Structural and compositional divergencies in the extracellular matrix encountered by neural crest cells in the white mutant axolotl embryo

During embryonic development, the pigment pattern of the vertebrate skin is established by migration of prospective pigment cells from their original position within the neural crest to distant subepidermal sites (Le Douarin, 1984; Newgreen and Erickson, 1986; Löfberg el al. 1989a). Tn order to reach these eventual locations, pigment cell precursors migrate dorsolaterally entering a cell-free space between the epidermis and somites. This subepidermal migratory pathway is filled with an intricate extracellular matrix (ECM) which is thought to support migration and participate in the guidance of neural crest cells as they move through the tissue environment (Löfberg et al. 1980; Spieth and Keller, 1984; Newgreen and Erickson, 1986; Martins-Green and Erickson, 1987; Dufour et al. 1987; Perris and Bronner-Fraser, 1989a; Löfberg et al. 1989a). More over, a previous study from our group has shown that a developmental alteration of the ECM surrounding the neural crest is a timing factor for the initiation of neural crest cell migration (Löfberg et al. 1985).

The white mutant axolotl is a naturally occurring variant of the species, which results from the homozygous recessive allele combination d/d. The pheno type of the normal dark embryo is determined by the dominant allele D, and hence homozygous (DD) as well as heterozygous (Dd) animals show the dark, wild-type skin pigmentation (Frost and Malacinski, 1980). In contrast, the white mutant axolotl exhibits a restricted skin pigmentation (Figs 1 and 2), arising from the inability of the embryonic epidermis to support dorso lateral migration of neural crest-derived pigment cell progenitors (Dalton, 1950; Bogomolova and Koroch kin, 1973; Keller et al. 1982). Medioventral migration of neuronal and neural support-cell precursors, however, does occur as revealed by the apparently normal formation of dorsal root ganglia (Borack, 1972), and development of the larvae to long-lived and fertile adults.

Recently, we discovered that the defect of the white axolotl epidermis, which is particularly noticeable in the trunk region, resides in the ECM produced by this tissue (Epperlein et al. 1986; Löfberg et al. 1989b). Surprisingly, however, while the ECM of the subepider-mal space of the white embryo is nonpermissive for neural crest cell movement at developmental stages corresponding to the onset of migration, it acquires motility-promoting capacities at stages corresponding to advanced neural crest cell movement (Löfberg et al. 1989b). Thus, the genetically governed anomaly ap pears to functionally rest upon an asynchronous devel opment of the neural crest cells versus the subepidermal ECM, such that this matrix attains migration-stimu lating capacities when neural crest cells have lost their ability to respond to this environmental signal.

In the present investigation, we have undertaken an ultrastructural, immunohistochemical and biochemical characterization of the ECMs of normal dark and white mutant axolotl embryos prior to and during neural crest cell migration. The results suggest that the regional ECM of these two axolotl variants differs in structure as well as molecular composition, and that this divergent matrix organization may be responsible for the restric ted subepidermal migration in the white axolotl mutant.

Normal dark and white mutant axolotl embryos (Arnbystorna mexicanum) at developmental stages 29 – 31 (Bordzilovskaya and Dettlaff, 1979) were fixed in modified Kamovsky fixative, 1.5% glutaraldehyde and 1.5% paraformaldehyde in 0.1M phosphate buffer pH 7.2, for 5h at room temperature or overnight at 4°C. Fixed embryos were rinsed in 0.2 M cacody late buffer, cut transversely level with the 3rd trunk segment to allow satisfactory penetration of the postfixative, and incubated in cacodylate buffered 1 % OsO* containing 400 ppm ruthenium red, for 4 – 6 h at 4°C. Stained specimens were extensively rinsed, dehydrated in a graded series of ethanol and embedded in Epon according to conventional pro cedures. Semi-thin sections, about 2 – 3 μm, were cut on a LKB ultramicrotome and mounted on methanol-washed slides. In order to enhance the contrast of ruthenium red-precipitated ECM structures, the sections were stained with silver and gold salts according to the method of Löfberg and Ahlfors (1978).

Dark and white mutant embryos were fixed in modified Kamovsky fixative at selected developmental stages (30– 35), with or without ruthenium red, or in periodate-lysine-paraformaldehyde (PLP; McLean and Nakane, 1974) fol lowed by postfixation with ruthenium red (see below), and prepared for scanning (SEM) and transmission (TEM) elec tron microscopy as whole or freeze-fractured specimens according to previously described procedures (Löfberg et al. 1980; 1985).

