Mesoderm-inducing factors and the control of gastrulation

While descriptions of cell behaviour during vertebrate gastrulation have become exquisitely detailed and accurate (see chapters by Bortier, Ho, Keller, Lawson and Stem), an understanding of the process at the levels of cell and molecular biology is lacking. In this article, we ask to what extent this problem might be approached by the use of mesoderm-inducing factors. In amphibian embryos such as Xenopus laevis, gastrulation is driven by cells of the mesoderm, and the mesoderm itself arises through an inductive interaction in which cells of the vegetal hemisphere of the embryo emit a signal that acts on overlying equatorial cells (Fig. 1; see review by Smith, 1989). In the last five years, several factors that induce mesoderm from presumptive ectoderm, or which modify the pattern of mesodermal differentiation, have been discovered. These include members of the TGFb superfamily such as activins A and B and bone morphogenetic protein 4 (BMP-4), members of the fibroblast growth factor family such as bFGF or eFGF, and members of the Wnt family such as Xwnt-8 (see reviews by Smith, 1989, 1992; Whitman and Melton, 1989; Dawid and Sargent, 1990; and papers by Smith and Harland, 1991; Sokol et al., 1991; Chakrabarti et al., 1992; Köster et al., 1991; Dale et al., 1992; Jones et al., 1992).

The types of mesoderm induced by each factor, and their effects in combination with each other, differ. Thus activin induces a range of mesodermal cell types, from organizer to tail, according to its concentration (Green et al., 1990; Green and Smith, 1990); BMP-4 induces posterior/ventral mesoderm (Kôster et al., 1991; Dale et al., 1992; Jones et al., 1992) and is capable of ‘ventralizing’ the response to activin (Dale et al., 1992; Jones et al., 1992); FGF also induces posterior/ventral mesoderm but has little ventralizing activity when applied in combination with activin (Kimelman and Kirschner, 1987; Slack et al., 1987; Green et al., 1990; Cooke, 1989). Members of the Wnt family do not induce mesoderm from presumptive ectoderm, but animal caps derived from embryos that have received injections of Xwnt-8 RNA respond to bFGF by forming dorsal, rather than ventral, mesoderm (Christian et al., 1992). Injection of Wnt RNA is also capable of inducing complete secondary axes in Xenopus embryos and of rescuing embryos that have been made completely ventral by ultra-violet light irradiation of their vegetal hemispheres before first cleavage (Smith and Harland, 1991; Sokol et al., 1991).

Previously, we have shown that mesoderm-inducing factors cause ectodermal tissue to undergo gastrulation movements (Symes and Smith 1987; Cooke and Smith, 1989; Smith et al., 1990). In this paper, we take advantage of this observation to attempt to answer four questions. How do cells know when to gastrulate? How do cells know what kind of gastrulation movement to undertake? What is the cellular basis of gastrulation? What is the molecular basis of gastrulation? We begin with a brief description of normal gastrulation, based on the work of Keller (1991; and this volume), which highlights the processes that we seek to understand.

Gastrulation in Xenopus converts the radially symmetrical blastula into a three-layered structure with anteroposterior and dorsoventral axes. The first external sign of gastrulation is visible about 10 hours after fertilization, when the blastoporal pigment line forms on the dorsal side of the embryo. The pigment line arises because bottle cells in this region undergo apical constriction and, as this happens, they displace the deep mesoderm associated with them inwards and upwards, thus initiating involution. Although bottle cell formation and involution begin on the dorsal side of the embryo, the movements then spread laterally and ventrally, reaching the ventral side of the embryo about 2 hours later.

Once involution has begun, gastrulating cells exhibit three distinct types of behaviour. The first mesoderm to involute undergoes cell migration, using the fibrillar matrix on the inner surface of the blastocoel roof as a substratum. On the dorsal side of the embryo, these cells eventually contribute to the head mesoderm (see Winklbauer and Selchow, 1992). Cell migration, however, contributes little to the total force required for gastrulation; if the blastocoel roof, the substratum for migration, is removed, gastrulation proceeds relatively normally (Keller et al., 1985). The main driving force for gastrulation derives from later-involuting cells which, on the dorsal side of the embryo, go on to form notochord and somite. These cells undergo ‘convergent extension’, during which the marginal zone constricts to form a smaller ring (convergence) and lengthens along the anteroposterior axis (extension). The greatest degree of convergent extension occurs in this dorsalmost section of the marginal zone; lateral and ventral regions converge to similar extents, but the degree of extension decreases in progressively more ventral regions, such that cells in the ventral midline can be said only to converge (Keller and Danilchik, 1988).

