Xbra functions as a switch between cell migration and convergent extension in the Xenopus gastrula

Gastrulation is the process by which the basic body plan of the embryo is established from the provisional geometry of the egg. A concerted set of cell movements rearranges the germ layers to their final positions relative to one another, with endoderm on the inside, ectoderm on the outside, and mesoderm between the two. In Xenopus embryos, the Spemann organizer, or the upper dorsal blastopore lip, is crucial for the generation of the basic body plan. The anterior region of the Spemann organizer involutes first and forms prechordal mesoderm, while the later involuting region forms chordamesoderm and gives rise to the notochord. These regions can be distinguished by their early gene expression and type of cell motility during gastrulation, as well as their respective fates. Prechordal mesoderm cells exhibit active cell migration, while cells in the chordamesoderm undergo convergent extension. These two behaviors have been studied extensively on a descriptive level,particularly in Xenopus (Gerhart and Keller, 1986; Winklbauer,1990; Keller and Winklbauer,1992; Winklbauer and Selchow,1992; Keller et al.,2000).

Active cell migration of the prechordal mesoderm is characterized by the ability of single cells to spread and crawl upon a fibronectin substrate. During gastrulation, the cells generate bipolar, actin-rich lamellipodia and actively migrate upon the extracellular matrix secreted by the blastocoel roof. Interaction with a fibronectin substrate is necessary, as blockade of cell surface interactions with fibronectin, using either peptides that mimic fibronectin (GRGDSP) or fibronectin-blocking antibodies, abolishes migration(Winklbauer, 1990). Prechordal mesoderm cells migrate in vitro at the same rate as they do in the embryo,about 2.5 μm/minute (Wacker et al.,1998; Winklbauer,1990). Migrating cells come to rest at the future anterior end of the embryo, where they play a crucial role in patterning the anterior neurectoderm (Sive et al.,1989). Inhibition of active cell migration of the prechordal mesoderm, as caused by fibronectin-blocking antibodies, can lead to a variety of head defects (Marsden and DeSimone,2001).

In contrast to prechordal mesoderm cell migration, convergent extension is not so much a migrating process as a cell sorting process. Groups of cells rearrange and intercalate to change the overall shape of a tissue. During gastrulation, chordamesoderm cells rearrange from a largely isotropic organization, intercalate and generate a rod-like structure that will differentiate into notochord. Cell-cell adhesion, rather than cell-substrate adhesion involving fibronectin, is crucial for this process. Unlike the case of active cell migration, peptides that mimic fibronectin (GRGDSP) have no effect on the convergence and extension movements of isolated explants(Ramos and DeSimone, 1996;Ramos et al., 1996;Winklbauer and Keller, 1996). Cell-cell adhesion molecules such as cadherins and protocadherins are important for convergent extension. Because cell-cell contact is required for convergent extension, this behavior cannot be observed in isolated single cells, rather, it is analyzed in the context of an intact explant. Defects in this process result in shortened trunks with a fully formed head(Wallingford and Harland,2001).

These two behaviors are discrete and non-overlapping. Prechordal mesoderm exhibits active cell migration, but not convergent extension; conversely,chordamesoderm undergoes convergent extension, while only a few cells dissociated from the chordamesoderm actively migrate on fibronectin(Wacker et al., 1998). Gastrulation in Xenopus laevis probably presents the most accessible system to study control of these fundamental processes of cell and tissue morphogenesis. Though many growth factors, transcription factors and signaling molecules have been identified in the process of cytodifferentiation, the control of cell behavior and cell motility is not well understood.

To understand better each cell behavior from both a developmental and cell biological standpoint, we have examined embryonic tissues in explants. The animal cap, a tissue explant whose unperturbed fate is ventral ectoderm, can be induced to express prechordal or chordamesoderm markers via treatment with the TGFβ factor activin (Symes et al., 1994; Gurdon et al.,1996; Gurdon et al.,1999). Depending on the conditions of the treatment, either migration or convergent extension can be induced. We have used this system to study factors that are important in the induction or regulation of either behavior.

VegT, a transcription factor of the T-box family,is necessary for both cell behaviors. By contrast, Xbra, another T-box transcription factor, although required for convergent extension, inhibits cell migration. This cell migration block can be partially rescued by inhibiting convergent extension downstream of Xbra. In addition, Xbra appears to inhibit cell migration by specifically inhibiting adhesion to fibronectin. We propose that Xbra functions as a fundamental switch to keep cell migration and convergent extension as mutually exclusive behaviors in adjacent domains during gastrulation.

