Versican V0 and V1 Guide Migratory Neural Crest Cells*

We previously showed the selective expression of the chondroitin sulfate proteoglycans versican V0 and V1 in barrier tissues that impede the migration of neural crest cells during embryonic trunk development (Landolt, R. M., Vaughan, L., Winterhalter, K. H., and Zimmermann, D. R. (1995) Development 212, 2303-2312). To test for an active involvement of these isoforms in the guidance process, we have now established protocols to isolate intact versican V0 and V1 in quantities sufficient for functional experiments. Using stripe choice assays, we demonstrate that pure preparations of either a mixture of versican V0/V1 or V1 alone strongly inhibit the migration of multipotent Sox10/p75NTR double-positive early neural crest stem cells on fibronectin by interfering with cell-substrate adhesion. We show that this inhibition is largely core glycoprotein-dependent, as the complete removal of the glycosaminoglycan chains has only a minor effect on the inhibitory capacity. Our findings support the notion that versican variants V0 and V1 act, possibly in concert with other inhibitory molecules such as aggrecan and ephrins, in directing the migratory streams of neural crest cells to their appropriate target tissues.

To test for an active involvement of these isoforms in the guidance process, we have now established protocols to isolate intact versican V0 and V1 in quantities sufficient for functional experiments. Using stripe choice assays, we demonstrate that pure preparations of either a mixture of versican V0/V1 or V1 alone strongly inhibit the migration of multipotent Sox10/p75NTR double-positive early neural crest stem cells on fibronectin by interfering with cell-substrate adhesion. We show that this inhibition is largely core glycoproteindependent, as the complete removal of the glycosaminoglycan chains has only a minor effect on the inhibitory capacity. Our findings support the notion that versican variants V0 and V1 act, possibly in concert with other inhibitory molecules such as aggrecan and ephrins, in directing the migratory streams of neural crest cells to their appropriate target tissues.
The highly precise and coordinated migration of neural crest cells during early phases of embryonic development is controlled by the differential expression of permissive substrates and non-permissive/inhibitory molecules within the pathways and the bordering tissues, and the set of membrane receptors present on the moving cells (reviewed by Refs. [1][2][3][4]. The journey of the multipotent neural crest stem and progenitor cells begins in the dorsal neural tube from where they emerge shortly after its closure. In the trunk region the cells are initially guided along a ventral trajectory before a second wave starts to invade the dorsolateral tissue underneath the ectoderm. Whereas the ventrally migrating populations differentiate into neurons and glia of the sensory and the sympathetic nervous system, the laterally progressing cells give rise to the melanocytes of the skin. On their route, neural crest cells pass through highly permissive extracellular matrices, which allow rapid cellular movements. These pathways are flanked by tissues that block neural crest cell immigration and thus provide the directional information. These barrier tissues, previously identified by microsurgical manipulations, include the posterior sclerotomes (5), the perinotochordal region (6), and for a short period also the subectodermal matrix prior to melanocyte precursor invasion (7). Consequently, the streams of neural crest cells, which originally emigrate in an unsegmented fashion from the dorsal neural tube, are on their ventral path canalized into the anterior sclerotome strictly avoiding the posterior somitic halves and the more ventrally localized perinotochordal zone. This particular migration behavior finally leads to the characteristic segmental pattern of the forming sensory and sympathetic ganglia.
Since the major migration promoting substrates, fibronectin and laminin (8), are uniformly expressed in both halves of the somites (9), the guidance of the migratory neural crest cells appears to depend mainly on inhibitory cues. Several extracellular matrix and cell surface components match the candidate profile for a migration blocking function as they are selectively expressed in non-permissive tissues. Molecules consistently absent from the pathways, but highly expressed within the barriers, include chondroitin 6-sulfate proteoglycans, peanut agglutinin (PNA) 4 -binding glycoproteins (10,11), F-spondin (12), semaphorin3A (13), T-cadherin (14), collagen IX (15), and, except for the dorsolateral path, ephrins (16 -18).
For some of these molecules, like semaphorin3A (13) and ephrins (ephrin-B1 in avian (16) and ephrin-B2 in mammalian embryos (18)), inhibitory activities on neural crest cell migration have been demonstrated in vitro using neural tube or whole trunk explant culture systems. Unexpectedly, however, neither the gene inactivation of semaphorin3A (19) nor of ephrin-B2 (20) and the corresponding Eph receptors on neural crest cells (21,22) resulted in the mutant mice in aberrant migration patterns through the somites. These observations suggested that a concerted action of multiple inhibitory and some attractive cues are required to guide trunk neural crest cells in vivo (1,17,23,24).
