 mM iodacetamide / 0.02% sodium azide / 25 mM Tris, Ph 8.0 at 4°C. The suspension was sonicated and cleared of bacterial debris by centrifugation. The supernatant was diluted with 1.5 volumes of 67 mM sodium phosphate / 0.3 M NaCl / pH 8.0 and directly applied to a 3.5 ml Ni2+-NTA agarose affinity column (Qiagen, Basel, Switzerland). After three wash steps with 67 mM sodium phosphate / 0.3 M NaCl at pH 8.0 and pH 6.0 followed by 0.1 M sodium acetate / 0.3 M NaCl, pH 3.8, the column was neutral-ized again with 67 mM sodium phosphate / 0.3 M NaCl, pH 8.0. Elution of the fusion protein was achieved by including 0.25 M imidazol in the latter buffer. Peak fractions were analyzed by SDS-polyacrylamide gel electrophoresis on 12.5% Phastgel (Pharmacia, Uppsala Sweden) and compared with bacterial extracts prepared by boiling fusion protein expressing cells in SDS sample buffer.
In addition, control extracts of M15[pREP4] cells transfected with the expression vector alone were obtained by sonication in 0.15 M NaCl / 0.2 mM PMSF / 2 mM iodacetamide / 0.02% sodium azide / 25 mM Tris, pH 8.0. After centrifugation at 10,000 g for 20 minutes, the proteins in the supernatant were precipitated with 3 volumes of ice cold ethanol.
Preparation and purification of polyclonal antibodies
A Chinchilla rabbit was immunized according to standard protocols with fusion protein emulsified in incomplete Freund’s adjuvant (Harlow and Lane, 1988). Polyclonal antibodies were purified on a Sepharose 4B column (CNBr-activated; Pharmacia) coupled with 5 mg/ml bacterial protein from the control extract, followed by absorption on a column coupled with 2.5 mg/ml isolated fusion protein. Elution was performed with 0.2 M glycine, pH 2.5. The eluate was neutralized with 0.5 volume of 1 M Tris, pH 8.0. The antibodies were precipitated by addition of (NH4)2SO4 to 40% saturation and redissolved in TBS / 1% BSA / 0.02% sodium azide. Pre-immune serum was treated analogously and used for negative controls.
Cell culture
Primary fibroblast cultures were prepared by trypsinizing stage 31 chick trunks. The dissociated cells were cultured in DMEM containing 10% FCS. Confluent cultures were passaged up to 4 times in a ratio of 1:4. HNK-1 hybridoma cells (ATCC / TIB 200) were grown in RPMI 1640 medium containing 20% FCS, 0.02 mM β-mercaptoethanol, 100 U/ml penicillin, 100 μg/ml streptomycin and 1 μg/ml fungizone as proposed by the supplier.
Isolation of proteoglycans
Versican-enriched fractions were prepared from the culture super-natant of primary fibroblasts by anion exchange chromatography. After addition of PMSF, EDTA and sodium azide (final concentrations 1 mM, 10 mM and 0.02%, respectively), the conditioned culture medium was incubated with 5 ml Q-Sepharose (Pharmacia) per liter on a shaker at 4°C overnight. After centrifugation at 3,000 g for 5 minutes, the supernatant was removed and the resin was packed into a column and washed with 6 M urea / 10 mM EDTA / 1 mM PMSF/ 50 mM Tris, pH 8.0 and with the same buffer containing 0.3 M NaCl. The proteoglycan-enriched fraction was eluted by raising the NaCl concentration to 1 M. Fractions containing versican were identified by immunoblotting and precipitated with 7 volumes of ethanol : H2O, 6:1. Aliquots were digested with chondroitinase ABC (ICN, Costa Mesa, CA) overnight at 37°C, using 2 U/ml of enzyme and 10 μg/ml ovomucoid (Sigma, Buchs, Switzerland) in 40 mM Tris-acetate, pH 8.0.
