Ectodermal fragments from normal frog gastrulae condition substrata to support normal and hybrid mesodermal cell migration in vitro

One of the most interesting problems in studies of gastrulation in amphibian embryos is the mechanism of directed cell locomotion of prospective mesodermal cells along the inner aspect of the roof of the blastocoel (Nakatsuji, 1983). The roof of the blastocoel is a primitive epithelium in amphibian gastrulae. This epithelium is not bounded in the gastrula stage by a basement membrane containing many layers of collagen fibres, proteoglycans and glycoproteins in a dense meshwork. Instead, it is underlain by a gossamer network of extracellular fibrils with a diameter of approximately 0·1 µm. The extracellular fibrils have been observed in several species of urodele gastrulae (Nakatsuji, Gould & Johnson, 1982; Nakatsuji, 1983; Nakatsuji & Johnson, 1983α) and several species of anuran gastrulae (Nakatsuji & Johnson, 19836), including Xenopus laevis and two species of normal Rana embryos (Rana pipiens and R. sylvatica). In an earlier study, we reported an absence of extracellular fibrils in Rana gastrulae (Nakatsuji & Johnson, 19836) but we now know that this report was partially in error. With improvements in fixation methods, we have now shown that extracellular fibrils on the basal surface of the blastocoelic epithelium also occur in R. pipiens and/?, sylvatica embryos. This improvement in fibril fixation has allowed us to make tests of the hypothesis that extracellular fibrils promote cell locomotion in vivo and in vitro.

In the present study, we have made use of large numbers of normal and hybrid embryos by collecting normal embryos or by fertilizing the eggs of R. pipiens with the sperm of several different species of Rana. In these interspecific hybrid embryos, two kinds of developmental events most often follow, if fertilization and cleavage are initiated normally (Moore, 1955). In one, formed by fertilizing the eggs of R. pipiens with the sperm of a closely related but distinct species, R. palustris, gastrulation and subsequent development occur in a completely normal fashion, although at a rate slightly slower than the rate of gastrulation in normal control embryos. In all other gamete combinations, fertilization and cleavage occur normally with a normal rate, but an abrupt developmental arrest occurs at the onset of gastrulation. These arrested embryos fail to progress through the gastrula stages and subsequently die between 1 day and 7 days after arrest. The time of death after arrest at the early gastrula stage is highly variable for different gamete combinations but is an invariant and consistent feature of any given combination. For example, arrested hybrid embryos formed by fertilizing the eggs of R. pipiens with the sperm of R. calesbeiana die 1 day after arrest (Johnson, 1976). In contrast, arrested hybrid embryos formed by fertilizing the eggs of R. pipiens with the sperm of R. sylvatica or R. temporaria die 7 days after arrest and undergo considerable cytodifferentiation before arrest (Johnson, 1971; Johnson & Adelman, 1984).

It is convenient to use shorthand notation to designate different kinds of gamete combinations: normal R. pipiens embryos will be called R. pip. Normal Rana sylvatica embryos will be called R. syl; R. pipiens eggs fertilized with R. palustris sperm will be called pal., R. pipiens eggs fertilized by R. sylvatica sperm will be called syl; R. pipiens eggs fertilized by R. temporaria sperm will be called temp; R. pipiens eggs fertilized by R. catesbeiana sperm will be called cat; R. pipiens eggs fertilized by R. clamitans sperm will be called clam.

We have discovered that extracellular fibrils are absent in blastulae and increase in amount during gastrulation in R. pip, R. syl embryos and pal embryos. Explanted fragments of the roof of the blastocoel of all three gastrulating embryos will deposit extracellular fibrils onto plastic substrata under appropriate conditions in vitro. Substrata conditioned in this manner by fragments from either normal embryos or pal embryos will subsequently support the attachment and locomotion of presumptive mesodermal cells from either homologous normal embryos or from heterologous normal embryos (Table 1). We have also found that extracellular fibrils are generally absent in arrested hybrid blastulae. They either fail to appear at all after arrest or appear in small numbers, depending upon the particular gamete combinations under study. Explanted fragments from arrested hybrid embryos have few if any fibrils and they deposit few fibrils on plastic substrata in vitro. Furthermore, when seeded with presumptive mesodermal cells from either normal or hybrid embryos, substrata ‘conditioned’ by fragments from fibril-deficient arrested hybrid embryos support neither strong attachment nor cell locomotion. These results give further support to the hypothesis that oriented fibrils promote oriented movement of presumptive mesodermal cells by contact guidance.

