Localized synthesis of the Vg1 protein during early Xenopus development

Formation of mesoderm in the marginal zone of the early amphibian embryo is thought to involve inductive interactions whereby mesoderm is induced in the ani-mal hemisphere by signals emanating from the vegetal hemisphere (Dale & Slack, 1987; Gurdon et al. 1985; Nakamura et al. 1971; Nieuwkoop, 1969; Smith et al. 1985). Regional specification of the mesoderm into dorsal mesoderm, characterized by notochord and muscle and into ventral mesoderm characterized, by kidney and blood, is thought to arise from differences in the signals produced by dorsal and ventral vegetal blastomeres, respectively (Boterenbrood & Nieuw-koop, 1973; Dale & Slack, 1987; Dale et al. 1985). The signals that mediate mesoderm induction have been shown to be diffusible factors in trans-filter experiments (Grunz & Tacke, 1986; Gurdon, 1989; Slack et al. 1987). Recently, it has been determined that soluble growth factors can induce mesoderm in isolated animal pole explants (Godsave et al. 1988; Grunz et al. 1988; Kimelman & Kirschner, 1987; Rosa et al. 1988; Slack et al. 1987; Smith, 1987).

Basic fibroblast growth factor (bFGF), from mam-malian sources, has previously been shown to induce mainly ventral mesoderm in animal pole explants (Grunz et al. 1988; Kimelman & Kirschner, 1987; Slack et al. 1987). More recently, Xenopus bFGF has also been shown to have comparable effects (Kimelman et al. 1988; Slack & Isaacs, 1989). More dorsal mesoderm, such as notochord, is rarely found in animal pole explants treated with bFGF, and so it is probable that another molecule, alone or together with bFGF, is responsible for inducing dorsal mesoderm. A clue to potential nature of the dorsal inducing signal was given by the observation that human transforming growth factor-β1 (TGF-β1) can act synergistically with bFGF to give increased levels of muscle in animal pole explants (Kimelman & Kirschner, 1987). Furthermore, a closely related molecule, TGF-β2 (but not TGF-β1), can alone induce mesoderm as judged by muscle actin expression (Rosa et al. 1988). Thus a combination of bFGF and TGF-β may be sufficient to induce both dorsal and ventral mesoderm.

A number of TGF-β-like molecules have been ident-ified in Xenopus. XTC-MIF (Xenopus tissue culture-mesoderm-inducing factor) is a factor isolated from a Xenopus tissue culture cell line that induces large amounts of dorsal mesoderm (Smith, 1987; Smith et al. 1988). Interestingly, antibodies to TGF-β2 but not TGF-β1 can reduce the efficacy of XTC-MIF action suggesting that XTC-MIF is a TGF-β2-like molecule (Rosa et al. 1988). A novel TGF-β-related mRNA, called TGF-β5, has recently been identified in Xenopus embryos (Kondaiah et al. 1989). TGF-β5 is highly expressed at the onset of neurulation and is abundantly expressed by the cell line that secretes XTC-MIF. In a screen for mRNAs that have spatially restricted distri-butions within the egg, another mRNA related to TGF-β was isolated (Rebagliati et al. 1985; Weeks & Melton, 1987). This mRNA, called Vgl, becomes localized during oogenesis to form a tight subcortical shell in the vegetal hemisphere of stage VI oocytes (Melton, 1987a). Upon egg maturation and subsequent fertiliz-ation, this tight shell of Vgl mRNA is released such that all the vegetal blastomeres in the cleaving egg inherit Vgl mRNA (Rebagliati et al. 1985; Weeks & Melton, 1987).

The greatest similarity between Vgl and the TGF-β family lies in the last 120 carboxyl terminal amino acids, and includes 7 conserved cysteines which are presum-ably used for dimer formation. A pair of basic amino acids which are recognized as a cleavage site for the release of a small secreted peptide in the TGF-βs are also conserved in the putative Vgl protein (see Massagué, 1987; Rizzino, 1988; Sporn et al. 1987 for reviews of TGF-β structure and function). The putative Vgl protein has both a TV-terminal hydrophobic signal se-quence for insertion into the endoplasmic reticulum and three putative N-linked glycosylation sites suggesting that it may be secreted.