SDS-PAGE and immunoblotting
Extracts of bacteria expressing fusion protein and proteoglycan fractions were run under reducing conditions on 12.5 or 4-15% SDS-polyacrylamide gels (Phastgels, Pharmacia), respectively, and stained with Coomassie blue or processed for diffusion blotting according to the manufacturer’s recommendations. The blotting efficiency was tested by including prestained molecular mass standards (Bio-Rad Laboratories, Richmond, CA) in each electrophoresis run. Blots were blocked with 3% low fat milk in TBS for 30 minutes. Affinity purified primary antibodies were used at a dilution of 1:1,000. Subsequent steps were performed with a ProtoBlot AP western blot detection system (Promega Biotec, Madison, WI).
Immunohistochemical staining
White leghorn chicken embryos at developmental stages ranging from 16-35 (Hamburger and Hamilton, 1951) were used for this study. After removal of the surrounding membranes, embryos were fixed in Bouin’s fixative (saturated picric acid in H2O / 37% formaldehyde / glacial acetic acid 15:5:1) for 1 hour at room temperature, dehydrated in a graded series of ethanol and embedded in parablast. Immunohis-tochemical staining of 2-3 μm thick sections was achieved using an immunoperoxidase Vectastain ABC kit (Vector Laboratories, Inc., Burlingame, CA). Incubation with 1:10 diluted first antibodies was performed overnight at 4°C in a moist chamber. The subsequent steps were done according to the manufacturer’s instructions. All sections were counterstained for 15 seconds in Mayer’s hematoxylin solution. The control experiments included blocking of immunoreaction by addition of 50 μg/ml fusion protein to the antibody solution or replacing the first antibodies with processed pre-immune serum.
Neural crest cells were detected with the monoclonal antibody HNK-1 (Tucker et al., 1984). For immunostaining, undiluted super-natant of HNK-1 hybridoma cell cultures was applied to rehydrated tissue sections. Detection was performed with biotinylated goat anti-bodies recognizing mouse IgM (Vector Laboratories) and a Vecta-stain ABC kit.
RESULTS
Preparation of a recombinant versican fragment in a bacterial expression system
For the preparation of versican-specific antibodies, a unique portion of the chicken versican core protein (Fig. 1) was expressed in a bacterial expression system. The corresponding DNA fragment, coding for a part of the GAG-β domain, was PCR-amplified from genomic DNA. The sequence of the 1.1 kb PCR-product was identical with the published cDNA sequence of chicken versican/PG-M (Shinomura et al., 1993), except for an in frame deletion of a repetitive element between bases 9662 to 9673. Due to its histidine-rich leader peptide the recombinant core protein fragment could be isolated in a single chromatography step on a metal chelating column charged with nickel. 10.5 mg of fusion protein (Fig. 2) was obtained from one liter of bacterial culture.
Localization of the versican core protein fragment used for the preparation of polyclonal antibodies (shaded area). Models of versican splice-variants V0 and V1 are modified from Dours-Zimmermann and Zimmermann (1994). HABR, hyaluronan-binding region; GAG-α and GAG-β, glycosaminoglycan attachment domains; EGF/Lec/CRP, epidermal growth factor-, lectin- and complement regulatory protein-like elements.
Localization of the versican core protein fragment used for the preparation of polyclonal antibodies (shaded area). Models of versican splice-variants V0 and V1 are modified from Dours-Zimmermann and Zimmermann (1994). HABR, hyaluronan-binding region; GAG-α and GAG-β, glycosaminoglycan attachment domains; EGF/Lec/CRP, epidermal growth factor-, lectin- and complement regulatory protein-like elements.
Characterization of polyclonal antibodies against chick versican. Bacterial extracts (EFP), affinity-purified fusion proteins (FP) and proteoglycan preparations from embryonic fibroblast cells (Fbl-PG) were separated by SDS-PAGE on 12.5% and 4-15% Phastgels, respectively. Gels were either subjected to Coomassie blue staining or processed for western blotting. The purified antibodies against the recombinant core protein fragment recognize the V0- and V1-splice variants of chick versican (arrows). (+) and (−) indicates whether proteoglycan samples were digested with chondroitinase ABC. The Mr of standard proteins are indicated.