Convergence and extension are driven by cell intercalation (see Keller, 1991). Convergence occurs predominantly through radial intercalation, while extension is the result of both radial and mediolateral intercalation. These processes can be studied in isolated expiants of the marginal zone region, where, in contrast to the ventral marginal zone, the dorsal marginal zone undergoes dramatic elongation. A detailed description of cell intercalation during convergent extension is given in the chapter by Keller.

In what follows, we discuss to what extent the use of mesoderm-inducing factors might bring about an understanding of these different aspects of gastrulation.

The timing of gastrulation in normal Xenopus development is remarkably precise; members of a batch of embryos fertilized at the same time will form a dorsal blastopore lip, after about 10 hours of development, within 20 minutes of each other (see review by Cooke and Smith, 1990). The nature of the timer responsible for this synchrony is unknown. One obvious suggestion, that it is timed by reference to the mid-blastula transition (MBT), is not correct, because if the timing of the MBT is changed, in axolotl, newt, sturgeon or Drosophila embryos, by changing the nuclear to cytoplasmic ratio, gastrulation still starts at the normal time (see Cooke and Smith, 1990; Yasuda and Schubiger, 1992). Other potential timers have also been ruled out; it is not, for example, simply a question of counting numbers of cell cycles, because gastrulation will proceed on time if cell division is inhibited at the late blastula stage (Cooke, 1973). Furthermore, the ‘pseudogastrulation’ movements displayed by some unfertilized amphibian eggs begin at the same time, and take the same time, as the normal movements in their fertilized siblings. The ‘timer’ that controls these movements must be cytoplasmic and independent of transcription, because they will occur in enucleated oocytes (see review by Satoh, 1982), and this is consistent with results in Caenorhabditis elegans, which show that gastrulation movements start at the normal time in the presence of α-amanitin (results of L. Edgar, cited by Yasuda and Schubiger, 1992). Unfortunately, in spite of all this circumstantial, albeit interesting, evidence, there is no idea as to the molecular basis of the gastrulation clock.

Is it important that gastrulation is so accurately timed a process? In normal development, gastrulation begins at dorsal lip of the blastopore and movements spread around to the ventral side. Surprisingly, this aspect of the timing of gastrulation does not seem to be essential. If the dorsoventral timing of gastrulation is reversed, by applying a temperature gradient across the embryo, development is normal and resulting embryos are indistinguishable from controls (Black, 1989). Under conditions where the metabolic rate is uniform throughout the embryo, however, there is evidence that the proper timing of gastrulation is important in setting up the correct spatial pattern of gene expression and differentiation.

Ultraviolet irradiation of the vegetal hemisphere of the fertilized Xenopus egg results in embryos that lack anterior and dorsal structures (see chapter by Gerhart, this volume, and Cooke and Smith, 1987). Such embryos begin gastrulation movements later than controls, with extreme ‘ventralized’ cases beginning gastrulation at the time of appearance of the ventral blastopore lip in synchronous normal embryos. By contrast, embryos treated with LiCl at the 32- -cell stage develop as extreme anterior-dorsal body patterns. Gastrulation movements in these embryos begin on time, but the blastopore lip appears synchronously around the entire marginal zone, rather than spreading from the dorsal to the ventral side. It is difficult to answer the question of whether these abnormal temporal patterns of gastrulation are an early manifestation of the ventralized and dorsalized phenotypes of the embryos or whether they are the cause of them. One experiment, however, is consistent with the suggestion that the total time that cells spend gastrulating influences their anteroposterior pattern of differentiation. Treatment of gastrulating embryos with polysulphonated compounds such as trypan blue or sodium suramin arrest convergent extension movements, and the resulting tadpoles have body axes that are truncated at different anteroposterior levels depending on the time of application of the compound; the earlier the treatment, the more posterior the body axis that develops, whereas if treatment is delayed until the end of gastrulation, the embryos that develop are normal (Gerhart et al., 1989). These results are consistent with the suggestion that the shorter the time that cells spend gastrulating, whether because they start late or because their movements are interrupted by sodium suramin, the more ventral the cell types they form.