The role of the Vgl protein in early development is unknown, but two observations suggest Vgl is a candi-date for a natural mesoderm inducer. First, the putative Vgl protein bears great similarity to molecules that are known to be active in mesoderm-inducing assays, and secondly, Vgl mRNA is localized within the cells of the early embryo that emits inductive signals. Therefore, we have proposed that the Vgl protein may be secreted, by vegetal cells to induce animal cells to form meso-derm (Weeks & Melton, 1987). In this paper, we show that Vgl is a glycoprotein synthesized both in oocytes and in the vegetal blastomeres of early embryos. We also show that a small Vgl peptide (M_r= ∼17x10³) can be cleaved from the full-length Vgl protein (M_r=∼43x10³). Additionally, we show that the full-length Vgl protein can be secreted in vitro. These results are consistent with the proposed function of the Vgl protein as a mesoderm-inducing signal within the early amphibian embryo.

Xenopus embryos were obtained as described in Krieg & Melton (1985). All embryonic stages were according to Nieuwkoop and Faber (1967). Embryos were dissected in lx MMR (Newport & Kirschner, 1982) on top of a layer of 1 % agarose using watchmakers forceps. To label proteins in developing embryos, [³⁵S]-Trans-label (ICN) was concen-trated 5-to 10-fold before microinjection. Synthetic mRNAs were dissolved at a concentration of 200 ngμ1^-1 in sterile water before microinjection (20 nl per egg). Embryos injected with mRNA 30 min before the first cleavage furrow formed were labelled for protein by a second injection of concen-trated Trans-label 20-30 min later. Oocytes were staged as previously described (Dumont, 1972). Oocyte proteins were labelled essentially as before (Harvey et al. 1986; Melton, 1987b) except that Trans-label was substituted for [³⁵S]-methionine. Oocytes were placed in 96-well microtitre plates coated with I mg ml^-1 BSA to prevent the non-specific absorption of secreted proteins.

A chimeric Vgl gene, consisting of Vgl cDNA sequences from the BamHl site at position +81 to the BstEII at position + 1106 and genomic sequences from the Pstl site at position +6 to the BamHl site at position +81 (see Weeks & Melton, 1987 for numbering), was cloned into the transcription vector pSP64T (Krieg & Melton, 1984). After linearization, at the appropriate restriction site, synthetic capped mRNAs were generated as in Krieg & Melton (1984). Enzymes and other reagents were supplied by New England Biolabs, Promega Biotec and Amersham International.

Non-yolk Xenopus proteins were isolated essentially as de-scribed previously (Lane et al. 1979; Melton, 1987b). SDS-polyacrylamide gel electrophoresis, fluorography and other general methods were carried out as described in Goding (1986) and Harlow & Lane (1988).

The Alul to Rsal fragment of the Vgl coding region (Weeks & Melton, 1987) encompassing the region related to TGF-β was cloned into the T7 expression vector pET3a and transformed into the bacterial strain pLysS (Rosenberg et al. 1987). Cultures of transformed bacteria were grown and induced as suggested (Rosenberg et al. 1987). Bacteria were pelleted by centrifugation and boiled in SDS-sample buffer before load-ing on to a preparative SDS-polyacrylamide gel. After electrophoresis, proteins were visualized with cold 4 m-sodium acetate (Higgins & Dahmus, 1979) and the recombinant Vgl protein band was excised. The acrylamide gel slice was macerated in 10 vol. of 50mm-Tris pH8·0, 150mm-NaCl, 5mm-DTT, 0·lmm-EDTA, 0·1% SDS and protein eluted overnight at room temperature. Elutate was removed and proteins precipitated overnight at —20°C by addition of 5 vol. of cold acetone. Recombinant Vgl protein was pelleted by centrifugation, dried and dissolved in PBS.

To generate rabbit polyclonal antibodies against Vgl protein, 100pg of the recombinant Vgl protein, made as above, was mixed with an equal volume of complete Freund’s adjuvant and injected directly into popliteal lymph nodes (Sigel et al. 1983). 10 and 14 weeks later the rabbits were boosted with a subcutaneous injection of recombinant Vgl protein (100 μg) in Freund’s incomplete adjuvant. Immune sera was collected 1 week after each boost. The IgG fraction was purified by ammonium sulfate precipitation and DEAE Affi-Gel blue (Bio-Rad) column chromatography as described by the manu-facturer.

