Presence of basic fibroblast growth factor in the early Xenopus embryo

Much interest has been aroused recently by the possibility of identifying the morphogens responsible for mesoderm induction in the early amphibian embryo. This is the process by which the mesoderm is induced from the animal hemisphere of the blastula by signals from the vegetal region (Nieuwkoop, 1969; Dale et al. 1985; Gurdon et al. 1985; Jones & Woodland, 1987). Recently we reported that a small group of heparin-binding growth factors (HBGFs) are active as mesoderm-inducing agents and the relatively high degree of specificity suggested to us that at least the ventral mesoderm induction within the embryo was mediated by an HBGF (Slack et al. 1987). However, it has now been found that another type of growth factor, TGFβ-2, is active (Rosa et al. 1988) and a mesoderm-inducing factor recently purified from a Xenopus cell line is also thought to belong to the TGFβ family (Smith, 1987; Smith et al. 1988). So which, if any, of these factors is really involved in mesoderm induction in the early embryo? A clue was provided by Kimelman & Kirschner (1987) who detected an mRNA of the bFGF type in Xenopus embryos. We now show that a mesoderm-inducing factor can be isolated from Xenopus blastulae. It is similar to bFGF by biochemical and immunological criteria and has a similar mesoderminducing activity in vitro. The quantity present is small but sufficient for it to be responsible for induction in vivo. This is consistent with, although does not yet finally prove, the involvement of bFGF as a morphogen in mesoderm induction.

Eggs and embryos were dejellied with 2% cysteine hydrochloride pH7·9 and stored at —70°C before use, as were whole ovaries from adult females. The final protocol adopted was as follows: 25−100 g of ovary, unfertilized eggs or embryos were homogenized in 2−3 vols of 0−15 M-ammonium sulphate containing Imm-PMSF and 100 nM-pepstatin A. Homogenization was by Ultra turrax blender for 3x1 min at maximum speed, the temperature being kept below 10°C with ice. All subsequent steps up to the Amicon concentration were carried out at cold room temperature. The pH was adjusted to 4·5 with acetic acid and the homogenate was stirred for 1 h on ice. It was then centrifuged at 150000g for one hour and lipid was removed by straining through nylon mesh. The supernatant was neutralized with ammonia and fractionated by adding, successively, 0·24 g ml^-1 (40%) and 0 ·2g ml^-1 (70%) solid ammonium sulphate followed by centrifugation at 12000g for 10min. The 40−70% cut was dialysed overnight against 0·6m-NaCl, 20mm-Tris pH 7·0 and then applied to a 2 ml column of heparin Sepharose (Sigma or Pharmacia) equilibrated in the same buffer. The column was washed in the same buffer and eluted with 2m-NaCl, 20 mm-Tris-HCl pH 7·0. Ovalbumin (Sigma, mol. wt std) was added as carrier and the product was concentrated to 1 ml with an Amicon ultrafiltration cell and YM10 membrane. This material is referred to below as the ‘heparin-bound fraction’.

Earlier preparations differed in that the initial homogenate was centrifuged less vigorously (25 000g, 30 mins), a 40−80 % ammonium sulphate precipitate was taken, and also that a CM-Sephadex column at pH 6 was employed before the heparin column.

The samples were separated using 15 % SDS-polyacrylamide gels and transferred to nitrocellulose using standard tech-niques. The blots were blocked for Ih in 0·15m-NaCl, 0·1 m-Tris-HCl pH 7·4, 1% milk powder. They were stained overnight in the same solution with 30 ng ml^-1 anti-bFGF (protein A purified) or 2·5 μg ml^-1 anti-bFGF 16−30 (affinity purified) with or without 25 μg ml^-1 of BSA-peptide conjugate or free peptide. Visualization was with the Vector Alkaline phosphatase ABC kit and number 2 substrate, following the manufacturer’s protocol.

0·75 mm thick SDS gels were silver stained by the method of Heukshoven & Dernick (1985).

