The role of fibroblast growth factor in early Xenopus development

Work in experimental embryology has given us a fairly detailed picture of the processes of regional specification occurring in the Xenopus embryo prior to gastrulation. These processes are collectively called ‘mesoderm induction’ because they lead to the formation of a ring of mesodermal tissue around the equator of the blastula (Nieuwkoop, 1969; Dale et al. 1985; Gurdon et al. 1985; Jones and Woodland, 1987). This knowledge has made it possible to ask meaningful biochemical questions about the nature of the signals and the responses and about how they can lead to the formation of a spatial pattern of specified regions in two or three dimensions.

Briefly, we believe that the egg is divided into three cytoplasmic zones by the onset of the first cleavage: animal, ventrovegetal and dorsovegetal. The animal hemisphere will form epidermis in the absence of inductive signals, but also has the competence to form mesodermal and probably endodermal tissues in response to such signals. The vegetal hemisphere consists of a large ‘ventral inducing’ zone and a small ‘dorsal inducing’ zone comprising less than 90° of latitude around the dorsal meridian (Dale and Slack, 1987b). During the blastula stages, these two regions emit signals which induce, respectively, an extended region of ventral mesoderm around most of the equator, and a small organizer region on the dorsal side. The signals are quite short range, their influence extending only a few cell diameters (Gurdon, 1989), but, because of simultaneous migration of cells down into the equatorial zone, about 40% of the animal hemisphere eventually becomes recruited into the mesoderm (Dale and Slack, 1987a). These signals are the first two of the ‘three-signal model’ which our group has advanced to explain mesodermal patterning, the third being a dor-salization of the mesoderm as a function of distance from the organizer (Slack and Forman, 1980; Smith and Slack, 1983; Dale and Slack, 1987 b). This model may need to be revised as new data come in but at present we believe that it still provides the best unified account of the known facts.

This understanding naturally leads us to ask three questions: (1) What is the molecular nature of the inducing substances? (2) Are dorsal and ventral signals qualitatively or quantitatively different? (3) What is the molecular nature of the competence of the animal hemisphere cells? Two critically important clues were provided by recent experiments on signal transmission. Grunz and Tacke (1986) showed that the signals could pass through a nucleopore filter in the absense of cell processes, and Warner and Gurdon (1987) showed that the signals could pass from vegetal to animal cells even when gap junction communication had been blocked. These biological experiments greatly narrowed the possible range of mechanisms and firmly pointed towards signals that consisted of secreted extracellular substances. In this paper, we describe our recent work on the role of fibroblast growth factor in mesoderm induction. Work on TGFβ-like factors is described in the accompanying paper by Smith and his colleagues.

Although sources of mesoderm-inducing factors (MIFs) were discovered many years ago, they tended to excite little interest. This was for three reasons: they came from heterologous sources; they were assayed as grafted pellets, a method that precludes quantitative biochemistry; and most were very crude extracts. The best characterized was the ‘vegetalizing factor’ of Tiedemann (1982) isolated from late chick embryos, but even this did not inspire confidence in the wider scientific community. We started work on the subject in 1984, following our reinvestigation of the basic mesoderm induction phenomenon, and commenced by establishing an assay procedure for MIFs which was quantitative and which worked in solution. Briefly, this consists of treating animal pole explants with serial dilutions of the test substance and defining the minimum concentration required to provoke an induction as 1 unit ml⁻¹. The full procedure is described in Godsave et al. (1988; see also Cooke et al. 1987). We then attempted to extend Tiedemann’s work on the chick embryo factor using our improved assay, but, following the report by Smith (1987) of inducing activity secreted by a Xenopus cell line, we turned our attention to an investigation of known growth factors. In our initial screen, we tested a wide range of factors and found only three that were active. These were basic fibroblast growth factor (bFGF), embryonal carcinoma derived growth factor (ECDGF) and acidic fibroblast growth factor (aFGF), all of which belonged to a small group of heparin-binding growth factors (Slack et al. 1987). More recently, we have examined some of the FGF-like oncogenes that have recently been discovered (Paterno et al. 1989). We have done this by in vitro transcription of cDNAs from plasmids containing SP6/T7 bacteriophage promoters followed by translation in a rabbit reticulocyte lysate. The lysate can then be assayed directly by treating ectoderm explants with a series of dilutions, and the specific activity determined by measurement of the concentration of the translated protein. So far, we have examined kFGF, which is the product of the human ks and hst oncogenes (Delli-Bovi et al. 1987; Taira et al. 1987), and INT-2, the product of the murine int-2 oncogene (R. Smith et al. 1988). Both are active as mesoderm-inducing factors. The specific activity of the kFGF is very similar to that of the a and bFGF, while the specific activity of INT-2 is very much lower. Considering the factors as a group, there is a good correlation between their mesoderm-inducing activity and their mitogenic activity when tested on mammalian fibroblasts. This suggests that similar signal transduction machinery is being used for the two processes. It should be emphasized that MIFs do not have any mitogenic effect on Xenopus blastula ectoderm cells, which are already cleaving every 30 min in the absence of growth factors and are probably incapable of further stimulation.