R. pipiens females and various Rana male species (R. palustris, R. sylvatica, R. temporaria, R. catesbeiana and R. clamitans) were obtained from Nasco (Fort Atkinson, Wisconsin) and C. D. Sullivan (Nashville, Tenn.). R. sylvatica embryos from natural spawnings were collected in the field in Fairfax County, Virginia. They were stored in plastic boxes at 4°C in tap water supplemented with antibiotics and salts (Johnson & Adelman, 1981). Females of R. pipiens were ovulated by pituitary injection (Rugh, 1962). Fertilization was carried out by stripping eggs into sperm suspensions made from macerated testes from various Rana species males in 10% Steinberg’s solution (SS). Fertilization for temp embryos has been described previously (Johnson & Adelman, 1984). Developmental stages of normal embryos were determined according to Shumway (1940). Jelly coats were first removed manually with Dumont no. 5 forceps, and the remaining thin layer of jelly was dissolved by incubating eggs for 10 min in 0·7% sodium thioglycolate in 50% SS (pH8·6), followed by 10 rinses in 10% SS.

The composition of the medium used in this study has been published in detail previously (Nakatsuji & Johnson, 1982). Briefly, the basic salt solution is a modified Stearn’s solution (MSS) (Stearns & Kostellow, 1958) buffered with 5mM-HEPES (Sigma). Mesodermal cells were dissociated in 0·02M-sodium citrate (Feldman, 1955) in Ca²⁺-, Mg²⁺-free MSS. For explants of the ectodermal layer, we used the culture medium used in the previous study (Nakatsuji & Johnson, 1982), with a pH of 8.0, a Ca²⁺ concentration of 110µM, and containing 0·5% bovine serum albumin (Sigma). At this Ca²⁺ concentration, ectodermal fragments adhere to the plastic substratum and remain flat, as in the case of conditioning by Ambystoma gastrula ectoderm layers (Nakatsuji & Johnson, 1983a). The dissociated mesodermal cells were cultured in the same culture medium with a Ca²⁺ concentration of 100 µM.

Procedures for conditioning culture substrata have been described in detail (Nakatsuji & Johnson, 1983a). Rectangular pieces of the ectodermal cell layer were cut from the dorsal side of early gastrulae (stage 11; Shumway, 1940) of normal or hybrid embryos, in MSS, using fine forceps and hair loops. The whole blastocoelic roof with a round shape (instead of the rectangular pieces) was used for some of the conditioning experiments in which the conditioning effect was compared, using normal and hybrid conditioning fragments and cells, without examination of the alignment of the cell trails along the blastopore-animal pole axis.

The ectodermal pieces from normal and hybrid embryos were explanted in plastic tissue-culture dishes (35 mm diameter, Falcon Plastics, no. 3001) with their inner basal surface covered by fibrils (in the case of normal embryos and pal hybrid embryos) resting on the dish surface. Two explants were placed in a dish, and cultured for 4 h at 22–24 °C. For the examination with scanning electron microscopy, the explants were cultured on plastic coverslips for tissue culture (Thermanox tissue coverslips, Miles Laboratories) placed in plastic Petri dishes. The margin and blastopore-animal pole axis, in the case of rectangular pieces of the explant, were marked on the dish or coverslip surface by scratching with the tip of Dumont no. 5 forceps.

Explants were removed by gently flushing them away with a Pasteur pipette, from a direction perpendicular to the blastopore—animal pole axis in the case of rectangular pieces, to avoid possible artificial stress alignment of the fibrils along this axis. The conditioned surface was rinsed with two changes of the culture medium, and seeded with the dissociated mesodermal cells from middle gastrulae (stage 12) of either normal or hybrid embryos using a Pasteur pipette. When the ectoderm and mesoderm from normal embryos are used for the experiment, the cells on the conditioned area start to attach and locomote in 20 min. There was no cell attachment outside the conditioned area. The large size of amphibian embryonic cells allows direct counting of the attached cells using a higher magnification (50×) of the dissecting microscope. Unattached cells are spherical and displace with the slightest agitation of the dish. Attached cells were flattened and bipolar or polygonal with lamellar substratum adhesions at the corners of stretched cells. The attached cells were counted and scored on each conditioned area. Time-lapse 16 mm films (Kodak, Plus-X Reversal) were taken with 2O× or 10× phase-contrast objective lenses at 8-s intervals, controlled by a Nikon cine-autotimer CFMA for 1 h or longer within 1-3 h after the seeding of cells.