Prime candidates for a cooperative partnership with the cell surface contact inhibitors of neural crest motility are extracellular matrix components belonging to the chondroitin sulfate proteoglycans (CSPGs) and PNA-binding glycoproteins (10,11,25,26). Especially, the chondroitin sulfate proteoglycans versican (27) and aggrecan (28) appear to be functionally involved in the inhibition. Both proteoglycans are members of the hyalectan family forming large complexes through interactions with hyaluronan and link proteins (29). They play key roles during development, as the constitutive abrogation of their expression leads in homozygous mice to early intra-uterine (versican) or perinatal death (aggrecan) (30,31).
At least four different isoforms of versican (V0 to V3) exist as a result of alternative splicing of two exons encoding the central glycosaminoglycan carrying domains, glycosaminoglycan (GAG)-␣ and GAG-␤ (32,33). The largest splice variants of versican, V0 and V1, are highly * This work was supported in part by grants from the Swiss National Science Foundation, the Hartmann Mü ller, the Lydia Hochstrasser, and the Velux Foundation (to D. R. Z.), and grants from the Swiss National Science Foundation and the National Center of Competence in Research "Neural Plasticity and Repair" (to L. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  expressed during embryogenesis and are frequently associated with little adhesive, fast-proliferating tissues displaying a high extracellular matrix turnover rate (34,35). Versican V2, a smaller central nervous system-specific isoform is in contrast produced late during neurogenesis (36). It is a potent inhibitor of axonal growth and hence, seems to participate in restricting the structural plasticity of myelinated fiber tracts and impede regeneration in the mature central nervous system (37).
We have previously shown that the expression of versican V0 and V1 is during neural crest cell migration tightly associated with the formation of barrier tissues in the trunk of chicken embryos (35). Versican V0 and V1 are selectively deposited in the posterior half of the sclerotome, transiently present within the dorsolateral subectodermal tissues and to a lesser extent expressed in the perinotochordal tissue where aggrecan appears to be the prominent hyalectan (38,39). This intriguing codistribution of versican V0 and V1 with barrier tissues prompted us to isolate the intact forms of these large versican proteoglycan variants and to explore their potential guidance function in neural crest cell migration during early embryonic trunk development.

MATERIALS AND METHODS
Antibodies-Polyclonal antibodies against recombinant fragments of the GAG-␣ or GAG-␤ domains of bovine, human, and mouse versicans were affinity purified from rabbit antisera previously prepared in our laboratory (32,37,40,41). Monoclonal antibodies CS-56 recognizing intact chondroitin sulfate chains and ⌬Di-6S specific for chondroitin 6-sulfate "stubs" exposed after chondroitinase ABC digestion were obtained from Sigma and Seikagaku, respectively. Polyclonal antibodies against p75 NTR were purchased from Chemicon and a monoclonal antibody specific for Sox10 (42) was a kind gift of Michael Wegner (University of Erlangen, Germany). For immunostaining of fibronectin the monoclonal antibody IST-4 (Sigma) was used.
Isolation of Versican Isoforms-Intact versicans V1, V2, and a mixture of V0 and V1 were isolated from calf aorta, bovine spinal cord, and the spent culture medium of the human glioma cell line U251MG, respectively. Our previously developed procedure for the isolation of brain versican V2 was adapted for this purpose (40). Fig. 1 summarizes sources and steps involved in the protein-chemical purification of the individual isoforms.
Bovine tissues obtained from the local abattoir were homogenized in 6 volumes of ice-cold extraction buffer containing 4 M guanidine hydrochloride (GdnHCl), 50 mM sodium acetate, pH 5.8, 2 mM EDTA, 10 mM N-ethylmaleimide, 1 mM Pefabloc, 25 mM 6-aminocaproic acid, and 5 mM benzamidine and extracted overnight. After filtration through cheesecloth, the extract was cleared by centrifuging at 100,000 ϫ g for 45 min. The supernatant was dialyzed extensively against 50 mM Tris, 10 mM EDTA, pH 7.0. The precipitate formed during dialysis was removed by centrifugation at 27,000 ϫ g for 30 min. Subsequently, ammonium sulfate was added to the supernatant to reach 20% saturation. Proteins were allowed to precipitate overnight at 4°C. Following centrifugation at 27,000 ϫ g for 45 min the ammonium sulfate concentration was raised to 60% saturation. After an additional overnight incubation at 4°C the precipitate was collected and resuspended in urea buffer (6 M urea, 0.25 M NaCl, 50 mM Tris, 10 mM EDTA, pH 6) followed by dialysis against 10 volumes of the same buffer. This crude extract was batch absorbed on Q-Sepharose FF (Amersham Biosciences) overnight and then packed into a column. The bound proteins were first washed with urea buffer containing 0.25 M NaCl and then eluted with a linear NaCl gradient to 1 M. 10-l aliquots of each fraction were tested for the presence of versicans by slot blotting using antibodies against bovine GAG-␣ (V2) or GAG-␤ (V1) epitopes (40). Samples positive on the blot were pooled, dialyzed against 0.5 M NaCl, 20 mM Tris, 10 mM EDTA, pH 8, and batch-absorbed overnight to hyaluronan-Sepharose, which had previously been prepared according to the method of Tengblad (43) by coupling hyaluronan from bovine trachea (Sigma) to EAH-Sepharose 4B (Amersham Biosciences). Separate hyaluronan-Sepharose batches were used for each versican isoform preparation to avoid cross-contamination. The hyaluronan-Sepharose affinity column was subsequently packed and washed with a gradient of 0.5 to 3 M NaCl. Finally, hyalectans were eluted with 4 M guanidinium buffer (4 M GdnHCl, 20 mM Tris, 10 mM EDTA, pH 8). After analysis with slot blotting, the versican-containing fractions were pooled and concentrated to a volume of 500 l using Biomax 100/Ultrafree-15 filters (Millipore).
