Heparan sulfate proteoglycans are required for mesoderm formation in Xenopus embryos

Progressive determination of mesodermal tissues involves several distinct steps. Initially, presumptive mesoderm is induced in animal cells by signals produced by vegetal cells of the early blastula. During induction, mesoderm acquires dorsoventral polarity, such that cells that will form dorsal tissues (notochord and muscle) become different from cells destined to develop into ventrolateral tissues (kidney, blood, mesenchyme) (Slack, 1993; Smith, 1993; Gerhart et al., 1989). Somitic, notochordal, head, heart and lateral plate mesoderm primordia segregate during gastrulation (Gerhart and Keller, 1986; Keller, 1991). Subsequently, a major differentiation of mesodermal tissues ensues.

Although the participation of secreted growth factors of the FGF and TGFβ families in mesoderm induction has been well established (see Slack, 1993, and Smith, 1993, for reviews), very little is known about the molecules involved in the control of morphogenetic movements during gastrulation and in later steps of mesoderm formation. Gastrulation, which is critical for establishing the vertebrate body plan, begins with formation of bottle cells at the dorsal blastopore followed by head mesoderm migration along the blastocoel roof and by involution of marginal zone cells. These movements may be driven by differential cell adhesion, such that cells located in different regions of the embryo would express different adhesion molecules and, consequently, behave differently during gastrulation. Despite substantial progress in characterization of cell adhesion ligands and receptors (Hynes and Lander, 1992) and the implication of fibronectin and its receptors in amphibian gastrulation (Keller and Winklbauer, 1990), the molecules that provide regional differences to embryonic cells and regulate gastrulation remain largely unidentified.

Among different classes of cell adhesion molecules, proteoglycans (PGs) have unique properties. First, PGs mediate cellsubstrate and cell-cell adhesion, interacting with a variety of cell surface and ECM molecules, including fibronectin, thrombospondin, collagens, NCAM and integrins (see Toole, 1991; Esko, 1991, for reviews). Second, PGs are implicated in migration and differentiation of a variety of cell types (Ruoslahti, 1988; Esko, 1991; Toole, 1991). For example, inhibitors of PG synthesis block migration of mesenchymal cells during sea urchin gastrulation (Solursh et al., 1986; Lane and Solursh, 1988), interfere with proper cardiac looping in Xenopus (Yost, 1990), and disrupt epithelial-mesenchymal interactions during mouse kidney development (Lelongt et al., 1988). These data suggest that PGs may be important regulators of developmental processes.

The expression of PGs is consistent with their proposed role in morphogenesis (Toole, 1991; Esko, 1991; Rapraeger, 1993). Distinct PGs are expressed transiently and in spatially restricted patterns during development. For example, syndecans are expressed in the developing tooth mesenchyme, kidney and limb bud of mammalian embryos (Vainio et al., 1989a,b; Solursh et al., 1990). In the mouse, mesodermal expression of syndecan 1 is unequal along the anteroposterior axis with stronger expression in the posterior region (Sutherland et al., 1991).

An interesting property of PGs is their ability to interact with various secreted growth factors from the FGF, Wnt and TGFβ families (Burgess and Maciag, 1989; Bradley and Brown, 1990; Massagué, 1992). Such interactions appear to be important in regulating growth factor activity as has been demonstrated for FGF and TGFβ (Yayon et al., 1991; López-Casillas et al., 1993; Yamaguchi et al., 1990). FGF does not bind to its high affinity receptor in mutant CHO cells with an impaired ability to synthesize heparan sulfate chains (Yayon et al., 1991). The TGFβ type III receptor is a proteoglycan known as betaglycan (López-Casillas et al., 1991; Wang et al., 1991). Binding of TGFβ to betaglycan and to another proteoglycan, decorin, is known to modulate the growth factor activity (López-Casillas et al., 1993; Yamaguchi et al., 1990).