For fixation of the subepidermal ECM in situ, a piece of the lateral flank epidermis (between 4th and 9th somite) was excised from the embryo following a previously devised microsurgical procedure (Löfberg et al. 1985; Penis and Löfberg, 1986). Ruthenium red-containing Kamovsky fixa tive was then flushed with a Pasteur pipette into the exposed subepidermal region in order to obtain instantaneous fixation of the local ECM. Alternatively, to improve the preservation of carbohydrate moieties of matrix components, the subepi dermal compartment of living embryos was flushed with PLP prior to the ruthenium red-precipitation. PLP-flushed em bryos were kept in the same ice-cold fixative for 2– 6 h, rinsed in 0.1M cacodylate buffer and further fixed overnight or longer in 1.0 % glutaraldehyde in the same buffer at 4°C. The embryos were then rinsed in buffer and postfixed in cacody-late-buffered 1 % OsO₄ containing 0.05 % ruthenium red for 1 h at 4°C, and finally prepared for SEM according to routine procedures. In some specimens, individual somite blocks were dissected out and mounted separately to allow SEM examination of the perineural ECM. Subepidermal and peri neural ECMs of 7 embryos of each genotype (D/- and dd) and stage (30 and 35) were subjected to morphometry by computerized image analysis. In each embryo, 10 areas each of subepidermal and perineural interstitial ECM were chosen at random and photographed (magnification ×7800) in the SEM. The micrographs were then analyzed with a Crystal image processor (Quantel, Ltd). In each series of 10 micro graphs, the projected area and length of the fibrils were measured for 20– 26 individual fibril segments. The ratio between projected fibril area and fibril length was adopted as a parameter indicating the degree of granulation of the fibrils. Since normal distribution could not be assumed, the results are presented as median values with 95% confidence limits. The non-parametric Mann-Whitney U-test was used in pair wise comparisons to establish whether median values signifi cantly differed between groups.

Embryos to be treated with enzyme for characterization of ECM components in ultrathin sections were fixed in modified Karnovsky fixative at 4°C overnight. They were then rinsed in 0.2 M phosphate buffer for 1– 2 h at room temperature and cut transversely to allow penetration of the enzyme. Once rinsed, the embryos were incubated in 0.1M phosphate buffer, pH 5.2, containing 100 TRU ml⁻¹Streptomyces hyaluronidase (Miles; Derby and Pintar, 1978) for 2h at 37°C. Control embryos were incubated in the same buffer but lacking enzyme. During the subsequent rinsing steps, treated and control embryos were transferred from the phosphate buffer to 0.1M cacodylate buffer pH7.4 and postfixed 1% OsO₄/ ruthenium red. Counterstaining was performed using uranyl acetate and lead citrate. Ultrathin sections from specimens treated with enzyme, and control specimens were cut 200-300 pm caudal to the cut edge of the embryos in order to obtain sections from comparably stained and digested trunk areas. Sections were cut at an approximate thickness of 50– 70 nm.

The various antibodies used in this study were obtained as follows: affinity-purified antisera against axolotl, Xenopus and human fibronectins from Drs Jean-Claude Boucaut, Laboratoire de Biologie Experimentale, Universite Rene Descartes, Paris, France, Richard Hynes, Department of Biology, MIT, Cambridge, MA, and Antti Vaheri, Depart ment of Pathology, University of Helsinki, Finland; affinity-purified antisera against collagen types I and 111 isolated from bovine and chick from Drs Daniel Hartmann, Centre de Radioanalyse, Institute Pasteur de Lyon, Lyon, France and Charles Little, Department of Anatomy and Cell Biology, University of Virginia, Charlottesville, VA; affinity-purified antisera recognizing the globular domains of human collagen type IV from Dr Jdrgen Wieslander, Biocarb AB, Lund, Sweden. Cross-reactivity of the anti-collagen antisera for the corresponding antigens in the axolotl was ascertained by Western blot analysis which showed that: the anti-collagen type I antiserum recognized two major bands at 95000, and 145000 M_r, and a doublet at 195000-220000 AJ_r; the anti collagen type Ill antiserum recognized a single band at 105 000 M_T- the anti-collagen type IV antiserum to the NCI domain labelled a major band at 175000 M_T. An antiserum against laminin (purified from EHS mouse tumor) was received from Dr Brigid Hogan, Medical Research Council, Mill Hill Laboratories, London UK. Mouse monoclonal antibodies reactive with the chondroitin-6-sulfate, chondroitin-4-sulfate, unsulfated chondroitin and keratan sulfate moieties of proteo glycans were donated by Dr Bruce Caterson, Department of Biochemistry, West Virginia University, Morgantown, WV.

Normal dark and white mutant embryos at developmental stages 30 and 35 were precooled for 1 h at 4 °C and sub sequently immersed in ice-cold Saint-Marie fixative (Saint-Marie, 1962), for 18h at 4°C. Fixed embryos were rinsed in 0.1M phosphate buffer pH 7.4 and cut transversely at the level of the 3rd trunk segment in order to facilitate penetration of the embedding medium. Embryos were then dehydrated through a graded series of cooled ethanol washes, passed through precooled xylene, brought to room temperature and embedded in paraffin wax. Serial transverse and longitudinal sections were cut on LKB-Historange microtomes at an approximate thickness of 6 pm. Transverse sections were cut simultaneously from dark and white mutant embryos on two different microtomes, ordinarily between the 4th and 8th segment. ‘Dark and white’ sections were mounted adjacent to each other on albuminized slides.

