Guidance of mesoderm cell migration in the Xenopus gastrula requires PDGF signaling

The migration of mesodermal cells away from their site of internalization is a conserved feature of vertebrate gastrulation, although the importance of this process on the mechanics of gastrulation may differ between taxonomic groups (Winklbauer, 1994). Available evidence suggests that migration is essential for gastrulation in the chicken (Harrisson et al.,1993), and contributes to germ layer formation in carp(Gevers et al., 1993) and zebrafish (Yamashita et al.,2002). In the mouse, null mutations for various integrins and matrix proteins have not yet provided conclusive evidence for the functional relevance of mesoderm migration (De Arcangelis and Georges-Labouesse, 2000; Gustafsson and Fassler, 2000). In amphibians, blocking migration impedes gastrulation completely in the urodele Pleurodeles (Boucaut et al.,1984; Darribere et al.,1990), and less severely, though distinctly, in anurans such as Xenopus or Rana (Smith et al., 1990; Saint-Jeannet and Dawid, 1994; Winklbauer and Keller, 1996).

In Xenopus, migration of dorsal anterior mesoderm cells across the ectodermal blastocoel roof (BCR) relies on their interaction with a network of fibronectin (FN) containing fibrils(Winklbauer and Keller, 1996),the extracellular matrix of the BCR(Nakatsuji and Johnson, 1983a; Winklbauer, 1998). FN contributes to the adhesion of mesoderm cells to the BCR, and, most importantly, it elicits the formation of lamellipodia, the locomotory organelles of these cells (Winklbauer,1990; Ramos and DeSimone,1996; Winklbauer and Keller,1996).

In pioneering work, Nakatsuji and Johnson(Nakatsuji and Johnson, 1983b)showed that in addition to serving as permissive substratum, the BCR extracellular matrix contains guidance cues that determine the direction of migration. When BCR is explanted to deposit its matrix on an inert surface,dispersed mesodermal cells from several amphibian species migrate on this conditioned substratum directionally, recognizing cues that guide them to the animal pole region (Nakatsuji and Johnson,1983b). In Xenopus, migration in vitro is also directional, although only when aggregates of mesodermal cells are used to probe conditioned substratum, instead of single cells(Fig. 1A). Moreover, in the Xenopus embryo, anterior mesoderm cells extend locomotory processes towards the animal pole, confirming that migration is directional in vivo(Winklbauer and Nagel,1991).

To identify the nature of the guidance cues that direct mesoderm migration has been a goal since their discovery two decades ago. Previously, we have shown that signals emanating from the vegetal part of the embryo, which employ fibroblast growth factor (FGF)- and activin-like signaling pathways for propagation, polarize the BCR layer. This polarization is a prerequisite for the deposition of a correctly oriented extracellular matrix(Nagel and Winklbauer, 1999). However, the molecular basis of the matrix-borne directional cues remained elusive. We address this long-standing unresolved issue, and show that platelet-derived growth factor (PDGF) signaling plays an essential role in mesoderm guidance during Xenopus gastrulation.

PDGF signaling is involved in a variety of processes, including early embryonic development, formation of the nervous system, heart and lung development, and angiogenesis. It has also been implicated in several pathological conditions, such as oncogenesis, atherosclerosis, lung and kidney fibrosis, and wound healing. Major target processes of PDGF signaling are the regulation of proliferation, and the control of cell motility and migration(Ataliotis and Mercola, 1997; Heldin and Westermark, 1999; Betsholtz et al., 2001).

The PDGF family comprises four members: PDGFA and PDGFB, which form homo-and heterodimers (Heldin and Westermark,1999), and the CUB-domain containing PDGFC and PDGFD(Li et al., 2000; Bergsten et al., 2001). In the Xenopus gastrula, PDGFA is expressed in the BCR, whereas PDGFB is barely detectable, suggesting the exclusive presence of PDGFAA homodimers(Mercola et al., 1988). The A-chain occurs in two splice variants, a long form (lfPDGFA) that contains a C-terminal site involved in extracellular matrix binding, and a short form(sfPDGFA) lacking the matrix-binding motif(Raines and Ross, 1992; Andersson et al., 1994). RNAs of both variants are present in the Xenopus gastrula(Mercola et al., 1988). Of the possible PDGF receptor dimers (PDGFRαα, PDGFRαβ and PDGFRββ), only PDGFRαα recognizes PDGFAA(Ataliotis and Mercola, 1997; Heldin and Westermark, 1999),and the PDGFRα isoform is in fact expressed in the Xenopusmesoderm at gastrulation (Jones et al.,1993).

