Nuclear translocation of fibroblast growth factor during Xenopus mesoderm induction

In Xenopus laevis blastulae, mesoderm formation from marginal zone cells is thought to be induced by growth factors belonging to the FGF and transforming growth factor beta (TGF-β) superfamilies, secreted by presumptive endoderm of the vegetal pole (reviews, Smith, 1989; Whitman and Melton, 1989). These growth factors can mimic the effect of vegetal tissue on responsive cells in vitro (Slack et al. 1987; Kimelman and Kirschner, 1987; Paterno et al. 1989; Rosa et al. 1988; Asashima et al. 1990; Smith et al. 1990; van den Eijnden-Van Raaij et al. 1990), and several FGF and TGF-βfamily members are also known to be present in the embryo (Kimelman et al. 1988; Slack and Isaacs, 1989; Thomsen et al. 1990). However, it has not yet been demonstrated that any of these factors can induce mesoderm in vivo. Spatial localization of growth factors within the embryo at successive developmental stages may provide vital clues regarding their function.

bFGF is known to be present in Xenopus oocytes and early embryos at picogram levels (Kimelman et al. 1988; Slack and Isaacs, 1989). In order to detect these low levels, we used a freeze-substitution method of fixation designed to maintain native molecular structures and preserve growth factor epitopes (Michael et al. 1984). Advantage also was taken of the high sensitivity and resolution of Immunogold–silver staining (Holgate et al. 1983; Danscher and Norgaard, 1983), where antibody binding is visualized in the form of non-diffusible black metallic silver grains. With these procedures, specific immunocytochemical staining was obtained with low background. Unexpectedly, both aFGF and bFGF antibodies stained early embryos, and a change in the subcellular distribution of FGFs from cytoplasm to nucleus appeared concurrently with induction of mesoderm.

Ovary and dejellied Xenopus laevis embryos of various stages (Nieuwkoop and Faber, 1967) were collected as described (Godsave et al. 1988). Tissues were preserved by freezesubstitution (Michael et al. 1984; Pearse, 1980; Simpson, 1941) and embedded in polyester wax (Kusakabe et al. 1984). 7 μm sections were mounted on glycerin/albumin-coated slides and were processed by Immunogold–silver staining (Holgate et al. 1983; Danscher and Norgaard, 1983). Five rabbit polyclonal IgGs were used as primary antibodies: (1) protein A-purified anti-native basic FGF from bovine brain (R&D systems, Minneapolis, USA); (2) affinity-purified anti-bovine basic FGF, peptide 16–30 (J.M.W. Slack, Oxford, UK); (3) protein A-purified anli-Xenopus basic FGF, complete sequence (Slack); (4) affinity-purified anti-bovine basic FGF, peptide 1–24 (Gonzalez et al. 1990) and (5) protein A-purified anti-native acidic FGF from bovine brain (R&D Systems). Three mouse monoclonal antibodies (subclass IgG₁ also were used: (1) 1215–00, anti-bovine acidic FGF (Genzyme, Boston, USA); (2) Tub 2.1, anti-tubulin (ICN, Lisle, IL, USA); (3) MCA 151A, anti-mitochondria (Serotec, Oxford, UK). Negative control antibodies included chromatographically purified rabbit IgG (Zymed, So. San Francisco, USA) and mouse IgG₁ (Chemicon, Temecula, USA). Other negative controls omitted one of the following: primary antibody, secondary antibody–colloidal gold conjugate and silver enhancement step.