Xenopus embryos were obtained from X. laevis frogs(NASCO). They were fertilized by in vitro fertilization, dejellied and cultured at 14-18°C in 0.1× Marc's Modified Ringer's (MMR)(Peng, 1991). Embryos were staged according to (Nieuwkoop and Faber,1967). For injection with RNAs and DNAs, embryos were placed in 0.1× MMR containing 5% Ficoll and 50 μg/ml gentamycin. Embryos were cultured in 0.1× MMR containing 50 μg/ml gentamycin at 14-18°C until they reached the desired stage. For animal cap explants, both cells of a two-cell embryo were injected superficially in the animal hemisphere. All injection amounts reported are per blastomere.

Prechordal and chordamesoderm explant preparation: at stage 10, the vitelline was removed and incisions made on either side of the dorsal blastopore lip. A third incision was made at the blastopore lip, leaving an explant of involuted prechordal mesoderm, chordamesoderm (preinvolution) and neural tissue. Involuted prechordal mesoderm was lifted off of the explant as an intact piece by inserting an eyebrow knife into Brachet's cleft. The remaining explant was trimmed to remove neural tissue, leaving a chordamesoderm explant. For the mesendoderm extension assay, explants were prepared as described (Davidson et al.,2002).

Sibling animal caps were dissected at stage 9. Half of the caps were left intact and treated with activin protein for one hour (1 U/ml in 1×MMR). They were then transferred from activin solution into 1×MMR and allowed to heal and grow until stage 19, when they were scored for convergent extension. The rest of the caps were dissociated in CMFM (Ca²⁺ and Mg²⁺ free medium) and then treated with activin protein (1 U/ml in 1×CMFM) for 1 hour. The dissociated cells were subsequently plated in Modified Barth's Solution (MBS) into fibronectin-coated chambered coverslips(VWR). Coverslips were coated with 0.1 mg/ml fibronectin (Sigma, diluted to the appropriate concentration with MBS) for 2 hours at room temperature, and then blocked with bovine serum albumin (BSA; 50 mg/ml in MBS).

For the migration assay, cells were analyzed in the following manner: a field of cells was randomly chosen, and images were captured at 30 second intervals for at least 20 minutes (15 minutes for control samples:uninjected/untreated cells and uninjected/Activin-treated cells). Images were assembled into timelapse form (using Openlab software) and played back at a speed of either 10 or 20 frames per second. A cell was scored as positive for migration if it both translocated across the substrate and exhibited active protrusions.

Timelapse analysis of dissociated cells was performed using a Hamamatsu C2400 CCD Camera attached to a Zeiss Axiovert 135. Images were acquired and timelapse files assembled using Openlab software (Improvision). Fluorescence images were acquired using a Princeton Instruments cooled CCD Camera. Color images were acquired with a Sony 3CCD Color Camera mounted onto a Zeiss Stemi SV11 Stereoscope.

Plasmid DNAs were linearized overnight and purified using the Qiagen PCR Purification Kit. All RNAs were synthesized using the mMessage mMachine kit(Ambion) for capped RNA, purified using the Qiagen RNeasy Mini Kit, and subsequently precipitated in ethanol and resuspended in RNAse-free water.

XbraΔDNABD is a delection in amino acids 206-229(Kispert and Hermann, 1993),Xdsh mutants are all as in(Rothbacher et al., 2000). VegT-EnR was constructed as described previously(Horb and Thomsen, 1997). Xbra-EnR was constructed as described previously(Conlon et al., 1996). dn Wnt11 was constructed as previously (Tada and Smith, 2000).

All constructs are in a pCS2 backbone(Rupp et al., 1994;Turner and Weintraub, 1994),except for Xenopus Activin, which is in pSP64T.

Xenopus oocytes were harvested by manual defolliculation, and injected with 30 ng of capped RNA encoding full-length Xenopusactivin. Oocytes were cultured in OR2 solution for oocyte storage (+ 0.5 mg/ml BSA) in 96-well plates for 2 days at 18°C (five oocytes per well in 200μl culture medium). The supernatant was collected, filtered, aliquotted and stored at -80°C. Activity of each batch was tested, and was consistently found to be 20 U/ml [units as defined by Green et al.(Green et al., 1992)].