For the V1 preparation from aorta, an additional gel filtration step on Sepharose CL-4B was required to remove partially degraded versican. For this purpose, the column was equilibrated with 3 bed volumes of the GdnHCl buffer before loading of the sample. The column was run at a slow flow rate of 0.4 ml/min. Every second fraction was tested for the presence of versican V1 on a slot blot. Antibody-reactive samples were pooled, dialyzed against PBS, and concentrated.
For the isolation of a mixture of versicans V0 and V1 from U251MG cell supernatants the proteoglycans were first precipitated from conditioned medium by adding ammonium sulfate to 70% saturation followed by purification with anion exchange and hyaluronan affinity chromatography as described above. Domain-specific antibodies recognizing the human versican homologues were used for the slot blot detection instead.
Protein Electrophoresis and Immunoblotting-Versican samples were separated on 4 -15% PHAST SDS-polyacrylamide gels (Amersham Biosciences) under reducing conditions and stained with either Coomassie Blue (NOVEX colloidal blue staining kit, Invitrogen) or with silver stain (PHAST GEL Silver Kit; Amersham Biosciences). For immunoblotting, chondroitinase ABC-digested samples were resolved on 4 -15% PHAST polyacrylamide gels followed by diffusion transfer onto Immobilon-P membranes (Millipore) at 70°C for 30 min. The blots were blocked with 3% dry milk in PBS for 30 min followed by incubation with the primary antibodies (all diluted 1:1000 in the blocking buffer) overnight. After washing, alkaline phosphatase-conjugated goat anti-rabbit or anti-mouse Ig secondary antibodies (BIOSOURCE International, CA; diluted 1:15000 in PBS) were allowed to bind for 1 h. The color reaction was performed with Western blue substrate solution (Promega).
Preparation of Primary Cultures of Early Neural Crest Stem Cells (eNCSCs) and Embryonic Fibroblasts-eNCSCs cultures were derived from neural tubes of embryonic day 9 (E9) mice according to the protocol of Kléber et al. (44,45). Briefly, embryos were gently squeezed out of the uterus. After cleaning with Hanks' balanced salt solution, the trunk region caudal of the heart extending to the most posterior somite was cut from the rest of the embryo with the help of forceps and tungsten needles. The isolated trunks were pooled and transferred to a digestion solution containing 0.4 units/ml Dispase (Roche Diagnostics) in Hanks' buffer without Ca 2ϩ and Mg 2ϩ . The trunks were slowly triturated and then kept at 4°C for 6 min. This procedure was repeated 3 times using a fresh digestion mixture. Finally the neural tubes were triturated again very gently to completely free them of all other tissues and then transferred to Dulbecco's modified Eagle's medium containing 10% fetal bovine serum to stop the digestion reaction. The explants were subsequently grown on various substrates in SN1 medium (SN medium with 10 ng/ml of bovine fibroblast growth factor) while keeping the cultures in a gas tight modular incubator chamber flushed with 1% O 2 , 6% CO 2 and balanced N 2 (44). Embryonic fibroblasts were prepared from the trunk of E14.5 mouse embryos as described by Talts et al. (46).
Stripe Choice Assays-Stripe choice assays were done as described previously (37) (modified protocol from Vielmetter et al. (47)). 20 ϫ 20-mm glass coverslips were coated in alternating stripes of test and control substrates for migratory eNCSCs. The coating solutions of the test substrate contained variable concentrations of versican isoforms (0 to 100 g/ml) admixed to 100 g/ml human fibronectin, whereas the control lanes were treated with 20 g/ml fibronectin (Roche Diagnostics) alone. To exclude major differences in the coating efficiency of fibronectin upon addition of versican to the test substrate, ratios of immunofluorescence staining intensities of control and substrate lanes were determined using the image software analySIS (Soft Imaging System/Olympus; region of interest: 480 m 2 , n ϭ 10 each) (Fig. 2).