The complementary pattern of expression, with PDGFA ligand in the BCR and its receptor in the adjacent mesoderm(Ataliotis et al., 1995),suggests a role for paracrine PDGFA signaling in mesoderm development. Disruption of PDGF function by means of a dominant-negative PDGFRα leads to aberrant movement of the mesoderm(Ataliotis et al., 1995), and treatment of mesoderm aggregates with PDGFAA protein stimulates spreading on FN in vitro (Symes and Mercola,1996). However, it is not known which of the several mesoderm-driven gastrulation movements in Xenopus is actually controlled by PDGFA.

We identify a distinct in vivo function for PDGFA signaling in early Xenopus development. We show that interfering with PDGF function randomizes mesoderm migration on conditioned substratum in vitro, and disrupts the orientation of migratory cells in the embryo. We propose that matrix-bound PDGFAA could be instructive in mesoderm guidance during gastrulation.

Xenopus laevis embryos from induced spawnings(Winklbauer, 1990) were staged according to Nieuwkoop and Faber(Nieuwkoop and Faber, 1967). Operation techniques and buffer (Modified Barth's Solution, MBS) have been described previously (Winklbauer,1990).

BCR explants were held against the bottom of Greiner tissue culture dishes(TC grade) by strips of coverslip resting on both ends on silicone grease. Before removing BCR explants by aspiration with buffer, their outlines and orientation were marked on the dish. Substrata were saturated with 50 mg/ml bovine serum albumin (BSA) and used in the migration assay(Winklbauer and Nagel, 1991)(Fig. 1A). FN-coated substratum was prepared by incubating Greiner tissue culture dishes with bovine plasma fibronectin solution (Sigma) at 200 μg/ml for 30 minutes, followed by saturation of the substratum with 50 mg/ml BSA.

Displacement of anterior mesoderm explants (about 200 cells per explant) on conditioned substratum (one explant per substratum) was observed on an inverted microscope. Positions of explants were recorded at the start of the experiment, and 1 hour later. The positions after 1 hour are indicated by circles, with starting points all superimposed at the origin. Movement towards the position of the animal pole of the BCR was assigned positive values on the y-axis. Using a one-sided sign test, the null hypothesis that explants move animally and vegetally with equal frequency was tested against the alternative hypothesis, that aggregates move preferentially towards the animal pole.

Embryos were fixed in 4% formaldehyde, BCRs were removed with needles,blocked in 5% BSA and stained with rabbit antibody to Xenopus FN(Winklbauer, 1998), and FITC-conjugated secondary antibody. Anterior mesoderm cells on conditioned substratum were fixed after 1 hour in 4% formaldehyde/0.1% Triton-X-100, and stained with 5 μM TRITC-conjugated phalloidin (Sigma) for 30 minutes.

Embryos were fixed in 2.5% glutaraldehyde in MBS, and BCRs were removed with needles. Specimens were postfixed in 1% OsO₄, dehydrated in a graded ethanol series, critical point dried and sputter coated with gold-palladium.

In situ hybridization was modified from the methods of Harland(Harland, 1991). Alkaline phosphatase substratum was BM Purple (Roche). For the antisense Xbradigoxigenin probe, plasmid DB30 (gift of M. Sargent, London) was linearized with BgIII and used as a template for transcription with T7 RNA polymerase. chordin and gsc probes(Sasai et al., 1994; Cho et al., 1991) (gifts from H. Steinbeisser, Heidelberg) were generated by linearizing clones 59 and H7,respectively, in pBluescript SK(–) with EcoRI and transcription with T7 polymerase.