The heparin-bound fraction was diluted to 0·5m-NaCl with 20 mm-Tris-HCl pH 7·5, applied to a Heparin 5PW HPLC column (Anachem) and eluted with a gradient of NaCl from 0·5 to 1·5M in 20mm-Tris-HCl pH7·5. A portion of each fraction was assayed and another portion run on a gel, transferred to nitrocellulose and reacted with one of the antibodies.

Assays were carried out by the end-point dilution method for inducing factors described by Godsave et al. (1988). Briefly, expiants of ectoderm from the animal pole region of Xenopus blastulae are exposed to serial dilutions of the preparation. Induction is assessed by the formation of swollen vesicles after 2−3 days culture (such vesicles always contain mesodermal tissues). The titre of the preparation is the reciprocal of the highest dilution that retains activity, i.e. if 1/64 is the highest active dilution then the activity is 64 units ml^-1

Methods for fertilization, composition of salines, histological techniques and inducing factor treatments are all given in Godsave et al. (1988).

a and bFGF were prepared in the laboratory from bovine brain, using methods described previously (Slack et al. 1988). kFGF was an in vitro translation mixture prepared by G. Paterno in this laboratory (see Paterno, Gillespie, Slack and Heath, submitted). Antibodies to a and bFGF were protein A purified IgG donated by Dr D. Gospodarowicz (UC Medical Center, San Francisco). The anti-bFGF(16-30) was prepared in the laboratory against a thyroglobulin conjugate of the peptide HFKDPKRLYCKNGGF, and was affinity purified using a column carrying a BSA conjugate of the peptide. Although the peptide is from the bovine sequence (Abraham et al. 1986), it differs from the Xenopus sequence only in that it has H instead of S at the first position (D. Kimelman, personal communication). TGFβ-2 and anti-TGFβ were purchased from R&D Systems Inc. Anti-kFGF was donated by Dr D. Rogers (Genetics Inst.). Heparin was type I purchased from Sigma.

Preliminary experiments using ovary showed that soluble mesoderm-inducing activity could be extracted by homogenizing in a suitable buffer and stirring in the cold room for 1h. Activity was assayed by observing mesoderm induction in expiants of ectoderm isolated from Xenopus blastulae and exposed to serial dilutions of the fractions (Godsave et al. 1988). Similar yields were obtained using low (0·15 M) or high (1 M) salt and low (4·5) or neutral (7·4) pH and the standard method chosen was that previously used for HBGFs, namely 0·15 M-ammonium sulphate at pH 4·5 (Gospodarowicz et al. 1984; Esch et al. 1985; Lobb et al. 1986). The solubilized activity is relatively stable in crude extracts and is only reduced by about fourfold by acidification (0-1% trifluoroacetic acid) or boiling. Although there are losses during subsequent steps, most or all of the remaining activity was found to bind to CM-Sephadex at pH 6 and, more significantly, to bind to heparin Sepharose in 0·6m-NaCl at pH 7, thus satisfying the definition of an HBGF.

Preparations were made from ovary, unfertilized eggs and blastula-stage embryos and, in all cases, virtually all of the recoverable activity was found to bind to heparin. A protocol from a 50 g embryo preparation is shown in Table 1. The total content of inducing activity, judged from the activity of crude extracts, was 7 units g^-1 wet weight in packed dejellied embryos (average of 5 preparations), compared to about 1000unitsg^-1 of bovine brain, the standard source for HBGFs. The actual values for the five preparations were: 16, 13, 5, 2 and 1 units g^-1 wet wt, the two lowest figures coming from small scale preparations using about 0·25−0·5 g embryos. If these two were excluded on the grounds that the scaled-down protocol might promote losses, then the average would be 11 units g^-1 wet wt. The serial dilution assay has an inherent uncertainty of a factor of 2 but from these results it seems unlikely that the true figure would exceed 30 or be less than 5 units g^-1 wet wt. The content in unfertilized eggs was about the same as for embryos, but in ovary was substantially higher (109 units g^-1 wet wt, average of 8 preparations of which max. 197 and min. 68 unitsg^-1 wet wt). It should be remembered that ovary contains follicle cells, connective tissue, nerves, blood vessels etc. in addition to oocytes, and at present we do not know how much of the activity is present in the oocytes and how much in the other cell types.