Characterization of polyclonal antibodies against chick versican. Bacterial extracts (EFP), affinity-purified fusion proteins (FP) and proteoglycan preparations from embryonic fibroblast cells (Fbl-PG) were separated by SDS-PAGE on 12.5% and 4-15% Phastgels, respectively. Gels were either subjected to Coomassie blue staining or processed for western blotting. The purified antibodies against the recombinant core protein fragment recognize the V0- and V1-splice variants of chick versican (arrows). (+) and (−) indicates whether proteoglycan samples were digested with chondroitinase ABC. The Mr of standard proteins are indicated.
Polyclonal antibodies directed against the recombinant fragment recognize intact chicken versican
The antiserum obtained from a rabbit immunized with the fusion protein reacted strongly with the recombinant core protein in immunoblots (Fig. 2). Test blots with protein extracts of bacteria expressing another fusion protein demonstrated that the affinity-purified antibodies neither recognized proteins of the M15 host cells nor the amino-terminal fusion sequence (data not shown). The polyclonal anti-fusion protein antibodies reacted specifically with versican secreted into the conditioned culture medium of primary chicken fibroblast. Whereas the intact versican molecule remained in the loading slots of 4-15% SDS-polyacrylamide gels, two high molecular mass core proteins were recognized on immunoblots after digestion with chondroitinase ABC (Fig. 2). The two bands correspond to the V0 and V1 splice variants of chicken versican. Cultured embryonic fibroblasts express mainly the V0 isoform as judged from the intensities of these two core protein bands. A third, clearly smaller component recognized by the antibodies most likely represents a proteolytic degradation product of versican.
In immunohistochemical experiments, the polyclonal anti-bodies reacted with versican in cryo-preserved (data not shown) as well as in Bouin-fixed and parablast embedded tissues. The binding could be blocked to completion by addition of 50 μg/ml fusion protein to the antibody solution (Fig. 5B). Similarly, no immuno-staining was observed in sections, where the anti-versican antibodies were replaced with processed pre-immunserum (Fig. 5I).
Versican expression in somite tissues inversely correlates with the migration pathways of neural crest cells
The segmental development of the paraxial mesoderm in the trunk region of the chicken embryo takes place in an anterior-posterior sequence. At Hamburger and Hamilton stages between 16 and 20, early and advanced forms of somitic tissues coexist in the same embryo. In stage 20 embryos (Fig. 3), versican expression is first seen in association with the basement membrane surrounding the newly formed somites. The staining is particularly intense between the somites and the ectoderm. The expression constantly decreases in the basement membrane of the more rostral somites. Where the intersomitic fissure becomes evident, the paucity of versican staining at the anterior edge of the somitic tissue is noteworthy (Fig. 3A).
The expression of versican inversely correlates with the pathways of neural crest cells and extending axons. Parasagittal (A,B,G,H) or oblique frontal to parasagittal sections (C,D,E,F) of stage 20 chick embryos (41 somite-stage) were stained with either polyclonal antibodies against versican (A,C,E,G) or with the monoclonal antibody HNK-1 recognizing neural crest cells (B,D,F,H). Rostral is to the left, caudal to the right. In somites −7 to −4 (numbers refer to the position rostral from the last forming somite), versican staining is associated with the basement membrane of the early somites and underneath the ectoderm (A). Within the caudal somites, versican appears in the forming sclerotome of somite −5 and gradually increases in more rostral somites. First neural crest cells migrate in the intersomitic cleft between somite −5 and −6 along the anterior edge of somite −5, where versican is nearly absent (arrowheads in A,B). In somite −6, a few neural crest cells are detected within the rostral sclerotome. In somites −15 to −12 (C,D) versican staining is markedly increased in caudal sclerotome halves. HNK-1 immunoreactivity reveals a large number of neural crest cells that migrate through the rostral sclerotome. At the level of the forelimb (E,F; somites −25 to −22) the staining patterns of HNK-1 and anti-versican antibodies are clearly complementing each other. Diffuse versican staining is seen within the lateral neural tube and at its border with the sclerotome tissue. In somites located more rostrally (somites −33 to −28) the dermatome is dispersing (G,H). Versican expression relocates to pre-chondrogenic areas and is therefore also found in the rostral somite between DRG and myotomes. Versican is, however, completely absent in regions that bind HNK-1 antibodies. nt: neural tube. Scale bars, 100 μm