Several heparin-binding growth factors from the FGF, activin and Wnt families are implicated in mesoderm formation (Slack, 1993; Smith, 1993), and heparin has been reported to block mesoderm induction in Xenopus (Slack et al., 1987). Although heparin is not normally present in the embryo, these observations suggest that the endogenous heparan sulfate proteoglycans (HSPGs), that contain polysaccharide chains related to heparin, may participate in determination of mesoderm during early development. To test the potential role of PGs in Xenopus embryogenesis, we studied differentiation of embryonic explants from which glycosaminoglycan (GAG) components of PGs were specifically eliminated by enzymatic treatment. We found that removal of heparan sulfates from blastula cells interferes with morphogenetic movements of embryonic explants and selectively inhibits mesodermal differentiation.

Eggs were obtained by injecting Xenopus laevis females with 700 units of human chorionic gonadotropin (Sigma). Fertilization and embryo culture were done as described (Newport and Kirschner, 1982). Staging was according to Nieuwkoop and Faber (1967).

LCM (low calcium medium) was formulated empirically to support the culture of embryonic explants under semi-dissociative conditions. LCM contains 76 mM NaCl, 1.4 mM KCl, 0.2 mM CaCl₂, 0.1 Mm MgCl₂, 0.5 mM Hepes, 1.2 mM sodium phosphate (pH 7.5), 0.6 mM NaHCO₃ and 0.06 mM EDTA. Animal caps were excised from mid-blastulae (stage 8-8.5) and cultured in LCM with or without growth factors and in the presence or absence of heparinase/chondroitinase until stage 11, when the explants were transferred to 0.7× MMR (Newport and Kirschner, 1982) for further culture or lyzed for northern analysis of early mesodermal markers. Marginal zone explants were excised from dorsal and ventral sides of midblastula (stage 8) or early gastrula (stage 10). Embryos with a clear dorsoventral pigment pattern were chosen for dissecting midblastula marginal zones. The position of the dorsoventral axis in the early gastrula was determined by the appearance of the dorsal blastopore. Marginal zone explants were cultured in LCM with or without heparinase until stage 11, when the explants were transferred to 0.7× MMR or collected for RNA preparation. Human recombinant activin A was used at a concentration of 5 ng/ml. Xenopus recombinant bFGF (a gift from D. Kimelman) and human recombinant bFGF (Gibco) were used at concentrations of 100 and 150 ng/ml. Heparinase 3 (a gift from M. Herndon; Herndon and Lander, 1990), recombinant heparinase 1 (a gift from R. Sasisekharan; Sasisekharan et al., 1993) and chondroitinase ABC (Sigma) were used at concentrations 8 μg/ml, 1.25 μg/ml and 0.2 U/ml, respectively (with the exception of the experiments presented in Fig. 3). The effects of both heparinase 3 and purified recombinant heparinase 1 were indistinguishable in our experiments. All experiments were repeated from 3 to 7 times. Enzymatic activity of heparinase 1 and chondroitinase was measured as described (Linker and Hovingh, 1972) using commercial preparations of heparin (Sigma) and chondroitin sulfate B (Sigma). For rescue experiments (Fig. 4A), heparan sulfate (Sigma) was used at concentrations of 0.5 μg/ml and 50 μg/ml.

Capped Xwnt8 mRNA was synthesized in vitro with SP6 polymerase as described (Sokol, 1993). 2-cell embryos were injected in the animal pole with 0.1 ng of Xwnt8 RNA and were cultured until stage 10, at which time animal caps were isolated. The animal caps were treated with heparinase in LCM until stage 12, and were either collected for Northern analysis of early markers or cultured further in 0.7× MMR to assay muscle-specific actin expression at stage 28.

Embryonic explants were fixed in MEMFA (0.1 M Mops, 2 mM EGTA, 1 mM MgSO₄ and 3.7% formaldehyde) (Hemmati-Brivanlou and Harland, 1989) for 1 hour and prepared for histological analysis as described (Sokol et al., 1990).