Monoclonal antibodies were prepared from RBF/DNJ mice (Jackson laboratory) previously immunized subcu-taneously with the recombinant Vgl protein (50 pg) in Freund’s adjuvant. Immune spleen cells were fused with the FoxNY myeloma cell line (Taggart & Samloff, 1983) using standard polyethylene glycol-mediated fusion methods (Goding, 1986; Harlow & Lane, 1988), and hybridomas were selected for in Optimem media (Gibco) with adenine, amino-pterin and thymidine (Taggart & Samloff, 1983). Positive hybrids were identified by ELISA assay (Engvall, 1980) using Immulon II assay plates (Dynatech Laboratories Inc.) coated with 25 ng per well of recombinant fusion protein. An anti-mouse IgG antibody conjugated with alkaline phosphatase (Cappell) was used as the secondary antibody and after washing three times with 150mm-NaCl, 50mm-Tris pH 7·6. positive wells were identified by development with 1 mg ml^- para-nitrophenol phosphate in 1mm-ZnCl₂, lmm-MgC1₂, 0·lM-glycine pH 10·4. Positive cultures were cloned at least two times by limiting dilution. Peritoneal tumors were induced by injection of 10⁶ hybridoma cells into BALB/c mice previously primed with pristane 1 week beforehand. Mice were also gamma irradiated 2 days before injection of the hybridoma cells to encourage formation of the tumor. After 1-2 weeks ascites fluid was taken from the mice and purified by ammonium sulfate precipitation and passage over protein-A sepharose as described in Harlow & Lane (1988).

Protein samples were separated on 10 or 15 % gels and transferred to nitrocellulose sheets (Towbin et al. 1979). Blots were reversibly stained with 3 % Ponceau S (Sigma) in 0·5 % TCA and the required regions of the blot cut out. Blots were blocked in 1 % low fat milk powder in 150mm-NaCl, 50-mm-Tris pH 7·6 for 2h before incubation with primary antibody solution (10-50 μg ml^-1) in the same solution for 2h. After washing in 150mm-NaCl, 50mm-Tris pH 7·6, antibody binding was visualized with Vector Laboratories alkaline phosphatase ABC kit using NBT/BCIP substrates (Promega Biotec).

Immunoprecipitation techniques were modified from Lee et al. (1984). Labelled protein samples were diluted to 1 ml with cold lx IP buffer (150mm-NaCl, 1% NP40, 0·5% sodium deoxycholate, 0·1% SDS, 50mm-Tris, pH7·6) and antibody (2 μl of polyclonal serum or 1·2 μg of monoclonal antibody) added and left overnight at 4°C. 25μ1 of protein-A sepharose beads (which had been preabsorbed to an oocyte protein extract) were added and rocked gently for lh at 4 °C. Beads were pelleted and then washed three times (5-10 min each wash) in lx IP buffer. After removal of all the liquid, an equal volume of SDS-sample buffer was added to the beads and samples boiled for 5 min before loading directly on to protein gels. For analysis of glycosylation, beads were washed an additional three times with 150mm-NaCl, 50mm-Tris pH 7·6 before resuspending in an equal volume of 100 mm-sodium acetate pH5-2, 40ITIm-EDTA, 0·4% Triton X-100, 0·5mm-PMSF. 5mU of endoglycosidase H (Boehringer Mannheim) or 250 mU of endoglycosidase F (Boehringer Mannheim) were then added and incubated overnight at 37 °C. Samples were boiled in SDS sample buffer before loading onto gels as described above.

Synthetic mRNAs were translated in rabbit reticulocyte lysates as described by the manufacturer (Promega Biotec) except that Trans-label (ICN) was used as the radioactive label. In vitro processing was carried out with canine pancre-atic microsomal membranes (Promega Biotec) as suggested by the manufacturer. Some samples were treated with 100 μg ml^-1 proteinase K (Boehringer Mannheim) on ice for lh in either the presence or absence 0·1% Triton X-100. Proteinase K was inactivated by the addition of PMSF to ImM. Samples were immunoprecipitated in lx IP buffer as described above. Determination of the subcellular location of the translated products in the membranes was carried out essentially as described by Teixidd et al. (1987).