Sections to be immunolabelled were deparaffinized, washed thoroughly in 0.1M phosphate buffer pH 7.4, and either incubated directly with antibodies or treated with different enzymes prior to immunostaining. For staining with antisera directed against collagen types I-IV, sections were incubated with 200i.u.ml⁻¹ of bovine testicular hyaluroni dase (Sigma) in Earle’s balanced salt solution for 1 h at 37°C. After subsequent washing, primary antibodies were applied at dilutions 1:20– 1:40. Incubation with anti-laminin antibodies was preceded by digestion with collagenase type III (Sigma) in Earle′s balanced salt solution, 250i.u.ml⁻′ for Ih at 37°C. The anti-laminin antiserum was applied at 1:40– 1:80 di lutions. The proteoglycan epitopes identified by antibodies directed against the characteristic chondroitin isomers were generated by treating sections with 0.2i.u.ml⁻¹ chondro-itinase ABC (Sigma) in 0.1M bicine buffer (Calbiochem) pH8.0 for 1 h at 37°C (Caterson et al. 1987). Antibodies against the proteoglycan determinants were used at final dilutions of 1:50. The anti-fibronectin antiserum was applied at concentrations of 1:80– 1:160. All incubations with primary antibodies were performed at 4°C overnight. Biotinylated or fluoresceinμhodamine-conjugated secondary antibodies specific for the various immunoglobulin types (DAKO-PATTS, Denmark; EY-Laboratories Inc.; Bio-Yeda; Amer-sham) were subsequently applied at dilutions 1:80-1:100 for 1 h at 37°C. When biotinylated secondary antibodies were employed, visualization of the antibody-antigen binding was accomplished by further incubation of the sections with a streptavidin-fluorescein complex (Amersham), 1:160 for 1 h at room temperature. Stained sections were finally mounted in either a glycerol/glycine-Cl-buffered mixture containing 2mgml⁻¹ l,4,-diazabicyclo(2,2,2)octane (DABCO; Sigma), pH8.0, or in a mounting medium prepared by mixing 2.4g Mowiol 4-88 (Calbiochem), 6.0g glycerol, 2ml 0.2M Tris-HCl buffer and 6ml H₂O to give the final pH8.5. Controls were performed by preincubation with antigens (when available), omission of primary antibodies in the first incubation step, or replacement of primary antibodies with appropriate nonimmune globulins.

Pieces of sterile polycarbonate filter membranes (Nuclepore; pore size 0.04μm), measuring about 0.2×0.4mm were cut from whole filter sheets and used as microcarriers. These were inserted microsurgically into the neural crest migratory path ways of living embryos according to a previously described procedure (Löfberg et al. 1985; Perris and Lbfberg, 1986). After a conditioning period in the embryo, the membrane microcarriers bearing regional ECM that had become adsorbed onto their surfaces in situ were explanted and transferred to a labelling buffer (140 mM NaCl, 5mM KC1, 6 mM NaHCO₃, 1.5mM CaCl₂×2 H₂O; pH7.5).

For one-dimensional gel electrophoresis, ECM-covered microcarriers were centrifuged and resuspended in 10 pl of the same labelling buffer. A volume of 10μl of ³⁵S-labelling reagent (³⁵SLR, Amersham), dissolved in benzene and corre sponding to an activity of 10 μCi, was then added to the matrix samples. The samples were incubated for 30 min with ³⁵SLR, and thereafter another 20μ1 volume of labelling buffer was added, followed after 5 min by 200μ(1 Eagle’s balanced salt solution. The labelled microcarriers were then centrifuged and the ECM materials attached to their surfaces solubilized in a small volume of SDS-containing buffer (0.8 g Tris, pH6.8; 10.0g glycerol; 5.0g SDS; 5.0g mercaptoethanol in 100ml H₂O). Solubilized material, corresponding to matrix material adsorbed in vivo and explanted on 15– 20 microcar riers, was run on 0.1 mm thick 7.5 % polyacrylamide gels at a constant current of 2.0mA for 1.5h. After separation, the gels were fixed for 10 min in 45% ethanol/10% acetic acid, washed in 50% ethanol and dried. Finally, gels were exposed to X-ray films (Hyperfilm B-max; Amersham) at —70°C for a period of 1– 3 weeks.