Cells were fixed for 20 minutes in MBS + 3.7% formaldehyde. Samples were permeabilized with MBS + 0.1% Triton X-100 for 5 minutes, and then blocked in MBS + 0.1% Triton X-100 + 2% BSA for 10 minutes. Cells were stained for 20 minutes using 1 μg/ml rhodamine-phalloidin (Sigma) in blocking solution. All incubations were done at room temperature.

Cells dissociated in CMFM (as described above) were plated into fibronectin- or poly-L-lysine-coated chambered coverslips (VWR). Fibronectin substrates containing a `high' concentration of fibronectin were prepared:coverslips were coated with 0.2 mg/ml fibronectin (Sigma, diluted with MBS)for 3 hours at room temperature, and then blocked with BSA (50 mg/ml in MBS)for 30 minutes at room temperature. Poly-L-lysine-coated coverslips were prepared: coverslips were coated with a solution of 1 mg/ml poly-L-lysine(Sigma) in water for 30 minutes at room temperature. The coverslips were then rinsed 10 times with MilliQ H₂O, and finally blocked with BSA (50 mg/ml in MBS) for 30 minutes at room temperature.

For the dose response experiment, coverslips were coated as described here with the following concentrations of fibronectin: low (0.1 mg/ml), medium(0.15 mg/ml) and high (0.2 mg/ml).

After plating, cells were allowed to recover and spread on the substrate for 2 hours, at which time they were fixed and processed for rhodamine-phalloidin staining. A cell was scored as spread if it took on a spread morphology: flattened with irregular borders and exhibiting actin-rich protrusions extending along the substrate.

The mesendoderm extension assay was performed as described(Davidson et al., 2002). Coverslips were coated with 0.1 mg/ml fibronectin (Sigma, diluted with MBS)overnight at 4°C. Coverslips and plastic petri dishes were blocked with Danilchik's For Amy [DFA, containing 1 mg/ml BSA, as described previously(Davidson et al., 2002)] for at least 1 hour before explants were mounted. Timelapse analysis was performed using a Hamamatsu C2400 CCD Camera attached to a Zeiss Stemi SV11 Stereoscope. Images were captured and assembled into a timelapse file using Openlab software (Improvision).

To assay cell migration, we dissociated animal cap cells, treated them with activin protein, plated them onto fibronectin-coated coverslips and observed them by time lapse microscopy. Fibronectin is the natural and required component of the extracellular matrix upon which prechordal mesoderm cells crawl (Winklbauer and Keller,1996). Animal cap cells not exposed to activin assume a characteristic round shape and remain stationary, although many exhibit circus movement, the rapid circling of an actin-rich protrusion around the cell(Fig. 1A)(Johnson, 1976). Activin-treated animal cap cells attach to the fibronectin substratum; they possess two flat bipolar lamellae similar to dissociated prechordal mesoderm cells (Fig. 1B,C). Prechordal mesoderm cells crawl at an average speed of 2.5 μm/minute(Winklbauer, 1990), comparable with activin-treated animal cap cells(Wacker et al., 1998). As the cells are plated onto a simple (uniform) fibronectin substrate, this assay does not measure directional cell migration; a cell is scored positive for migration if it exhibits both active protrusions and the ability to translocate upon the fibronectin substrate. A cell is scored negative for migration if it lacks the ability to translocate across the substrate, whether or not it has active protrustions.

Convergent extension, the other major motile activity in early mesoderm,was assayed in the intact animal cap by examining its asymmetric extension along one dimension. In our standard assay, animal caps are treated with activin protein for 1 hour and allowed to heal and develop until stage 19,when convergent extension-like movements are first easily recognizable. In experiments described here, cell migration and convergent extension were assayed in parallel, although it should be pointed out that these assays are not contemporaneous: cell migration can be assayed within 3 hours of plating,while convergent extension is not noticeable for ∼ 18 hours. It should be noted that concentrations of activin used here are comparable with amounts used in many previous studies, in which markers of prechordal and chordamesoderm were induced in each respective assay(Green and Smith, 1990;Green et al., 1992;Symes et al., 1994;Gurdon et al., 1997;Gurdon et al., 1999).