The complete amino acid sequence of the predicted translation product of the vegetally localized Vgl mRNA has been deduced from its DNA sequence (Weeks & Melton, 1987; A. Fields, H. O’Keefe and D. Tannahill unpublished results). These data showed that the Vgl protein is related to the TGF-β family of secreted growth factors. From the predicted protein sequence, it was anticipated that Vgl protein would be processed from a full-length form of M_r = ∼42·l x 10³ to a M_T= —13·1 x10³ secreted form by cleavage at a pair of basic amino acids. This data also suggested that the Vgl protein could be glycosylated and could undergo dimer formation either with itself or with another, as yet uncharacterized, TGF-β family member (see Fig. 1 for predicted protein structure of Vgl). Given that the small secreted peptide of TGF-β is responsible for its biological effects in other systems (Massagué, 1987; Rizzino, 1988; Sporn et al. 1987) we decided this would be the most appropriate region of the Vgl protein with which to raise antibodies. Therefore, both polyclonal and monoclonal antibodies were made against the last 120 amino acids of the Vgl protein expressed in the pET bacterial expression system (Rosenberg et al. 1987).

Western blotting experiments show that the resulting antisera specifically react only with bacterially ex-pressed Vgl protein and not with an unrelated protein (N-CAM) expressed in the same vector (Fig. 2A and B). Another band is often seen migrating at the predicted size for a dimer of bacterial Vgl protein (Fig. 2B). A further proof of the specificity of the antisera is given by the observation that a protein of M_r = ∼35x 10^s can be immunoprecipitated from in vitro translation reactions directed with synthetic full-length Vgl mRNA using polyclonal anti-Vgl antisera but not with control preimmune antisera (Fig. 2C). The Vgl protein made in vitro has a faster mobility in SDS-PAGE compared to the relative molecular mass predicted from the Vgl amino acid sequence. This does not arise through premature termination of translation in vitro since the deglycosylated in vivo form of Vgl protein also has this faster mobility (see Fig. 5). Four independent polyclonal antisera and three independent monoclonal antibodies against the Vgl protein behave similarly in the above assays (data not shown).

To determine whether the polyclonal and monoclonal antibodies recognize embryonically synthesized Vgl protein, fertilized eggs were injected with [³⁵S]-Trans-label and allowed to develop until stage 8. Proteins labelled within this period were isolated and analyzed by immunoprecipitation techniques. Fig. 3A shows that a protein of M_r = ∼43x10³ can be immunoprecipitated from labelled blastula extracts by both the monoclonal and polyclonal anti-Vgl antibodies (lanes 1 and 2) but not with preimmune sera (lane 3). Fig. 3B shows that the protein immunoprecipitated by the anti-Vgl anti-body can be competed with recombinant Vgl protein (lane 1). Note that the samples in Fig. 3A and B do not resolve a doublet of Vgl protein that can often be seen on lower percentage gels (see Figs 3C and 5).

By analogy to TGF-β we expect to find a smaller peptide of M_r = —13-lx 10^s to be cleaved from the full-length Vgl protein (see Fig. 1), however, we are usually unable to detect such a product by immunoprecipitation (Fig. 3B). We sometimes observe a very weak band at M_r = ∼17x 10³ which may represent the cleaved form of Vgl protein (see the results of mRNA injection de-scribed experiments below). above analyze the Vgl protein that is synthesized embryonically, i.e. made after fertilization. By per-forming western blot experiments on unlabeled pro-teins we have analyzed total Vgl protein (both matern-ally inherited and embryonically synthesized). Fig. 3C shows a Western blot of early Xenopus blastula proteins probed with an anti-Vgl monoclonal antibody (the polyclonal anti-Vgl antibodies react only weakly in this assay). The blot was split into two and the left half was probed with anti-Vgl antibody and the right half with anti-actin antibody. This shows that the anti-Vgl anti-body recognizes two bands of M_r = ∼42x10³ and 44X10³. Both of these are close to the predicted size of the Vgl protein and suggests a post-translational modi-fication of the Vgl protein given that the in vitro translated Vgl protein is some M_r= — 8X10³ smaller in size (Figs 2C and 4). The anti-Vgl antibody does not react with actin which is an abundant protein of a similar molecular weight.

To estimate the amount of Vgl protein in early blastula, known amounts of recombinant Vgl protein and early blastula proteins were compared by Western blotting using an anti-Vgl monoclonal antibody. We estimate that a single early blastula contains approxi-mately 0-5-1-Ong of full-length Vgl protein (data not shown).