What can mesoderm-inducing factors tell us about the timing of gastrulation? Treatment of animal pole regions with activin A causes the tissue to undergo a dramatic elongation, very similar to the elongation of isolated dorsal marginal zone regions (Symes and Smith, 1987; see Keller et al., 1985 for the behaviour of dorsal marginal zones). Additional evidence that the elongation represents gastrulation movements comes from the observation that the rates of elongation of the induced animal cap explants and the dorsal marginal zone regions are similar and, most significantly, that they begin at the same time. The time of onset of the gastrulation-like movements in induced animal caps is independent of the stage at which the cells are exposed to activin (Symes and Smith, 1987), and this rules out another suggestion for the timing of gastrulation, that it occurs a certain time after mesoderm induction.

It is harder to judge the time of onset of gastrulation-like movements in response to FGF, because, consistent with the suggestion that FGF induces ventral-type mesoderm (Green et al., 1990), the gastrulation movements seen in response to this factor are less dramatic. This makes it more difficult to decide when they bègin. An alternative approach to this problem was introduced by Cooke and Smith (1989), who microinjected mesoderm-inducing factors into the blastocoels of host embryos. This causes the cells of the roof of the blastocoel to undergo a transformation in behaviour that resembles the transformation seen in the mesoderm of the marginal zone. By fixing and dissecting embryos at different times after injection, it is relatively simple to judge the time of onset of gastrulation-like movements. When activin was injected into the blastocoel, irrespective of the time of injection (over a 3.5 hour range), or the concentration of factor (over a 250-fold range), this transformation began at the same time, shortly after the onset of gastrulation at the dorsal side of the marginal zone. This result confirms that of Symes and Smith (1987), but also indicates that concentration of inducing factor is also irrelevant to the timing of gastrulation. When FGF is injected into the blastocoel, the time of the transition in cell behaviour is also independent of the time of injection and of the concentration of factor. However, this time is approximately 1.5 hours later than that seen in response to activin, and corresponds more to the time of appearance of the ventral lip of the blastopore (Cooke and Smith, 1989). It is possible, therefore, that at least one aspect of the control of the timing of gastrulation is influenced by the nature of the factor(s) that induce the mesoderm, the tissue that drives gastrulation.

What do these experiments with mesoderm-inducing factors tell us about the timing of gastrulation? Firstly, they shed no light on the nature of the cytoplasmic clock, save to indicate that the clock is present and running even in cells of the animal pole of the embryo, which would not normally need to refer to it. They do, however, indicate that gastrulation is not timed by reference to the time of mesoderm induction nor to the concentration of mesoderm-inducing factor. This latter point is of interest because diferent concentrations of activin do induce different cell types (Green et al., 1990; Green and Smith, 1990) with higher concentrations inducing progressively more dorsoanterior tissue types. If, therefore, the duration of gastrulation does determine anteroposterior position in the embryo, one prediction would be that gastrulation movements in response to low concentrations of activin stop before those seen in response to higher concentrations. This has not been investigated directly, although lower concentrations of activin do cause less elongation than higher ones (unpublished observations). It is not known, however, whether this is due to a slower rate of elongation or to an earlier cessation. Overall, unless clues come from other organisms such as Drosophila, it seems likely that the best route to understanding the nature of the gastrulation ‘clock’ will be to work backwards from the molecular changes that drive the process. These might include, for example, changes in expression of cell adhesion molecules (see below).

At least three distinct types of cellular behaviour drive gastrulation: cell migration, convergence and extension. Different types of behaviour are followed by cells in different regions of the embryo at different times. How do cells ‘know’ which type of behaviour they should undertake? We have studied the responses made by animal pole tissue to different mesoderm-inducing factors, alone and in combination, to discover how, in principle, cell behaviour might be specified. We compare the gastrulation-like responses made by the cells with the types of tissue induced by each factor or factors.

Our experiments have not yet allowed us to study convergence and extension separately, so here we refer only to convergent extension. The cells that undertake the most vigorous convergent extension are those of the dorsal marginal zone, which go on to form head mesoderm and notochord, and which express genes such as goosecoid (Cho et al., 1991). Notochord, and goosecoid expression, are induced from animal pole tissue by activin (Green et., 1990; Cho et al., 1991) and, similarly, activin induces dramatic convergent-extension-like movements from this tissue (Symes and Smith, 1987). As discussed above, these movements also resemble those of convergent extension in that they begin at about the same time that convergent extension begins in the normal embryo. By contrast, FGF does not induce dramatic elongation from animal pole regions, although it does elicit a more subtle change in shape (Slack et al., 1987), which might be interpreted as being due to convergence in the absence of extension. This is, of course, consistent with the proposed role of FGF as an inducer of ventral/posterior mesoderm (Green et al., 1990; Amaya et al., 1991).