Plasmids pCS2 containing wild-type Xenopus PDGFRα, a dominant-negative version of this receptor, PDGFR37, or a PDGFRαmissense construct (Ataliotis et al.,1995), were linearized with NheI and transcribed with T7 polymerase. pGHE2 containing the wild-type long form of Xenopus PDGFA(Mercola et al., 1988),lfPDGFA, was linearized with NheI and transcribed with T7 polymerase. A C-terminally truncated form of lfPDGFA lacking the matrix-binding motif(amino acid residues 198-227) was generated by PCR using pGHE2-PDGFA(Mercola et al., 1988) as a template (Fig. 1B). The truncated form of PDGFA (trPDGFA) was subcloned into the BamH1/EcoR1 site of pCS2 and used for in vitro transcription with SP6 polymerase after linearization with XhoI. Plasmid pGHE2 harboring a processing defective mutant of mouse PDGFA that acts as a dominant-negative mutant in Xenopus(Mercola et al., 1990) was linearized with NheI and transcribed with T7 polymerase. Embryos were injected at the four-cell stage into all four blastomeres, marginally with 250 pg mRNA per embryo (PDGFRα, PDGFR37, missense PDGFRα), or animally with 400 pg mRNA per embryo (lfPDGFA, trPDGFA) or 200 pg mRNA per embryo(dnPDGFA).

Human recombinant PDGFAA protein, long form (TEBU, Germany) was used at 50 ng/ml. Tyrphostin AG 1296, an inhibitor of PDGF receptor tyrosin kinases, and AG 43 control tyrphostin (Calbiochem-Novabiochem) were dissolved in DMSO (10 mg/ml) and used at a final concentration of 5 μM, Wortmannin (Sigma) at 100-200 nM. Anti-PDGFA morpholino antisense oligonucleotide (GeneTools,5′-AGAATCCAAGCCCAGATCCTCATTG-3′)and 5 bp mispaired control morpholino (exchanged bases underlined) were injected at 60 ng per embryo.

Migration of cells from the Xenopus gastrula is studied in vitro in simple, serum-free medium. To compare migratory behaviors of mesoderm aggregates, their translocation was observed on three different substrates:plain tissue culture plastic, an artificial substratum of purified FN, and BCR-conditioned substratum (Fig. 2A-C). In the absence of FN, explants do not move(Fig. 2A). Coating dishes with FN suffices to stimulate translocation but, as expected, the direction of movement is random relative to an arbitrary coordinate system(Fig. 2B). By contrast,aggregates move directionally towards the position of the animal pole on conditioned substratum, and migration is also more vigorous(Fig. 2C), in agreement with previous results (Winklbauer and Nagel,1991; Nagel and Winklbauer,1999). During substratum conditioning, FN is deposited, which promotes migration; however, additional components are apparently present that guide mesoderm explants to the animal pole.

To examine whether PDGF signaling is involved in this guidance, a dominant negative PDGFRα construct, PDGFR37(Ataliotis et al., 1995), was expressed in mesoderm explants migrating on conditioned substratum. Under these conditions, translocation is randomized and resembles that on plain FN substratum (Fig. 2D). Co-injection of a wild-type receptor with the dominant-negative construct rescues the directionality of migration significantly(Fig. 2E). Expression of a PDGFRα missense mutation in the migrating mesoderm does not affect directionality (Fig. 2F).

The inhibitory PDGFR37 does not impair mesoderm induction in Xenopus (Ataliotis et al.,1995). However, in order to ensure that its effect on migration is not due to subtle alterations of mesoderm development, PDGFRα signaling was blocked directly in migrating explants taken from normally developed mesoderm, by treatment with the cell permeable inhibitor, AG1296(Kovalenko et al., 1997). This inhibition (Fig. 2G), but not treatment with a similar control substance, AG43(Fig. 2H), also abolished directionality, suggesting that PDGFR signaling is required during actual migration, and not during formation or patterning of the mesoderm. As the previously reported effect of PDGFRα signaling on mesoderm motility was mediated by PI3 kinase (Symes and Mercola,1996), aggregates on conditioned substratum were also treated with an inhibitor of phosphatidylinositol 3 (PI3) kinase, Wortmannin. In its presence, migration was non-directional(Fig. 2I), implicating this kinase in the pathway that regulates directionality of movement.

