The RhoGEF Pebble is required for cell shape changes during cell migration triggered by the Drosophila FGF receptor Heartless

The spatiotemporal pattern of cell migration in the embryo is crucial for cells to receive and send out signals essential for correct specification and differentiation of tissues. The ability of cells to become motile and to migrate into certain target territories of the organism must be appropriately controlled in order to prevent misplacement of cells(Forbes and Lehmann, 1999; Lauffenburger and Horwitz,1996; Montell,1999; Winklbauer and Keller,1996). Cell migration in the developing organism is likely to be regulated at multiple levels and involves the integration of signal recognition, signal transduction, cell substratum interactions and the cytoskeleton. Fibroblast-growth-factor receptors (FGFR) have been implicated in early steps of directional cell migration in development(Szebenyi and Fallon, 1999). One function of FGFRs in vivo is to recognize chemotactic gradients of FGF that direct cell migration in the embryo(Sato and Kornberg, 2002; Sutherland et al., 1996; Yang et al., 2002).

How are signals from FGFR activation transduced to the cytoskeletal and adhesive systems of the cell to elicit the migratory behavior of the cell? One of the best-characterized FGFR signaling pathways involves the Ras GTPase,which among other targets promotes activation of MAP kinase(Campbell et al., 1998). MAP kinase phosphorylates a variety of protein targets including transcription factors, other protein kinases, phospholipases or cytoskeletal proteins. For example, MAP kinase promotes phosphorylation of myosin light chain (MLC)kinase, a regulator of the contractile actin-myosin system(Klemke et al., 1997). Although many downstream components of FGFR signaling pathways have been identified, the targets that regulate cell migration are not well understood.

Migratory cells form a variety of characteristic cellular protrusions, most prominently filopodia and lamellipodia(Lauffenburger and Horwitz,1996). The formation of these protrusions in response to extracellular stimuli is controlled by small GTPases of the Rho family: Rho,Rac and CDC42 (Hall, 1998). Mammalian cell culture systems have provided ample evidence that Rho GTPases represent the key molecules in transducing extracellular signals to the actin cytoskeleton (Schmidt and Hall,1998). Guanine nucleotide exchange factors (GEFs) act upstream of Rho GTPases and promote their local activation in the cell. Although FGFRs might be required and sufficient for the formation of filopodial protrusions,as shown for tracheal cell migration in Drosophila(Ribeiro et al., 2002), the molecules that connect FGFR signaling pathways to the regulators of the cytoskeleton during morphogenesis are yet to be defined.

Mesoderm migration in the early Drosophila embryo provides a good model system to study FGFR-dependent cell migration(Wilson and Leptin, 2000). Migration of the mesoderm is an important precondition for the regional specification of different mesodermal derivatives. The most dorsal mesodermal fates are controlled by combinatorial action of multiple signaling pathways and intrinsic factors. These inputs govern the formation of a subset of dorsal mesodermal derivatives, marked by the expression of even skipped(eve) (Carmena et al.,1998; Halfon et al.,2000; Knirr and Frasch,2001). In mutants in which mesoderm migration is affected, e.g. in embryos mutant for the FGFR homolog Heartless (HTL), mesoderm cells are unable to receive appropriate signals. As a consequence, the dorsal eve-expressing mesoderm cells are not specified in htlmutant embryos (Beiman et al.,1996; Gisselbrecht et al.,1996; Shishido et al.,1997).

Although the ligand of HTL is yet unknown, the synthesis of proteoglycans is required for HTL activation (Lin et al., 1999). The signal transduction pathway downstream of HTL involves the product of the downstream of FGF (dof; stumps – FlyBase) gene(Imam et al., 1999; Michelson et al., 1998a; Vincent et al., 1998). Genetic experiments suggest that DOF functions as adaptor protein and operates downstream of the receptor and upstream of signals triggered by RAS1. Furthermore, activation of MAP kinase was observed in a subgroup of migrating mesoderm cells and was shown to depend on HTL and DOF function(Gabay et al., 1997; Vincent et al., 1998).