RNA was extracted from explants and embryos with proteinase K/SDS buffer, fractionated in formaldehyde-agarose gels, transferred to GeneScreen nylon membrane and hybridized to the ³²P-labeled antisense riboprobes as described (Sokol et al., 1990). RNA probes were prepared by in vitro transcription with SP6, T7 or T3 polymerases. Plasmids containing Xenopus goosecoid, brachyury (Xbra), 1A11, Xlim1, Xwnt8 and fibronectin (Blumberg et al., 1991; Smith et al, 1991; Greene et al., 1993; Taira et al., 1992; Christian et al., 1991; Yisraeli et al., 1990) were used to analyze the expression of early markers at stage 11 or stage 12. Probes for Xenopus cardiac actin, NCAM and EF1α (Dworkin-Rastl et al., 1986; Kintner and Melton, 1987; Krieg et al., 1989) were used for northern analysis of explants and embryos at stage 28. After each experiment involving RNA analysis in explants cultured for the same period, the membrane was hybridized with multiple probes. The only exception was Fig. 2B, where Xlim1 expression data were derived from an independent experiment. When necessary, the same blot was stripped by boiling in water for 5 minutes and reprobed.

2-cell-stage Xenopus embryos were injected into each blastomere with 20 nl (0.8 μCi) of sodium [³⁵S]sulfate. When injected embryos reached stage 8, cells from the animal half of the embryo, including marginal zone, were dissociated in Ca, Mg-free NAM (Slack, 1984) and incubated for an additional 5 hours in the absence or in the presence of heparinase, chondroitinase, or both. Cells were then lysed in 50 mM Tris-HCl (pH. 8.0), 0.15 M NaCl, 0.1% Triton X-100. Lysates were loaded onto DEAE-Sephacel (Pharmacia) columns previously equilibrated in 50 mM Tris-HCl (pH. 7.5), 150 mM NaCl, 1 mM EDTA, 0.03% Triton X-100. After extensive washing of the columns with equilibration buffer, PGs were eluted with 1 M NaCl, 0.03% Triton X-100, 30 mg/ml BSA, 5 mg/ml dextran sulfate and precipitated with 2 volumes of ethanol. Samples were redissolved in SDS-PAGE sample buffer and electrophoresed through a 5-15% gradient gel (Harlow and Lane, 1988). The gel was treated with 1 M sodium salicylate for 30 minutes (Chamberlain, 1979), dried and exposed to Kodak X-Omat AR X-ray film at —80°C for 3 days.

When Xenopus blastula ectodermal explants (animal caps) are cultured in a simple salt solution, they differentiate into ciliated epidermis. In contrast, when these explants are treated with activin, they undergo morphogenetic movements and differentiate into a variety of mesodermal and neural tissues (Slack, 1993; Smith, 1993). We found that elongation of animal caps in response to activin was strongly inhibited by heparinase, which removes heparan sulfate chains from PGs, but not by chondroitinase, which eliminates chondroitin sulfate chains (Fig. 1A). Furthermore, histological examination revealed that explants treated with activin alone or with activin and chondroitinase developed notochord, muscle and neural tissue, whereas formation of mesoderm, but not neural tissue, was blocked in explants treated with activin in the presence of heparinase (Fig. 1B).

Inhibition of mesoderm formation was confirmed at the molecular level by northern analysis. Removal of heparan sulfate from activin-treated animal caps strongly inhibited the expression of muscle-specific actin (Dworkin-Rastl et al., 1986) mRNA but did not change the expression of NCAM (Kintner and Melton, 1987), a neural tissue marker (Fig. 2A). This effect was consistently observed with several independent preparations of heparinase 3 and with purified recombinant heparinase 1, which also cleaves heparan sulfate, but with different specificity. These results establish that HSPGs are required for elongation and mesoderm formation in activin-treated animal caps.

To monitor the effects of the enzymes on the PG composition in Xenopus blastula, each blastomere of 2-cell-stage embryos was microinjected with 0.8 μCi of ³⁵S-labeled sodium sulfate. When injected embryos reached midblastula stage, animal hemisphere cells were isolated and treated with heparinase, chondroitinase, or with both enzymes for 5 hours. Proteoglycans were purified from cell lysates on a DEAE-cellulose column and analyzed by SDS-PAGE (Fig. 1C). A broad smear of high molecular mass sulfated PGs was detected in lysates of untreated cells, but not in lysates of cells treated with heparinase. Chondroitinase eliminated only some lower molecular mass PGs. Virtually all PGs are eliminated when both enzymes are applied (Fig. 1C). Thus, our data indicate that the majority of sulfated proteoglycans in Xenopus late blastulae/early gastrulae are HSPGs, as opposed to chondroitin sulfate PGs, which is consistent with earlier reports for Rana pipiens blastulae (Kosher and Searls, 1973) and Xenopus neurulae (Yost, 1990).