PDGF signaling was also disrupted by blocking PDGFA in the BCR. Expression of the dominant negative dnPDGFA (Mercola et al., 1990) in the BCR leads to non-directional migration of normal mesoderm on substratum conditioned with injected BCR(Fig. 3A). Likewise, impeding translation of endogenous PDGFA mRNA in the BCR by morpholino antisense oligonucleotide randomizes translocation(Fig. 3B), whereas injection of a 5-mispaired control morpholino has no such effect(Fig. 3C). To further demonstrate specificity of the morpholino, we employed the fact that mesoderm differentiation can be induced by exogenous PDGFA in animal caps expressing PDGFRα (Ataliotis et al.,1995). In contrast to untreated caps, those co-injected with receptor and with Xenopus PDGFA mRNA containing the morpholino target site develop transparent bulges indicative of ventral-type mesoderm. Additional injection of anti-PDGFA morpholino completely blocks this effect,but it can be rescued by treatment of explants with PDGFA protein(Fig. 3I). This suggests that the morpholino specifically inhibits Xenopus PDGFA function.

Injection of mRNA coding for the matrix-binding lfPDGFA together with the anti-PDGF morpholino does not substantially rescue directionality of migration(Fig. 3D). A rescuing effect is not to be expected, however, as overexpression of lfPDGFA alone suffices to reduce the directionality of translocation significantly(Fig. 3E). Thus, both diminishing and increasing PDGFA activity in the BCR adversely affect directionality, consistent with an instructive function for this factor in mesoderm guidance.

To guide cells on conditioned substratum, PDGFA should be associated with the extracellular matrix. To see whether the matrix-binding capability of lfPDGFA is essential for its interference with directional migration, we deleted the C-terminal matrix-interacting motif from the wild-type lfPDGFA. When this truncated form, trPDGFA, is expressed in the BCR, directionality of migration on respective conditioned substratum is not affected(Fig. 3F). Apparently, only matrix-binding PDGFA can interfere with mesoderm guidance. This implies that it is the endogenous matrix-binding lfPDGFA that is involved in vivo in determining the direction of migration on the BCR.

When trPDGFA is expressed in the mesoderm, migration on normal conditioned substratum is randomized (Fig. 3G). This shows that the trPDGFA construct is functional, and that its ectopic expression in the mesoderm interferes with the orientation of aggregates on normal substratum. In addition, exposure of mesoderm aggregates to human recombinant PDGFAA protein during migration on normal conditioned substratum abolishes directionality (Fig. 3H). This confirms that proper PDGF signaling is required during migration, and not earlier, as, for example, during patterning. Furthermore,as in the previous experiment, it suggests that the abundant, uniform presence of soluble PDGFAA interferes with the ability of migrating cells to read the cues on the conditioned substratum, arguing also in favor of an instructive role of matrix-binding lfPDGFA in cell guidance.

By the mid-gastrula stage, vegetal rotation(Winklbauer and Schurfeld,1999) has moved the anterior dorsal mesoderm against the BCR. The blastocoel floor is concave, and the mesoderm exhibits a pointed leading edge that is apposed to the BCR (Fig. 4A). Both the expression of inhibitory dnPDGFA(Fig. 4B) and the overexpression of wild-type lfPDGFA (Fig. 4C) in the BCR affect gastrula morphology in a similar way: the mesoderm does not develop a pointed leading edge, and the blastocoel floor does not become concave (Fig. 4B,C). This morphology is consistent with mesoderm movement across the BCR being impeded while vegetal rotation continues to internalize vegetal cell mass.

The BCR-apposed surface of the mesoderm displays a shingle arrangement:cells underlap adjacent cells in front of them, and are underlapped by processes from cells behind, indicating an orientation of cells in the direction of the animal pole (Winklbauer and Nagel, 1991) (Fig. 4D). In dnPDGFA-expressing embryos, but also upon overexpression of lfPDGFA, this shingle arrangement is abolished(Fig. 4E,F). Quantitative analysis shows that normally, cytoplasmic processes point almost always animally, but that protrusions in PDGF signaling-defective embryos are directed with equal frequency towards and away from the animal pole (arrows in Fig. 4E,F; Fig. 5A), indicating disruption of cell orientation.

The frequency of protrusions is also affected in dnPDGFA and lfPDGFA expressing embryos, respectively. In both cases, it is significantly reduced to almost half its normal value (Fig. 5A). However, this does not reflect a direct control of protrusive activity by PDGF signaling. On substratum conditioned by control BCR, and by BCR injected with dnPDGFA or lfPDGFA mRNA, respectively, the frequency of cells extending processes does not vary as strongly as in the embryo, although slightly fewer protrusions form on dnPDGFA substratum(Fig. 5B). Apparently, it is under the conditions of mesoderm movement in the embryo where proper PDGF signaling is required to increase protrusion formation.