VegT, also known as Xombi, Brat and Antipodean,is a member of the T-box family of transcription factors, and is a direct downstream target of activin signaling(Horb and Thomsen, 1997;Lustig et al., 1996;Stennard et al., 1996;Zhang and King, 1996). Ectopic expression of VegT can induce both endoderm and mesoderm. In addition, oocyte depletion (Zhang et al.,1998) shows that maternal VegT RNA is required for both endoderm and mesoderm development. We tested whether VegT is required for cell migration and convergent extension. A dominant inhibitory form ofVegT (VegT-EnR) was constructed by fusing its DNA-binding domain to the repressor domain of the Drosophila engrailed gene; this form of VegT will repress its normal downstream targets. Previously,VegT-EnR has been demonstrated to inhibit expression of mesodermal markers (Gsc, Xbra and Xlim-1) and mesodermal patterning;VegT-EnR also inhibited blastopore formation(Horb and Thomsen, 1997). As might be expected, expression of VegT-EnR blocks both cell migration and convergent extension (Fig. 2). Cells injected with an inhibitory dose of VegT-EnR look like untreated animal cap cells. These data suggest that VegTactivity is necessary for both activin-induced cell migration and convergent extension.

Xenopus brachyury (Xbra), a second T-box transcription factor, is also induced by activin. Like VegT, it functions as a transcriptional activator (Kispert et al.,1995). Xbra is expressed throughout the marginal zone beginning at stage 9 and persists through gastrulation in the circumblastoporal ring, including the chordamesoderm undergoing convergent extension. It is also required for convergent extension(Conlon and Smith, 1999). However, Xbra expression is excluded from the prechordal mesoderm,and presumably is not required for prechordal mesoderm migration. As the specific exclusion of Xbra in that region could itself be important,three different forms of Xbra were tested in the cell migration and convergent extension assays: wild type, an inhibitory form (Xbra-EnR)and a DNA-binding domain mutant that is inactive (XbraΔDNABD)(Kispert and Hermann, 1993). As previously demonstrated, Xbra-EnR completely blocks convergent extension, but has minimal effects on cell migration(Fig. 3). By contrast,wild-type Xbra does not inhibit convergent extension, but partially inhibits cell migration in a dose-dependent manner. Convergent extension movements are not inhibited in caps injected with wild-type Xbra,although the caps look qualitatively different, thinner, than uninjected caps.XbraΔDNABD has virtually no effects on either cell migration or convergent extension, indicating that these phenotypes are mediated by downstream transcriptional targets of Xbra. Although Xbrafunctions to promote convergent extension, these data suggest that the specific exclusion of Xbra from the prechordal mesoderm could be important in allowing cell migration. This is consistent with the idea thatXbra is specifically repressed in cells undergoing cell migration in response to activin (Gurdon et al.,1999; Symes et al.,1994). The regulation of Xbra expression may therefore determine whether gastrula stage mesodermal cells undergo cell migration or convergent extension. This is in contrast to VegT, which is required for both cell migration and convergent extension.

To determine whether Xbra inhibits cell migration by promoting convergent extension, we blocked convergent extension with a secreted Wnt inhibitor. Specifically, it has been reported that Wnt11 is a direct transcriptional target of Xbra, and is necessary for convergent extension (Tada and Smith,2000). As Xbra-EnR inhibits Wnt11 expression both in activin-treated animal caps, as well as in the endogenous mesoderm(Tada and Smith, 2000), we constructed a dominant-negative Wnt11 (dn Wnt11) that lacks the C-terminal 70 amino acid cysteine-rich domain (Tada and Smith, 2000). As shown in Fig. 4, Xbra inhibits cell migration in a dose-dependent manner. However, co-expression of dn Wnt11 partially rescues cell migration. These data suggest that blocking convergent extension impairs the ability ofXbra to inhibit cell migration. Therefore, we propose thatXbra inhibits cell migration by actively promoting convergent extension. In this view, cell migration is a default state of activin stimulation, and Xbra acts as a switch between cell migration and convergent extension.

Because dominant-negative Wnt11 partially restores cell migration inhibited by Xbra, it is possible that the Wnt pathway regulates both convergent extension and cell migration. Therefore, a role for Wnt11 in cell migration was tested. As demonstrated previously, dn Wnt11 partially blocks convergent extension (Fig. 5B)(Tada and Smith, 2000);however, cell migration is unaffected (Fig. 5A). This partial inhibition correlates with the partial effects seen in counteracting the Xbra-mediated inhibition of cell migration.

Levels of Wnt11 signaling seem to be crucial for convergent extension, as reported previously (Djiane et al.,2000). It has been shown that both overexpression and inhibition of Wnt11 signal inhibit convergent extension. As demonstrated previously,overexpression of Wnt11 blocks convergent extension(Fig. 5B)(Tada and Smith, 2000);however, cell migration is unaffected (Fig. 5A). Because Wnt11 is not sufficient to inhibit cell migration, it cannot solely account for the inhibition of cell migration by Xbra. Wnt11 may act in conjunction with other factors in order to block cell migration and promote convergent extension.