One way in which the extent, if not the nature, of these movements might be modified in different regions of the embryo is through different concentrations of inducing factor; lower concentrations of activin, for example, cause less elongation than higher ones (see above). Another possibility is that additional factors modify the putative activin and FGF signals that operate in the embryo. Recently, two such additional signals have been discovered. mRNA for the Xenopus homologue of bone morphogenetic protein 4 (XBMP-4) is expressed at low levels maternally, and the gene is then strongly activated throughout the embryo at the mid-blastula transition (Koster et al., 1991; Dale et al., 1992). Injection of mRNA for XBMP-4 into the fertilized egg of Xenopus causes the resulting embryo to become severely ‘ventralized’, a phenotype that may be due to the abilities of XBMP-4 to induce strong expression of Xhox3 (see Ruiz i Altaba and Melton, 1989), and to override the effect of activin, causing animal caps incubated even in high concentrations of activin to differentiate as ventral cell types (Dale et al., 1992; Jones et al., 1992). This ventralization of the response to activin is also observed at earlier stages, during gastrulation, where the activin-induced elongation of animal pole regions is strongly inhibited by XBMP-4 (Fig. 2). This effect may be important in regulating the extent and duration of convergent extension, but without detailed knowledge of the spatial and temporal distribution of XBMP-4 protein, it is not possible to speculate further.

Another factor that can influence convergent extension is Xwnt-8, a member of the Wnt family of growth factors. Injection of mRNA encoding int-1, Xwnt-8 or wingless protein into the ventral side of Xenopus embryos results in duplication of the embryonic axis, while injection into the dorsal side has little effect (McMahon and Moon, 1989; Smith and Harland, 1991; Sokol et al., 1991; Christian et al., 1991; Chakrabarti et al., 1992; reviewed by Smith, 1992). In this respect, the effect of Xwnt-8 is the opposite of that of XBMP-4, in that it ‘dorsalizes’ ventral mesoderm; consistent with this, if animal caps derived from embryos that have received injections of Xwnt-8 RNA are treated with FGF, they undergo dramatic elongation, as if they had been treated with activin (Christian et al., 1991). Again, the significance of this observation in the absence of knowledge about the spatial distribution of Xwnt-8 protein is not clear, but it does define another route by which gastrulation might be regulated. Finally, as the transcriptional responses to different inducing factors are elucidated, it may become possible to manipulate gastrulation movements by overexpression of intracellular responses to induction rather than through the extracellular signals (see below).

The other type of cell behaviour observed during gastrulation is cel) migration. In the intact Xenopus embryo, this is most easily studied in the earliest-involuting cells, which go on to form head mesoderm. These cells do not undergo convergent extension and may be important in ensuring that exogastrulation does not occur (see above). In vitro, cell spreading may be studied by dissociating prospective mesodermal cells and seeding them onto fibronectin (FN)-coated tissue-culture plates. Mesodermal cells from all regions of the marginal zone spread and migrate on this substratum, as will prospective endodermal cells, but cells derived from the animal pole region do not adhere, and remain as loosely attached spheres (Nakatsuji, 1986; Winklbauer, 1988). Prospective ectodermal cells can, however, be induced to spread and migrate on fibronectin if they are treated with activin (Smith et al., 1990; see Fig. 3). This is a rapid response, and one that can be induced in single cells; it does not, therefore, require a ‘community effect’ (Gurdon, 1988).

Recently, we have gone on to investigate whether other mesoderm-inducing factors also cause animal pole cells to spread and migrate on fibronectin. Our first experiments used bFGF, in response to which factor animal pole cells attach strongly to the FN substratum, but do not spread or migrate. This was slightly surprising, because fibroblast growth factor has been suggested to induce mesoderm of ventral character (Slack et al., 1987; Green et al., 1990), and cells derived from the ventral marginal zone of the early gastrula spread on FN in a manner very similar to that of cells from the dorsal marginal zone (J. E. H., unpublished observations). It remains possible that other factors induce cell spreading in the ventral marginal zone, and we are now investigating this by treating cells simultaneously with FGF and activin, and by observing the behaviour of cells derived from animal pole regions that have received injections of Xwnt-8 or XBMP-4 mRNA and are cultured in the presence, or absence, of activin and FGF. Preliminary results indicate that FGF inhibits activin-induced spreading on FN, and it is possible that this provides a way of regulating cell migration during gastrulation.