We show that HTL is necessary for protrusive activity of migrating mesoderm cells. The early attachment of mesoderm cells to the ectoderm is affected in htl mutants, suggesting that HTL is required for the adhesion of mesoderm cells to the ectoderm. Our data provide evidence that HTL exerts a permissive function in mesoderm migration. In a genetic screen, we identified the Rho GEF Pebble (PBL) as an essential molecular component of mesoderm migration. Like HTL, PBL is required for specific cell shape changes and protrusive activity of the mesoderm cells during migration. The function of PBL in cell migration is independent of the well-known function of PBL in cytokinesis (Hime and Saint,1992; Lehner,1992; Prokopenko et al.,1999). Overexpression of HTL or a constitutively active form of HTL using the twist (twi) promoter, are sufficient to trigger late cell shape changes in the mesoderm, but are unable to initiate early cell shape changes in pbl mutants. These results indicate a novel function of PBL in HTL triggered cell migration in the Drosophila gastrula.

The following stocks were used and obtained from the Bloomington stock Center unless otherwise indicated: Tp(3;2)C309/TM2, C(3)se,Df(3L)pbl^NR/TM3(ftz::lacZ),Df(3L)pbl^XI/TM3(ftz::lacZ), twi::Gal4(2x), twi::Gal4(2x);Dmef2::Gal4 (Ranganayakulu et al.,1998), UAS::pbl^3.2(Prokopenko et al., 1999), C(1)DX;twi::Gal4;Y (N. Brown, Cambridge, UK), twi::CD2 cn,htl^AB42/TM3(ftz::lacZ), htl^YY26/TM3(ftz::lacZ),pbl³/TM3(ftz::lacZ),pbl^11D/TM3(ftz::lacZ) (C. Lehner, Bayreuth, Germany), pbl⁵/TM3(ftz::lacZ) (Rob Saint, Canberra,Australia, stg^7M/TM3(ftz::lacZ),UAS::Rho1^N19, UAS::Rho1^V14 (M. Mlodzik,New York, USA), UAS::RAS1^V12, UAS::λhtl,UAS::htl^M. More information on translocation stocks and compound chromosomes used in the screen are provided elsewhere(Müller et al.,1999).

Embryos were obtained, staged, fixed and immunolabeled as described(Müller and Wieschaus,1996). For f-actin staining, embryos were fixed with 37%formaldehyde/heptane for 5 minutes followed by devitellinization using 80%ethanol. For cross-sections, fluorescently labeled embryos were embedded in Technovit 7100 (Hereaus, Germany) following the manufacturer's instructions. Sections (7 μm) were obtained with a Reichert Jung Microtome and mounted on slides. As anti-bleaching procedure, sections were incubated for 20 minutes in 50 mM DABCO in PBS and mounted in DABCO/MOWIOL. Embryos were genotyped using balancer chromosomes carrying lacZ transgenes. Fluorescence microscopy was performed with a Leica-TCS NT confocal microscope. For whole-mount staining embryos were double labeled with appropriate secondary antibodies conjugated to alkaline phosphatase or biotin. Biotinylated secondary antibodies were detected using the ABC-kit from Vectastain (Vector,USA). Whole-mount stained embryos were dehydrated in ethanol and acetone and embedded in araldite. Embryos were then either mounted and oriented in araldite on microscope slides or embedded for sectioning. Sections were cut at 5 μm and mounted in araldite. Light microscopy was performed on a Zeiss Axiophot. Images were processed using Adobe Photoshop on an Apple Computer.

Preparation of embryos for transmission electron microscopy was performed as described by Müller and Wieschaus(Müller and Wieschaus,1996) with few modifications. For genotyping, the embryos were prefixed at the interphase of 25% glutaraldehyde in 0.1 M phosphate buffer pH 7.4 and heptane for 25 minutes. The embryos were then hand-peeled and rinsed in X-Gal staining buffer (0.15 M NaCl; 1 mM MgCl₂; 0.01 M sodium phosphate buffer, pH 7.2; 3.3 mM K₃Fe(CN)₆; 3.3 mM K₄Fe(CN)₆). For X-Gal staining, embryos were incubated overnight with 10% 5-Br-4-Cl-3-Indolyl-β-D-galactoside in X-Gal buffer at 16°C. The embryos were then sorted and processed as described. Micrographs were taken at a Zeiss EM109 transmission electron microscope.

The following antibodies were used: mouse anti-dpERK (Sigma); rabbit anti-Twist (Siegfried Roth, Cologne); rabbit anti-βGal (Cappel); mouse anti-βGal (Promega); mouse anti-EVE, mouse anti-Engrailed (EN), mouse anti-Neurotactin (NRT; DSHB, Iowa); mouse anti-CD2 (Serotec, Germany),Phalloidin-Alexa568 (Molecular Probes, USA), rat anti-DE-Cadherin and rat anti-Dα-Catenin (Hiroki Oda, Japan).