As the control of cell migration is largely unexplored in embryos, the downstream targets through which Xbra inhibits cell migration remain largely unknown. Downstream transcriptional targets could act on any number of processes in order to inhibit cell migration, for example, extracellular signaling, intracellular signaling to the cytoskeleton, actin polymerization or cellular adhesion. To define better the mechanisms by which Xbrainhibits cell migration, we examined the response of the cells to small GTPases involved in the cytoskeleton and adhesion.

The Rho family of small GTPases has been shown in cultured cell lines to have specific effects on cell morphology and the actin cytoskeleton. Activated Cdc42 injected into cells induces filopodia, while activated Rac induces lamellipodia (Nobes and Hall,1995). Xenopus animal cap cells respond similarly to these activated GTPases. To avoid early embryonic phenotypes, V12 Cdc42 or V12 Rac were injected as expression plasmids into both blastomeres at the two-cell stage, thereby limiting expression to post mid-blastula transition. Under these conditions, V12 Cdc42 induces filopodia and V12 Rac induces elaborate lamellipodia (Fig. 6A). When wild-type Xbra is co-injected with the activated form of either GTPase, the cell shape changes are inhibited(Fig. 6B); the cells have the same morphology as those that have not been injected. By contrast,co-injection of Goosecoid (Gsc), a transcription factor expressed in the prospective prechordal mesoderm, has negligible effects (data not shown).

Because Xbra can inhibit cell shape changes induced by activated Cdc42 or Rac, this suggests that the effect of Xbra is not restricted to the signaling pathways upstream of the GTPases. More likely, its effect is on the signaling pathways downstream of Cdc42 and Rac that lead to regulated actin polymerization and cell morphology changes. However, this assay cannot distinguish between inhibition of actin polymerization and inhibition of cell adhesion to the fibronectin substrate.

To distinguish between these two possibilities, cell spreading assays were performed. When a twofold higher concentration of fibronectin is used to coat coverglasses (see Materials and Methods), dissociated, uninjected animal cap cells (not treated with activin) spread upon the substrate. The cells are flattened, exhibiting an irregular border and few small spiky protrusions(Fig. 6C). By comparison, cells plated on a poly-L-lysine substrate also spread, exhibiting an irregular border and, often, large flat lamellae(Fig. 6C). Because the injection of activated GTPases is not necessary for the cells to exhibit these morphological behaviors, this represents a simple assay for adhesion and spreading upon a substrate. Injection of Xbra inhibits cell spreading upon fibronectin, such that the cells assume a round, unspread, unadherent morphology (Fig.6C,6D). However, expression of Xbra has no effect on spreading upon poly-L-lysine. Fig. 6E shows dose response to Xbra in the cell spreading assay upon fibronectin. Three fibronectin concentrations were tested: low, medium and high. Low is equal to the amount used in previous migration assays, and high is equal to the amount used in Fig. 6D for the previous cell spreading assay (see Materials and Methods for details). These data suggest that in a simple assay to test adhesion and cell spreading,Xbra can inhibit adhesion specifically to fibronectin. Thus, the effects mediated by Xbra are less likely to be general effects upon the cytoskeleton and contractility, and may be specific to the process of adhesion to fibronectin.

It is worth noting that Xbra-mediated inhibition of cell spreading in this assay is much more potent than its inhibition of activin-induced cell migration (compare Fig. 3A withFig. 6D,E). Comparable amounts of Xbra inhibit activin-induced cell migration by 50-60%, but inhibit cell spreading on fibronectin almost completely. This difference is likely due to the effect of activin treatment. It has been demonstrated that activin, in inducing cell migration, upregulates adhesion to fibronectin(Wacker et al., 1998). This upregulation of adhesion probably accounts for the relative lower potency ofXbra in the migration assay, and higher potency in the cell spreading assay, in which cells are not treated with activin.

In addition to cell spreading assays, visual actin polymerization assays were performed to test the ability of small GTPases to induce actin polymerization in extracts (Ma et al.,1998). Small scale extracts were made from animal caps from either uninjected embryos or embryos injected animally with Xbra (500 pg RNA) at the two-cell stage. GTPγS-loaded Cdc42 and GTPγS-loaded Rac induced actin polymerization equally well in extract from uninjected animal caps as in extract from Xbra-injected animal caps (S. Eden,K.M.K. and M.W.K., unpublished). Taken together, these data suggest thatXbra inhibits adhesion, not actin polymerization, and, more specifically, adhesion of cells to a fibronectin substrate.