For each assay, three to four neural tubes were placed perpendicular to the stripe pattern, covered with little SN1 culture medium, and first incubated for 45 to 50 min at 37°C in a 5% CO 2 atmosphere allowing the attachment of the explants. Once the tubes adhered to the substrate, the dishes were gently flooded with medium and kept at reduced oxygen levels. For this purpose, the dishes were placed in a gas-tight modular incubator chamber, which was flushed for 3 to 5 min with a gas mixture containing 1% O 2 , 6% CO 2 and balanced N 2 to generate 3 to 6% actual levels of O 2 . eNCSCs were allowed to migrate out of the neural tube for 20 h and were then fixed with 3.7% formaldehyde in PBS for 10 min. They were either directly visualized by phase-contrast microscopy or processed for immunofluorescence staining. Alternatively, embryonic fibroblasts were allowed to emigrate onto the stripe pattern from the surface of small uncoated circular coverslips.
Neural Crest Cell Adhesion Assay-For the adhesion assays, eNCSCs were isolated from E10.5 rat embryos as described (48) and re-plated after 15 h in culture on plastic dishes coated with fibronectin alone (100 g/ml), versican (75 g/ml) alone, or a mixture of versican and fibronectin (75 and 100 g/ml, respectively). The time course and saturation of the attachment was monitored by fixing the cells after 5, 15, 30, or 45 min of incubation. The number of cells that attached after 45 min on fibronectin alone was taken as 100%.
Immunofluorescence Staining of Cells and Tissue Sections-eNCSCs were visualized by double immunofluorescence staining of fixed cultures with polyclonal antibodies against p75 NTR (1:300 dilution; Chemicon) and with a monoclonal antibody against Sox10 (1:3 dilution (42)). For this purpose, cells were first blocked and permeabilized with PBS containing 10% goat serum and 0.3% Triton X-100 for 10 min. Subsequently they were incubated with the primary antibodies for 2 h at room temperature and then labeled for fluorescence detection with Cy3-conjugated goat anti-rabbit IgG and fluorescein isothiocyanate-conjugated anti-mouse IgG secondary antibodies (both 1:200 dilution; Jackson Laboratories). Alternatively, the substrate coats were immunostained in some experiments with polyclonal antibodies against the GAG-␤ domain of versican (dilution: 1:100) and with the monoclonal antibody IST-4 recognizing human fibronectin.
For immunostaining of tissue sections, mouse embryos were formalin fixed, dehydrated, and embedded in paraffin. 2-m sections of the tissue were deparaffinized in xylene and re-hydrated by a dilution series of ethanol in water. Heat unmasking of the antigens was done in 10 mM Tris, 1.7 mM EDTA, 1 mM sodium citrate, pH 7.8, in a steam cooker for 2 min at 100°C. The sections were two times washed in PBS for 5 min and subsequently blocked, first in 0.2% gelatin and 0.5% bovine serum albumin in PBS for 30 min and then in blocking buffer from a M.O.M. kit (mouse-over-mouse, Vector Laboratories). The incubation with the monoclonal antibody against Sox10 (1:6) and the polyclonal antibodies specific for versican GAG-␤ (1:1000) was allowed to proceed overnight at 4°C. All antibodies were diluted in 0.5% bovine serum albumin, 0.2% gelatin, 0.02% NaN 3 in PBS. After washing in PBS, the sections were incubated with Alexa 488 goat anti-rabbit IgG and Alexa 594 goat antimouse IgG secondary antibodies (both Molecular Probes) diluted 1:200 in PBS. Counterstaining was done with Hoechst H33258 bis-benzimide stain (Invitrogen) for 2 min. Sections were finally washed in PBS and mounted in fluorescence mounting medium (Dako). Images were taken with an Olympus BX61 microscope equipped with a F-view II camera using the analySIS 3.2 software (Soft Imaging System). Red, green, and blue fluorescence were stored separately in the corresponding color channels in RGB-format. Brightness and contrast were adjusted with Photoshop 7.0.