In Drosophila the mesoderm originates from a ventral population of cells in the monolayered blastoderm epithelium(Costa et al., 1993). At the onset of gastrulation these cells are first internalized through an invagination of the epithelium. After internalization, the cells undergo mitosis, lose their epithelial characteristics, and start to spread as an aggregate between the central yolk sac and the basal cell surfaces of the ectoderm (Oda et al.,1998).

To follow cell shape changes of mesoderm cells, we used a transgene driving expression of the transmembrane protein CD2 from rat under the control of the twist (twi) promotor (twi::CD2)(Dunin-Borkowski and Brown,1995). twi::CD2 is already expressed during invagination and represents a cell-surface marker specific for the mesoderm. Mesoderm migration can be divided into three phases with characteristic cell shape changes (Fig. 1). After invagination, the mesoderm initially forms an epithelial tube(Fig. 1A,D). At phase 1 of migration, the surface of the mesoderm cells appears relatively smooth(Fig. 1D). After disassembly of the epithelial tube and mitosis, phase 2 begins, in which the mesodermal aggregate migrates out in dorsolateral direction(Fig. 1B,E). Cells at the leading edge of the aggregate are stretched along the dorsoventral axis and extend multiple cellular protrusions. The longest cellular protrusions often measure half to two-thirds the size of a cell diameter (to a length of 10-15μm in fixed samples). Cross-sections revealed that not only the leading edge cells exhibit this polarized morphology, but that the cells immediately following the leading edge cells frequently also extend in dorsolateral direction (Fig. 1M). In this work, we will be using the term `protrusive activity' to describe the formation and/or the dynamics of the filoform and lamelliform protrusions that we observed in our fixed preparations. The protrusive activity is specific for the migratory phase, because when the cells have reached their final positions(phase 3) and form a coherent monolayer, large extensions are absent and only few filoform protrusions were observed(Fig. 1C,F).

Examination of embryos homozygously mutant for htl revealed that HTL is required for cell shape changes during phase 1 and 2 of mesoderm migration. In phase 1, the mesodermal epithelial tube extends further into the interior of the embryo when compared with wild type(Fig. 1A,G; see below). During phase 2, the leading edge cells did not extent dorsolaterally(Fig. 1H,K). This phenotype is not simply explained by the possibility that htl mutant mesoderm cells were not able to contact the ectoderm, because cells directly apposed to the ectoderm also failed to extend (Fig. 1H). In phase 3, mesoderm cells of htl mutant embryos did not establish a monolayer configuration(Fig. 1I). Interestingly,during and after phase 3, htl mutant mesoderm cells exhibit directional protrusions suggesting that some migration might occur at these stages (Fig. 1L,N). This result indicates that HTL is not generally required for protrusive activity of the mesoderm cells. As reported earlier, this late migration in htlmutants is never able to rescue the defects in mesoderm differentiation, most probably because of a second requirement of HTL for mesoderm differentiation(Michelson et al., 1998b).

Our observations suggest that HTL is required for the early interaction of the mesoderm with the ectoderm. By analyzing sections, we observed that wild-type cells at the base of the mesodermal tube are attached to the basal surfaces of the ectoderm (Fig. 2A). By contrast, htl mutant mesoderm cells at the respective stage and position failed to establish contact to the ectoderm(Fig. 2C). This phenotype correlates well with a misalignment of the mesodermal tube in htlmutants (Fig. 2B,D; Fig. 1J). We conclude that HTL is required for the effective attachment of mesoderm cells to the ectoderm,which might promote the protrusive activity of mesoderm cells during migration.

htl and dof represent the only zygotically expressed genes that have thus far been described to be essential for mesoderm migration. To obtain a better insight into the genetic control of mesoderm migration, we performed a genetic screen to identify zygotically expressed genes involved in mesoderm migration (T.G. and H.A.J.M., unpublished).

Three loci mapped to the third chromosome and were characterized using chromosomal deletions and available point mutations. Two loci corresponded to the genes htl and dof, respectively. Embryos lacking the chromosomal interval 61A to 68 (based on breakpoints of the chromosomal translocation T(2;3)C309) displayed defects in mesoderm migration(Fig. 3A-C). Genetic mapping revealed that small overlapping chromosomal deletions, which exhibited the phenotype, all removed the gene pbl (data not shown). Analysis of a strong loss-of-function point mutation in pbl, pbl³,indicated that pbl is required for mesoderm morphogenesis. Embryos homo- or hemizygously mutant for pbl³ show a dramatic reduction in the number of EVE-positive mesoderm cells at the extended germband stage (Fig. 3D,E). These results demonstrate a thus far unrecognized function for pbl in mesoderm differentiation.