The expression of versican inversely correlates with the pathways of neural crest cells and extending axons. Parasagittal (A,B,G,H) or oblique frontal to parasagittal sections (C,D,E,F) of stage 20 chick embryos (41 somite-stage) were stained with either polyclonal antibodies against versican (A,C,E,G) or with the monoclonal antibody HNK-1 recognizing neural crest cells (B,D,F,H). Rostral is to the left, caudal to the right. In somites −7 to −4 (numbers refer to the position rostral from the last forming somite), versican staining is associated with the basement membrane of the early somites and underneath the ectoderm (A). Within the caudal somites, versican appears in the forming sclerotome of somite −5 and gradually increases in more rostral somites. First neural crest cells migrate in the intersomitic cleft between somite −5 and −6 along the anterior edge of somite −5, where versican is nearly absent (arrowheads in A,B). In somite −6, a few neural crest cells are detected within the rostral sclerotome. In somites −15 to −12 (C,D) versican staining is markedly increased in caudal sclerotome halves. HNK-1 immunoreactivity reveals a large number of neural crest cells that migrate through the rostral sclerotome. At the level of the forelimb (E,F; somites −25 to −22) the staining patterns of HNK-1 and anti-versican antibodies are clearly complementing each other. Diffuse versican staining is seen within the lateral neural tube and at its border with the sclerotome tissue. In somites located more rostrally (somites −33 to −28) the dermatome is dispersing (G,H). Versican expression relocates to pre-chondrogenic areas and is therefore also found in the rostral somite between DRG and myotomes. Versican is, however, completely absent in regions that bind HNK-1 antibodies. nt: neural tube. Scale bars, 100 μm
Within the somites, versican staining first appears in the caudal half of the fifth somite rostral of the segmental plate. The staining is present around the cells clustered within the somite, but is absent from the outer epitheloid cell layer. In the more rostrally located somites that are reorganizing into dermomyotome and sclerotome tissues, the patchy versican staining gradually becomes more intense, but remains confined to the caudal somitic tissues. The expression of versican in the posterior half of the forming sclerotome precedes the invasion of neural crest cells as demonstrated by immunohistochemical staining with the HNK-1 antibody. The first migrating neural crest cells are observed in the inter-somitic fissure between the fifth and the sixth somite rostral from the segmental plate (Fig. 3B). Interestingly, they follow the anterior border of the somite, which lacks versican expression. The neural crest cells only enter the rostral half of the somitic tissues once the transition of the somites into der-mamyotome and sclerotome is well underway. The complementary staining pattern of anti-versican and HNK-1 anti-bodies becomes even more prominent in the sclerotome tissues localized more rostrally (Fig. 3C,D). Migratory neural crest cells are clearly absent in tissues that express versican. As the DRG form and motor axons extend from the ventral neural tube, the sharp discontinuity between versican staining in the posterior and the absence of staining in the anterior half of the sclerotome prevails (Fig. 3C). Only in the most rostrally localized somites could versican-specific immunoreactivity be seen in the anterior half of the sclerotome between the DRG and the myotome tissue (Fig. 3G).
Apart from in the caudal half of the sclerotome, versican staining is also observed between the ectoderm and the der-mamyotomes of the 12 to 15 caudal-most somites of stage 20 embryos (Fig. 3A,C). This unsegmented expression of versican gradually decreases at later embryonic stages and finally dis-appears (Fig. 4D,E) at a time when neural crest cells migrate along the dorsolateral path (at the hindlimb level slightly later than stage 22).