All procedures were carried out at room temperature. Sections were dewaxed and rehydrated in a graded ethanol series (100%–70%), and they were treated with Lugol’s iodine solution (1g iodine crystals plus 2 g potassium iodide/200ml ethanol) for 5 min in order to complex heavy metals for removal. After rinsing in deionized water, slides were immersed for 3 min in a 2.5 % w/v sodium thiosulphate aqueous solution in which iodine–metal complexes are soluble. Slides were rinsed again in water, and they were treated for 1h in 100mM glycine, 100 DIM NH4CI in TBS (50 mM Tris–HCl, 150 mM NaCl, pH 7.6) to convert silver-reducing endogenous aldehydes to non-reactive amino groups. After washing in TBS for 10 min, sections were covered with 3 % normal goat serum (NGS) in TBS for 1 h to block non-specific protein binding. Then primary antibodies (1 μg IgG ml^-1,1 % NGS in TBS) were applied for 24 h. After a 30 min wash in TBS, goat anti-rabbit IgG- or goat antimouse IgG–gold particle conjugates (1 nm diameter; Amersham, UK) were diluted 1:500 in TBS, 1 % NGS containing 0.5% v/v fish gelatin (Birrell et al. 1987), and the colloidal suspensions were applied to sections for 24 h.

Signal amplification of immunogold staining was performed by silver enhancement. Slides were washed for 30 min in TBS followed by 30 min in deionized water to remove chloride ions. Then they were immersed for 20 min in a physical developer solution (5.6mM silver lactate, 78 mM hydroquinone, 133 mM citric acid, 80 MM trisodium citrate, pH 3.8 in 20% gum arabic) under photographic safelight (Danscher and Norgaard, 1983). After development, slides were extensively washed in deionized water. They were dipped in 1 % acetic acid for 1 min, rinsed in water, and were treated for 3 min in 2.5% sodium thiosulphate solution to remove . unreacted silver ions. Slides were given a final rinse in water and were dehydrated by sequential dipping in a graded ethanol series (70%–100%) followed by two changes of 100% xylene. Coverslips were mounted with Merckoglas liquid coverglass medium (Merck, Darmstadt, Germany). Photomicrography was performed with an Olympus BH-2 microscope using 35 mm FUJI Neopan film, and black and white prints were made on high-contrast FUJIBRO WP FM4 paper.

Additional immunocytochemical controls for antibody specificities were done by staining tissue sections with polyclonal anti-Xenopus bFGF and anti-bovine bFGF (R&D) IgGs that were pretreated by solid-phase absorption with bFGF (Michael et al. 1984; Sternberger, 1986). Recombinant bovine bFGF (30 μg, Boehringer Mannheim) was covalently coupled to cyanogen bromide-activated Sepharose 4B (100 μl Pharmacia) according to the manufacturers instructions. A 20 μl portion of the resulting matrix was mixed with 200 μl of each anti-bFGF IgG (1 μgml^-1, 1% NGS in TBS), and the mixtures were incubated with gentle agitation for 18 h at 4°C. After low-speed centrifugation, supernatants were used for immunostaining. To rule out the possibility that antibodies were eliminated by non-specific effects, a control absorbent matrix composed of Sepharose-coupled keyhole-limpet haemocyanin (KLH) was constructed, mixed in the same ratio with each of the anti-bFGF IgGs, and treated similarly. Supernatants from absorptions with each kind of matrix were tested by using them as primary antibodies for immunostaining sections of mid-blastulae. Results were compared to positive control non-absorbed anti-bFGF IgGs and negative control normal rabbit IgG.