For two-dimensional polyacrylamide gel electrophoresis, ECM-covered microcarriers were solubilized in lysis buffer (8.5M urea, 5% mercaptoethanol, 2% Servalyte 5– 8), with or without addition of 2% Triton X-100. Non-equilibrium electrofocusing was performed in microtube gels prepared using microcapillaries (0.55 × 60mm). Gels were loaded with 3.0of matrix homogenates and run at a constant current setting of 20gA/gel for about 30min, followed by 2.5h at 220 V, and finally 30 min at 400 V. Matrix samples were then coupled to ³⁵SLR by the following procedure: focusing gels were fixed in 40% methanol/10% acetic acid for 10min, washed twice in distilled H₂O (2 min), then twice in labelling buffer. In parallel, a sample of 10 μl of ³⁵SRL solution in benzene (about 10μCi) per tube gel was pipetted into an Eppendorf reaction vial. The benzene solvent was dried off with a gentle stream of dry nitrogen, the reagent was redissolved in 10 μl of DMSO, and the ³⁵SRL solution was instantly added to the tube gel. The gel had meanwhile been immersed in a droplet (20 ul) of labelling buffer. The reaction was then allowed to progress for Ih at room temperature. Eventually, a volume of 100μI of labelling buffer was added, and after 5 min the gels were fixed and stained with Coomassie Brilliant Blue, R 250 as previously described (Neukirchen et al. 1982; von Boxberg, 1988). Tube gels were then run for the second dimension SDS gel electrophoresis as described above (Neukirchen et al. 1982; von Boxberg, 1988). Completed electrophoresis gels were fixed and analyzed autoradiographi-cally as described for the one-dimensional electrophoresis.

At stages of initial neural crest cell migration, the cell-free region surrounding the premigratory neural crest cord in dark and white embryos was filled with a dense granulo-fibrillar ECM network (Figs 5 and 6). In con trast, the intercellular spaces within the premigratory neural crest cord were conspicuously poor in extracellu lar material (Figs 7 and 8). Migrating neural crest cells protruded their leading filopodia through the fibrillar matrix network and could be observed forming special ized contacts with these structures (Figs 9 and 10). For the structural study of the ECM in the two embryos we used: (1) fixatives lacking cationic dyes to allow optimal examination of oriented collagen fibrils at the ultra structural level (Löfberg et al. 1980); (2) fixatives containing ruthenium red to retract the whole proteo glycan molecules into granule-like structures without causing fragmentation of individual molecules (Hun ziker and Schenk, 1987) or causing clustering of proteo glycans and affiliated molecules (as induced by CPC; Tuckett and Morriss-Kay, 1988), and minimizing masking of aligned collagen fibrils; and (3) PLP fixatives (McLean and Nakane, 1974) in combination with ruthenium red to enhance the retention of proteoglycans during tissue processing.

The general ultrastructural appearance of the inter stitial ECM of dark and white embryos following ruthenium red/PLP-ruthenium red precipitation is illustrated in Figs 11 and 13. At stages of initial neural crest cell migration, the subepidermal and perineural basal laminae were continuous in both dark and white mutant embryos and local association of extracellular material indicated the progressive appearance of base ment membranes-like structures. Typically, collagen-based fibrils oriented in arrays adjacent to the lamina densa (Fig. 12) from which they were separated by amorphous, flocculent material (Sanders, 1979; Mar tins-Green and Erickson, 1987; Inoue and Leblond, 1988). The basement membranes had a thicker appear ance in the more dorsal regions of the embryo, indi cating that development of complex basement mem branes occurred in a dorsoventral succession. Numerous, 3 – 5 nm thick fibrils were observed anchored between cells/basement membranes and the collagen ous interstitial matrix fibrils through branched arms (Figs 9, 11, 14, and 15). These fine fibrils had the morphological appearance of hyaluronate-proteo-glycan complexes precipitated by (PLP)μuthenium red (Scott, 1985; Turley et al. 1985; Hunziker and Schenk, 1987). Consistent with their morphology, these microfi brils were found to be susceptible to Streptomyces hyaluronidase digestion which degraded them and caused an overall reduction in ruthenium red-precipi tated ECM material (see below).

No differences could be disclosed morphologically be tween subepidermal basement membranes of dark and white mutant embryos, whereas interstitial matrices diverged in overall structure. The interstitial matrix fibrils suspended in the subepidermal spaces of both dark and white mutant larvae varied considerably in length and thickness, and appeared irregularly organized. Alignment of subepidermal matrix fibrils in the direction of cell movement was observed in both embryos in connection with neural crest cells migrating individually; probably as a result of traction by the moving cells. Although no overall difference in orien tation of matrix fibrils deposited ahead of the moving neural crest cell front could be observed between dark and white matrices, a striking dissimilarity was detected in the amount of electron-dense granules that decorated individual matrix fibrils (Figs 16 and 17). To quantify this difference and determine whether it was significant, we carried out pairwise statistical examinations of the extent of granulation on interstitial matrix fibrils by employing computer-aided image processing and mor phometry. The quotient of projected fibril area/fibril length of individual fibril segments analyzed by SEM was adopted to denote the degree of granulation. The median values of these quotients for fibrils of the regional matrices of the two embryos are summarized in Fig. 22. In pairwise comparisons (e.g. D/— subepider mal versus dd subepidermal ECM at stage 30 and 35), the morphometric analysis demonstrated that granu lation was significantly lower (P>0.001 according to Mann-Whitney U-test) in the subepidermal ECM of white embryos at both initial and advanced stages of neural crest cell migration.