Although we demonstrated that removal of heparan sulfates inhibited mesoderm formation in animal cap explants treated with activin, it was unclear which developmental processes required HSPGs. Since PGs can directly modulate binding of FGF and TGFβ to their receptors in tissue culture cells (Yayon et al., 1991; Rapraeger et al., 1991; Nurcombe et al., 1993; López-Casillas et al., 1993), we analyzed whether heparinase affects the early transcriptional response of animal cap cells to activin.

Animal caps were treated with activin alone or with activin in the presence of heparinase or chondroitinase. When control embryos reached stage 11, RNA was extracted from explants for northern analysis. Interestingly, whereas both morphogenetic movements and muscle differentiation were blocked, early mesodermal markers were affected selectively. Expression of a general mesodermal marker, Xbra (Smith et al., 1991), which is activated in the entire marginal zone, was inhibited by heparinase. However, levels of dorsal-specific markers such as goosecoid (Blumberg et al., 1991) and Xlim1 (Taira et al., 1992) mRNAs and levels of Xwnt8 mRNA, a ventrolateral marker (Christian et al., 1991), did not change significantly (Fig. 2B). These findings reveal that activin signals via at least two distinct pathways, which differ in their requirements for HSPGs.

Our finding that activin signaling involves both HSPG-dependent and HSPG-independent pathways led us to study whether heparan sulfates are required for mesoderm formation mediated by FGF (Slack et al., 1987; Kimelman et al., 1988). When added to animal caps, FGF activates transcription of several mesoderm-specific genes, including Xbra, 1A11 and Xwnt8 (Smith et al., 1991; Taira et al., 1992; Christian et al., 1991), but does not induce goosecoid and Xlim1 (Cho et al., 1991; Taira et al., 1992). We observed that expression of both early (Xbra, 1A11) and late (cardiac actin) markers was inhibited by heparinase in animal caps stimulated with FGF (Fig. 2C). Interestingly, while Xwnt8 mRNA levels were not affected by heparinase in animal caps stimulated with activin (Figs 2B, 4B), its expression decreased in heparinase-treated animal caps incubated in the presence of FGF (Fig. 4B and data not shown). These observations indicate that mesodermal marker expression, induced by FGF, depends on the presence of HSPGs, consistent with their proposed role in modulating FGF binding to its high affinity receptor (Yayon et al., 1991; Rapraeger et al., 1991; Nurcombe et al., 1993).

To compare biological effects of heparinase with its enzymatic activity, animal caps were treated with activin in the presence of different amounts of heparinase and the inhibition of meso-dermal marker expression was evaluated (Fig. 3A). The inhibitory effect of heparinase was dose dependent and correlated with its ability to digest heparin, but not chondroitin sulfate (Fig. 3A, B). In contrast, chondroitinase digested chondroitin sulfate, but not heparin (Fig. 3C). These data show that the enzymes used in the present study function in a dosedependent and substrate-specific manner. A small increase in goosecoid expression observed in these experiments may be explained by a decrease in cell-cell adhesiveness in heparinasetreated animal caps (K. I. and S. S., unpublished observations) followed by the more efficient penetration of activin as compared with explants not treated with the enzyme. Moreover, we failed to detect significant protease activities in our preparations of heparinase and mesoderm formation was not restored by addition of protease inhibitors (data not shown). These observations, in combination with the data presented in Fig. 1C, suggest that the enzymatic activity of heparinase is responsible for the observed inhibition of mesoderm formation.