Approximately 4000 unfertilized Xenopus eggs (UFE) and 2000 blastula stage 8.5–9 embryos (Bl) were homogenized in 5 ml and 2.5 ml of extraction buffer (10mM Hepes, pH7.4_; 0.5 M NaCl, 2mM EDTA, 1 mM EGTA, 1 mM PMSF, 2 μtg ml^-1 trypsin-chymotrypsin inhibitor, 20 KlUmU¹ aprotinin, 5 μgml^-1 protease inhibitor), respectively. Extracts were stirred for 1h and centrifuged at 15 000g for 15 min. Heparin–Sepharose CL-4B (0.5 ml) was mixed with each supernatant, agitated at 4°C for 1h, and packed into microcolumns. Columns were washed with 10 mM Hepes, 0.5 M NaCl, pH 7.4 and were eluted with 10mM Hepes, 2 M NaCl, pH 7.4. Eluates were maximally concentrated to 1.1ml at 440 μgml^-1 (UFE) and 0.5ml at 960 μgml^-1 (Bl) by centrifugation through Centrisart I tubes (Sartorius). 15 μl samples were electrophoresed in precast SDS– polyacrylamide gels (10–20 % gradient) (Laemmli, 1970), and proteins were transblotted to Immobilon P (Millipore) membranes (Towbin et al. 1979). Nonspecific binding sites were blocked by immersing blots in 5 % non-fat dried milk in Tris-buffered saline pH7.6 (TBS) for 1h. Then blots were incubated in primary antibodies overnight at 4 °C. Rabbit antibovine bFGF (1:400) (R&D) and mouse monoclonal antibovine aFGF (1:400) (Genzyme) were diluted in 5% NGS, 1 % bovine serum albumin in TBS. Other anti-bFGF antibodies and control IgG were diluted similarly for separate blots (not shown). Primary antibody binding was detected by the peroxidase anti-peroxidase (PAP) method (Sternberger et al. 1970).

Antibodies to bFGF and acidic FGF (aFGF) were immunoreactive with epitopes in both oocytes and early embryos. The results are summarized in Table 1. All anti-bovine as well as antl-Xenopus bFGF antibodies, whether anti-peptide or anti-complete sequence, showed very similar patterns of localization, differing only in their relative staining intensities. As an example, reactivity of anti-bovine bFGF (R&D Systems) is depicted in Fig. 1. In ovary (Fig. 1I,J), there was no significant immunocytochemical staining of previtellogenic oocytes, although some follicle cells surrounding them were immunoreative for bFGF. By contrast, in vitellogenic oocytes, both FGFs appeared to be localized between yolk platelets in the cytoplasm of both the animal and vegetal hemispheres as well as in the cytoplasm of enveloping follicle cells. bFGF antibodies also stained follicle cell nuclei. After fertilization, antibodies to bFGF and aFGF showed similar patterns of immunoreactivity up to stage 8. At cleavage stages (Fig. 1A,E), staining was still cytoplasmic between yolk platelets, but it was concentrated in the marginal zone and vegetal pole. Very little antigen was detected at the animal pole, suggesting that FGF is redistributed following fertilization to become localized in the mesoderm- and endoderm-forming regions of the embryo. These regional differences persisted into blastula and gastrula stages, and the staining continued to be predominantly intracellular.

In mid-blastula stages, a change in subcellular distribution of FGFs became apparent. By stage 8 (Fig. 1B,F), many vegetal hemisphere cells showed perinuclear staining, and in some cells immunoreactivity was present within the nucleus. This is around the time of the mid-blastula transition (MBT) and onset of mRNA transcription (Newport and Kirschner, 1982), so it is possible that the change in staining pattern represents either the appearance of FGFs synthesized from new transcripts or a redistribution of existing FGFs. At the MBT the rate of cellular division slows, making it unlikely that FGFs are stimulating proliferation at this time. Alternatively, it is possible that FGFs act as transcription factors, perhaps activating mesen-dodermal gene expression. During stage 8 the staining patterns of bFGF and aFGF became distinct. bFGF was predominantly nuclear and remained in this location during gastrula (Fig. 1D,G) and neurula stages (not shown), ectoderm always being more lightly stained than other regions. At stages 9 (Fig. 1K) and 10 (Fig. 1L,O,P), aFGF was mainly cytoplasmic, although many nuclei showed some level of immunoreactivity. Presumptive ectoderm was again only lightly stained.

Antibodies to tubulin and mitochondria were used as positive controls, and it was possible to demonstrate that the fixation and staining methods produced the predicted patterns of subcellular distribution for these antigens. Anti-tubulin antibody strongly stained the mitotic apparatus of rapidly dividing blastula cells (Fig. 1M), and anti-mitochondrial antibody stained the cytoplasm but not the nuclei of these cells (Fig. 1N).

