Embryonic cell invasiveness: an in vitro study of chick gastrulation

A number of early embryonic cell populations show characteristic invasive properties in which they undergo epithelial-to-mesenchymal transformation followed by infiltration of a second cell population or tissue space (Sanders, 1989). In this process they may, like neoplastic cells, penetrate a basement membrane in order to gain access to the adjacent tissues. One of the earliest morphogenetic events in amniotes, and one that exemplifies this behaviour, is gastrulation, during which the twolayered embryo is converted to a three-layered structure by means of an epithelial-to-mesenchymal transformation in the area of the primitive streak (Sanders, 1986). In this process, cells of the epithelial epiblast layer ingress through the primitive streak and emerge into the underlying tissue space as fibroblastic and migratory mesoderm cells (Bellairs, 1986). Despite the apparent simplicity of this sytem of a limited number of interacting cell types, and its apparent similarity to the neoplastic invasion process, the mechanisms of cell transformation and basement membrane penetration are not understood.

Among the approaches to these problems are searches for cell surface alterations that precede, accompany or follow the ingression of the cells, and searches for localized enzymic activity that could account for the basement membrane dissolution. The former approach has indicated, by means of monoclonal antibody techniques (Thorpe et al. 1988; Loveless et al. 1990) and lectin binding studies (Griffith and Sanders, unpublished) that identifiable changes in cell surface glycoconjugates do indeed accompany gastrulation. The latter approach has been less satisfactory, although a role for plasminogen activator has been discounted (Sanders and Prasad, 1989a) and preliminary evidence suggests a possible role for hyaluronidase (Stem, 1984).

Because of the localized nature of these changes, and the relatively small numbers of cells involved, investigation of such factors in vivo is difficult. However, the development of an in vitro model for the invasionary characteristics of primitive streak cells (Sanders and Prasad, 1989b) has opened up the possibility of experimental perturbation of the process. This model, using the invasion of a reconstituted basement membrane gel as a paradigm, has indicated that epiblast cells that are about to undergo ingression at the time of explantation retain this ability in the gel, as do fully ingressed mesoderm cells, but epiblast cells remote from the primitive streak lack this ability. The penetration of the gel is accompanied by ultrastructural evidence of some disruption of the matrix texture.

In the present work we have used this culture system to investigate factors influencing the invasionary characteristics of mesoderm cells. We have examined the role of cell surface glycoconjugates by application of inhibitors of glycosylation and oligosaccharide processing to the cultures. In addition, the effects on the invasion of cellbinding peptide fragments from laminin (YIGSR) and fibronectin (RGDS) have been determined. Possible enzymic activities have been investigated by application of inhibitors of various classes of protease. We find evidence that invasion is dependent on normal cell surface N-linked glycosylation, and to a lesser extent on changes in the processing of the N-linked oligosaccharide chains. No evidence was found for the activity of YIGSR or RGDS receptors or for the activity of enzymes from a variety of classes.

Chick embryos at stage 4–5, of Hamburger and Hamilton (1951), were removed from their yolk and rinsed in Tyrode’s saline. The thin endoblast layer was removed from the area pellucida using fine tungsten needles and then mesoderm cells from the region immediately adjacent to the primitive streak were dissected away from the overlying epiblast. Dissection was performed without the aid of prior enzymic digestion.