We next examined whether the defects in mesoderm differentiation in pbl mutants are based upon a requirement of pbl for cell migration. While internalization of the mesoderm occurred normally, the characteristic cell shape changes during mesoderm migration were blocked in pbl mutants (Fig. 3F,G). As in htl mutants, the cells at the base of the mesodermal tube were unable to attach to the ectoderm. In phase 2, pbl mutant cells were in close apposition to the ectoderm, but failed to extend dorsolaterally and no protrusions were observed. Distinct from htl mutants, however, protrusive activity was also blocked during phase 3 in pbl mutants, suggesting that PBL represents an essential component of the mechanisms driving protrusive activity of the cells(Fig. 3H). To further characterize the cellular protrusions in our fixed preparations, we examined f-actin distribution in the mesoderm cells of wild-type and pblmutants. We found that the protrusions in wild-type embryos contain f-Actin. In pbl mutant mesoderm cells an f-actin rich cell cortex was equally distributed at the circumference of the cells with no protrusions present(Fig. 3I,J). These results indicate that PBL is required for cell shape changes during migration of the mesoderm cells, which might involve the reorganization of the cortical f-actin cytoskeleton.

It remained possible that the defects in pbl mutants in cell migration might be a secondary consequence of a failure in the epithelial/mesenchymal transition of the mesoderm after invagination. During epithelial/mesenchymal transition cells loose their epithelial polarity and gain mesenchymal character. We compared the presence of adherens junctions in the mesoderm cells in wild type, htl and pbl mutant embryos by electron microscopy. Apical adherens junctions, a hallmark of apicobasal polarity in epithelial cells, are lost from the mesoderm cells during phase 1 in the wild type as well as in embryos homozygous for htl or pbl (see Fig. S1). Similarly, immunolabeling with antibodies against the ectodermal epithelial marker DE-cadherin or the adherens junction marker Dα-Catenin showed that downregulation of epithelial characteristics occurs normally in pbl and htl mutants (see Fig. S2). In summary these results indicate that PBL is not required for the loss of epithelial character of mesoderm cells, but that PBL is essential for the gain of mesenchymal characteristics of mesoderm cells after invagination.

pbl encodes a RHO1-GEF most similar to the vertebrate ect2 proto-oncogene (Miki et al.,1993; Prokopenko et al.,1999; Tatsumoto et al.,1999). Both pbl and ect2 are required for the assembly of the contractile actin ring during cytokinesis. Interfering with the function of PBL or ECT2 results in a failure of cytokinesis and the generation of multinucleate cells (Fig. 3G,H). Because mutations in pbl affect cell shape changes before mitoses in the mesoderm occur, we suspected that the requirement of pbl for mesoderm migration might be independent from its cytokinesis function. To determine, whether the defects in mesoderm migration in pbl mutants are direct rather than a secondary consequence of the failure in cytokinesis, we analyzed the pbl phenotype in division-defective embryos.

Postblastoderm mitotic divisions are controlled by zygotic expression of the cell cycle regulator String (STG)(Edgar and O'Farrell, 1990). As mesoderm migration and specification of EVE-positive mesoderm cells occur normally in stg mutant embryos, this mutation provides a genetic condition to assay cytokinesis-independent functions of pbl(Fig. 4B,E)(Carmena et al., 1998; Leptin and Grunewald, 1990). The cytokinesis defect of pbl was completely blocked by stg(Fig. 4F). In pbl stgdouble mutant embryos, migration of the mesoderm and specification of dorsal mesodermal derivatives was impaired similar to pbl single mutants(Fig. 4C,F). Moreover, cell shape changes in phase 2 occur normally in stg mutant embryos, but protrusive activity of the mesoderm cells was blocked in pbl stgdouble mutants (Fig. 4E,F). These results indicate that the activity of pbl is required for mesoderm migration even in the absence of mitosis and thus in the absence of cytokinesis defects. We therefore conclude that PBL has independent functions in cytokinesis and cell migration, respectively.

As the functions of PBL can be separated genetically, we wanted to investigate whether the two functions might involve different molecular mechanisms.