To study how Vgl protein is processed, synthetic Vgl mRNA was translated in vitro in either the presence or absence of canine microsomal membranes, and the products analyzed by immunoprecipitation. The results in Fig. 4 show that a protein of =—35 x 10³ is translated from Vgl mRNA in the absence of micro-somes (lane 2). This is about M_r= 8X10³ smaller than the predicted relative molecular mass of Vgl protein, therefore full-length Vgl protein has a faster than expected mobility on SDS-PAGE. A protein of M_r = ∼31x10³ is also translated in vitro from Vgl mRNA (Fig 4, lane 2). This protein arises from down-stream initiation of translation that occurs in vitro (data not shown). Translation of Vgl mRNA in the presence of microsomes results in a number of slower mobility forms which indicate processing (e.g. glycosylation) of the Vgl protein (lane 4). These slower mobility forms of Vgl protein are secreted into the lumen of the membranes as shown by their resistance to proteinase K degradation (lane 5) which is abolished by the addition of detergent (lane 6). Endoglycosidase H (endo H) treatment of Vgl protein translated in the presence of microsomes converts it to a protein of relative molecu-lar mass similar to that of the unprocessed Vgl protein (lane 11). These data show that Vgl is glycosylated at Ñ-linked sites.

The results described above do not distinguish whether Vgl is an integral membrane protein or a secreted protein. To address this issue the Vgl protein translated in the presence of microsomes was treated with or without alkaline bicarbonate and the membrane and cytosol fractions were assayed for Vgl protein. Alkaline bicarbonate treatment opens microsomes and allows release of their contents into the cytosol (Teixidó et al. 1987), thus secreted Vgl should now be found in the cytosol. If Vgl was a membrane protein it should remain with the membrane fraction regardless of alka-line bicarbonate treatment. As shown in Fig. 4 (lanes 7 and 8), the Vgl protein translated in the presence of microsomes is found entirely in the cytosol fraction after treatment with alkaline bicarbonate buffer. This shows that glycosylated Vgl protein is secreted and has presumably undergone signal sequence removal. Frac-tionation in the absence of alkaline bicarbonate treat-ment results in the association of the glycosylated forms of Vgl with the membranes (lane 10), while the unglycosylated forms remain in the cytosol (lane 9) as expected. (Note that some mRNA is translated on free ribosomes as well as membrane-bound ribosomes in this system.) These experiments show that Vgl is a glycoprotein that can be secreted into the lumen of the endoplasmic reticulum.

Although the above experiments show glycosylation and secretion of Vgl protein in vitro, this does not provide direct evidence on how the Vgl protein is processed in vivo. To characterize the Vgl protein in embryos, we analyzed the endogenous Vgl protein and also the Vgl protein made in embryos injected with synthetic Vgl mRNA. Vgl protein was immunoprecipi-tated from stage 8 embryos previously injected with [³⁵S]-Trans-label and the immunoprecipitates treated with endo F or endo H. Endo H cleaves high-mannose glycans from TV-linked glycoproteins whereas endo F cleaves both high-mannose and complex glycans from N-linked glycoproteins (Elder & Alexander, 1982). As shown in Fig. 5, endo F (lane 2) and endo H (lane 3) treatment converts the endogenous Vgl protein to faster mobility forms relative to the untreated sample (lane 1). This proves that embryonically synthesized Vgl protein has TV-linked glycosylation. A further proof of the in vivo glycosylation of Vgl protein is given by the observation that Vgl protein made in embryos injected with Vgl mRNA (lane 4) also shows a compar-able sensitivity to endo F (lane 5) and endo H (lane 6). Vgl protein made in embryos from injected mRNA is therefore processed similarly to endogenous Vgl pro-tein. Maternal Vgl protein, detected by Western blot-ting, is also likely to be glycosylated since it has the same relative molecular mass as the embryonically synthesized Vgl protein.

Since Vgl protein is glycosylated in vivo, it is likely that it is present in the secretory pathway. Fig. 5 shows the result of an experiment where labelled early blas-tula proteins are separated into membrane and cytosol fractions and assayed for Vgl protein by immuno-precipitation. As shown, there is Vgl protein present in the membrane fraction (lane 7). Again, when Vgl mRNA is injected into embryos, the resulting Vgl protein behaves in a manner identical to that of the endogenous protein. The presence of some Vgl protein in the cytosol fraction may be due to rupturing of the membranes during isolation. Alternatively, it is poss-ible the Vgl protein has been secreted into the extra-cellular space. The faster mobility form detected in these experiments is enriched in the cytosol fraction relative to the membrane fraction and perhaps rep-resents an extracellular form of the protein. Overall, these results confirm that the endogenous Vgl protein is a glycoprotein and that it is associated with cell mem-branes, presumably the endoplasmic reticulum.