Meanwhile, work in other laboratories has shown that some factors belonging to the TGF β family are also active. These are TGF β -2 (Rosa et al. 1988) and the XTC-MIF of Smith (Smith, 1987; J. C. Smith et al. 1988) and so at the time of writing we have a total of seven active factors.

Obviously the minimum requirement for identification of an endogenous morphogen is that the substance should be present in the embryo at the developmental stage when the relevant events are happening, and in amounts that are capable of exhibiting the observed degree of biological activity. We have approached this problem directly by asking whether a MIF can be obtained from the Xenopus blastula, and which of the seven or more candidates it is. Our results show that it is possible to purify a MIF from Xenopus blastulae using heparin-affinity chromatography and that it consists of two proteins of M_r 19 and 14 ×10³ which react with antibodies against bFGF (Slack and Isaacs, 1989). The quantity in blastulae is about 10 ng ml ^{− 1} which is sufficient to account for the ventral but not the dorsal induction. The biological properties and specific activity of the Xenopus bFGF seem similar to the bovine bFGF which has been used for most of our experiments on the responses of animal cells. All the MIF activity in a crude embryo or ovary extract can be inhibited by a neutralizing antibody to bFGF, but not by antibodies to a or k FGF or TGF β-2. Parallel work by Kimelman et al. (1988) has also shown the presence of bFGF mRNA and protein in Xenopus blastulae. Their estimate of quantity is much greater than ours but, unlike ours, it is not based on the use of quantitative biological assay methods.

We would obviously predict that the bFGF would be secreted by the cells of the vegetal hemisphere. So far, immunolocalization on embryo sections has not proved successful, probably because of the small quantities present. We have shown that the MIF released by vegetal cells in transfilter experiments can be neutralized by heparin, as can both Xenopus and bovine bFGF, but not by anti-bFGF antibodies. This may mean that the bFGF is secreted as part of some complex not recognised by our neutralizing antibody, but further work is necessary to prove beyond doubt that the vegetal cells really secrete bFGF. One problem in this regard is the well-known fact that bFGF lacks a classical signal sequence for secretion (Abraham et al. 1986), and so there remains some uncertainty about its mechanism of release from cells.

In this work, it has been found that the properties of a and bFGF in their capacity as MIFs are very similar indeed. In what follows, ‘FGF’ will be used to refer to either form indifferently.

Untreated explants from around the animal pole of Xenopus blastulae develop into solid masses of epidermal cells. It can be shown by using antibodies to epidermal markers that 100 % of cells become epidermal (Fig. 1A-D). Mesoderm inductions can be provoked by FGF concentrations in excess of about 100 PM (Fig. 2A). After explants are exposed to FGF nothing much appears to happen for the first few hours, the explants round up with their blastocoelic surface inside and the cells continue to cleave just like untreated explants. However, it is the first 90min or so of exposure that are critical. After this time, the FGF can be withdrawn without affecting the course of subsequent events. Then, while control embryos are undergoing gastrulation, the explants elongate with the original closure point at one end and the original animal pole at the other (Fig. 1E). Within a batch, the degree of elongation depends on the applied dose, but between batches there is considerable variation. After 24-36 h of culture, the induced explants start to swell and soon become transparent (Fig. 1F). These vesicles invariably contain mesodermal tissues although the quantity and type depends on the applied dose (Godsave et al. 1988; Slack et al. 1988). At low doses inductions consist of small amounts of mesenchyme and mesothelium with the occasional wisp of muscle while at higher doses wesee increasing amounts of mesenchyme and increasing amounts of muscle (Fig. 1G,H; Fig. 2B). Notochord is sometimes observed following the higher dose treatments, particularly when in vitro translated bFGF is used, but its formation is not very predictable. This dose-response curve is significantly different from that obtained with XTC-MIF, which will induce notochord reliably at a low multiple of the minimum inducing concentration (J. C. Smith et al. 1988), however, it is probably rather similar to that of Xenopus bFGF (Fig. 3).