Although the evidence (see below) suggested that all the inducing activity was attributable to bFGF, we thought it possible that TGFβ-like factors might not be extracted or might not be active under these conditions. Accordingly, we extracted both blastulae and also the pellets left by the aqueous buffer extraction with acid ethanol, following the protocol of Assoian et al. (1983) for TGFβ. However, we were unable to detect any inducing activity in these preparations or any synergism with bovine bFGF. Furthermore, there was no growth inhibition activity using CCL64 mink lung cells which are sensitive to both TGFβ-1 and TGFβ-2 (data not shown). These measurements place an upper limit on TGFβ content of 0·8 inducing units g^-1 wet wt, about one tenth of the total observed activity.

Both crude extracts and those partially purified by heparin affinity chromatography were tested against a panel of reagents to characterize them. At the same time the specificity of the reagents was checked by testing them against pure factors. The factors were aFGF, bFGF, kFGF and TGFβ-2, and the reagents were antibodies to aFGF, bFGF, kFGF, TGFβ, and also heparin. Conditioned medium from XTC cells was also tested. In each case a concentration of about 10 units ml^-1 of the factor was tested against twofold serial dilutions of the reagents. If there is any inhibition this becomes lifted below a certain dilution thus providing a neutralization titre. These were reasonably reproducible, except for TGFβ-2/heparin, and showed that there was no cross inhibition of pure factors by antibodies to other factors and that the neutralization titres were in the range 1−50 μg of IgG per unit of factor. The collected results are shown in Fig. 1. It is clear that the Xenopus preparations behave like bFGF and not like the other factors and since complete neutralization can be achieved with anti-bFGF it seems probable that there are no other active factors present either in the crude extracts or in the heparin-purified material.

We considered the possibility that the residual activity remaining after acidification of crude extracts might be TGFβ-like. However, this too was inhibitable by anti-bFGF but not by anti-TGFβ, suggesting that the partial acid resistance simply represents protection by other materials in the extract.

Although this series of experiments showed no inhibition of aFGF or kFGF by heparin, there may be a slight effect at high heparin concentrations as these have sometimes been apparent in series performed by other workers in this laboratory. Furthermore, two out of four experiments suggested that TGFβ-2 could be inhibited by heparin as effectively as bFGF, while the other two experiments showed no inhibition. We have no explanation for this unreproducible behaviour.

It should be noted that none of the reagents tested had any effect on the activity of the XTC-conditioned medium. The anti-TGFβ used is supposed to neutralize both TGFμ-1 and -2 and this may suggest that XTC factor is less similar to TGFβ-2 than has been proposed by Rosa et al. (1988).

Two antibodies were used as Western blotting reagents in this study. The first was an anti-peptide, anti-bFGF(16-30), prepared in this laboratory, which is capable of detecting Ing of bovine bFGF with high specificity. The second is the anti-bFGF used for the neutralization experiments. This is not so specific, showing a number of cross-reactive bands in the higher molecular weight range, and could detect about 20 ng of bovine bFGF.

When the anti-bFGF(16−30) was used to stain blots of heparin-bound material from ovary or embryos, a band was always seen at a molecular weight of about 19K, slightly larger than bovine bFGF. It sometimes appeared as a single band (lanes 3 and 5) and sometimes as a doublet (lane 7). It was completely suppressed by inclusion of a 10-fold excess of free peptide or BSA-peptide conjugate in the reaction mixture, showing that the antibody binding was specific (Fig. 2A). When ovary preparations were further fractionated by chromatography on a heparin 5PW column, which permits gradient elution, the 19K reactive species eluted at about l·25m-NaCl, coinciding with a peak of inducing activity. However, larger amounts of inducing activity also eluted at lower salt concentrations and it was clear that the 19K band could not represent all of it.