Generally, granules were present on virtually all fibrils, but were most dense on thick fibrils and scarce or absent from the finest fibrils. These latter fibrils did not exhibit collagenous characters. TEM examination of the granules attached to interstitial matrix fibrils, as well as decorating cell surfaces, showed that they were composed of clustered microspheres presumably corre sponding to single proteoglycan molecules. The density of subepidermal matrix fibrils and their associated granules progressively decreased further down the flanks of both larvae, but remained particularly concen trated in the grooves formed at the intersomitic clefts. An abrupt diminution in the overall amount of inter-stitial ECM occurred level with the endodermal tissue, which corresponds to the unpigmented portion of the larval body.

On the dorsolateral aspect of the neural tube, well prior to neural crest cell migration, arrays of striated, 20 – 25 nm thick fibrils were oriented perpendicular to the neural axis. The alignment of the fibrils coincided with the putative migratory direction of neural crest cells (Löfberg et al. 1980). Moreover, the assembly and orientation of these fibrils, which was most evident in SEM specimens fixed in the absence of ruthenium red, was similar in dark and white mutant embryos and resembled that observed in the chick embryo (Allenspach and Clark, 1988; Newgreen, 1989). The fibrillar ECM suspended between neural tube and somites along the medioventral pathway displayed an identical network organization in embryos of both strains. Moreover, similarly to the subepidermal ECM, matrix fibrils of the perineural region showed a con siderable variation in size and extent of granulation. The degree of granulation on matrix fibrils of the perineural matrix was comparable in dark and white mutant embryos (Figs 18 and 19). In both types of embryos at initial and advanced phases of neural crest cell migration, the degree of granulation of the peri neural matrix fibrils was lower than that of fibrils in the subepidermal region (Fig. 22). By removing selected somites from SEM specimens fixed in the presence of ruthenium red, it was possible to unfold tufts of ECM fibrils that were anchored to both the lateral neural tube surface and the subepidermal basement membrane, and that converged within the intersomitic clefts (Figs 20 and 21). The perineural basal laminae/basement membranes of the two embryos were morphologically indis tinguishable, but the amount of material associated with the laminae was lower than that contiguous with the subepidermal laminae. The perinotochordal regions of both dark and white mutant embryos were filled with a dense and intricate fibrillar meshwork, which pro jected out radially from the notochordal capsule. The fibrils of this network were equivalently decorated with electron-dense granules in both embryos.

At advanced stages of neural crest cell migration, the density of matrix fibrils increased along the migratory pathways of both dark and white mutant embryos. The greatest accretion of matrix fibrils occurred in the subepidermal spaces. In contrast, the amount and size of electron-dense granules decreased considerably in these regions, with the proportionally greatest decrease in dark embryos. The subepidermal basement mem branes progressively increased in thickness and exhibited a continuous thick appearance in both dark and white mutant embryos. The perineural and perino tochordal basement membranes, however, maintained a broadly similar structure to that exhibited during initial neural crest cell migration. Similarly, the degree of granulation of the perineural ECM of the two embryos remained virtually constant (Fig. 22).

To optimize comparison, immunohistochemical stainings with each antibody were carried out simultaneously on adjacent sections from dark and white embryos. The immunohistochemical differences in the distribution of various matrix molecules are illustrated schematically in Fig. 39. Fibronectin, laminin, collagen types I and IV were all abundantly distributed in the ECM of the neural crest migratory pathways of dark and white mutant embryos (Figs 23 – 25, 31, and 34). At initial cell migration, both the subepidermal and medioventral migratory routes were strongly labelled in the two embryos. Labelling for fibronectin and collagen type 1 was pronounced in the interstitial ECM between the epidermis and somites, and the perineural ECM sus pended between somites and neural tube. Immunoreac tivity for these molecules was also observed around the notochordal capsule and its associated fibrillar matrix, and on the lateral side of the sclerotome. Both matrix components were extensively immunolocalized in as sociation with basement membranes and within the intersomitic clefts, whereas only fibronectin was detect able between premigratory neural crest cells. Labelling for fibronectin was equivalently strong in the inter somitic clefts, but was discontinuous on the dorsolateral somite surfaces of white mutant embryos. Immuno staining for collagen type I in white embryos differed from that in dark animals by being discontinuous along the neural tube basement membrane and lateral sclero tome.

Laminin and collagen type IV colocalized along basement membranes and were scarcely detectable in the interstitial matrices (Figs 29 – 31, and 34). The distri bution patterns of these ECM molecules were broadly similar in dark and white mutant embryos, but some local differences could be observed. Immunolabelling for collagen type IV and laminin was more pronounced in white mutant embryos, in the intersomitic clefts and along the subepidermal basement membrane respect ively.

Collagen type III was exclusively detected in the most dorsal portion of the subepidermal compartment of dark embryos (Fig. 26). In contrast, in white mutant embryos, collagen type III distributed uniformly along the entire basement membrane of the epidermis as well as around the premigratory neural crest cord, including the dorsolateral aspect of the neural tube (Fig. 27).

At advanced stages of neural crest cell migration, immunolabelling for fibronectin, and collagen types III and IV was virtually equivalent in the two embryos. In contrast, immunoreactivity for laminin was most accen tuated on the lateral somite surfaces of white embryos, whereas collagen type I labelling was most intense around the neural tube and sclerotome of dark em bryos.