The experiments described above support the hy-pothesis that full-length Vgl protein (M_r= -42·1 x10³) can be secreted from a cell, but as shown in Fig. 3 we do not observe significant amounts of a small cleaved Vgl protein (M_r = —13·1X10³). Possibly, we cannot easily detect the small Vgl protein because it is only made at very low levels relative to its large precursor. There-fore, we increased the total amount of Vgl protein made by injecting synthetic Vgl mRNA into fertilized eggs. Injected eggs were allowed to develop until the gastrula stage before analyzing for Vgl protein by Western blotting. Fig. 6 shows that there is at least a 10-fold increase in the amount of full-length Vgl protein (M_T = ∼43x10³) in Vgl mRNA injected embryos. A small protein of M_r= ∼17x 10³ is also detected at a low level in embryos injected with Vgl mRNA and thus arises by cleavage of the full-length Vgl protein. Immunoprecipitation experiments from many unin-jected embryos have also shown that a protein of M_r = —17X10³ can be sometimes detected at extremely low levels, which is difficult to photograph (data not shown). Since the small Vgl protein has the potential for TV-linked glycosylation, this might explain the size discrepancy between the predicted small Vgl protein (M_r= —13X10³) and the small Vgl protein (M_T = —17X10³) seen in injected embryos. Consistent with this is an observation that a protein of M_r= ∼17x10³ that is immunoprecipitated from Vgl injected embryos can be converted to a smaller protein of Al_r = ∼13x10³ by endo H treatment (data not shown).

In embryos injected with Vgl mRNA, the amount of the small Vgl protein (M_r = ∼17x10³) detected by Western blotting is at least 20 times less abundant relative to the full-length Vgl protein (M_r= —43X10³). By analogy, if only 5 % of the full-length endogenous Vgl protein were processed to the small product then it would be beyond the level of detection by Western blot analysis (which is around 0-4ng in our hands).

Oocytes are known to be efficient secretory cells; therefore, we decided to examine whether oocytes injected with Vgl mRNA could secrete Vgl protein. Oocytes injected with Vgl mRNA were labelled over-night before fractionation of the proteins and analysis by immunoprecipitation. Fig. 6B shows that a large amount of Vgl protein (M_r= ∼43x10³) is found in both the membrane fraction (M) and cytosol fraction (C) of oocytes injected with Vgl mRNA; however, no Vgl protein is found in the culture media (S). These results are obtained whether or not oocytes have had their follicle cells removed (data not shown). The lack of secreted Vgl protein in the oocyte culture media is not due to inadequate recovery of material from the media since we can easily obtain secreted interferon from the media of oocytes injected with interferon mRNA (Fig. 6C). Hence these results indicate that an oocyte does not secrete Vgl protein.

To examine the expression of the Vgl protein during oogenesis and early development, a monoclonal anti-Vgl antibody was used to probe Western blots of different developmental stages (Fig. 7). The synthesis of Vgl protein is first detected at low levels in stage IV oocytes (this does not reproduce well in the photo-graph) and then increases throughout oogenesis and maturation. Vgl protein cannot be found before stage IV of oogenesis, even when more oocyte equivalents are loaded on the gel (data not shown). Therefore, it appears that the majority of Vgl protein is synthesized during oogenesis after the mRNA has become localized to the vegetal pole (Melton, 1987a; Yisraeli & Melton, 1988).

Fig. 7B shows the expression of total (maternally inherited plus embryonically synthesized) Vgl protein during early embryonic development. From the newly fertilized egg until the early gastrula, there is a slight increase in the amount of Vgl protein present. From neurula to swimming tadpole, there is a significant decrease in protein detected with the anti-Vgl anti-body. Given the observation that the majority of Vgl mRNA is degraded during gastrulation (Rebagliati et al. 1985), this suggests that Vgl protein is rather stable.

To examine when Vgl protein is synthesized during embryogenesis, embryos were injected with label at different developmental stages and allowed to develop for 5h before analyzing for Vgl protein by immuno-precipitation. As shown in Fig. 7C, the majority of Vgl protein is made prior to the gastrula stage. Each consecutive labelling period has less Vgl protein than the previous one, and no Vgl protein appears to be made after neurulation. These results are consistent with the fact that most of the Vgl mRNA is degraded by the late gastrula stage (Rebagliati et al. 1985).