After treatment of tissue culture cells with heparinase, binding of FGF to its high affinity receptor is impaired (Rapraeger et al., 1991). Addition of soluble heparin or heparan sulfate to treated cells was shown to restore FGF binding to cells and its biological activity (Yayon et al., 1991; Rapraeger et al., 1991; Ornitz et al., 1992). To examine whether addition of exogenous heparan sulfate can rescue activin-mediated mesodermal differentiation in heparan sulfate-deficient ectodermal explants, animal caps were excised at stage 8 and pretreated with heparinase for 2 hours. After washing off the enzyme with culture medium, explants were cultured with activin and 0.5 μg/ml (or 50 μg/ml) of heparan sulfate for another 2 hours. When sibling embryos reached stage 28, the explant RNA was extracted for northern analysis. Although heparin has been shown to block mesoderm induction in animal-vegetal conjugates (Slack et al. 1987), doses of heparan sulfate, used in our experiments, did not inhibit response of animal caps to activin (data not shown).

Pretreatment of animal caps with heparinase inhibited muscle actin expression in induced animal caps (Fig. 4A). When heparan sulfate and activin were added to heparinasetreated animal cap explants, the expression of muscle-specific actin was restored to control levels (Fig. 4A). In contrast, soluble chondroitin sulfate failed to rescue induction of muscle actin mRNA by activin (data not shown). These results extend the earlier studies of the tissue culture cell response to FGF (Yayon et al., 1991; Rapraeger et al., 1991) and demonstrate that soluble exogenous heparan sulfate can substitute for HSPGs in mesoderm formation.

To address the question whether heparinase affects muscle and notochord differentiation by interfering with the early steps of mesoderm induction, at the level of growth factor binding to their receptors, animal caps were stimulated with activin or FGF for 2 hours, rinsed briefly with excess of LCM and incubated in the presence of heparinase for another 3 hours. Even when heparinase was applied 2 hours after growth factor treatment, when the growth factor no longer needs to be present in the medium for the complete effect (Smith, 1993); Xbra mRNA levels were still substantially decreased in heparinasetreated explants at the equivalent of stage 11 (Fig. 4B). Interestingly, Xwnt8 expression was inhibited by heparinase in FGF-treated caps, but not in activin-treated caps (Fig. 4B). Furthermore, heparinase inhibited elongation movements (data not shown) and muscle actin mRNA expression in both FGF- and activin-treated explants (Fig. 4B). These results suggest that HSPGs are required for morphogenetic movements and for induction of both early and late mesodermal markers even after the initial stages of mesoderm induction.

Wnts are signaling polypeptides, which can bind heparin (Bradley and Brown, 1990) and are thought to modify effects of mesoderm-inducing factors on embryonic cells (Christian et al., 1992; Sokol and Melton, 1992; Sokol, 1993). When animal caps are isolated at stage 8 from Xwnt8 mRNA-injected embryos, they differentiate into atypical epidermis. However, when these animal caps are isolated at stage 10, they form mesodermal tissues in the absence of exogenous growth factors (Sokol, 1993).

To test whether removal of heparan sulfate alters Wnt effect on mesoderm formation in animal caps (Sokol, 1993), animal hemisphere of 2-cell-stage embryos was injected with Xwnt8 mRNA. When control embryos reached stage 10, animal caps were excised from the Xwnt8 mRNA-injected embryos and incubated in the presence of heparinase. While animal caps isolated from uninjected embryos formed atypical epidermis, the animal caps from Wnt-injected embryos elongated (data not shown) and expressed Xbra and muscle actin RNA (Fig. 5). Heparinase treatment inhibited these morphogenetic movements (data not shown) and levels of Xbra and muscle actin mRNAs (Fig. 5), suggesting that HSPGs may participate in Wnt signaling.

Although the animal cap assay is a useful model system to study mesoderm induction, in normal embryogenesis, meso-dermal tissues originate from the marginal zone (Keller, 1991). Therefore, we decided to test whether HSPGs are needed for differentiation in dorsal marginal zone explants in the absence of exogenous growth factors. We found that, similar to the activin-treated animal caps, both autonomous morphogenetic movements and muscle actin mRNA expression in the early gastrula marginal zone explants were inhibited by heparinase, while NCAM expression remained unaffected (Fig. 6A,C). In these experiments, however, inhibition of morphogenetic movements was not complete, possibly due to limited accessibility of the enzyme to cells in the explant. While animal caps have only two to three layers of cells, the marginal zone explants that we isolated are thicker. Since heparinase affected mesodermal differentiation in the stage 10 explants (Fig. 6C), in which early mesodermal markers have already been activated, HSPGs also function in the later stages of mesoderm formation, after initial interactions of mesodem-inducing factors with their receptors, are over (see Slack, 1993; Smith, 1993, for reviews).