We have used activin-induced animal caps to recreate, as best as possible,conditions in the marginal zone, where endogenous populations of cells undergo either migration or convergent extension. We therefore wished to test whetherXbra has the same effects in the marginal zone cells themselves. For this, we injected Xbra into the marginal zone of the two dorsal cells of a four-cell embryo. At stage 10, head mesoderm was dissected from the resulting gastrulae, and the dissociated cells assayed. As shown inFig. 7A, expression ofXbra inhibits the ability of the prechordal mesoderm cells to migrate, in a dose-dependent manner. Sibling embryos were cultured until early tailbud stages, and, as shown, overexpression of Xbra in the dorsal marginal zone causes anterior truncations(Fig. 7B). Therefore, even though migration is inhibited in only 50% of prechordal mesoderm cells, this is clearly enough to cause severe defects in head development. Another assay was used to test prechordal mesoderm migration. Davidson et al. have developed an assay to test the rate and extent of mesendoderm extension, in which the head mesoderm migrates upon a fibronectin substrate as an intact mantle(Davidson et al., 2002). Timelapse analysis was performed to compare mesendoderm extension in dorsal marginal zone explants (see Materials and Methods) fromXbra-injected, uninjected and Xbra-EnR-injected embryos(explants numbered 1, 2, and 3, respectively inFig. 7C). As shown, theXbra-injected explant initially extends at a similar rate as the other explants. However, by about 7.5 hours(Fig. 7C), theXbra-injected explant has reached its maximum extension, while the uninjected and Xbra-EnR-injected explants continue to extend for another 2.5 hours (Fig. 7C). Thus, even as an intact mantle, inhibition of migration by 50% causes a clear decrease in the extent of mesendoderm migration. Meanwhile, migration of theXbra-EnR-injected explant was indistinguishable from that of the uninjected explant, not surprisingly, as expression of Xbra-EnR has no effect in the in vitro cell migration assay.

In a complementary set of experiments, the effect of blocking Xbrafunction in the chordamesoderm was tested. Xbra-EnR was injected into the marginal zone of the two dorsal cells of a four-cell embryo, and the chordamesoderm region of the dorsal marginal zone was dissected from the resulting gastrulae. As reported previously, expression of Xbra-EnR in the dorsal marginal zone inhibits its ability to undergo convergent extension (Fig. 8A). Embryos cultured to early tailbud stages exhibit a phenotype indicative of inhibition of convergent extension: short trunk and tail with a fully formed head(Fig. 8B).

We then asked whether inhibition of Xbra activity (by expression of Xbra-EnR) promoted cell migration. Cells from the dorsal marginal zone, which undergo convergent extension, do not migrate processively,although a fraction of them spread on a fibronectin substrate(Fig. 1D)(Wacker et al., 1998). However, expression of Xbra-EnR in these cells causes an increased fraction of them to undergo cell spreading and migration(Fig. 8C). This effect can be rescued with equal amounts of wild type Xbra RNA. Therefore, in this context, blocking Xbra function leads to increased adhesion to fibronectin and cell migration. This suggests that a normal function ofXbra expression in the marginal zone is to inhibit adhesion to fibronectin and, therefore, cell migration.

We wondered to what extent the effects of Xbra inhibition in the marginal zone could be recapitulated by inhibition of the Wnt signaling pathway, specifically the planar cell polarity pathway which has been shown to be necessary for convergent extension(Wallingford and Harland,2001). To determine whether the effect of Xbra upon cell behavior can be largely attributed to downstream effects on cell fate or a more direct effect upon cell motility, different Dishevelled mutants were tested within the chordamesoderm for effects on cell migration.Dishevelled (Dsh) is an intracellular mediator of Wnt pathway signal transduction. It is a modular protein with multiple functions:the DIX domain mediates cell fate decisions via the Wg/Wnt pathway, while the PDZ domain mediates effects on cell behavior via the planar cell polarity(PCP) pathway (Rothbacher et al.,2000; Wallingford et al.,2000). We tested whether the effects of Xbra-EnR on chordamesoderm cell behavior could be recapitulated with mutants ofDsh that specifically block the planar cell polarity pathway downstream of a Wnt signal, in this case specifically Wnt11.XdshΔDIX (deletion of the DIX domain), XdshΔPDZ(deletion of the PDZ domain) and Xdd1 (deletion of most of the PDZ domain and part of the following region) were tested for effects on both convergent extension and cell migration within the chordamesoderm.