To exclude the possibility that disorientation or absence of protrusions are due to defective FN fibril network formation(Winklbauer and Nagel, 1991; Winklbauer and Keller, 1996),the BCR was stained with FN antibody. A normal FN fibril network(Fig. 4G) is present in dnPDGF(Fig. 4H) or lfPDGFA overexpressing embryos (Fig. 4I), or after anti-PDGF morpholino injection (not shown). Together, these results suggest that orientation of migrating mesodermal cells and normal protrusive activity in the embryo depend on proper PDGFA signaling.

To determine what consequences disoriented migration has on gastrulation,mesoderm translocation was inferred from the expression pattern of the mesoderm markers Xbra (Smith et al., 1991,Smith et al.,1991), chordin (Sasai et al., 1994) and goosecoid (gsc)(Cho et al., 1991)(Fig. 6). Xbra is initially expressed in a subequatorial ring(Fig. 6A), and chordin, a marker of head and axial mesoderm, in the dorsal lip(Fig. 6D). Expression of dnPDGFA or overexpression of lfPDGFA in the BCR does not abolish this pattern(Fig. 6B,C,E,F). This supplements the finding that inhibition of PDGF receptor function does not affect mesoderm regionalization (Ataliotis et al., 1995).

In the mid-gastrula, Xbra expression labels the blastopore and the notochord region, which has begun to extend anteriorly(Fig. 6G). In dnPDGFA, lfPDGFA and PDGF-morpholino-injected embryos, extension of the notochord region is reduced (Fig. 6H-J). At this stage, chordin is normally expressed in a broad anterior area,marking the head mesoderm, and in a narrow strip behind, indicating the extending notochord region (Fig. 6K). In dnPDGFA, lfPDGFA and PDGF morpholino injected embryos, the chordin expression domain is compressed along the anteroposterior axis, showing that its anterior end has moved less far animally(Fig. 6L-N). Likewise, compared with controls (Fig. 6O), the gsc domain has moved less far from the dorsal lip in injected embryos, and the archenteron is less well developed(Fig. 6P-R). Taken together,this suggests that mesoderm movement is attenuated in PDGF signaling-defective embryos. Explanted dorsal lips from control(Fig. 6S), dnPDGFA injected(Fig. 6T),lfPDGFA-overexpressing (Fig. 6U) or PDGF-morpholino-injected(Fig. 6V) gastrulae all elongate similarly, showing that convergent extension is not affected. This suggests that mesoderm migration is the primary target of PDGF signaling, and that impairment of convergent extension in the intact embryo is indirect.

Deficient mesoderm movement most probably contributes to the aberrant later development of PDGF signaling-defective embryos. As previously described(Ataliotis et al., 1995),expression of inhibitory PDGFR37 results in diminished head structures, a shortened, bent axis and split tails (Fig. 7A). A comparable phenotype was obtained after expression of dnPDGFA or wild-type lfPDGFA in the BCR(Fig. 7C,D). Phenotypes were less pronounced after injection of PDGFA morpholino(Fig. 7B), 5-mismatch control morpholino had no effect (Fig. 7E).

It had been shown previously that inhibiting PDGFA signaling in the Xenopus embryo by expression of PDGFR-37 does not disrupt mesoderm formation and patterning, but interferes with gastrulation movements(Ataliotis et al., 1995). Presently we have demonstrated that blocking PDGFA/PDGFRα function specifically affects a guidance mechanism that orients mesoderm migration through directional cues associated with the extracellular matrix of the BCR.

We have determined that mesodermal expression of the same dominant negative PDGFR-37 used by Ataliotis et al.(Ataliotis et al., 1995)interferes with directional movement on conditioned substratum. This suggests that the previously observed gastrulation defects can at least in part be explained by an impaired guidance mechanism. Further, our results show that for directional translocation, PDGFA/PDGFRα signaling is required during actual migration, and not during mesoderm development: treatment of mesoderm explants with various agents during the 1 hour period of migration on conditioned substratum is sufficient for interference with directionality, and normal mesoderm explants move non-directionally on substratum conditioned with PDGFA-defective BCR.