Isolation of the Proteoglycan Isoforms of Versican (V0, V1, and V2)-
The preparation of the large isoforms of versican for functional studies has notoriously been difficult, as they are, except for versican V2, predominantly expressed in embryonic tissues yielding only very small quantities of the intact proteoglycans (49). In contrast, adult tissues such as blood vessels contain frequently moderate amounts of immunohistochemically detectable versicans (50). At this stage, however, versican is largely present in fragments resulting from a physiological cleavage process involving ADAMTS proteases and/or other matrix degrading metalloenzymes (51). We therefore sought for new sources of intact versicans in tissues of young animals (calves) and in supernatants of cells in culture. Adapting our previously developed protein-chemical protocol to purify versican V2 from bovine brain (40), we were able to isolate a mixture of versican V0/V1 from the culture supernatant of the glioma cell line U251MG, versican V1 with trace amounts of V0 from calf aorta, and for biochemical comparison, versican V2 from bovine spinal cord (Fig. 1). Practically all the isoforms were intact revealing their characteristic high molecular mass core protein bands on SDS-PAGE only after removal of the chondroitin sulfate side chains ( Fig. 3; V0, Ϸ650 kDa; V1, Ϸ520 kDa; and V2, 400 kDa). Degradation products were nearly absent. Only two minute bands migrating in the size range of the V2 isoform became weakly apparent on immunoblots of the U251MG and aorta preparations (Fig. 3C). Their immunoreactivity with GAG-␤-specific polyclonal antibodies clearly excluded an identity with the V2 variant, which lacks this domain. Except for some low molecular mass components later introduced by the treatment with chondroitinase ABC, no other contaminating proteins were detectable in Coomassie Blue and silver-stained gels. The yields of these preparations were 0.25 mg of versican V0/V1 per liter of culture supernatant, 0.2 mg of versican V1 per 100 g of aortic tissue, and 0.3 mg of versican V2 per 100 g of spinal cord, respectively (referred to the protein content).
Versicans V0/V1 and V1 Specifically Restrict the Migration of Early Neural Crest Stem Cells in Vitro-Having sufficient amounts of intact versicans at hand, we could now study their effect on the migration behavior of eNCSCs in stripe-choice assays. These in vitro assays simulate the metameric expression pattern of versican V0 and V1 within the posterior sclerotome during the active neural crest cell migration period occurring around stage 20 in chick (35) and around E9.5 in mouse embryos (Fig. 4A). Consequently, neural tubes from E9.5 mouse embryos were used as the stem cell source for the in vitro experiments. The explants were cultured on coverslips coated with alternating stripes of the migration-promoting extracellular matrix protein fibronectin and the versican test substrate admixed to fibronectin. Like in vivo, the eNCSCs emigrated from the neural tube as sheets and moved onto the substrate-coated surface. This migration remained uniform in assays, in which the coverslips were exclusively coated with 20 and 100 g/ml fibronectin in a stripe pattern (Fig. 4C). The picture changed, however, dramatically, when eNCSCs were confronted with lanes of fibronectin alternating with stripes treated with a mixture of versican V0 and V1 plus fibronectin (Fig. 4B). In these experiments, the continuous sheets of the migratory neural crest cells divided into separate streams shortly after leaving the neural tube subsequently advancing only on the versican-free surfaces. This selective movement greatly differed from the behavior of embryonic trunk fibroblasts, which did not display a substrate preference. In a comparable stripe choice assay, the fibroblasts migrated uniformly, even at a high versican coating concentration (Fig. 5).
In contrast to the fibroblasts, early embryonic neural crest stem cells avoided versican V0/V1 containing substrates already at coating concentrations of versican as low as 25 g/ml (Fig. 6, B and F), clearly overriding the migration promoting effect of fibronectin (100 g/ml) also present in these lanes. Whereas the inhibition was not yet complete at this low versican V0/V1 level, no crossing of the cells could be observed anymore at higher coating concentrations (Fig. 6, C and G and  D and H). Of note, parallel experiments with intact versican V1 from bovine aorta gave very similar results in this experimental setting (data not shown).
In contrast to the strong effect on migration behavior, the contact with versican substrates appeared to have no influence on cellular differentiation. Double immunofluorescence staining with the markers p75 NTR and Sox10, characteristic of an early neural crest stem cell phenotype (42,48), revealed throughout the versican concentration range tested a rather homogenous population along the entire migration path (Fig. 6, E-H).