Antibodies recognizing proteoglycans bearing pre dominantly unsulfated chondroitin chains stained pri marily the perinotochordal region of dark embryos at stages corresponding to initial neural crest cell mi gration. In contrast, distribution of these proteoglycans in the white mutant embryo at equivalent developmen tal stages was more widespread (Fig. 28). In white embryos, immunoreactivity for proteoglycans rich in unsulfated chondroitin was sparsely distributed along the basal side of the epidermis and more prominently around the neural tube. The anti-chondroitin antibody also decorated the notochordal capsule and its associ ated fibrillar network. The intersomitic staining was similarly pronounced in the two embryos. At stages of advanced neural crest cell migration, staining for pro teoglycans rich in unsulfated chondroitin chains demar cated the entire neural tube and notochord surfaces of dark embryos, and a discontinuous labelling was also discernible around and within the prospective sclero tome (Figs 37 and 38). In white mutant embryos of comparable stages, unsulfated chondroitin staining was more homogeneously distributed along both migratory pathways of neural crest cells; i. e. the subepidermal pathway was more strongly stained than in dark em bryos.

At developmental stages corresponding to initial neural crest cell migration, the chondroitin-4-sulfate proteoglycan subunit was not immunohistochemically detectable in either normal dark or white mutant embryos. A dense perinotochordal staining for proteo glycans bearing this determinant appeared in both embryos first at later phases of neural crest cell mi gration, with the staining pattern being more irregular in the white mutant embryo.

The immunoreactive pattern of proteoglycans rich in chondroitin-6-sulfate differed in dark and white mutant embryos in a fashion similar to that of proteoglycans with unsulfated chondroitin (Figs 29 and 30). At initial phases of cell movement in dark embryos, chondroitin-6-sulfate-bearing proteoglycans traced discontinuously the basement membrane of the epidermis as well as the perineural and perinotochordal regions, and were sparsely distributed around and within the somites (Fig. 29). Conversely, in white embryos a uniform staining was observed along both migratory routes of neural crest cells, including the intersomitic clefts (Fig. 30). Later in development, immunolabelling in the dark embryo was regionally comparable to that observed in white mutant embryos, with the exception of the dorsolateral aspect of the somites which was more prominently stained in the white mutant embryo. Staining for chondroitin-6-sulfate-rich proteoglycans was accentuated in the subepidermal space and inter-somitically (Fig. 36), particularly in dark embryos, and around the periodic pigment cell aggregates that formed along the premigratory neural crest cord (Epperlein and Löfberg, 1984). In spaces between these aggregates, staining appeared patchy around the few neural crest cells still in their premigratory position.

Keratan sulfate-bearing proteoglycans were solely immunolocalized along the medioventral migratory pathway of the neural crest, and in analogous distri butions in the two embryos. At initial neural crest cell movement, the staining on the lateral surface of the neural tube appeared more homogeneous in white mutant embryo than in the dark. At later developmen tal stages, however, immunostaining for keratan sulfate was identical in both types of embryo.

Utilizing membrane microcarriers, region-specific ECMs were adsorbed in vivo from neural crest mi gratory pathways of dark and white mutant embryos (Löfberg et al. 1985; Perris and Löfberg, 1986). Sub sequent electrophoretic micro-analyses were restricted to ECM isolated from embryos at initial stages of neural crest cell migration (stages 30-31). Although a fully representative repertoire of ECM macromolecules can not be expected to be retained on microcarriers, the technique offers the advantage of isolating regional ECMs synthesized and secreted in the proper location in the embryo. Ten to fifteen ECM-covered microcar riers were found to provide enough material for each electrophoretic analysis. A total of seven one-dimen sional separations of ECM samples analyzed in parallel according to this procedure revealed only a minor compositional difference between the subepidermal and perineural ECMs of the two embryos. This con sisted in a broad (subepidermal) versus narrow band (perineural) at about 55 – 60000 M_t (data not shown). Extraction of the matrix samples with Triton X-100 largely prevented this smearing but in addition elimi nated several bands appearing without Triton X-100 extraction. Extraction with other detergents had similar effects.

Two-dimensional electrophoretic separation (5 cases) of subepidermal and perineural ECMs disclosed several differences. These consisted in the appearance of seven faint spots in gels of the dark subepidermal ECM in the molecular weight region corresponding to 30 – 60 000 M_r, which were absent in the corresponding molecular weight region of gels of the white mutant subepidermal ECM (Fig. 40). Gels from white mutant subepidermal ECM, on the other hand, showed a distinct spot at about 40000 not contained by the parallel gel of dark subepidermal ECM. In addition, two prominent spots appeared exclusively in the white subepidermal ECM in the 130000 region.