Versican is expressed within tissues interfering with neural crest cell migration as well as motor and sensory axon locomotion. Transverse sections at the hind limb level document the different expression pattern between rostral (A,D,G) and caudal halves of somites (B,E,H) and the distribution of versican in the hindlimb bud (C,F,I). Corresponding sections were taken from stage 20 (A-C), stage 22 (D-F) and stage 25 embryos (G-I). Versican is seen mainly in the ventral portion of the caudal sclerotomes (B,E,H), in the pelvic girdle precursor and in the hindlimb (C,F,I). Versican is, however, absent from pathways of extending motor and sensory axons. Growth cones extend from the ventral horn of the neural tube and the DRG (A), pass the anterior sclerotome and spread out in the plexus region, but are temporarily restrained from entering the hindlimb (D). At stage 25 motor and sensory axons have traversed the gaps in the pelvic girdle precursor and extended into the limb (G). Whereas versican is uniformly distributed in stage 20 hindlimb (C), expression is confined in stage 25 to areas of condensing mesenchyme and to subectodermal tissue (I). nt, neural tube; m, motor axons; drg, dorsal root ganglia; n, notochord; p, plexus region; pg, pelvic girdle precursor. Scale bars, 100 μm.
Versican is expressed within tissues interfering with neural crest cell migration as well as motor and sensory axon locomotion. Transverse sections at the hind limb level document the different expression pattern between rostral (A,D,G) and caudal halves of somites (B,E,H) and the distribution of versican in the hindlimb bud (C,F,I). Corresponding sections were taken from stage 20 (A-C), stage 22 (D-F) and stage 25 embryos (G-I). Versican is seen mainly in the ventral portion of the caudal sclerotomes (B,E,H), in the pelvic girdle precursor and in the hindlimb (C,F,I). Versican is, however, absent from pathways of extending motor and sensory axons. Growth cones extend from the ventral horn of the neural tube and the DRG (A), pass the anterior sclerotome and spread out in the plexus region, but are temporarily restrained from entering the hindlimb (D). At stage 25 motor and sensory axons have traversed the gaps in the pelvic girdle precursor and extended into the limb (G). Whereas versican is uniformly distributed in stage 20 hindlimb (C), expression is confined in stage 25 to areas of condensing mesenchyme and to subectodermal tissue (I). nt, neural tube; m, motor axons; drg, dorsal root ganglia; n, notochord; p, plexus region; pg, pelvic girdle precursor. Scale bars, 100 μm.
Transition of versican expression from barrier tissues to pre-chondrogenic areas. Transverse sections of stage 20 (A,B), 22 (C,D), 25 (E,F), 30 (G,H) and 35 embryos (I,K) at the hindlimb (A,B,C,E,G) and the forelimb level (D,F,H,I,K). Sections A, C-H and K were stained with anti-versican antibodies, sections B and I were subjected to control experiments which included: blocking of the anti-versican immunoreaction by addition of 50 μg/ml fusion protein (B) or replacing the first antibody with processed pre-immune serum (I). In the course of neural crest cell migration and peripheral nerve outgrowth (until stage 25), versican is mainly present in barrier tissues (A-D). In the trunk of stage 25 to stage 30 embryos, versican expression relocates to regions where chondrogenesis takes place (E-H). At stage 35, when cartilageous vertebrae and ribs have formed, versican has nearly disappeared (K). Scale bars, 100 μm (A-F); 300 μm (G-K).
Transition of versican expression from barrier tissues to pre-chondrogenic areas. Transverse sections of stage 20 (A,B), 22 (C,D), 25 (E,F), 30 (G,H) and 35 embryos (I,K) at the hindlimb (A,B,C,E,G) and the forelimb level (D,F,H,I,K). Sections A, C-H and K were stained with anti-versican antibodies, sections B and I were subjected to control experiments which included: blocking of the anti-versican immunoreaction by addition of 50 μg/ml fusion protein (B) or replacing the first antibody with processed pre-immune serum (I). In the course of neural crest cell migration and peripheral nerve outgrowth (until stage 25), versican is mainly present in barrier tissues (A-D). In the trunk of stage 25 to stage 30 embryos, versican expression relocates to regions where chondrogenesis takes place (E-H). At stage 35, when cartilageous vertebrae and ribs have formed, versican has nearly disappeared (K). Scale bars, 100 μm (A-F); 300 μm (G-K).