The coverslips were fixed for scanning electron microscopy by transferring them into a fixation solution of 2·5% glutaraldehyde in 0·05M-PIPES (Sigma) buffer (pH7·3) supplemented with 5 mM-CaCl₂. The same buffer with Ca²⁺ was used for the post-fixation with 1 % OsO₄, followed by dehydration through an ethanol series, critical-point drying through liquid CO₂, and sputter coating with gold-palladium. The samples were examined with a JEOL JSM35 scanning electron microscope. Whole embryos were fixed and processed in the same manner, except that they were cut open in the prefixation solution with a razor blade.

The method of drawing cell trails from films was described in detail previously (Nakatsuji & Johnson, 1982). Briefly, we projected the film at 4-min intervals and marked the centre of the cell body. For the further analysis of the cell trails, we used an image-analysis system with a microcomputer (Graphics Tablet connected to Apple II Plus Computer), and a computer program that we developed and described previously (Nakatsuji & Johnson, 1983α). It gives the alignment parameter R of each cell trail, as well as its length. The R value is log 2r, where r = ∑Δy/∑Δx and Δy is the vertical component of a vector tangential to a point on a cell trail and Δx is the horizontal component of a vector tangential to a point on a cell trail. A random, unaligned cell trail would give an R value of zero, while alignment along the blastopore-animal pole axis of the ectoderm explant that conditioned the surface, or perpendicular to that axis, would give positive, or negative, values of R, respectively. For example, an R value of +1-0 means that the cell has made twice as much displacement along the blastopore-animal pole axis as perpendicular to that axis. An R value of—1 ·0 means the opposite. An R value of 0 means random alignment or no net translocation. (See Nakatsuji & Johnson, 1983α for details.)

In an earlier publication, we reported that extracellular fibrils were extremely sparse on the inner surface of the roof of the blastocoel in R. pipiens gastrulae. We have subsequently made dramatic improvements in our ability to fix extracellular fibrils in R. pipiens gastrulae and hybrid embryos thanks largely to helpful suggestions received from Drs J. LeBlanc and I. Brick of New York University. Preliminary experiments with fixation conditions revealed that the Ca²⁺ concentration in fixative solutions was critical. Using the PIPES buffer described above with 1 mM-Ca²⁺ added, there is a modest improvement in fibril preservation although few fibrils are observed in most cases. If the Ca²⁺ concentration is increased to 5 mM, there is a dramatic improvement in fibril fixation. Further increases in Ca²⁺ concentration to 10 mM do not improve fibril fixation, as judged by the frequency of fibrils encountered in the scanning electron microscope (SEM). Fibrils are either very sparse or entirely absent in stage 8 and stage 9 (Shumway, 1940) R. pip blastulae (Fig. 1A). Fibrils are also extremely sparse in normal R. syl embryos and in gastrulating interspecific hybrid pal in stage 8 and 9 blastulae. We have examined a number of other hybrid embryos at these stages as well and have found that there are few if any fibrils present prior to the onset of gastrulation in controls or at gastrular arrest in arrested hybrid embryos.

At the onset of gastrulation (stage 10), there is a sudden dramatic increase in the number of extracellular fibrils lining the inner aspect of the roof of the blastocoel in both normal R. pip (Fig. 1 B) and R. syl embryos and in the gastrulation hybrid pal as well. As gastrulation proceeds through stage 11 and 12, there is a further increase in the number of fibrils in both kinds of normal embryos (Fig. 1c) and in the pal hybrid. These changes, however, are not seen in arrested hybrid embryos. Two basic results were obtained in this study. (1) For the temp, syl and clam arrested hybrids there are a few fibrils present at stages 10, 11 and 12 (Fig. 2B-D). It is important to realize here that these arrested hybrid embryos do not undergo gastrulation and consequently never progress beyond stage 10. When we describe an arrested hybrid embryo as stage 11 or stage 12, we mean that the control normal embryos for that cross have reached stage 11 and 12, respectively. (2) The second result we obtained, using cat embryos, was quite different. In these arrested hybrids, essentially no fibrils are formed at any stage after arrest at stage 10 (Fig. 2A). Some typical results from our studies so far are shown in Figs 1 and 2.