Negative immunocytochemical controls, employing solid-phase absorption of anti-bFGF IgGs on a matrix composed of bovine bFGF covalently coupled to Sepharose 4B, demonstrated the removal of nuclear and cytoplasmic staining in cells of mid-blastulae (Fig. 2A). Positive control anti-bFGF IgG absorbed with inappropriate KLH-Sepharose showed neither reduction in immunostaining intensity nor change in the staining pattern in these embryos as compared to unabsorbed antibody (Fig. 2B and 2C). Anti-bovine bFGF (R&D) and anti-Xenopus bFGF gave identical results. Other immunocytochemical controls (see methods and Fig. 1H) also indicated that the staining was specific.

Antibody specificities were also tested on western blots (Fig. 3). Anti-Xenopus bFGF (Slack, Fig. 3A) and antibovine bFGF (R&D, Fig. 3B) were strongly reactive with both bovine bFGF (M_r 18 ×10³, R&D) and recombinant Xenopus bFGF (M_r 17 ×10³, Kimelman et al. 1988), though both apparently contained some degradation products at M_r 14 ×10³. This cross-species reactivity is consistent with the 84 % amino acid sequence identity between bovine bFGF (Abraham et al. 1986) and Xenopus bFGF (Kimelman et al. 1988) as determined by computer analysis using a local homology algorithm for optimal alignment (BestFit). In western blots of heparin-binding proteins from extracts of unfertilized eggs (Fig. 3A and 3B) and late blastula stage embryos (Fig. 3B), anti-bFGF recognized a band with a relative molecular mass of approximately 15 ×10³, in agreement with Kimelman et al. (1988), who analysed similar extracts. This is slightly lower than the value of M_r 17 ×10³ for the recombinant bFGF, prepared from the cDNA clone of Kimelman et al. (1988) and used as a control to show that the antibody recognized Xenopus bFGF. By contrast, Slack and Isaacs (1989) found bands at M_r 19 ×10³ and 14 ×10³, and in some preparations fainter bands were seen at M_r 15 to 16 ×10³. We have also occasionally seen a weak band at approximately M_r 19 ×10³ (Fig. 3A). Control blots prepared simultaneously and stained with normal rabbit IgG revealed several faint non-specific bands above M_r 29 ×10³ but none below.

Anti-aFGF (Genzyme) also recognized a band in both unfertilized egg and blastula extracts (Fig. 3C). This band had a relative molecular mass slightly higher than the approximately M_r 16 ×10³ bovine aFGF positive control (Thomas et al. 1985). No immunoreaction was observed between monoclonal anti-aFGF and bovine or Xenopus bFGF, and no bands were detected in control blots treated with normal mouse IgG. However, rabbit polyclonal anti-bFGF IgGs did show some reaction with bovine aFGF, presumably due to the 55 % amino acid sequence identity between the acidic and basic forms of bovine FGF (Thomas and Gimenez-Gallcgo, 1987). Nevertheless, immunocytochemical staining patterns were distinct, particularly with respect to the nuclei. For that reason, it seems likely that the immunostaining of anti-bFGF IgGs was the result of their binding to Xenopus bFGF in the tissue sections.

Basic FGF has been detected in early Xenopus embryo extracts (Kimelman et al. 1988; Slack and Isaacs, 1989), while Xenopus aFGF has not been described previously. Slack and Isaacs (1989) obtained coincident mesoderm-inducing activity and bFGF immunoreactivity, which could be neutralized by anti-bFGF but not anti-aFGF antibodies. However, these preparations were more highly fractionated than those reported here, and immunoblotting experiments were not performed for aFGF. Consequently, the present observation of aFGF in embryo extracts is not at variance with the current literature and merits consideration in future experiments.