Similar to ectodermal explants induced by activin, Xbra mRNA expression was inhibited in the midblastula (stage 8) explants treated with heparinase, but goosecoid and Xlim-1 (data not shown) mRNA expression did not change when compared to untreated controls (Fig. 6B). Thus, consistent with our data on the animal cap response to mesoderm-inducing factors (Fig. 2B,C), endogenous pathways activating Xbra transcription differ from pathways involved in induction of goosecoid and Xlim-1 RNA by their requirements for HSPGs.

We have shown that removal of heparan sulfate from HSPGs by heparinase treatment inhibits morphogenetic movements and mesoderm formation in Xenopus animal caps treated with growth factors and in the dorsal marginal zone explants in the absence of exogenous signals. In contrast, chondroitinase did not have this effect. Both enzymes exhibited dose-dependent and substrate-specific effects (Fig. 3). Addition of exogenous heparan sulfate to animal caps pretreated with heparinase restored the cell response to activin (Fig. 4A). Neither heparinase nor chondroitinase altered neural tissue differentiation (Figs 1, 2B, 6), suggesting that dorsal mesoderm cells do not lose their capacity to generate neural inducer(s) and that the response of ectodermal cells to neural induction is not changed. Taken together, these observations strongly suggest that HSPGs are involved in gastrulation and mesoderm formation in vivo.

Earlier studies have shown that different sulfate-containing molecules, including heparin (Slack et al., 1987; Mitani, 1989), trypan blue and suramin (Gerhart et al., 1991; Grunz, 1992), inhibit gastrulation and dorsal mesoderm formation and often cause microcephaly (see Gerhart et al., 1991, for review). Although these highly charged compounds may interfere with HSPG function, the molecular mechanism by which they affect morphogenesis remains to be clarified. Yost (1992) has injected heparinase into Xenopus blastocoel and observed altered fibronectin fibril deposition on the blastocoel roof and frequent reversal of left-right asymmetry of heart and gut. In that study the effects of the enzyme on mesoderm formation were not observed, presumably due to poor accessibility of heparinase to marginal zone cells.

In our experiments, selective inhibition of Xbra, an early mesodermal marker, in animal caps treated with activin and heparinase implies that HSPGs are required for the initial steps of mesoderm induction. In contrast, heparinase did not significantly alter expression of Xlim1, goosecoid and Xwnt8 mRNAs (Fig. 2B), which indicates that activin signals through HSPG-dependent and HSPG-independent pathways. These two pathways also operate in vivo, since the same effects of heparinase on the expression of early mesodermal markers were observed in dorsal marginal zone explants, which express these markers in the absence of exogenous growth factors (Fig. 6B).

Previous reports have shown that the response of certain tissue culture cells to FGF depends on the presence of HSPGs (Yayon et al., 1991; Rapraeger et al., 1991). Here we demonstrate that FGFdependent transcriptional activation of Xbra, 1A11 and Xwnt8 in the animal caps (Figs 2C, 4B) are inhibited by heparinase, thus extending the earlier observations on FGF signaling in mammalian cell lines to mesoderm induction in Xenopus. Our finding that two different pathways operate in the marginal zone explants is consistent with the recent report of Amaya et al. (1993) showing that overexpression of dominant negative form of FGF receptor in Xenopus embryos abolishes Xbra and Xpo expression, but does not affect goosecoid expression. Thus, FGF may function in a HSPG-dependent manner to activate Xbra, but is not likely to be involved in induction of goosecoid expression.