Throughout all stages of somite development, versican is absent from the dermamyotome. In contrast, the lateral neural tube and the overlying basement membrane displayed some immunohistochemical staining.
Versican is expressed in tissues flanking the pathways of extending motor and sensory axons
The specific expression of versican in tissues that are avoided by growing sensory and motor axons is best documented in cross-sections at the hindlimb level of stage 20 to 25 chick embryos (Fig. 4).
In the beginning of motor axon outgrowth at stage 20, the neural tube is generally surrounded by a basement membrane, which stains with the anti-versican antibodies (Fig. 4A,B). This membrane is, however, absent at positions where the axons leave the neural tube. The motor axons extend into the rostral sclerotome mostly lacking versican expression and join the ventrally growing axons from the DRG. In the versican-rich matrix of the caudal sclerotome, neither motor nor sensory axons are present. At this stage a uniform versican staining is observed in the developing limb bud (Fig. 4C).
At stage 22, motor and sensory axons have invaded the plexus mesenchyme and begin to migrate along the anterior-posterior axis (Fig. 4D,E). Whereas versican is not expressed in the plexus, strong immunohistochemical signals are observed in the neighbouring pelvic girdle precursor and in the anterior sclerotome. At this time point, the staining with anti-versican antibodies in the developing limb bud becomes more intense underneath the ectoderm and in the pre-chondrogenic region including the pelvic girdle precursor (Fig. 4F).
Innervation of the limb bud is well underway in stage 25 embryos (Fig. 4G-I). The axons extend through versican-negative gaps in the pelvic girdle precursor and branch off into a dorsal and ventral ganglion. Versican has now completely disappeared from the tissues invaded by the migrating growth cones, it is however, strongly expressed in the boundaries of the pathways, namely the pre-chondrogenic area of the forming pelvic girdle and the femur, as well as in the developing dermis.
Throughout stages 20-25 (hindlimb level), weak versican staining is observed in the lateral neural tube and around the motor columns. In stage 20 and 22 embryos, a faint signal can be detected in the zone of the roof plate (Fig. 4A,B,D,E). In addition, versican is associated with the basement membranes surrounding the neural tube and the notochord. The initially strong versican staining in the perinotochordal mesenchyme near the newly formed somites (data not shown) has mainly disappeared at the hindlimb level of stage 20-25 embryos (Fig. 4). However, versican expression again becomes more pronounced ventrally to the floor plate, when notochord and neural tube get separated by a layer of mesenchymal tissue (Figs 4G,H, 5E,F).
Versican expression switches from barrier tissues to pre-chondrogenic areas after completion of axon outgrowth
Beginning from about stage 25 of development, versican expression is evident in condensing mesenchyme that forms the vertebrae and ribs (Fig. 5). The first staining of versican in these prechondrogenic tissues is observed dorsolaterally between the neural tube and the differentiating myotome as well as in the mesenchyme adjacent to the notochord (Fig. 5E,F). During the transition from mesenchyme to cartilage tissue, versican expression is down-regulated and becomes restricted to the periphery of the newly formed cartilage, where chondrogenesis is still underway (Fig. 5G,H,K). At the end of this process, versican is virtually absent from the dorsal trunk of the chick embryo, except for an area underneath the dorsal ectoderm where the dermal tissues are forming (Fig. 5K).