One important feature of mesoderm induction is that the signal(s) for induction emanate from the vegetal hemisphere of the egg. The results presented in Fig. 8 show that Vgl protein is localized in the vegetal hemisphere of eggs and blastula. Embryos were injected with label at the 1-cell stage and allowed to develop until either the 64-cell stage or until a stage-8 blastula before dissection into animal and vegetal halves and subsequent analysis for Vgl protein by immunoprecipitation (Fig. 8A, left-hand 6 lanes). The Vgl protein made between the 64-cell stage and stage 8 was similarly analyzed, by injecting label into the 64-cell embryo (right-hand 3 lanes). As shown, the embry-onically synthesized Vgl protein made during both early and later cleavage is confined to the vegetal pole. These results are in keeping with the vegetal localiz-ation of Vgl mRNA (Rebagliati et al. 1985; Weeks & Melton, 1987).

Although the embryonically synthesized Vgl protein is localized to the vegetal hemisphere, it is possible that maternally inherited Vgl protein has a uniform distri-bution within the embryo. Fig. 8B shows a Western blot assaying the distribution of total Vgl protein in the fertilized egg, 64-cell blastula and the stage-8 blastula. Two Vgl protein bands are detected: the more abun-dant Vgl protein (M_r= ∼44x10³) is localized to the vegetal pole in all these stages, and the less abundant Vgl protein (M_r= ∼42x10³) is more uniformly distrib-uted. Also shown below is an identical blot probed with an anti-actin antibody, showing that there are approxi-mately similar amounts of protein in each of the dissected embryo halves. The uniformly distributed form of Vgl protein (M_r= ∼42x10³), present in the animal hemisphere, is likely to be maternal in origin since embryonically synthesized Vgl protein is found only in the vegetal pole. It is possible that only one these large Vgl proteins can be processed and/or secreted, but at present, we have no data on differential processing.

We have previously proposed that the Vgl protein is a natural mesoderm-inducing molecule since the mRNA is localized to the vegetal pole and encodes a protein with sequence similarity to known inducers (Weeks & Melton, 1987). In this paper, we show that the Vgl protein is synthesized before gastrulation only in the vegetal hemisphere of the early blastula. In addition, we have shown that the Vgl protein is glycosylated and associated with the membrane fraction of embryos and that it can be secreted into the lumen of pancreatic microsomes. Hence, Vgl protein is synthesized both at the time when embryos are known to be competent for mesoderm induction (Dale et al. 1985; Gurdon et al. 1985; Jones & Woodland, 1987; Nakamura et al. 1971), and in the place where induction signals are known to arise (Gurdon, 1989; Gurdon et al. 1985; Nakamura et al. 1971; Nieuwkoop, 1969; Smith et al. 1985). These results are consistent with, but do not prove, the hypothesis that Vgl is indeed an endogenous meso-derm-inducing factor.

By analogy to TGF-β, a small secreted form of Vgl protein is likely to be the active molecule. Currently, we have only indirect evidence that the endogenous Vgl protein is processed to yield a small protein (Af_r = ∼17x10³) that is secreted. Our in vitro studies on glycosylation and secretion suggest that the endogenous protein will be similarly processed. In addition, Vgl mRNA injections provide strong evidence for cleavage of the full-length Vgl protein into a small ‘growth factor’-size molecule. It is possible that we do not reproducibly detect a small secreted form of the en-dogenous Vgl protein because it is generated in low amounts and/or is rapidly turned over. However, we do see a low amount of cleaved Vgl protein (M_r = ∼17x10³) in embryos injected with Vgl mRNA, although we do not know if it is secreted. If the amount of non-yolk protein in a blastula is taken as 25 μg and the volume as 1-0 μI (Gurdon & Wickens, 1983), this means the endogenous large Vgl protein (M_r = ∼43 x 10³) is present at a concentration of 0·5-1·0 μg ml^-1 (10-20 nM) within a blastula. However, if a small secreted Vgl protein was produced at a level 10-to 20-fold times less than its large precursor, then the concentration of a small secreted Vgl protein might be similar to the concentrations at which other growth factors have effects on Xenopus embryos. In all, our data highlight the fact that the expression of the gene is regulated in several ways: localization of mRNA and protein, cleavage and glycosylation of full-length pro-tein and secretion.