When the neutralizing anti-bFGF was used to stain Western blots of heparin-bound material from ovary or embryos, the 19K band was again seen but now it was accompanied by another band at about 14K (Fig. 2B). Using this antibody, there was quite a good fit between immunoreactivity and biological activity on the heparin 5PW fractionation, with the 14K band corresponding to a peak of activity eluting at about 0-8 M-salt and the 19K band to the peak eluting at 1·25 M (Fig. 3). After the HPLC separation, it was possible to identify a silver-staining band corresponding to each of the immunoreactive bands, although the 14K silver band extended over more fractions than the biological activity and so probably still does not represent a pure product (Fig. 4). The intensities of the blot and silver bands allowed us to estimate that the specific activities of the pure factors must be in the region of 10⁶ units mg^-1 which is the same as for bovine bFGF.

Because of the chemical and immunological similarities of our factor to bFGF, it is tempting to assume that it is the Xenopus homologue of mammalian bFGF. It will accordingly be referred to as such.

A number of ectoderm expiants were examined histologically after treatment with various doses of Xenopus bFGF and culture for three days to allow differentiation. The results were identical whether embryo-, egg- or ovary-derived material was used, the preparations tested being heparin-bound fraction from each and, in the case of the ovary, also the two active peaks from the heparin 5PW gradient separation. At low doses, of about 1−4 units ml^-1, the inductions were typically ventral in pattern, containing loose mesenchyme surrounding a mesothelial layer with a few pycnotic cells in the centre and sometimes a small wisp of muscle (Fig. 5A). At higher doses, from 8−256 units ml^-1, the inductions were mainly ‘intermediate’ (see Dale & Slack, 1987), containing abundant loose mesenchyme and large muscle blocks (Fig. 5B). However, two cases out of 20 treated with high doses contained notochord, which is perhaps more than would be expected from comparable experiments with bovine FGFs. The external appearance of the expiants after induction was the same as previously described for bovine bFGF, usually consisting of swollen vesicles with some contents (Fig. 5C,E). Cases inhibited by antibodies or heparin were indistinguishable from untreated ones (Fig. 5D).

In order to qualify as a morphogen a molecule has to satisfy a number of criteria. These have recently been the subject of much discussion in the developmental biology community and a consensus set might read as follows. Numbers 1−6 apply to any instructive inducing factor, and nos 7−9 specifically to cases where a morphogen gradient is thought to specify a spatial pattern.

1. The candidate molecule should produce the appropriate types of induction when tested on isolated responding tissue.

2. It should be present in the embryo at the stages when the normal process is taking place.

3. The amount should be adequate for the biological activity observed in vivo.

4. It should be exported from the region identified as the signalling centre.

5. It should be transmitted to the responding tissue.

6. Inhibition of its synthesis, transport or mode of action by mutation or by specific chemical antagonists should inhibit the response in vivo.

7. A gradient of concentration should exist within the responding tissue.

8. Different structures should be induced by different concentrations in vitro.

9. The order of structures in terms of decreasing distance from the signalling centre shall be the same as the order in terms of increasing concentration required to induce.

So far, only three substances approach the satisfaction of all these criteria. These are the bicoid product in the egg of Drosophila (Driever & Niisslein-Volhard, 1988), retinoic acid in the developing limb bud of the chick (Thaller & Eichele, 1987) and the slime mould morphogen, DIF (Morris et al. 1987). In each of these cases, a number of uncertainties remain about their role in vivo, but these would take too long to discuss here.