Plastic culture chambers (Sterilin) were half filled with cold basement membrane ‘Matrigel’ (Collaborative Research Inc.). The Matrigel, culture chambers and pipettes were maintained at 4°C until use, and then placed in a 37°C incubator for about 30 min to induce gelling. Uniformly sized small pieces of dissected tissue (30-50 /on in diameter) were then placed on top of the gel, and the chamber was filled with culture medium consisting of nine parts of medium 199 plus one part of foetal bovine serum with gentamycin. The chamber was then incubated at 37 °C in a moist chamber for about 16 h in order to allow the tissue to attach to the gel. After this time the culture medium was replaced by medium containing the test agent. The chamber was sealed with a coverslip and vaseline and incubated for a further 3 days. Observations were made at daily intervals up to 3 days. The degree of invasion of the gel by the tissue was assessed quantitatively by measuring the diameter of the outgrowth and comparing with the bevaviour of controls run concurrently using normal medium. Outgrowth in control cultures after 2–3 days typically measured about 1.5 mm in diameter, but routine comparisons were made using arbitrary units with a grid overlay in the field of view. Recovery was monitored by sliding off the coverslip and replacing the test medium with control medium.

Invasionary outgrowth was clearly distinguishable from the spreading of cells on the surface of the gel, as described earlier (Sanders and Prasad, 19895). The latter behaviour was not seen with the mesodermal tissue used here, being characteristic of epithelial tissues. The inhibitors of glycosylation and of carbohydrate processing were: tunicamycin, castanospermine, deoxy- mannojirimycin and swainsonine (all from Boehringer Mannheim Ltd). Cell-binding fragments from laminin, YIGSR, and from fibronectin, RGDS, were obtained from Peninsula Laboratories Inc. or Bachem Inc. The following enzyme inhibitors were all obtained from Sigma Chemical Company: aprotinin, p-nitrophenylguanidinobenzoate, e-amino-n-caproic acid (serine protease inhibitors); leupeptin (thiol protease inhibitor); pepsta- tin (carboxyprotease inhibitor); 1,10-phenanthroline (metalloprotease inhibitor). The cathepsin L inhibitor, Z-Phe-Ala-CH₂F₁ was a kind gift from Dr A. Warner (Yagel et al. 1989c). Finally, in an effort to stimulate ‘latent’ procollagenase activity, mersalyl acid (salyrganic acid, Sigma Chemical Company) was used (Sellers et al. 1977). A minimum of three repeat experiments were performed for each agent used.

The general characteristics of mesoderm cell cultures in Matrigel have been described previously (Sanders and Prasad, 1989). In the present experiments, there was no outgrowth of cells during the 16 h incubation period prior to the addition of the agents. Cell outgrowth commenced after 1–2 days in culture as narrow invasive tongues of cells, and by 3–4 days in culture, the outgrowth had formed an extensive network of individual cells that invaded the matrix in three dimensions (Fig. 1). The observations recorded in the tables of results were made after 2 days in the presence of the agent.

The only treatment that consistently resulted in complete inhibition of invasion was tunicamycin at concentrations of 0.5–5.0 μg ml⁻¹ (Fig. 2). Reversal of inhibition and beating of the tissue (see below) were used as evidence for the non-toxic nature of the treatment. By contrast, castanospermine, swainsonine (10–50 μg ml⁻¹; Fig. 3) and deoxymannojirimycin (2mgml^-1) produced inhibition in the range of 25 to 50%.

Addition of neither the laminin cell-binding fragment YIGSR (5×10⁻³M, 10⁻³M), nor the fibronectin cellbinding fragment RGDS (10⁻²M, 10⁻³M), had any effect on the invasion of the matrix by the mesoderm cells.

Because proteases have been reported to be essential for the penetration of basement membranes prior to neoplastic cell invasion (Murphy et al. 1989; Rifkin et al. 1989; Yagel et al. 19896; Liotta, 1990), a number of inhibitors directed at different classes of proteolytic enzymes were introduced into the mesoderm cultures. No effect was seen with inhibitors to serine proteases (aprotinin, p-nitrophenylguanidinobenzoate, ϵ-aminocaproic acid); thiol proteases Oeupeptin); carboxyproteases (pepstatin); or cathepsin L (Z-Phe-Ala-CH₂F). In the case of the metalloprotease inhibitor, 1,10-phenanthroline, results were negative at a concentration of 5×10⁻⁵ M; however, at increasing concentrations up to 10⁻³M, inhibition was noted, but this proved to be irreversible.