Accurate and detailed descriptions of cell behaviour during gastrulation have been made by Keller and colleagues (see chapter by Keller in this volume), and it might appear that studying the gastrulation-like movements seen in response to mesoderm-inducing factors would have little to add to these. As with the other aspects of gastrulation discussed in this paper, however, the advantage of studying gastrulation in response to defined factors is that the situation is simplified: a narrower range of cell behaviours will occur, and the existence of one type of behaviour will not confuse analysis of another.

Keller has suggested that the dramatic extension produced by convergent extension movements can be achieved through relatively short-range cell intercalation. We have investigated this question in activin-induced convergent extension of animal caps by microinjecting a lineage tracer into Xenopus embryos at the two-cell stage such that half the embryo becomes labelled (DeSimone et al., 1991). Such embryos were allowed to develop to the mid-blastula stage, when animal pole regions were dissected and cultured in the presence or absence of activin. Histological analysis of such caps immediately after dissection showed that little mingling of labelled and unlabelled cells had occurred during early development, in agreement with the results of Wetts and Fraser (1989). We therefore went on to examine cell mixing in explants allowed to develop until the equivalent of stage 12. 5, by which time animal caps exposed to activin had undergone considerable elongation, while controls remained spherical.

In both induced and uninduced explants, blastomeres mixed with each other only to a very limited extent. In both. a few individual cells could be found away from the main group, but overall the impression, in agreement with Keller, is that convergent extension does not involve long-range cell migration but a short-range, and perhaps directed, exchange of neighbours. Keller and Hardin (1987) suggested that such exchange of neighbours might occur through ‘jostling’ of adjacent cells, and this idea is supported by the appearance of induced and uninduced cells in the scanning electron microscope. At the equivalent of the early gastrula stage, induced and uninduced explants appear similar, with rounded cells at the surface of the explant that give the impression of being motile. By the late gastrula stage, however, the external surface of uninduced animal caps is very smooth, with the cells forming a flat epithelial-like sheet; the induced explants, by contrast, still have a motile appearance.

We have made little progress in the analysis of the cellular activities involved in spreading and migration, although it seems likely that these will involve major changes in the cytoskeleton and in proteins associated with the cytoskeleton. Rather, we have concentrated on changes in expression of cell adhesion molecules such as the integrins.

To understand the molecular basis of gastrulation, it is necessary to discover the sequence of events between the receipt of a mesoderm-inducing signal such as activin and the onset of gastrulation behaviour. One way of going about this is to follow the events in the order in which they occur. This approach is being followed already in an effort to understand the patterning of the Xenopus mesoderm, and receptors both for FGF and activin have been demonstrated in the early embryo (Gillespie et al., 1989; Musci et al., 1990; Friese1 and Dawid, 1991; Amaya et al., 1991; Kondo et al., 1991; Mathews et al., 1992), progress has been made in understanding second messenger pathways (Maslanski et al., 1992), and several mesoderm-specific genes such as XhoxS (Ruiz i Altaba and Melton, 1989) Xenopus Brachyury (Smith et al., 1991) and goosecoid (Cho et al., 1991) have been found. The next step in this approach is to ask whether overexpression of, for example, Brachyury can lead to mesoderm-like movements in animal pole cells, and such work is in progress (V. T. Cunliffe, J. E. H. and J. C. S., unpublished work).

The other approach is complementary to that outlined above. A very rapid response to mesoderm induction is the acquisition of the ability of cells to spread and migrate on FN. If the molecular events responsible for this change in behaviour can be elucidated, we can in principle work back-wards from them towards the initial inductive event. One class of molecules likely to be involved is the integrin family. Indeed, there is already good evidence that FN-integrin interactions play an important role in amphibian gastrulation. This comes from work on Pleurodeles, which has shown that involution and migration are blocked by injection of anti-fibronectin (FN) antibodies into the blastocoel of the embryo (Boucaut et al., 1984), by intrablastocoelic injection of peptides corresponding to the cell binding site of FN (Boucaut et al., 1985) and by intracellular injection of an antibody targeted to the cytoplasmic domain of integrin βI (Darribère et al., 1990). Intrablastocoelic injection of antibodies raised against the extracellular domain of integrin pi also inhibits gastrulation by blocking formation of the matrix over which the cells migrate. Thus it is likely that the migration element of gastrulation is, at least in Pleurodeles, dependent on the interaction of FN with integrin Pi. Recently, we have investigated whether the same is true in Xenopus, and, if so, whether mesoderm-inducing factors such as activin influence the expression of integrins in Xenopus.