We have performed a large number of experiments in which we have conditioned artificial substrata with explanted fragments of the roof of the blastocoel from normal embryos, gastrulating hybrid embryos and arrested hybrid embryos. We have seeded such conditioned substrata with dissociated presumptive mesodermal cells from normal embryos, gastrulating hybrid embryos and arrested hybrid embryos. Three interesting results were obtained from these studies. First, we were able to show that explanted ectodermal fragments from R. pip and R. syl normal embryos promote the attachment and rapid translocation of homologous presumptive mesodermal cells (Fig. 3A) (Table 1). For example, R. pip conditioned substrata support translocation at a rate of 4· 1 µm/min ± 1 ·2µm/min S.D. (n = 84) for R. pip cells. We have also measured heterologous combinations of ectoderm and mesodermal cells. R. pip ectoderm also conditions for R. syl mesodermal cells with extensive cell attachment and cell translocation at a rate of 3.3 µm/min ± 0·8 µm/min S.D. (n = 7). Ambystoma maculatum ectoderm has a modest conditioning effect for R. pip cells, with cell attachment and cell translocation at a rate of 2·3 µm/min ± 0.5 µm/min S.D. (n = 9). A. maculatum ectoderm can also condition for A’, laevis mesodermal cells (Nakatsuji & Johnson, 1983α). Second, we found that ectoderm from R. pip would condition substrata such that even cells from arrested hybrid embryos would show moderate attachment and even substantial rates of translocation. R. pip conditioned substrata support translocation at a rate of 1 ·8 µm/min ± 0·6 µm/min S.D. (n = 53) for cat cells (Fig. 3c) (Table 1) and 0·6µm/min ± 0·2µm/min for temp cells. In other gamete combinations, attachment and translocation was slight (Table 1). Some examples of normal and hybrid cells attached to conditioned substrata are shown in Fig. 3. In all of the cultures that we examined, conditioning effects were limited to the area of the culture covered by the explanted ectodermal fragment. In areas of the culture dish outside the boundaries of the conditioning fragment, there was no cell attachment and no translocation for all combinations of conditioning ectoderm and presumptive mesodermal cells (Fig. 3B, Fig. 9). Third, we found that ectoderm from arrested hybrid embryos had no conditioning effects (Fig. 3D).

We have made a detailed study of cell translocation of normal and hybrid cells from time-lapse films (Fig. 4). We have found that normal cells show substantial rates of translocation (Fig. 4 and Table 1). In some preparations, we explanted rectangular fragments of the roof of the blastocoel where the long axis of the rectangle was parallel to the animal pole-blastopore axis of the gastrula. We then measured the orientation of the cell trail (R value) with respect to the animal pole-blastopore axis. When we conditioned substrata with R. pip ectoderm and seeded them with R. pip mesodermal cells, we found that the cell trails were extensive but only weakly oriented with respect to the animal pole-blastopore axis (R = 0 · 12) (Fig. 5) and not polarized (Fig. 6).

When we conditioned substrata with R. pip ectoderm and used arrested hybrid test cells, we found that cat cells showed moderate translocation that was substantially less than when we used R.pip test cells; and that temp cells showed very little movement at all (Fig. 4 and Table 1).

We found that/?, pip ectodermal fragments deposit fibrils and bits of cellular debris on plastic substrata after a 4-h conditioning period. Some of this debris appears to be the remains of small filopodia or retraction fibres that sheared off when conditioning fragments were washed away by a stream of culture medium expelled from a Pasteur pipette. The deposited fibrils, which formed anastomosing networks, were not as prominent on conditioned substrata as they were in intact fragments of embryos. When R. pip presumptive mesodermal cells were seeded on substrata conditioned by R.pip ectoderm, the test cells were flattened considerably and formed broad fan-like lamellipodia (Figs 7A-B, 8A) and were extensively elongated, assuming a distinct fusiform shape, presumably due to traction forces exerted by lamellipodia. In many instances, long filopodia were observed preceding these lamellipodia (see Trinkaus & Erickson, 1983). Cells from arrested hybrid embryos seeded on substrata conditioned by R. pip explants do not attach as extensively as cells from normal embryos. In the scanning electron microscope, hybrid cells are not nearly as flattened as normal cells (Fig. 7c, D). Instead, they remain more or less rounded up in culture and form smaller numbers of filopodia and associated lamellipodia. Attached hybrid cells are also not as fusiform as cells from normal embryos on conditioned substrata. Perhaps hybrid lamellipodia are not as capable of exerting traction forces or do not adhere as firmly as normal cells. We have examined normal and hybrid cells in cultures in which substrata had been conditioned but cells had been deposited outside the conditioned area. In these areas, both normal and hybrid cells remain rounded up due to a lack of substratum adhesion (Figs 3B and 9).