The protein patterns obtained after parallel, two dimensional separation of perineural ECMs of dark and white mutant embryos were slightly different from each other and clearly distinguishable from the corresponding ones of the subepidermal ECMs (Fig. 40). The major differences occurred in the molecular weight region corresponding to 40– 70000. A conspicuous difference consisted in the appearance of a eight-spot row, which localized at about 50-55 000 in the white mutant perineural ECM, and which was not present in the parallel gels of dark perineural ECM. Two single spots were detectable at about 90000 and 120000 in the white perineural ECM, but not in the matching matrix of the normal dark embryo.

Several previous studies have suggested that the topo graphical organization of ECM substructures may de termine the permissiveness of the matrix as a substrate for neural crest cell migration (Ebendal, 1977; Löfberg et al. 1980; Allenspach and Clark, 1988; Brauer and Markwald, 1988; Newgreen, 1989). On the basis of this suggestion it could be assumed that the inhibited subepidermal neural crest cell migration in embryos of the white axolotl mutant could be attributed to a transiently defective topological arrangement of its subepidermal ECM. To some extent our present ultra structural observations contradict this assumption by indicating that the matrix along the migratory pathways of neural crest cells of dark and white mutant embryos exhibits an indistinguishable configurational arrange ment. These observations led us to the alternative hypothesis that the defect in the subepidermal ECM of the white mutant embryo was more complex than simply an altered assembly of a constitutionally normal matrix.

We have previously correlated an increase in number of ECM fibrils surrounding the premigratory neural crest with initiation of cell movement (Löfberg et al. 1980; 1985). However, accumulation of extracellular material does not always imply a more favorable environment for cell migration since, for example, abundance of ECM in other embryonic areas seems to obstruct neural crest cell movement (Newgreen et al. 1986; Payette et al. 1988; Perris et al. 1990). In accord ance with these observations is our finding that the dense perinotochordal ECM of both dark and white mutant axolotl embryos is avoided by the migrating neural crest cells. Local matrix density can probably influence the arrest of migrating neural crest cells as indicated by the occurrence of tightly packed fibril bundles on the lateral neural tube surface in the salamander Hynobius, which seem to correspond to sites of coalescence of the dorsal root ganglia (Hirano and Shirai, 1986). The present ultrastructural studies and recent immunohistochemical findings in the chick (Perris et al. 19896) corroborate these observations and suggest that these fibrils attached to the neural tube at the level of the intersomitic septa are collagenous and are linked by hyaluronan-proteoglycans complexes.

The immunohistochemical data obtained in this study indicate that the distribution of fibronectin, laminin, collagen types I and IV, and proteoglycans bearing predominantly chondroitin-4-sulfate and keratan sul fate side chains differs marginally in the ECM of dark and white mutant embryos. Some of the minor immu nohistochemical differences in the distribution pattern of fibronectin and laminin could readily be due to differential masking effects. On the other hand, masking of immunogenic epitopes could reflect specific complexing and comformations of these molecules that prevent the occurrence of proper cell-matrix interactions during migration. In fact, there is in vivo evidence that neural crest cells require unrestricted access to extracellular fibronectin and to laminin complexed with heparan sulfate proteoglycan in order to accomplish their normal migration through the embryo (Boucaut et al. 1984; Bronner-Fraser, 1986; Bronner-Fraser and Lallier, 1988).

Although fibril density/arrangement was indis tinguishable, a striking difference between the subepi dermal ECMs of the two embryos was manifested by a significantly higher accumulation of electron-dense granules on the collagen fibrils of normal dark embryos. A similar difference in size and density of spherical elements associated with interstitial matrix fibrils has been described elsewhere (Spieth and Keller, 1984). In this previous investigation, however, the detergent cetylpyrinidium chloride was included in the fixative, and both cell membrane components (Tuckett and Morriss-Kay, 1988) and interstitial matrix structures were collapsed into heterogeneous spherical aggre gates. These aggregates are presumed to embody the hyaluronan microfibrils, which we suggest here to function structurally as multiple linkages between cells/ interstitial collagen fibrils and the basement mem branes, and to comprise clusters of proteoglycans and affiliated molecules. Thus, this fixation protocol is particularly suitable for retention of hyaluronan in tissues (Derby and Pintar, 1978; Perris et al. 1990) but is inadequate when examining the detailed structure of the ECM (Hunziker and Schenk, 1987).

Ruthenium red-containing and PLP fixatives can be used to stabilize matrix complexes including hyaluro nan, proteoglycan-like molecules and highly glycosyl ated proteins during tissue processing. After such structural preservation of the ECM, proteoglycan monomers retract to form electron-dense granules, the size of which largely corresponds to the size of the core protein and/or the glycosaminoglycan/protein ratio (Scott, 1985; Turley et al. 1985; Cidadao & David-Ferreira, 1986; Hunziker and Schenk, 1987). Ruthenium red and PLP may also bind to a variety of negatively charged molecules by nonspecific ionic inter action. The more pronounced distribution of electrondense granules in the subepidermal ECM of the dark embryo could indicate that larger numbers of proteo glycan-like molecules are deposited in the dark subepi dermal ECM. Alternatively, the same numbers of proteoglycan molecules could be present in the two embryos, but the relative size and protein/ glycosaminoglycan ratio or the structure of the glycosaminoglycan moieties could differ. Furthermore, the difference in granulation could also reflect a discrepancy in the relative proportion of hyaluronan and proteoglycan-like molecules able to interact with hya luronan, as well as a differential occurrence of link proteins causing destabilization of the complexes stained by the cationic dyes.