The putative stimulator of latent’ collagenase, mersalyl, had no effect at a concentration of 5 ×10⁻⁵M, but caused inhibition of invasion at increasing concentrations up to 2 ×10⁻⁴M. The latter effect was attributed to toxicity, since it also was irreversible.

The non-toxicity of the agents used was assessed by reversal experiments in which the agent was replaced by normal medium after 1 or 2 days followed by examination for cell outgrowth. The concentrations used were based on their documented non-toxicity, and all treatments were reversible, with the exception of tunicamycin at 5 μg ml⁻¹ and mersalyl at 2×10⁻⁴ M. Non-toxicity was also indicated by the ‘beating^’ of the tissue that occurred after 1 day in culture, since the mesoderm that was used frequently included pre-cardiac tissue. This tissue was often observed to be beating despite an inability to invade the gel under the influence of an inhibitor, indicating the general nontoxicity of these treatments. Examination of the cultures suggested that recovery was not restricted solely to the behaviour of a resistant sub-population, but to a generalized re-invasiveness.

The experimental system used here to study cell invasiveness is derived from similar models that have been developed for the assessment of tumour cell invasiveness and which involve the evaluation of cell penetration through a reconstitued basement membrane (Terranova et al. 1986; Welch et al. 1989). The reliability of this system for use with embryonic cells has been shown previously using primitive streak cells from the chick embryo (Sanders and Prasad, 1989b). In that study it was shown that cells from the region of the primitive streak – even those still within the epiblast – were able to invade the matrix, but that epiblast cells remote from the streak were not. Fully ingressed mesoderm cells were found to invade the basement membrane matrix as individual fibroblastlike cells, as illustrated in the present work.

Of the three classes of agent used in the present study, the only one that consistently inhibited the invasiveness of the mesoderm cells was the class that interfered with cell surface glycosylation. Tunicamycin, the inhibitor of the first step in the dolichol pathway, completely suppressed invasion at non-toxic concentrations, while the inhibitors of specific enzymes in the oligosaccharide- trimming pathway (McDowell and Schwartz, 1988) partially did so. The consequent implication that glycosylation is important to mesoderm cell invasiveness is supported by observations indicating that the differentiation of these cells is accompanied by changes in their complement of surface oligosaccharides at the time of passage through the primitive streak (Griffith and Sanders, unpublished; Loveless et al. 1990). Indeed, in view of the failure of the other classes of inhibitor to influence invasiveness reliably, it appears that cell surface modulation accompanying gastrulation may be the primary determinant of ingression and the epithelial-to- mesenchymal transformation.

This conclusion is paralleled to some extent by that reached in similar studies of tumour cells. It seems clear that normal glycosylation and trimming are important for expression of the metastatic phenotype (Humphries et al. 1986; Dennis and Laferte, 1987; Cornil et al. 1990), although it is not so clear that it is specifically the invasionary step of metastasis that is the target. For some tumour cells, tunicamycin, but not the oligosaccharide processing inhibitors, is an effective suppressant of invasion (Mareel et al. 1985; Bruyneel et al. 1990), while the latter are effective in other cases (Yagel et al. 1989a).