Versican Core Glycoprotein Retains the Inhibitory Capacity after Removal of the Chondroitin Sulfate Side Chains-To investigate, whether the inhibitory function of versican V0/V1 originates from the GAG moiety or from the core glycoproteins, we digested our preparations with chondroitinase ABC and tested them again in stripe choice assays (Fig. 7). Prior to these experiments, the efficient removal of the chondroitin sulfate side chains had been confirmed by slot and Western blot analysis with the monoclonal antibodies CS-56 against intact chondroitin sulfate and ⌬Di-6S recognizing the stubs of chondroitin 6-sulfate exposed next to the core protein linker region after GAG cleavage (Fig. 7, A and B). Because intact versican V0/V1 was in the CS-56 slot blot still detected at a 1:100 dilution, but no staining was observed in a 1:2-diluted sample after chondroitinase ABC treatment, we concluded that the digestion was more than 98% complete. The stripe choice experiments demonstrated that the versican core glycoproteins were still able to restrict the migration of eNC- SCs. Similar to the results with the intact proteoglycans, the effect was directly proportional to the increasing concentration of the GAG-free versican core glycoprotein, although some reduction of the inhibitory capacity was observed. At low coating concentrations (25 g/ml) of digested V0/V1, cells migrating out of the neural tube showed only a marginal stripe restriction (Fig. 7C). At concentrations of 50 g/ml (Fig. 7D) and higher, almost all the cells followed the stripe pattern migrating again on areas coated with fibronectin alone.
This set of experiments clearly showed that the capacity to inhibit eNCSC migration resides within the versican core glycoprotein. The chondroitin sulfate side chains may, however, be required to modulate this inhibitory function, because the coating concentration had roughly to be doubled after the GAG removal to achieve a similar effect as with the intact proteoglycan.
Versican Interferes with the Substrate Adhesion of eNCSCs to Fibronectin-To test, whether intact versican interferes with fibronectin-mediated cell adhesion, uniform Sox10 and p75-positive eNCSC populations were re-plated on culture dishes coated with versican V0/V1 alone (75 g/ml), fibronectin alone (100 g/ml), and a mixture of versican and fibronectin (75 and 100 g/ml, respectively). These experiments demonstrated that in relation to the number of cells binding to fibronectin, less then 10% adhered to the surface coated with versican V0/V1 plus fibronectin, whereas on pure versican V0/V1 (75 g/ml) no cell adhesion was observed, even long after the cell attachment had reached saturation in the control dishes (Fig. 8). Hence, these comparative in vitro assays provided strong evidence that versicans inhibit neural crest cell migration by negatively controlling the binding of neural crest cells to adhesion-promoting extracellular matrix substrates such as fibronectin.

DISCUSSION
The findings of our present and previous work (35) provide together clear evidence for a direct functional involvement of versicans V0 and V1 in the inhibition and guidance of neural crest cell migration during embryonic trunk development. These are in particular: 1) the selective, spatiotemporally coordinated expression patterns of versican V0 and V1 in barrier tissues of chicken and mouse embryos impeding the invasion of neural crest stem and progenitor cells; 2) the strict concentration-dependent exclusion of eNCSC movement from areas containing intact versican V0 and V1 in stripe choice experiments; 3) the specific, mainly core glycoprotein-associated inhibitory function of versicans interfering with eNSCS migration even after complete removal of chondroitin sulfate chains; and 4) the strong activity of isolated versicans in suppressing the adhesion of eNCSCs to their physiological substrate fibronectin.
Prerequisite for our functional studies has been the identification of suitable sources for the isolation of the intact versicans. This has in the past been hampered by the fact that the largest versican splice variants V0 and V1 are predominantly expressed during early phases of embryonic development (34,35), where they are subjected to a highly dynamic turnover most likely being controlled by the action of specific ADAMTS proteinases and matrix metalloproteinases (51,52). Consequently, the yields of intact versican are minimal, when embryonic tissues are used as a source (e.g. 30 g of core protein from the limb buds of 750 chick Phase-contrast (A) and immunofluorescence staining (B: versican, red; Hoechst nuclear dye, blue) demonstrates that embryonic mouse fibroblasts, which themselves express versican, show in contrast to eNCSCs no substrate preference in an analogous stripe choice migration assay. Bars, 100 m.

FIGURE 6. The inhibition of eNCSC migration by versican V0/V1 is concentration-dependent.
Neural tubes of E9.5 mouse embryos were placed on an alternating substrate stripe pattern coated with 20 g/ml fibronectin alone and a mixture of increasing concentrations of intact versican V0/V1 (A/E, 0 g/ml; B/F, 25 g/ml; C/G, 50 g/ml; D/H, 100 g/ml) and 100 g/ml fibronectin. Phasecontrast images (A-D) show that the eNCSCs emigrate uniformly from the neural tube, but rapidly divide into separate streams when they encounter versican. Whereas the cells moving at the front completely avoid versican V0/V1 at coating concentrations of 50 and 100 g/ml, some lane crossing can still be observed at 25 g/ml. The migratory cells maintain their multipotent phenotype all along the migration path as demonstrated by double immunofluorescence staining with the early stem cell markers Sox10 (green) and p75 NTR (red) (E-H). Bars, 100 m. embryos (49)). During maturation, the expression of the V0 and V1 isoforms is greatly reduced and the core proteins left in the adult organism are largely fragmented as demonstrated by protein-chemical analysis of tissue extracts and by immunological stainings of neo-epitopes exposed after physiological cleavage with ADAMTS proteinases (51,52). Despite this partial degradation, V0 and V1 fragments stay incorporated in various elastic tissues such as blood vessels (50,53) and skin (41) and can therefore be detected by immunohistochemical techniques. Whereas these fragments may still contribute to the mechanical properties of mature extracellular matrices, it appears likely that the limited proteolysis of versicans V0 and V1 is required for the abrogation of their functions in cell proliferation and migration during embryonic development.