Our biochemical data suggest that several uncharac terized, highly glycosylated proteins, some of which are probably contained in the ruthenium red/PLP-precipitated granules, occur differentially distributed in dark and white subepidermal matrices. In addition, immunohistochemistry revealed that proteoglycan sub sets bearing predominantly low sulfated and/or chon-droitin-6-sulfate side chains were more abundantly distributed in the subepidermal ECM of the white mutant embryo. The implication of the predominant distribution of these proteoglycan types in the white mutant embryo is highlighted by parallel studies on the effects of these proteoglycans on neural crest cell migration. For instance, chondroitin sulfate proteogly cans derived from cartilage are nonpermissive sub strates for chick neural crest cell-substrate adhesion (Newgreen, 1982), quail neural crest cell migration (Tan et al. 1987; Perris et al. in preparation), and axolotl neural crest cells migration in vitro (Perris and Johans son, 1987). In addition, this class of proteoglycans has been found to inhibit migration of neural crest cells over fibronectin, vitronectin, laminin and substrates of vari ous types of collagens by a mechanism mediated by cell surface hyaluronan (Perris and Johansson, 1987; 1990). Finally, studies in Taricha torosa embryos have pointed out the importance of chondroitin sulfate proteoglycans in pigment cell patterning (Tucker, 1986; Tucker and Erickson, 1986).

Collagens are generally weak supporters of axolotl neural crest cell migration in vitro, and in particular collagen type III has been found to be poorly permissive for neural crest cell movement (Perris and Johansson, 1990; Perris et al. in preparation). In the light of these experimental data, it could be assumed that the exten sive incorporation of collagen type III into interstitial fibrils of the white mutant embryo’s ECM could be a relevant factor for the restriction of neural crest cell migration.

That the structural/compositional anomaly of the white ECM is local for the subepidermal space was further evidenced morphologically by the fact that the ECM deposited along the medioventral pathways (where neural crest cell migration appears normal) showed equivalent morphology in the two embryos. Interestingly, however, the perineural ECM of both embryos was less granulated than the subepidermal ECM of the white mutant embryo and the dark subepi dermal ECM showed the overall highest granulation. Both immunohistochemistry and gel electrophoretic analysis of the perineural ECM of the two embryos disclosed additional molecular differences. The signifi cance of the dissimilarities in structural and molecular composition of the equivalently migration-permissive subepidermal and perineural ECMs of dark embryos, and differences between the perineural ECMs of the dark and white mutant embryos remain unclear.

Several spatiotemporal changes in morphology and composition of the ECM along the neural crest mi gratory pathways of the two embryos seem to occur during the course of migration. For instance, in the subepidermal ECM, there is an obvious increase in fibril density dorso-ventrally, whereas the overall extent of granulation decreases. In contrast, the perineural ECM maintains broadly the same spatial organization in both embryos. Intriguingly, the perineural ECMs, at both initial and advanced stages of neural crest cell migration, and the subepidermal ECMs at stages of advanced migration markedly deviate ultrastructurally from that of the dark embryo at stages of initial migration, but possess an ‘equivalent’ ability to trigger the onset of neural crest cell migration (Löfberg et al. 1989b). The differential distribution of characterized matrix components, including laminin, collagen types I/III, and proteoglycans bearing predominantly unsul fated and chondroitin-6-sulfate side chains adds to the series of changes that occur during the course of neural crest cell migration, and that preserve compositional differences in regional matrices of the two embryos, which are independent of their ability to support neural crest cell migration.

Finally, we conclude that the inability of the white subepidermal ECM to function as a substrate for neural crest cell migration (Spieth and Keller, 1984; Löfberg et al. 1989a,b) cannot solely be attributable to an anomaly in spatial organization since qualitative as well as quantitative differences in its structural/molecular composition are expressed at stages of initial neural crest cell migration. The divergent composition of the white subepidermal ECM appears to involve differential distribution of characterized matrix components and differential occurrence of a variety of uncharacter ized molecules. We suggest that these differences may cooperatively determine its assembly and the proper modelling of a migration-permissive environment.

We wish to express our gratitude to Drs Jean-Claude Boucaut, Richard Hynes, Antti Vaheri, Daniel Hartmann, Charles Little, Jôrgen Wieslander, Brigid Hogan and Bruce Caterson for the generous supply of antibodies. We thank Vibeke Nilsson, Lars-Erik Jônsson and Henrik Olsson for technical assistance, and Dr Marianne Bronner-Fraser for critical reading of the manuscript and financial support during the final phases of the work. The study was sponsored by the Swedish Natural Science Research Council (grant B-BU 3810 104).