We have examined the location of ¹²⁵I-labelled FGF in explants and find that it binds mainly to those plasma membranes that are exposed at the blastocoelic surface. There is little binding to the plasma membrane of the external surface (oocyte-derived or O-membrane) and little penetration into the cell mass (Darlington, 1989). The maximal response to the high doses consists of about 20% muscle by cell composition with an additional 10-20% of mesenchyme and this probably represents all the cells that were exposed on the blastocoelic surface of the explant at the time of treatment. The fact that many cells in induced explants are still epidermal may be entirely due to the limited penetration of the FGF since our studies of single cells leads us to believe that all cells without O-membrane are potentially inducible (see below).

Using animal-vegetal combinations from different stages, it has been shown that the competence of the ectoderm to respond to the natural signal(s) extends from about stage 6 (64 cells) to about stage (Jones and Woodland, 1987). We have studied the onset of competence to FGF in the ectoderm by exposing for a period of 90 min explants taken from different stages and this shows that competence begins at about stage 7. We have studied the loss of competence by permanent exposure of ectoderm explants taken from different stages and this shows that competence is lost between stages 9 and 10 (Slack et al. 1988). Furthermore, the degree of competence can be assessed by measurement of the amount of muscle formed by explants from different stages in response to a standard dose, and this shows a rise and fall with the peak at stage 8 (Darlington, 1989). So the competence for FGF seems to rise at about the same time as competence for the natural MIF(s) but falls rather earlier, since there are about 3 h between stage 9 and at 22 – 24°C. Competence to respond to XTC-MIF seems to persist into gastrulation, until stage according to our measurements (Darlington, 1989).

Since inductions arise in response to FGF concentrations in the pM range we expected that an essential molecular component required for competence would be a specific receptor. We have probed for a receptor on explanted tissues using ¹²⁵I-FGF and the cross-linking agent BS3. This has shown that a receptor is present and appears as two gel bands of M_r about 130 and 140 × 10³,similar to the mammalian FGF receptor (Gillespie et al. 1989). Binding studies show that about 70 – 80% of bound ¹²⁵I-FGF can be competed out by an excess of unlabelled FGF. Assuming that this represents binding to the specific receptor then the density is about 3 × 10⁸molecules mm ^{− 2} of cell surface which is within the range of values measured for mammalian cells. The binding curve shows a half-maximal value of about 3 – 4 nM and a plateau at about 10 nm, which is very similar to the dose-response curve for muscle formation. This suggests that the receptor binding is a limiting step in the response. If it were not, then a maximal response, in this case a maximal percentage of cells induced, would be obtained at an FGF concentration below that required to saturate the receptors.

The receptor density has been studied by binding of ¹²⁵I-aFGF to ectoderm explants taken from different embryonic stages. The competable binding rises by a factor of 10 between the early and middle blastula, and falls again to the starting level by the onset of gastrulation. This closely parallels the rise and fall of competence to respond to FGF and suggests that competence is indeed controlled by receptor density.

Competition experiments have shown that both a and bFGF bind to the same receptor but TGF β -2 does not. This again resembles the situation in mammalian cells and makes it probable that the extended period of competence that ectoderm explants show when treated with XTC-MIF is due to the presence of separate TGFβ receptors.

We have measured the regional distribution of FGF receptors in stage 8 blastulae by binding studies on explants (Gillespie et al. 1989). This shows, as predicted, that FGF receptors are present both in the marginal zone region, which normally responds to the signal in vivo, and in the animal pole region, which can respond in experimental situations but would not normally do so in vivo. There is a slight excess of receptor density in the marginal zone but this is only 50% more than the animal pole value, so it would seem that the normal extent of mesoderm induction is determined by the extent of the signal and not by the presence of a more highly competent tissue in the marginal zone. There is no difference in receptor density between dorsal and ventral regions of the animal hemisphere, so this cannot account for the difference between dorsal and ventral inductions. FGF receptor is also present in the vegetal region. We do not know whether these cells need FGF for their normal development since we cannot deprive them of it in the way that we can deprive the animal cells. However, since they do not normally turn into mesoderm, we can deduce that mesodermal competence consists of something more than the presence of FGF receptors on the cell surface.

Some other workers have noticed that isolated ectoderm cells will not differentiate into mesodermal cell types after induction, although their differentiation into epidermis may be suppressed (Symes et al. 1988). This phenomenon has been called the ‘community effect’ (Gurdon, 1988). We have found that this requirement can be met by a few simple macromolecular additives to the culture medium. Single internal blastula ectoderm cells can be induced if they are treated with FGF and then cultured in the presence of gamma-globulin on a surface coated with fibronectin and laminin. Usually they give rise to monotypic clones of muscle or an ‘epithelium’ which is a non-muscle, non-epidermal cell type, possibly a form of kidney. Sometimes mixed colonies are formed with more than one mesodermal cell type. When cells are treated for only 2 h with FGF, the colonies are always monotypic (Godsave and Slack, 1989). We are presently using this culture system to examine the specification of single cells isolated from different parts of the marginal zone of normal embryos, and have shown that mesodermal clones can be obtained from the marginal zone of midblastulae.