How does Xenopus bFGF rate as a candidate morphogen for the process of mesoderm induction? It is active in vitro, and we have now shown that it is present in the embryo at the appropriate stage. Since, by definition, a concentration of 1 unit ml^-1 is the minimum required for induction we can conclude that the amount in the embryo, although low, is adequate for it to be functioning as a morphogen in normal development, at least as an inducer of ventral mesoderm and particularly if it is concentrated within the equatorial belt of responding tissue. We may conclude that it has passed tests 1, 2 and 3.

An indication that conditions 4, 5 and 6 may be satisfied is the fact that passage of the signal in a transfilter apparatus can be inhibited using concentrations of heparin similar to those that inhibit bFGF in vitro (Slack et al. 1987). However, it is unlikely that heparin is a really specific reagent for bFGF as in the experiments reported here we also noticed inhibition of TGFβ-2 on some occasions. We therefore feel that more decisive evidence is required and we are continuing to work on this problem using the transfilter apparatus and other methods.

Xenopus bFGF approaches the satisfaction of condition 7 since low concentrations induce mesothelium and mesenchyme only, higher concentrations induce muscle, and the induction of notochord, although not common, has been observed. However, we have not so far been able to localize the bFGF within the embryo either by immunocytochemistry or by microscale extraction of dissected pieces, so we do not know whether there are any gradients in vivo. If we are correct in our assessment from the gel bands that the specific activity of Xenopus bFGF is similar to that of bovine bFGF then 7 units ml^-1 translates into 7 ng ml^-1 or 368 pM or 221 molecules μm⁻¹ . This is probably well below the detection limit of any in situ method and will make localization difficult. It is also unclear whether condition 9 is satisfied since the fate map data of Keller (1976) seems to suggest that the lateral plate arises from nearer the vegetal tissue than the somites, and this would not correspond to the observed concentration dependence of inductions in vitro, either for FGF or any other mesoderm-inducing factor.

So, at present, Xenopus bFGF scores about 4 out of 9 which in our opinion is not enough to be regarded as a proven morphogen. But the rapid pace of research around the world may enable us to reevaluate the situation quite soon.

The occurrence of bFGF in the unfertilized eggs indicates that not only mRNA, as shown by Kimelman & Kirschner (1987), but also protein must be synthesized before it is needed for signalling. The published cDNA clone for bovine bFGF lacks a signal sequence for secretion (Abraham et al. 1986) and since there is no cell death during the early stages of Xenopus development we must presume that some novel mechanism exists for getting it out of cells. Perhaps therefore its export is controlled by the synthesis of some other molecule at a later stage.

The difference of reactivity between our two antibodies is most easily explained by assuming that the 14K form has been severely truncated at the N-terminus and so lacks residues 16−30 against which the antipeptide is directed. In some, but not all, preparations, some fainter immunoreactive bands were seen around 15−16K with both antibodies, for example, see Fig. 2A lane 4. Also, the principal bands sometimes appeared as doublets and sometimes not. This sort of heterogeneity is found also in FGF preparations from other sources but it is not known whether it represents multiple endogenous forms or degradation during preparation .

As far as other factors are concerned we can place an upper limit on TGFβ content of about 10% of bFGF activity. However, this does not mean that TGFβ-like molecules are definitely absent since we may still not have found the optimal extraction procedures. Indeed, we expect other factors to be involved in mesoderm induction, partly because of the evidence advanced to support our own ‘three signal model’ (Dale & Slack, 1987; Slack et al. 1988), and partly because the XTC-MIF (Smith, 1987; Smith et al. 1988) is a material of impressive potency obtained from a homologous source, and it would be not be surprising if it too had a role to perform in the embryo.

We should like to thank Dr D. Gospodarowicz (UC Medical Center, San Francisco) for a gift of antibodies to a and bFGF; Dr G. Paterno (this laboratory) for the kFGF; Dr D. Rogers (Genetics Institute, Cambridge, Mass.) for permission to use the anti-kFGF; and Dr J. Rothbard (ICRF, London) for making the peptide and peptide conjugates.