The present results show that noπnal embryos and gastrulating hybrid embryos have extracellular fibrils in vivo. These fibrils are absent prior to gastrulation and appear in increasing numbers during gastrulation. Furthermore, these fibrils can be deposited in vitro, where they condition substrata so that presumptive mesodermal cells seeded on them can show extensive but unoriented and unpolarized cell movement. These fibrils are present in reduced numbers in some arrested hybrid embryos and are altogether lacking in other arrested hybrid embryos. Explanted ectodermal fragments from arrested hybrid embryos fail to deposit fibrils and to promote cell attachment and translocation. Finally, we have shown that ectodermal fragments from normal embryos can condition culture substrata in such a way as to promote moderate cell attachment, and in one instance even cell translocation of presumptive mesodermal cells taken from an arrested hybrid embryo (cat).

It appears now that extracellular fibrils on the inner aspect of the roof of the blastocoel of amphibian gastrulae are widely distributed among both the urodeles and anurans. We have already reported the existence of such fibrils in three species of urodeles, A. maculatum, A. mexicanum and Cynopspyrrhogaster (Nakatsuji, Gould & Johnson, 1982; Nakatsuji & Johnson, 19836). The fibrils in A. maculatum are significantly oriented along the animal pole-blastopore axis in vivo (Nakatsuji et al. 1982). They are also deposited in an oriented fashion in vitro and can promote oriented cell locomotion along the animal pole-blastopore axis of conditioning fragments. Furthermore, this oriented locomotion shows a slight polarization and cells move preferentially on them toward the animal pole rather than toward the blastopore, under certain conditions (Nakatsuji, 1983; Nakatsuji & Johnson, 1983α). We also reported previously the existence of fibrils in one anuran species, namely X. laevis (Nakatsuji & Johnson, 1983b); and now, since we have improved our fixation procedure, we find fibrils in two other species of anurans, R. pip and R. syl. Fibrils are also present in an interspecific hybrid that undergoes gastrulation (pal). Interspecific arrested hybrid embryos, however, have far fewer fibrils than normal embryos. It is safe to conclude that extracellular fibrils on the inner surface of the roof of the blastocoel in gastrulae are common features of both urodeles and anurans.

There is a striking difference in the number of fibrils in vivo in urodeles and anurans; the former invariably show many more fibrils. The number of fibrils deposited in vitro when urodele conditioning fragments are used is also considerably greater. The fibrils in urodeles are also more oriented, both in vivo and in vitro. The migrating mesodermal cells in urodeles often move as isolated cells or small cell clusters, as they migrate across the inner aspect of the roof of the blastocoel. In contrast, anuran cells form compact masses rather than loosely arranged groups of cells. We suggest that migrating mesodermal cells in both urodeles and anurans use these fibrils as a contact guidance system. Perhaps the cells in urodeles require a more formal fibrillar orientation mechanism because they are loosely arranged and thus require more oriented extracellular fibrils to guide them towards the animal pole. In contrast, perhaps the migrating mesodermal cells in anurans do not require such an organized system of extracellular fibrils because they form a more compact mass of cells.

The significance of our observations for the control of cell migration in amphibian embryos may be appreciated better if one considers the behaviour of an isolated migrating mesodermal cell in a urodele embryo. Once such a cell broke away from the pack of other migrating mesodermal cells, hypothetically it would loose contact inhibition of cell migration as an orienting mechanism. Once this orienting mechanism was lost, the cell would be forced to migrate at random so that the direction back towards the blastopore would be the same as the direction away from the blastopore. If there were a contact guidance system within the embryo, such as the oriented and polarized extracellular fibrils that we have described (Nakatsuji, 1983; Nakatsuji et al. 1982; Nakatsuji & Johnson, 1983a, b), then even an isolated cell would still have orienting cues in its environment and would continue to move directionally, i.e. away from the blastopore towards the animal pole. Such a system would have a greater efficiency than a system where only contact inhibition of locomotion was in effect. In contrast, in anuran embryos, the migrating mesodermal cells have never been observed to have loose cells where they have lost their contacts with neighbouring migrating mesodermal cells. In this kind of a system, the direction back towards the blastopore might be prohibited by other migrating mesodermal cells coming from behind by contact inhibition. Isolated presumptive mesodermal cells from/?.pip gastrulae show strong contact inhibition of lamellipodial activity and cell locomotion when they are studied in isolation in vitro (Johnson, 1976) and, most probably, they also exhibit this behaviour in vivo. Unfortunately, we are unable to make a definitive test of this hypothesis due to the opacity of amphibian embryos and the consequent impossibility of a direct analysis of cell movement in vivo.