In all these cases, migration is not inhibited, but randomized with respect to direction, ultimately resembling translocation on an artificial isotropic FN substratum. Coupled with the fact that blocked PDGFA/PDGFRα signaling has the same effect on directionality as a spatially homogeneous, artificial PDGF signal, this argues that PDGFA/PDGFRα activity is instructive, and directly involved in establishing the direction of migration. However, we cannot strictly rule out the possibility that PDGF signaling is permissive,but is required at exactly the normal level of activity. As expected for a matrix-dependent guidance cue, only the lfPDGFA that contains the matrix-binding motif is able to disrupt directional migration when overexpressed in the BCR.

Several possibilities exist to explain how substratum-dependent cues may guide translocation. One set of mechanisms involves the trapping of randomly migrating cells at a target site. For example, haptotaxis leads to the accumulation of cells in regions of increased substratum adhesiveness(Carter, 1965). However,BCR-conditioned substratum exhibits no spatial differences in adhesiveness(Winklbauer and Nagel, 1991). Other trapping mechanisms can be envisioned, for example, mechanisms based on the differential stimulation of motility by the substratum. However, these options are all restrained by the fact that in Xenopus, the mesoderm migrates as a coherent mass, where cells have limited freedom for independent random movement. On conditioned substratum, whole explants move persistently towards the animal pole, without random excursions to explore the substratum. By contrast, single mesoderm cells migrate randomly, but they are unable to accumulate in the animal pole area on conditioned substratum(Winklbauer et al., 1992). Together, these observations argue against a trapping mechanism of mesoderm guidance.

A different type of guidance mechanism relies on orienting each migrating cell towards its target, as in eukaryote chemotaxis. The hallmark of such a mechanism, the presence of polarized cells all pointing with their locomotory processes towards the target of migration, is actually observed in the Xenopus gastrula, in the shingle arrangement of anterior dorsal mesoderm cells (Winklbauer and Nagel,1991). This arrangement also forms when explants migrate directionally on conditioned substratum, but not when they move randomly on FN(Winklbauer et al., 1992). In the embryo, the shingle arrangement disappears when PDGFA function is compromised by inhibition or overexpression. In both cases, protrusions point with equal frequency towards and away from the animal pole. Apparently, a precisely controlled mode of PDGFA/PDGFRα signaling is essential for correct mesoderm cell orientation, such as when a gradient of matrix-bound PDGFA serves as a chemoattractant for the migratory cells. In situ hybridization data showing that PDGFA expression in the BCR is strongest at the animal pole, and decreases towards the marginal zone(Ataliotis et al., 1995), are consistent with this hypothesis.

Chemotaxis through PDGFRα signaling is well documented (e.g. Westermark et al., 1990; Kundra et al., 1994; Hansen et al., 1996; Ronnstrand et al., 1999), and PDGFRα promotes a chemotactic response at least in some cells (e.g. Hosang et al., 1989; Ferns et al., 1990; Rosenkranz et al., 1999; Yu et al., 2001), although probably through a different mechanism (for reviews, see Hayashi et al., 1995; Ronnstrand and Heldin, 2001). A great deal of work on PDGF-stimulated chemotaxis relates to cultured cells in vitro. Recently, an in vivo role for the Drosophila PDGF/VEGF receptor in guiding border cell migration during oogenesis has been demonstrated (Duchek et al.,2001). By showing that PDGFA/PDGFRα signaling is required for directional migration of mesoderm on its endogenous matrix substratum, as well as for cell orientation in the embryo, we demonstrate such a guidance function for this receptor-ligand pair for the vertebrate embryo.

Single mesoderm cells translocate vigorously in vitro. Their lamellipodia continuously extend, retract, divide and shift along the cell margin. When lamellipodia of two cells meet, they retract immediately, showing contact inhibition; however, lamellipodia are not prevented from underlapping another cell in a lamellipodia-free region(Winklbauer and Selchow, 1992; Winklbauer et al., 1992). In the embryo, the mesoderm forms a coherent mass, and random protrusive activity should lead to frequent encounters between processes and to their retraction as result of contact inhibition. This could explain the rarity of such protrusions in explants on artificial FN substratum(Winklbauer et al., 1992), but also in disoriented mesoderm on PDGFA-defective BCR. That cells extend processes more frequently in control embryos would then be an indirect consequence of their parallel orientation, which prevents collision of lamellipodia. Consistent with this, protrusion formation of single cells does not depend on PDGF signaling, and extension of cytoplasmic processes is not increased, but decreased on BCR overexpressing lfPDGFA, arguing against a direct role for PDGFA in the stimulation of protrusive activity.