As the proteolytic products of versican V0 and V1 rapidly accumulate postnatally, we have used bovine tissues from very young animals and supernatants from cell cultures to obtain sufficient quantities of highly purified and intact proteoglycans for our functional studies. In consequence, the proportion of degradation products could be minimized in these preparations. Nevertheless, minute amounts of versican fragments migrating on SDS-PAGE around 400 kDa and slightly above could still be detected on immunoblots apart from the intact core glycoproteins. Despite their size similarities with versican V2, they are most probably derived from versican V0 by cleavage within the GAG-␤ domain. This is concluded from the fact that they are, unlike versican V2, reactive with both, GAG-␣ and GAG-␤ domain-specific antibodies. Hence, from our tissue expression studies (32) and protein chemical analysis, we have currently no evidence that versican V2 is expressed outside of the central nervous system as has been indicated in a recent report (54).
Whereas the aorta extract contained predominantly intact versican V1 with only trace amounts of versican V0, a roughly equal proportion of these isoforms could be isolated from the glioma preparation. Both versican isoforms were after chondroitinase ABC digestion strongly reactive with the monoclonal antibody ⌬Di-6S (3B3) used in various immunohistological studies (10,11,25,26). Because all proteoglycan isoforms of versican also bind the peanut agglutinin, 5 it appears that at least parts of the CS-6 epitopes and PNA-binding carbohydrates are directly linked to the core proteins of versican V0 and V1 in tissues forming barriers to neural crest cell migration and axonal growth. The relationship between PNA-binding fragments of versican V0 and V1 and the previously described axon growth inhibitory glycoproteins isolated from the chick somites (55) is currently unknown.
Because primary cell cultures of dissociated embryonic trunks of chicken and mouse express the large versicans, V0 and V1, we have performed most of our functional studies with preparations containing a mixture of these isoforms. Nonetheless, versican V1 alone proved to be similarly effective in analogous experiments (data not shown). In all stripe choice assays, fibronectin was added to the versican coating solutions. This way, we have excluded that versicans not simply act as nonpermissive substrates for migratory neural crest stem cells, but rather function as active inhibitors, which suppress the migration-promoting   A and B, respectively). Different dilutions of digested and undigested versican V0/V1 samples were applied to Immobilon membranes and developed with monoclonal antibodies against the intact chondroitin sulfate side chains (CS-56) and against the stub epitopes (⌬Di-6S) being exposed after digestion. The complete disappearance of CS-56 immunoreactivity after chondroitinase ABC treatment (A) coinciding with a strong increase in ⌬Di-6S-specific staining (A and B) demonstrates the efficient removal of glycosaminoglycan side chains. Phase-contrast images (C and D) reveal that the migration preference of eNCSCs for the control lanes containing exclusively fibronectin is maintained, when test substrate stripes have been coated with at least 50 g/ml chondroitinase ABC-digested versican V0/V1 together with 100 g/ml fibronectin (D). At lower versican core glycoprotein concentrations (25 g/ml), only a marginal stripe pattern can be noted (arrowhead in C). properties of fibronectin. As fibronectin is in the embryonic trunk present in both pathways and barrier tissues (9), the stripe choice assays are closely reflecting the in vivo situation. In comparison to other inhibitory molecules that have previously been tested in a similar experimental set-up (13,15,18), only relatively low versican concentrations had to be applied to observe maximal inhibition. Considering molar relationships, versican V0 and V1 may even be the most potent inhibitors of neural crest cell migration.
The data of our functional study are in line with previous experiments demonstrating that the perturbation of chondroitin sulfate proteoglycan biosynthesis in mouse embryos leads to an abnormal migration behavior of neural crest cells through the posterior sclerotomes (56), which normally express large amounts of chondroitin sulfate proteoglycans and in particular the versican isoforms V0 and V1 (35). They are, furthermore, supported by observations made in splotch mice, which closely correlate the disruption of normal neural crest cell migration and the consequent failure of target tissue colonization with the ectopical expression of versican in the migratory pathways of these Pax3 mutants (57). Nevertheless, a controversy concerning the inhibitory potential of sclerotomal versicans on neural crest cell migration has been raised leaving a similar role for the notochord-associated aggrecan undisputed (54). In this previous paper (54) a long range attractive function of versicans has been postulated, whereas our experiments clearly demonstrate the contact inhibiting effect of the intact versicans V0 and V1. This discrepancy may be explained by the fact that partially degraded versican preparations from adult bovine aorta and chicken trunks were used in this earlier study. Because the specific cleavage of versicans with ADAMTS proteinases may form part of the physiological process to neutralize the inhibitory properties of versicans, these partly fragmented preparations may have been less active.