Intracellular localization of FGFs in both the marginal zone and vegetal pole suggests that FGF may confer a predisposition to form mesoderm and endoderm rather than acting as an inducer passed between producer and responding cells. However, both aFGF and bFGF are able to induce mesoderm in vitro (Slack el al. 1987; Kimelman and Kirschner, 1987; Paterno et al. 1989; Kimelman et al. 1988; Slack and Isaacs, 1989), FGF receptors capable of binding both forms of FGF are present on the surface of blastula cells (Gillespie et al. 1989; Musci et al. 1990)), and at least some mesoderm formation appears to depend upon induction (Dale et al. 1985; Godsave and Slack, 1991). Small amounts of FGFs could be transferred between cells to act as extracellular signals, but the gene encoding bFGF lacks the signal sequence required for classical mechanisms of protein secretion, although there may be alternative pathways to the extracellular environment. Conceivably, other FGF family members or distinct molecules (eg. TGF-β related growth factors) may mediate these cell-cell interactions. Activin B mRNA was recently shown to be expressed after the MBT (Thomsen et al. 1990). Since the protein is a very potent mesoderm inducer, it is currently a good candidate for a primary signalling molecule.

Establishment of mesodermal and endodermal cell lineages during Xenopus embryogenesis may depend upon the action of hormones and growth factors in the nucleus (Burwen and Jones, 1987; Logan, 1990). This could, for example, directly or indirectly involve changes in DNA methylation or changes in chromatin condensation and spatio-temporal co-ordination of transcription patterns of regulatory genes. The data presented here suggest that at least some FGFs are prepositioned within the egg in association with yolk in the vegetal hemisphere. They retain this distribution during cleavage and shift subcellular compartments from cytoplasm to nucleus in cells of the marginal zone at the time of mesoderm induction. The proximal cause that signals this translocation as well as the molecular mechanisms by which it is accomplished are unknown. One explanation may be that FGFs are bound to carrier molecules, possibly cytoplasmic FGF receptors, containing a nuclear localization sequence (Lee et al. 1989; Silver, 1991) for selective transport through pore complexes in the nuclear envelope.

In conclusion, the possibility that FGFs may have direct actions on the genome during embryonic induction is certainly worthy of further investigation. Intranuclear species of bFGF have recently been identified in cell lines derived from bovine endothelial cells (Bouche et al. 1987; Baldin et al. 1990), bovine cardiac muscle cells (Kardami and Fandrich, 1989), a human hepatoma cell line (Renko et al. 1990), and mouse, monkey, and hamster cell lines transfected with bFGF cDNAs (Renko et al. 1990; Bugler et al. 1991; Tessler and Neufeld, 1990). Likewise, aFGF, which has a nuclear localization sequence in its amino-terminal region (Imamura et al. 1990), has been found in nuclei of human mesenchymal cells (Sano et al. 1990). Along with cells from early Xenopus embryos, these differentiated mammalian cell types may serve as instructive models for molecular analysis. Finally, the methodology described is suitable for investigating the spatial localization of other polypeptide growth factors, such as activins, that are also under consideration as natural morphogens, so it offers a general approach for adding new pieces to the longstanding puzzle of mesoderm induction (Chuang, 1939).

We are very grateful to Dr J. M. W. Slack for recombinant Xenopus bFGF and for anti-Xenopus bFGF and anti-bovine bFGF (peptide 16–30) antibodies. We also thank Dr A. Baird for the generous gift of anti-bovine bFGF (peptide 1–24) antibodies, Dr M. Furusawa for helpful comments on the manuscript and Dr R. C. Strohman for valuable discussion. This work was funded by the Science and Technology Agency (STA) through an ERATO Fellowship to S.F.G. and by an STA fellowship to R.A.S. N.J. was a visiting scholar of the scientific exchange program between the Shanghai Institute of Biochemistry, Academia S ínica and the Tsukuba Life Science Center (RIKEN).