The immediate function of the PDGF-mediated guidance mechanism would then be to orient cell processes into a parallel array, to ensure that protrusions can form on every cell, and that processes do not mutually neutralize their effects by pointing in random directions. In turn, to engage all cells at the interface with the substratum in active migration seems essential if the mesodermal mass is to move as a whole.

Facilitating the shear movement between mesoderm and BCR is crucial, as is evident from the phenotypes of PDGF signaling-compromised embryos. That phenotypes are similar, regardless of whether PDGFA function is diminished or increased, argues that they are due to a common primary defect. Moreover,similar phenotypes are obtained when migration is blocked by FN antibodies(Ramos and DeSimone, 1996), or when the shear between BCR and mesoderm is mechanically alleviated by the fusion of the two layers (Winklbauer et al., 2001). This suggests that the common denominator is the attenuation of mesoderm advance across the BCR, as betrayed by marker gene expression patterns. Because in all cases, convergent extension(Keller, 2002) is not deficient in blastopore lip explants, shortened axes and the lack of blastopore closure must be secondary effects, caused by indirect inhibition of convergent extension in the embryo through arrested anterior mesoderm movement. PDGFA and its receptor are also expressed in a complementary pattern at later stages, e.g. in the cranial neural crest(Ho et al., 1994), and interfering with its late functions may contribute to the larval phenotype observed, for example, to head reduction.

As in the frog, PDGFA and PDGFRα show complementary expression in the mouse gastrula, with the ligand in the epiblast and visceral endoderm, and the receptor in the mesoderm (Palmieri et al.,1992; Orr-Urtreger and Lonai,1992). The role of PDGFA and PDGFRα in mouse mesoderm migration is not clear, however. The mutant patch includes the deletion of the PDGFRα (Smith et al., 1991,Smith et al.,1991; Stephenson et al.,1991). Homozygous patch embryos are defective in mesoderm development, but the primary cause for this has not been revealed(Orr-Urtreger et al., 1992). Embryos carrying a targeted null mutation for PDGFRα accumulate various defects during early development, but again, data bearing directly on mesoderm migration are lacking (Soriano,1997). The same is true for embryos homozygous for targeted null alleles of PDGFA (Bostrom et al.,1996).

In the zebrafish, PDGFA and PDGFRα are co-expressed in all cells of the gastrula (Liu et al.,2002a; Liu et al.,2002b). Despite this striking difference in expression, which suggests autocrine signaling, PDGFA is involved in the control of mesoderm migration, as in Xenopus. However, the role of PDGFA differs. In the fish, PDGFA signaling through PI3-kinase and PKB is needed for cell polarization and protrusion formation; inhibition of this pathway decreases protrusive activity and the velocity of migration, but cells are still oriented towards the animal pole (Montero et al., 2003). By contrast, Xenopus anterior mesoderm cells are intrinsically polarized(Winklbauer and Selchow, 1992; Wacker et al., 1998), and the frequency of protrusion formation is only indirectly affected in PGDFA misexpressing embryos. In Xenopus, PDGF signaling is required for the orientation of cells towards the animal pole, and for directional migration.

This work was supported by a Canadian Institutes of Health Research grant(MOP-53075) and by a Canada Foundation for Innovation grant to R.W.; National Institutes of Health (RO1 CA87375) to K.S.; Massachusetts Department of Public Health Breast Cancer Research Grant (DPH-34080066071) to E.T.; and aided by a grant from the American Cancer Society, Massachusetts Division to K.S. Part of the work was performed when R.W. and M.N. were at the Max-Planck-Institute for Developmental Biology, Tuebingen, Germany, and supported by the DFG. We are grateful to Drs Eddy De Robertis, Herbert Steinbeisser, Michael Sargent and Mark Mercola for plasmids. We thank the members of our laboratories for discussions; Hiromasa Ninomiya and Mark Makowiecki for suggestions to improve the manuscript; and Carl-Philipp Heisenberg for communication of a manuscript before publication.