The observation that mesodermal invasion of the matrix is unaffected by the presence of the laminin cell-binding peptide, YIGSR, suggests that these cells in vivo are not dependent for their movement on attachment to laminin in the overlying basement membrane. This result is supported by others in which it is shown that in vivo microinjection of YIGSR has no effect on the gastrulation process (Brown and Sanders, unpublished data). Since it is not yet clear whether the epiblast cells prior to ingression bear receptors for laminin, it is not possible to determine if the presence of such receptors is required for basement membrane penetration (as is postulated for tumour cells; Liotta, 1990) or if a loss of laminin receptors follows the epithelial-to-mesenchymal transformation. This would be feasible, however, since the basement membrane contains significant amounts of laminin, to which the epiblast could attach, even at this early stage of development (Mitrani, 1982; Boitier et al. 1989; Zagris and Chung, 1990). However, the apparent presence of laminin among the mesoderm cells themselves (Zagris and Chung, 1990) would seem not to influence the movement of these cells. The result presented here is at variance with that seen at later stages of development, where YIGSR has been shown to inhibit the migration of neural crest (Bilozur and Hay, 1988) and heart mesenchyme cells (Davis et al. 1989) in similar matrices. Inhibition of experimental metastasis of tumour cells, possibly as a result of the retardation of basement membrane penetration, has also been shown to occur after treatment with the YIGSR sequence of the laminin molecule (Iwamoto et al. 1987), although other cell-binding sequences may also be involved in this complex process (Kanemoto et al. 1990).

Fibronectin is believed to be a major influence on the in vivo attachment and movement of the mesodermal cells studied here (Sanders, 1986, 1989). The present results showed that the fibronectin cell-binding sequence RGDS had no effect on the movement of the cells in the basement membrane gel. This is in accordance with observations using neural crest cells in this matrix (Bilozur and Hay, 1988), and is perhaps to be expected, since fibronectin is not considered to be a major component of this matrix (Kleinman et al. 1982). This result does not therefore address the question of the role of cell attachment to fibronectin during or after ingression. However, the finding that laminin contains an RGD sequence that may influence cell attachment (Grant et al. 1989) strengthens the argument outlined above, that laminin is not a substratum for these mesoderm cells after ingression. The parallel situation in tumour cells, for the effect of RGDS on cell movement and penetration through basement membranes and matrices, seems to vary with cell type, being either inhibitory (Gehlsen et al. 1988; Yamada et al. 1990) or without influence (Welch et al. 1989).

The present results provide no evidence for the participation of any localized enzymic activity in association with mesoderm penetration through the basement membrane matrix. The failure of three different serine protease inhibitors to influence the invasion rules out a role for this class of enzyme, and is consistent with previous work showing the absence of plasminogen activator from the primitive streak (Sanders and Prasad, 1989a). The result with phenanthroline, a metalloprotease inhibitor, also seems to rule out the involvement of collagenase in the process, since this agent was ineffective at non-toxic concentrations. This still leaves the possibility of hyaluronidase activity (Stern, 1984), since it is unlikely that any of the inhibitors used would affect the activity of this enzyme. However, it seems improbable that hyaluronidase acting alone could completely degrade the basement membrane in the mid-line, even though hyaluronic acid is a significant constituent of the basement membrane at this time (Sanders, 1979). Hyaluronic acid is not an integral component of the network of type IV collagen that comprises the framework of basement membranes (Sanders, 1989).

The lack of clear evidence for enzymic activity associated with the epithelial-to-mesenchymal transformation of gastrulation is in contrast to the case in some other developmental situations (Brenner et al. 1989; Erickson and Isseroff, 1989), and in many tumour invasions (see, for example: Murphy et al. 1989; Rifkin et al. 1989; Yagel et al. 1989b,c). This is a major point of difference between gastrulation and tumour cell invasiveness, and supports an earlier contention that these two processes are not wholly comparable (Bellairs and van Peteghem, 1984).

If enzymic activity is not involved in the gastrulation process, it is necessary to ask what is responsible for the discontinuity in the basement membrane at the primitive streak that is seen in vivo. It seems most probable that this is a matter of differential synthesis of the extracellular matrix and that cells approaching the primitive streak, undergoing surface glycoprotein changes in preparation for ingression, are not capable of laying down the underlying matrix in the way that more lateral areas of the epiblast are able to do. Whether or not this speculation is reasonable awaits detailed studies on differential expression of the appropriate molecules in precise domains of the epiblast.

I thank Esther Cheung for careful technical assistance and the Medical Research Council of Canada for an operating grant in support of this work.