The dominant negative form of FGF receptor has recently been shown to interfere with activin signaling (Cornell and Kimelman, 1994; LaBonne and Whitman,1994), which opens the possibility that the effect of heparinase on activin-mediated induction is due to participation of FGF in the same pathway. This hypothesis is supported by the observation that induction of general mesodermal markers is more sensitive to inhibition by the injected dominant negative FGF receptor than activation of dorsal mesoderm-specific markers, Xlim1 and goosecoid (Cornell and Kimelman, 1994). However, Xwnt8 mRNA, a ventrolateral marker, was inhibited by heparinase in FGF-treated caps but not in activin-treated caps (Figs 2B, 4B), suggesting that activin and FGF induce Xwnt8 expression via two pathways, which differ in their requirements for HSPGs. Another finding, which is not consistent with the FGF pathway being stimulated by activin, is that MAP kinase is activated in animal cap cells in response to FGF, but not in response to activin (Graves et al., 1994; LaBonne and Whitman, 1994).

Alternative explanation of our results is that the cell response to activin may be modulated by follistatin, an activinbinding protein. Follistatin is associated with cell surface HSPGs (Nakamura et al., 1991), and follistatin RNA is present in Xenopus embryos (Tashiro et al., 1991). When heparan sulfate chains are eliminated by heparinase, interactions between follistatin and activin may be altered, thus, affecting the response to mesoderm induction.

Besides FGF and activins, Wnt gene products are also implicated in mesoderm formation in Xenopus. Wnt1 and Xwnt8 RNAs have a dorsal axis forming activity when injected into ventral side of Xenopus embryos (Sokol et al., 1991; Smith and Harland, 1991). Although Xwnt8 mRNA does not have a mesoderm-inducing activity on its own, in combination with FGF and activin, it modifies the character of the induced mesoderm from ventral type to dorsal type (Christian et al., 1992; Sokol and Melton, 1992). Also, Xwnt8 can synergize with endogenous signal spreading into animal caps to form mesoderm (Sokol, 1993). Interestingly, it has been shown that Wnt1 strongly binds heparin (Bradley and Brown, 1990). Our data indicate that HSPGs may participate in Wnt signaling pathway to cause elongation and mesoderm formation in animal cap explants cut at stage 10 from Xwnt8-injected embryos (Fig. 5). Although direct functional interaction of Wnt proteins and HSPGs remains to be demonstrated, it is tempting to speculate that Wnts modulate growth factor responsiveness via association with HSPGs.

Following early transcriptional activation of region-specific genes, presumptive mesoderm cells undergo convergent extension movements, which represent a major type of cell rearrangements during Xenopus gastrulation (Keller et al., 1985, Keller, 1991). Elongation of animal caps in response to growth factors mimics convergent extension of the marginal zone cells (Smith, 1993). We show that heparinase inhibits both elongation of animal caps in response to activin (Fig. 1A) and the autonomous morphogenetic movements in the dorsal marginal zone (Fig. 6A). Since removal of heparan sulfate may affect the earliest steps of growth factor signaling, inhibition of cell rearrangements may be a consequence of the heparinase effect on the growth factor binding to its receptor. However, taking into account our results that heparinase is still effective in the early gastrula marginal zone explants (stage 10), after the initial response to induction has already occurred (Fig. 6A,C), HSPGs may directly participate in convergent extension movements.

Thus, our results indicate that HSPGs may operate during early mesoderm induction and in cell rearrangements during gastrulation in Xenopus embryos. Although inhibition of Xbra transcription and blocking morphogenetic movements by heparinase strongly correlate with the late effects on mesoderm differentiation, it remains to be studied whether these events are causally connected. Further characterization of embryonic proteoglycans is necessary to define the molecular components of HSPG-dependent and HSPG-independent pathways for determination of mesoderm in the embryo.

We thank R. Rosenberg and M. Herndon for heparinase 3, R. Sasisekharan for heparinase 1, P. Wilson for Xbra plasmid, I. Dawid for Xlim1 and 1A11 plasmids, D. Kessler for goosecoid plasmid, D. Luxenberg for activin and D. Kimelman for bFGF. We also thank N. Moghal and B. Neel for helpful comments on this manuscript. K. I. is a fellow of the Human Frontier Science Program.