The differences between normal and hybrid embryos with respect to the system of extracellular fibrils provides further evidence for the importance of this fibrillar system in promoting cell migration. When extracellular fibrils are present in vivo, either in normal embryos or in interspecific hybrid embryos where gastrulation does in fact occur, mesodermal cell migration is extensive but unoriented. These fibrils are also deposited in vitro and promote cell attachment and translocation. When extracellular fibrils are reduced in number or absent in vivo in arrested hybrid embryos, mesodermal cell migration fails to occur. Explants from arrested hybrid embryos do not deposit fibrils in vitro and there is no promotion of cell attachment and translocation. Finally, the inability of cells from arrested hybrid embryos to attach and translocate in vitro can be partially restored by explants from normal embryos, which deposit fibrils. This represents the first evidence that behavioural abnormalities of hybrid cells can be corrected by factors derived from normal embryos and provides further substantial evidence that these extracellular fibrils promote cell adhesion and cell movement. It seems possible that the mechanism of this effect lies in the ability of extracellular fibrils to serve as an attachment site for filopodia, thereby promoting the formation of lamellipodia, which are attachment and locomotory organelles of many cells in vivo and in vitro (Trinkaus, 1976).

At the present time, we have little information concerning the chemical nature of these fibrillar materials. Closely related experimental studies in amphibia, however, suggest that these fibrils contain fibronectin. Boucaut & Darribere (1983α, ?) have found that gastrulae of Pleurodeles waltlii have an abundance of fibronectin lining the inner aspect of the roof of the blastocoel but much less in the same location at the blastula stage. Our earlier work (Nakatsuji etal. 1982; Nakatsuji & Johnson, 1983α,6) shows that much fibrillar material lines the inner surface of the roof of the blastocoel of urodele embryos but that these fibrils are sparse or absent prior to the onset of gastrulation. Gualandris, Rouge & Duprat (1983) have also shown a network of fine anastomosing fibrils lining the inner aspect of the roof of the blastocoel in P. waltlii embryos that were stained by fluorescent soy bean agglutin, a lectin with specificity for N-acetyl-α-D-galactosamine and galactose. The similarity in the appearance and pattern of these fibrillar networks between P. waltlii and other urodeles (Nakatsuji et al. 1982; Nakatsuji & Johnson, 19836) is striking.

Oriented fibrils of fibronectin presumably would have precisely the properties postulated for the fibrils in this study; namely, promotion of cell attachment and cell locomotion. Heasman et al. (1981) have shown that primordial germ cells fromX. laevis embryos migrate up the dorsal mesentery of the gut along a domain that is rich in fibronectin. Furthermore, they were able to show that explanted dorsal mesentery cells will deposit fibronectin in vitro, and promote elongation and migration of seeded primordial germ cells along lines parallel to the orientation of stretched mesentery cells and associated oriented fibrils of fibronectin. Normal R. pip embryos also show increases in synthesis of extracellular glycoconjugates during gastrulation (Johnson, 1977α-c, 1978), whereas arrested hybrid embryos show defects in extracellular glycoconjugate synthesis. In the future we plan to study fibronectin distribution in normal and hybrid embryos to determine whether the extracellular fibrils described in the present work contain fibronectin, laminin, proteoglycan, or some combination of similar basement membrane components.

We thank Dre Albert K. Harris and J. P. Trinkaus for their helpful suggestions on a draft of this manuscript. This research was supported by NIH grant HD 13419 to K.E.J.

Recentlly, Boucaut et al. (Nature, 307, 364—367, 1984) have prepared antibodies to Ambystoma mexicanum fibronectin. Fluorescent anti-fibronectin antibodies bind to a fibrillar network on the inner aspect of the roof of the biastocoel of A. mexicanum gastrulae. When transplantation experiments were performed so that the inner surface of the roof of the blastocoel projected outwards towards the vitelline membrane, mesodermal cell migration failed to occur over the patch of transplanted tissue. Finally, these authors showed that injection of monovalent Fab* fragments of antifibronectin antibodies prevent gastrulation but not neurulation.