Further experiments involving the induction of single cells have shown that cells bearing the oocyte-derived membrane (O-membrane) are non-inducible (Darlington, 1989). The most informative protocol has involved (1) labelling of donor embryos by injection with the lineage label rhodamine-dextran-amine (RDA), (2) isolating single labelled cells in Ca²⁺-free medium, (3) wrapping these in ectodermal jackets from unlabelled embryos, (4) inducing the whole sandwich with FGF or another MIF before it has sealed. When inner cells wholly surrounded by cleavage membrane (C-membrane) are used then many progeny of the labelled cell are found in the induction. However, when cells bearing O-membrane are used only a very few progeny are found to be induced. Close examination of these few shows that all of them are themselves wholly surrounded by C-membrane, and must therefore have arisen from the original cell by tangential cleavage. So they do not represent exceptions but rather they are important positive controls, showing that the culture conditions do not militate against mesoderm differentiation. A further control in these experiments is provided by the fact that the cells that do not form mesoderm do form epidermis, showing that the failure to form mesoderm is not due to some damage inflicted on the cells in the course of the manipulations.

It is generally agreed that the signal near the dorsal meridian differs from that around the remainder of the blastula circumference. Some workers have tended to think that it is qualitatively similar but more intense while others have leaned towards the view that it is qualitatively different. If we accept that bFGF is the ventral morphogen, then the quantitative view seems unlikely since notochord inductions are not reliably produced even by very high concentrations of FGF, and the uniform distribution of FGF receptor shows that the dorsal and ventral ectoderm will respond alike to similar concentrations of FGF. However, it has been shown that the effect of FGF can be modified by other factors. There is strong synergism between FGF and TGF β -1 (Kimelman and Kirschner, 1987) and between FGF and lithium ion (Slack et al. 1988). Neither TGF β-1 nor Li are active as mesoderm-inducing factors on their own and the synergism is usually manifested as an excess formation of muscle rather than by induction of notochord. TGFβ-2 does have mesoderm-inducing activity on its own, and like FGF does not usually induce notochord. However, the synergism between FGF and TGFβ-2 is strong enough to give reliable induction of notochord (E. Amaya, pers. comm.). We have seen above that the receptors for FGF and TGF β on Xenopus ectoderm are distinct but the synergistic effects suggest that there is a common intermediate at some level in the signal transduction pathway. This intermediate is presumably one whose level can be elevated by Li⁺.

A reasonable working hypothesis based on these data might be that bFGF is the ventral morphogen and bFGF+ TGFβ -2 the dorsal morphogen. We would further suppose that the FGF system is prelocalized in the vegetal hemisphere of the egg while the TGF/J system is activated on the dorsal side only as a result of the postfertilization cytoplasmic movements (see Fig. 4 and Gerhart et al. this volume). This would then explain the effects of UV radiation and Li⁺ on whole embryos. If the vegetal hemisphere of the egg is irradiated with a sufficient dose of UV light then the cytoplasmic movements are inhibited and a radially symmetrical ventral embryo is formed (Grant and Wacaster, 1972; Cooke and Smith, 1987). The simplest interpretation of this is that the FGF system is normally present in the vegetal hemisphere all around the circumference and is unaffected by the treatment while the TGF/3 system would depend on the postfertilization cytoplasmic movements which are blocked by the UV dose. Li⁺ treatment of the early embryo produces a symmetrical dorsalization (Kao el al. 1986; Cooke and Smith, 1988). Here the postfertilization movements have already happened but we presume that the Li can elevate the concentration of a signal transduction intermediate and so mimic the effect of a uniform dorsal stimulus. It has been shown that lithium will dorsalize isolated ventral marginal explants to the level of large muscle masses (Slack et al. 1988; Kao and Elinson, 1988).

In fact, we have no evidence at present that the Xenopus homologue of TGF β -2 is present in the early embryo and it may be that some other TGF β -like molecule is doing the job. An obvious candidate is the XTC-MIF of Smith since this has chemical properties resembling TGF β and is currently the most active of all the MIFs and the only one that will induce notochord on its own. Another possibility is the Vg1 product. Here we know that the mRNA is present in the embryo and localized in the vegetal hemisphere (Weeks and Melton, 1987; Yisrael et al., this volume). However, there does not appear to be any preferential localization on the dorsal side, and perhaps more seriously there is as yet no indication of any biological activity shown by the protein. Clearly more data is needed in this area and in particular data on the presence, distribution and activity of TGF β -like molecules in the early Xenopus embryo.