As reported previously, chordamesoderm explants expressingXdshΔPDZ and Xdd1 fail to elongate; explants expressingXdshΔDIX are also partially inhibited from undergoing convergent extension (Fig. 9A). As shown in previous figures, explants expressing either dn Wnt11 or Xbra-EnR are also inhibited in convergent extension movements. However, these same chordamesoderm explants, when assayed for cell migration, reveal that overexpression of specific Xdsh mutants (as well as dn Wnt11), causes an increased fraction of cells to undergo cell spreading and migration on a fibronectin substrate(Fig. 9B). It is notable that the mutants that inhibit the Wnt PCP pathway do so without altering cell fate in the dorsal mesoderm, as measured by staining for differentiated muscle and notochord (using the antibodies Tor70, MZ15 and 12/101), as well as Northern blot analysis of the genes Xlim1, Xnr3, Gsc, Xotx2, Chordin andXbra (Sokol, 1996;Wallingford and Harland,2001). These data suggest that specific inhibition of the Wnt PCP pathway, which does not alter cell fate, is sufficient to partially switch dorsal mesoderm cells from convergent extension behavior to cell migration. In addition, cell migration may be a default state within the dorsal mesoderm,which requires active signals to inhibit components of the migration machinery, such as adhesion to fibronectin, and promotes convergent extension.

During early development in Xenopus embryos, the regions destined to give rise to the head and trunk organizers are juxtaposed. Gene expression differences arise between these two regions: differences that not only define much later cell and tissue fates, but which also give rise to immediate differences in cell behavior during gastrulation. Cells in the anterior-most region migrate, spreading out over the blastocoel roof. Cells immediately behind them converge toward the midline, ultimately leading to extension of the body axis and forming notochord.

Xbra is a transcription factor that appears early in the trunk compartment of the organizer, where it has long been known to have a role in trunk mesoderm development. We demonstrate here that Xbra actively inhibits cell migration, both in activin-induced animal cap cells, and in the endogenous prechordal mesoderm. This inhibition of cell migration can be rescued by inhibiting convergent extension. These experiments suggest that in the embryo, Xbra acts as a morphogenetic switch not only promoting convergent extension, but also actively repressing cell migration. As expected for such a switch, blocking Xbra activity in the dorsal mesoderm via dominant-negative Xbra (Xbra-EnR) increases the number of cells in the chordamesoderm that spread on fibronectin and actively migrate. Under these conditions, the default behavior of the tissue appears to be cell migration. Therefore, one unappreciated role of endogenous Xbra is to inhibit cell migration within the chordamesoderm, while fostering convergent extension movements. Xbra, which has a role in the eventual cytodifferentiation of the chordamesoderm, also has an immediate dual role in morphogenetic movements.

Wnt11 is a direct downstream target of Xbra which is necessary for convergent extension (Tada and Smith,2000). Although Wnt11 signaling is not required for cell migration(Fig. 5A), expression of dominant-negative Wnt11 reverses the inhibitory effects of Xbra on activin-induced cell migration. These same effects can be produced in the dorsal mesoderm, with dn Wnt11 as well as deletion mutants ofDishevelled, which inhibit the Wnt planar cell polarity pathway(DshΔPDZ, Xdd1). Notably, these Dsh mutants have no effect on cell fate (Sokol,1996; Wallingford and Harland,2001).

Taken together, these data suggest that there exist within the dorsal mesoderm two cell behaviors, with cell migration as a default state. Expression of Xbra within the prospective chordamesoderm produces a domain where cell migration is inhibited. The ground state may be restored by inhibition of Xbra itself or the downstream Wnt planar cell polarity pathway. In the embryo, Xbra is excluded from the prechordal mesoderm, allowing these cells to exhibit active cell migration. Xbraacts as a cell behavior switch that is crucial for the proper anteroposterior development of the embryo. Inhibition of Xbra activity in the chordamesoderm results in embryos with a shortened trunk, a consequence of a failure of convergent extension; these cells now actively migrate. Misexpression of Xbra in the prechordal mesoderm results in embryos with head truncations, as a consequence of a failure of cell migration.