DISCUSSION
Versican is a member of the growing family of proteoglycans that interact with hyaluronan and thus form large aggregates. The binding is mediated by a link protein-like domain in the highly homologous amino-terminal portion of their core proteins. In addition to the hyaluronan-binding domain, aggregating proteoglycans share similar protein structures at their carboxyl-terminal ends. The sequence identities among corresponding elements of the different proteoglycans (currently aggrecan, versican, neurocan and brevican) range between 55 and 70%. This close structural similarity among the different core proteins imposes a major problem for the preparation of specific antibodies, as shared epitopes may cause cross-reactivity. We therefore selected a unique portion of the GAG-β domain of chicken versican, which most likely lacks glycosaminoglycan side-chains, and expressed it in a prokaryotic expression system. As anticipated, the polyclonal antibodies prepared against this recombinant core protein fragment specifically recognize intact as well as chondroitinase ABC treated versican V0 and V1 on immunoblots and tissue sections.
Using these antibodies in immunohistochemical experiments, we have studied the tissue distribution of versican during the trunk development of chicken embryos and observed that versican is generally absent from the path of migrating neural crest cells and extending axons. It is however, highly expressed in the flanking tissues at time points when active cell or axon locomotion takes place.
The onset of versican expression occurs shortly after the closure of the neural tube. Versican appears first in the tissues surrounding the notochord and at the periphery of the newly formed somites. This early expression precedes the ventral migration of neural crest cells that initially enter the inter-somitic space between the epithelial somites (Loring and Erickson, 1987). Already at this time point, it is noteworthy that the few migrating crest cells seem to follow the posterior side of the intersomitic fissure, which is free of versican. This complementary pattern between tissues expressing versican and migration pathways becomes more prominent, when neural crest cells invade the rostral sclerotome after the epi-thelial-mesenchymal transition of the somites has taken place. Again, the expression of versican in the caudal sclerotome precedes the immigration of neural crest cells into the rostral half. The third example for exclusion of migrating neural crest cells from matrices rich in versican is provided by the dorso-lateral tissue between ectoderm and the newly formed somite. This subectodermal matrix contains significant amounts of versican at the beginning of somite formation. The expression however, gradually decreases. When the dermatome is dis-persing, versican is virtually absent from this tissue. The time point coincides with the initiation of neural crest cell migration along the dorsolateral path (Serbedzija et al., 1989; Erickson et al., 1992).
This apparent exclusion of versican from migration pathways can also be observed during the outgrowth of axons from the ventral neural tube and from the DRG. Both motor axons and sensory axons are consistently absent from versican-rich tissues such as the posterior half of the sclerotome and the pelvic girdle precursor. The innervation of the hindlimbs only takes place when versican-free gaps in the pelvic girdle precursor have formed and the initially high, uniform versican expression in the limb bud is restricted to the pre-chondrogenic mesenchyme and the developing dermis.
This close association of versican expression with neural crest cell and axon migration may indicate a guidance function for this large aggregating proteoglycan during the formation of the peripheral nervous system. Previous investigations provided convincing evidence that the region-specific migration of neural crest cells and motor axons is mainly dictated by the combination of migration-promoting substrates in the pathways and migration-inhibiting molecules in the adjacent boundary tissues and not by migratory cells them-selves (reviewed by Tosney, 1991; Bronner-Fraser, 1993; Bronner-Fraser, 1994). For instance, microsurgical inversion of the rostrocaudal polarity of the somites also reverses the migration preference of neural crest cells and extending axons from the rostral to the caudal half of the sclerotomes (Keynes and Stern, 1984; Bronner-Fraser and Stern, 1991). In vitro experiments demonstrated that the growth cones of motor axons turn away from posterior sclerotome cells after surface contact, whereas they exhibit a selective affinity for anterior cells (Oakley and Tosney, 1993). Other migration-inhibiting tissues comparable to the caudal sclerotome have since been identified. They include the perinotochordal mesenchyme (Newgreen et al., 1986; Pettway et al., 1990; Tosney and Oakley, 1990), the pelvic girdle precursor (Tosney and Landmesser, 1984) and the extracellular matrix between ectoderm and somites prior to the dispersion of the dermatome (Erickson et al., 1992).