How the signal initiated through the contact of moving neural crest cells with versican is translated into the inhibition of cellular migration is currently still open. Versicans may either directly activate a versicanspecific receptor on the surface of neural crest cells or act indirectly by sterically interfering with the interaction between migration promoting substrates and their integrin receptors (58,59). The sterical hindrance model is supported by the observation that the complete removal of the chondroitin sulfate side chains leads only to a modest reduction, but not to the abolition of the inhibitory activity. This relatively small effect could be caused by a partial collapse of the core protein in the absence of the highly sulfated glycosaminoglycan side chains, greatly diminishing the hydrodynamic size of the versican molecule and possibly allowing the re-establishment of a few interactions between the cell surface and the permissive substrate. Along this line, it appears plausible that the extent of inhibition is in vivo controlled by the expression of specific splice forms varying in core protein size and number of chondroitin sulfate side chains covalently attached to them.
Despite the fact that no versican-specific transmembrane signaling molecule has been identified as yet, the versican-receptor hypothesis should, however, not be discarded. Only recently, we have shown that the central nervous system-derived versican V2 efficiently blocks axonal growth (37). Schweigreiter et al. (60) could subsequently demonstrate that this inhibition is mediated by a signaling cascade involving two members of the Rho family of small GTPases, RhoA and Rac1. This study has shown that the axonal contact with versican V2 activates RhoA and inactivates Rac1 in cerebellar granule cells. Consequently, the existence of an unknown versican receptor has been postulated that triggers this reaction finally leading to a growth cone collapse. An analogous signaling mechanism could also be responsible for the inhibition of neural crest cell migration by versican V0 and V1. During cellular movement, the activation of Rho is in general associated with retraction of migratory cells, whereas the activation of Rac promotes the formation of membrane protrusions at the leading edge (61). Integrin receptors are key mediators of these responses (62). For instance, the activation of RhoA and its downstream target, ROCK, directly affects the migration of leukocytes by decreasing the affinity of the ␣4␤1 integrin to its extracellular ligands (63). Interestingly, the same fibronectin-binding integrin is known for its central role in the locomotion of neural crest cells (64,65). Hence, a RhoA-mediated signaling process engaged upon contact with versican V0 and V1 could also negatively regulate the integrin-dependent migration of neural crest cells during embryonic development.
The suppression of cell-matrix interactions by versican V0 and V1 may not be sufficient to completely abrogate neural crest cell invasion of barrier tissues in vivo, as some neural crest cells could eventually switch from a collective chain-type movement (66) to a random amoeboid migration pattern (reviewed in Ref. 67). For a few cells, such an erratic locomotion within the caudal versican-containing portion of the scle- FIGURE 9. Inhibitors of neural crest cell migration selectively expressed in barrier tissues during trunk development. Routes of neural crest cell migration are indicated by green arrows, barrier tissues are marked in red, whereas pathways are colored in green. Versicans V0 and V1 carry both CS-6S epitopes and PNA-binding carbohydrates. They have therefore been grouped together in this scheme modified from Ref. 2: F-spondin (12), ephrins (16,18), Sema3A (13), aggrecan (38,68), collagen IX (15), T-cadherin (14), and CS-6S/PNA (10,11). rotome has indeed been observed in chicken trunk explants after disrupting the interaction between the EphB3 receptor on migratory neural crest cells and ephrin-B1 on the surface of posterior sclerotomal cells by addition of soluble ephrin-B1 (1,16). Nonetheless, constitutive inactivation of the corresponding ephrin or Eph receptor genes alone has not led to an aberrant phenotype in regard to trunk neural crest cell migration (20). This indicates that the cell surface-bound ephrins and the extracellular matrix-embedded versicans V0 and V1 may together direct the migration of neural crest stem and progenitor cells through the rostral sclerotome. Thus, future investigations will possibly have to rely on complex mouse models carrying combinations of multiple constitutively and/or conditionally inactivated genes to elucidate, how versicans, ephrins, and the other putative migration inhibitors (summarized in Fig. 9) join in vivo their functions to regulate the highly precise migration patterns of the various neural crest cell subpopulations in a concerted action.