It has been reported that Gsc and Mix.1, two transcription factors expressed in the dorsal marginal zone of gastrula stage embryos, synergize to repress the expression of Xbra(Latinkic and Smith, 1999). Expression of dominant inhibitory forms of either transcription factor results in inappropriate expression of Xbra in the prechordal mesoderm, and anterior truncation of the embryos. From the results presented here, it is likely in those experiments that derepression of Xbra in the prechordal mesoderm inhibited prechordal mesoderm migration, producing anterior truncations.

Niehrs and colleagues have reported on the gene anti-dorsalizing morphogenetic protein (ADMP), expressed in the chordamesoderm, which also seems to be part of a network that maintains the subdivision between prechordal and chordamesoderm (Dosch and Niehrs, 2000). It will be interesting to explore the relationship between ADMP and Xbra. It has already been shown that ADMP can induce the expression of Xbra (Moos et al., 1995), but whether Xbra also affects the expression of ADMP is unclear. Similarly, the effects of ADMP specifically on cell behavior have yet to be determined. Although ADMP is expressed solely in the chordamesoderm, Xbra is expressed throughout the mesoderm in a circumblastoporal ring and is specifically excluded from the prechordal mesoderm. Xbra may be expressed beyond the chordamesoderm to control various degrees of convergent extension required throughout the mesoderm to ensure proper blastopore closure via radiolateral convergent extension movements.

The fact that Xbra expression is maintained in chordamesoderm but not paraxial mesoderm may be important for establishing an axis and polarity for mediolateral convergent extension. In terms of inducing convergent extension in paraxial mesoderm, prolonged Xbra expression may not be required. There may be other signals, specifically from the chordamesoderm,that propagate mediolateral convergent extension behavior, and there may be other genes induced within the paraxial mesoderm itself that play a role in the behavior of its cells.

Just how does Xbra inhibit cell migration? Ectopic expression ofXbra has several effects: inhibition of cell migration, inhibition of the morphological effects elicited by activated Rac and Cdc42 in dissociated animal cap cells, and inhibition of adhesion to fibronectin in a simple cell spreading assay. The inhibition of cell migration depends on the ability ofXbra to bind DNA and activate transcription, which suggests that these effects are mediated by one or more downstream transcriptional targets of Xbra. Wnt11, although required for convergent extension, is not sufficient to inhibit cell migration. Therefore, Xbra must induce the expression of other factors which can, at the very least, inhibit adhesion to fibronectin, either alone or in conjunction with Wnt11. It will be interesting to learn the downstream target genes of Xbra that may be involved in cell adhesion and movement.

In both the activin-induced animal cap system and the endogenous chordamesoderm, expression of Xbra or Xbra-EnR alters expression level of many markers as early as stage 10, including Xbraitself, Mix.1, Xwnt-8, Gsc, Pintallavis, Xnot, Chordin, Noggin and Wnt11 (Conlon and Smith, 1999;Tada and Smith, 2000). Although changes in any of these genes may represent a change in cell fate, it is notable that some of the effects described here can be recapitulated using mutants of Dsh that do not alter cell fate in the dorsal mesoderm.

In summary, vertebrate embryos possess two distinct cell behaviors that pattern the dorsal side of the embryo. Convergent extension, which can be assayed only in populations of cells, leads to notochord formation and patterns the trunk. Cell migration, which can be studied in individual cells,leads to different cell fates and inductive effects, which are confined to the anterior neural plate. Distinct as they are, in Xenopus, these two behaviors arise from adjacent cell populations; both can be produced downstream of the same signaling protein, activin. The distinction in cell behavior depends on the domain of brachyury expression.Brachyury actively inhibits cell migration while inducing the tissue to undergo convergent extension via the Wnt planar cell polarity pathway. However, Wnt11 is not sufficient to inhibit cell migration. Therefore, the downstream targets of Xbra that inhibit cell migration are unknown,but are capable of inhibiting cell shape changes induced by small GTPases and adhesion to fibronectin in a simple cell spreading assay. These findings raise two further questions: what is the mechanism by which Xbra regulates convergent extension and cell migration, and what establishes the domains of cell behavior? The key to both of these is the regulation of Xbraexpression and its downstream targets.

We thank Robert Davis, Licio Collavin, Kris Kroll, Malcolm Whitman,Michelle Lee and Keuna Cho for critical reading of the manuscript, and the rest of the Kirschner laboratory for thoughtful advice. Special thanks in particular to Robert Davis and Licio Collavin for extensive discussions and support throughout this work. This work was supported by NIH grant#HD37277.