Versican is present in all of these barrier tissues, except for the perinotochordal mesenchyme, which contains versican only during early stages of the trunk development and again later in the pre-chondrogenic phase of vertebrae formation. During the period of motor axon outgrowth, versican is largely absent from the perinotochord, but seems to be replaced by the closely related proteoglycan, aggrecan (Oettinger et al., 1985; Yamagata et al., 1993b). In addition to versican and aggrecan, another large chondroitin sulfate proteoglycan (CTB-proteo-glycan) has been identified in barrier tissues, namely in the caudal sclerotome and the early notochord (Tan et al., 1987). The structural relationship between versican and CTB-proteo-glycan is currently unknown. The core protein of the CTB-pro-teoglycan has a Mr of 280×103, significantly smaller than versican V0 and V1. It seems possible however, that the CTB-proteoglycan corresponds to the shorter V2 splice variant of versican (Dours-Zimmermann and Zimmermann, 1994; Ito et al., 1995) or to a proteolytic fragment of the longer versican isoforms.
A number of other reports have described the close association of chondroitin sulfate proteoglycans with barrier tissues formed to restrict neural crest cell migration and axonal outgrowth (Oakley and Tosney, 1991; Perris et al., 1991; Oakley et al., 1994) during development. As monoclonal anti-bodies recognizing only the glycosaminoglycan chains were used, no assignment of the staining patterns to a particular pro-teoglycan core protein could be made. Our results reveal now that the combined expression of versican and aggrecan in the early chick trunk correlates well with the spatial and temporal distribution of chondroitin-6-sulfate and peanut agglutinin binding glycoproteins, both of which have been proposed as markers for barrier tissues (Oakley and Tosney, 1991; Oakley et al., 1994). Indeed, it seems likely that at least some of the carbohydrates detected by PNA and antibodies against CS-6 are covalently attached to the versican core protein, since limb bud versican from stage 22-23 chick embryos carries about 60% chondroitin-6-sulfate (Kimata et al., 1986) and binds to PNA affinity columns (Shinomura et al., 1990).
Other molecules that are selectively expressed in the caudal sclerotome include T-cadherin (Ranscht and Bronner-Fraser, 1991) and two PNA-binding proteins with molecular masses of 48×103 and 55×103 (Davies et al., 1990). In vitro experiments suggested that the PNA-binding glycoproteins induce growth cone collapse of dorsal root ganglion cells (Davies et al., 1990). As versican is often isolated from tissues in a proteolytically degraded form (unpublished observation) the question arises, whether the PNA-binding proteins may be related to fragments of the versican core protein. Davies and co-workers demonstrated that the 48K and 55K proteins do not bind to hyaluronan (Davies et al., 1990). This observation led them to exclude a similarity with hyaluronectin, a glycoprotein that most likely originates from the amino-terminal portion of versican by a proteolytic process. However, these findings do not rule out a relationship between the PNA-binding glyco-proteins and other portions of the versican core protein.
In fact, large chondroitin sulfate proteoglycans themselves are potent modulators of cell adhesion and migration. For instance, the migration of neural crest cells and the outgrowth of neurites from dorsal root ganglion cells on fibronectin or laminin are efficiently blocked by addition of chondroitin sulfate proteoglycans (Perris and Johansson, 1987, 1990; Snow et al., 1990, 1994). In vitro experiments demonstrated that versican/PG-M interferes with the attachment of embryonic fibroblasts to various substrates including fibronectin, laminin and collagens (Yamagata et al., 1989). It seems likely therefore that versican also modulates fibronectin-, laminin- and collagen-mediated cell-matrix interactions in vivo. Thus the specific expression of versican in barrier tissues may be associated with a putative guidance function directing neural crest cells and outgrowing axons to their target regions.
ACKNOWLEDGEMENTS
We thank Ana Sestak and Christine Frankenfeldt for their excellent technical assistance, Parvin Saremaslani and Peter Weber for the useful hints, Norbert Wey and Hannes Nef for helping with the photographs and Professors Jakob Briner and Philipp U. Heitz for their continued support. This work was supported in part by grants from the Swiss National Science Foundation (grant no. 31-28882.90), from the Ida de Pottère-Leupold Foundation and from the Krebsliga des Kantons Zürich to D. R. Z.