While we have tested for glycosylation and secretion of the Vgl protein, it is possible that other post-translational controls exist in generating active Vgl product. For example, it is known that TGF-βs are made in latent forms which can be activated by protease treatment or low pH (Lyons et al. 1988; Miyazono et al. 1988; Wakefield et al. 1988). Recently, it has also been shown that endo F can also activate TGF-β1 (Miyazono & Heldin, 1989). It is thus interesting that Vgl is sensitive to both endo F and endo H. Similarities to the TGF-β family suggest that Vgl would function as a dimer, therefore the formation of either homodimers or heterodimers to yield active Vgl protein may also be regulated.

Vgl protein is first detected in midstage oocytes but the maximal levels are found in larger oocytes. There-fore the localization of Vgl mRNA into a tight subcorti-cal shell (Melton, 1987a; Yisraeli & Melton, 1988) does not interfere with the translation of Vgl protein. Injected Vgl mRNA is also translated in oocytes, but none of the Vgl protein synthesized is secreted. Since oocytes can secrete a large number of proteins, this suggests a potentially interesting control mechanism whereby Vgl protein secretion is tightly regulated.

During early development we can detect synthesis of new Vgl protein until the gastrula stage. This embry-onically synthesized Vgl protein is found only in the vegetal hemisphere of blastula. Total (maternally in-herited plus embryonically synthesized) Vgl protein is also largely confined to the vegetal hemisphere of blastula although a maternally inherited component (M_r = ∼42x10³) is also present in the animal hemi-sphere. While the Vgl protein is present in the early embryo at a time and place consistent with acting in mesoderm induction, it is noted that the Vgl protein persists long after mesoderm induction is complete. Presently, we do not have any data that suggest a function for this persistent portion of the Vgl gene product.

Recently, Dale et al. (1989) have independently characterized the Vgl protein by immunoprecipitation using anti-Vgl antibodies. Although essentially similar, our results differ from Dale et al. in several respects. Since our monoclonal anti-Vgl antibodies work in Western blots, this has allowed us to show that Vgl protein is rather stable throughout development. More importantly, we have provided evidence that the Vgl protein is processed into a small peptide (M_r = ∼17x10³) by overexpressing Vgl in embryos. Finally, Fig. 8B shows that the egg inherits some of the faster mobility form of Vgl protein (M_r= ∼42x10³) in the animal pole. Considering the observation of Dale et al. that some Vgl protein can diffuse into the animal pole of oocytes, this suggests that diffusion of the larger Vgl protein (M_r = ∼44x10³, see Fig. 8B) into the animal pole of eggs is restricted or that it is relatively unstable in the animal pole.

The dorsal mesoderm-inducing signal may involve a molecule related to TGF-β, given that TGF-β1 acts synergistically with bFGF to give large amounts of muscle (Kimelman & Kirschner, 1987) and given that the TGF-β-like factor XTC-MIF can induce dorsal mesoderm (Smith, 1987). The dorsal region of the vegetal hemisphere releases the dorsal-inducing sig-nals) (Boterenbrood & Nieuwkoop, 1973; Dale & Slack, 1987; Dale et al. 1985). Therefore, these obser-vations, together with the vegetal localization of Vgl protein and sequence similarity of Vgl to TGF-β, raise the possibility that Vgl acts alone or together with bFGF to induce dorsal mesoderm. Indeed, preliminary results suggest that Vgl can act synergistically with bFGF to induce mesoderm. Further work is in progress to characterize this effect in more detail. If Vgl is involved in inducing dorsal or dorsal/anterior meso-derm, then the Vgl protein may be synthesized or processed differentially in dorsal versus ventral blasto-meres.

We are greatly indebted to Dan Kiehart and Jim Burkhart for help with antibody production, Chris Kintner for N-CAM expressing plasmid, Jim Lessard for anti-actin monoclonal antibody and Sean Munro for anti-interferon antibody. We thank Andy Fields for technical help. Also we would like to thank Charles Jennings, Kim Mowry, Sean Munro, Heather O’Keefe, Ariel Ruiz i Altaba, Sergei Sokol, Jerry Thomsen, Malcolm Whitman, Tod Woolf and Joel Yisraeli for sugges-tions on this work and comments on the manuscript. This work was supported by a grant from the N.I.H. to D. M. and a SERC/NATO postdoctoral fellowship to D.T.