Our initial experiments used an anti-integrin Pi antiserum (Marcantonio and Hynes, 1988) and a fibronectin affinity column to identify FN-binding proteins on the cell surfaces of mid-gastrula stage Xenopus embryos (Howard et al., 1992). This work indicated that the major fibronectin-binding protein on the surface of Xenopus gastrula cells contains integrin pi, and we therefore went on to analyze the expression of this molecule during gastrula and later stages; previous studies had demonstrated the presence of integrin pi mRNA and protein up to and during blastula stages (DeSimone and Hynes, 1988; Smith et al., 1990; Gawantka et al., 1992). Immunoprecipitation from surface-labelled gastrula extracts revealed very little integrin Pi, as previously reported (Krotoski and Bronner-Fraser, 1990; Gawantka et al., 1992), although expression was much higher by neurula stages. This expression pattern was confirmed by immunoprecipitation of integrin pi from metabolically labelled cells, which revealed a 105 β 10³M_rpolypeptide at all stages, representing the precursor form of integrin Pi, while mature species of M_t116-120 β10³were only clearly detectable after neurulation. Immunocytochemical studies performed in our laboratory and elsewhere (Krotoski and Bronner-Fraser, 1990) reflect this pattern of synthesis, with low levels of expression of integrin μ1 - so low as to be undetectable with many antisera - until notochord and somite formation.

The low levels of expression of the mature functional form of integrin Pi cast some doubt on its role in gastrulation in Xenopus. The anti-integrin β ₁ antiserum used for the expression studies was raised against a C-terminal peptide (Marcantonio and Hynes, 1988), but intracellular microinjection of this antibody did not disrupt function (J. E. H., unpublished observations). We therefore raised antisera against the extracellular domain of the mature molecule using XTC cells, which we knew to express integrin β₁ at high levels, as an immunogen. The resulting sera were indeed targeted to the external portion of integrin pi (Fig. 4), and were used to demonstrate that the spreading of newly induced mesodermal cells is dependent on integrin β₁ (Fig. 5; Howard et al., 1992). This is true not only for activin-induced animal cap cells but also for cells excised from the dorsal marginal zone, providing us with further evidence that activin closely mimics the action of the endogenous inducer. Furthermore, as in Pleurodeles, gastrulation is disrupted if this anti-integrin Pi serum is injected into the blastocoel, apparently because the matrix of the blastocoel roof is disrupted (Fig. 6; Howard et al., 1992). These data do not contradict the work of Keller et al. (1985), which shows that gastrulation in Xenopus can occur even if the blastocoel roof is removed; rather, they suggest that an overlying matrix is required only in the very early stages of involution and that the coordinated force of convergent extension is then sufficient to complete the process in the absence of further directional cues. In Pleurodeles, involuting cells migrate more independently, convergent extension is therefore necessarily less significant, and cells require continued directional cues.

These results indicate that integrin β ₁is involved in gastrulation in Xenopus, and to understand gastrulation it will be necessary to understand how the function of integrin β ₁ is regulated. One approach to this question is to use meso-derm-inducing factors such as activin, which, as we describe above, causes animal pole cells to undergo both convergent extension and cell migration, the latter being dependent on integrin β ₁ An understanding of how the function of this molecule is regulated by activin in this context could therefore provide insight into the intracellular events which initiate gastrulation. In our attempts to investigate this, we have, unfortunately, been unable to show any increase in expression of integrin β ₁ in response to activin, either at the level of rate of synthesis (Fig. 7; Smith et al., 1990) or of cell-surface expression (J. E. H., unpublished observation). This suggests that the mechanism underlying this very rapid response is a more subtle modulation of integrin function. This might involve a change in lipid environment (Conforti et al., 1990; Hermanowski-Vosatka et al., 1992), an altered interaction with the cytoskeleton, or a different dimerization partner (Dedhar, 1990). This last modification cannot be studied until α-specific antisera become available in Xenopus. In future work, we intend to address these problems.

Gastrulation is a fundamental problem in developmental biology and, to arrive at an understanding of the process, it will be necessary’ to combine the techniques of experimental embryology, cell biology and molecular biology. In this paper, we have described the use of mesoderm-inducing factors such as activin A in the study of gastrulation. The ability of factors such as these to induce naive cells to undergo gastrulation movements provides a powerful tool for coming to understand what makes cells behave in particular ways at particular times.