Genetic data indicate different requirements of the PBL protein for its two functions. We detected a significant difference in the strength of the migration phenotype of two distinct mutant alleles of pbl. Embryos homozygous for the strong loss of function allele pbl³exhibit on average only one hemisegment that contains EVE-positive mesoderm cells, compared with 22 EVE-positive hemisegments in the wild type(Fig. 5A,B; Table 1). By contrast, embryos mutant for pbl^11D exhibit an average of 12 hemisegments that contain EVE-positive cells and few pbl^11D homozygous embryos (2/45) even contained the wild-type number of EVE-positive cells(Fig. 5C; Table 1). Importantly, pbl³ and pbl^11D homozygotes exhibit equally penetrant defects in cytokinesis(Fig. 5D,E) (see also Cui and Doe, 1995; Weigmann and Lehner, 1995). These allele-specific differences demonstrate distinct requirements of the PBL protein in cell migration and cytokinesis, and suggest that the migratory function of PBL might be acting through a mechanism distinct from the cytokinesis function.

During cytokinesis, PBL performs a conserved function as a GEF for RHO1 in the formation of the contractile ring(Prokopenko et al., 1999; Prokopenko et al., 2000b; Tatsumoto et al., 1999). Expression of a constitutively active form of RHO1(Rho1^V14) in the mesoderm cells of pbl mutants was unable to rescue the migration defects in pbl mutant embryos(data not shown). This negative result is difficult to interpret, because expression of Rho1^V14 did not rescue the cytokinesis defects in pbl mutants either. We therefore tested the role of RHO1 in mesoderm migration directly by expressing a dominant-negative version of RHO1 (Rho1^N19) in the mesoderm(Strutt et al., 1997). Although expression of UAS::Rho1^N19 with twi::Gal4 led to cytokinesis defects of mesoderm cells similar to pbl mutants, specification of EVE-positive mesoderm cells remained unimpaired (Fig. 5F,H). Importantly, mesoderm cells expressing Rho1^N19 still showed the characteristic protrusive activity during phase 2 of migration(Fig. 5G). The fact that expression of Rho1^N19 led to a fully penetrant cytokinesis defect suggests that the function of endogenous RHO1 is efficiently blocked by the expression of the dominant-negative form(Fig. 5H). Although it might be possible that small amounts of RHO1 activity might be sufficient to support cell migration, further raising the level of Rho1^N19 in the mesoderm, by using twi::Gal4 in combination with a Dmef2::Gal4 driver did not affect the presence of EVE-positive mesoderm cells (Fig. 5I,J).

These data support our conclusion that the defects in cell migration in pbl mutants are not a secondary consequence of a failure in cytokinesis. The finding that cell migration is not affected in Rho1^N19-expressing cells suggests that RHO1 function is dispensable for mesoderm migration. In summary, we propose that the two functions of PBL are independent and might involve distinct molecular mechanisms in cytokinesis and cell migration, respectively.

Mesoderm migration depends on the mesoderm-specific expression and activity of both the HTL receptor and its putative cytoplasmic adaptor DOF(Beiman et al., 1996; Gisselbrecht et al., 1996; Michelson et al., 1998a; Shishido et al., 1993; Vincent et al., 1998). By contrast, pbl is expressed in all cells of the embryo and is required in all cells for cytokinesis in postblastoderm embryos(Hime and Saint, 1992; Lehner, 1992; Prokopenko et al., 2000a).

We sought to assess the cell-type specific requirements of PBL during mesoderm migration by rescuing the pbl mutant phenotype in a tissue-specific manner. A UAS::pbl transgene containing the full-length pbl cDNA was expressed in the mesoderm cells in a pbl mutant background using twi::Gal4. The cytokinesis defect and the mesoderm migration defect of pbl mutants was completely rescued by mesoderm-specific expression of pbl(Fig. 6A-C). Moreover,expression of EVE in dorsal mesoderm cells was restored(Fig. 6D,E). The cells in the ectoderm of such embryos still exhibited cytokinesis defects, indicating that the activity of the transgene was specific for mesoderm cells and that the function of PBL in the ectoderm is not essential for mesoderm migration(Fig. 6B). We therefore conclude that pbl acts in a mesoderm autonomous fashion to allow proper migration of the cells.

Because the function of pbl in mesoderm migration is specific for mesoderm cells, we next investigated how PBL function relates to signals derived from HTL. Molecular and genetic epistasis has shown that HTL activates the conserved Ras/Raf/MAP kinase pathway during mesoderm migration and differentiation (Gabay et al.,1997; Michelson et al.,1998a; Michelson et al.,1998b; Vincent et al.,1998). To assess the function of PBL in HTL-dependent migration,we asked whether pbl is required for the activation of downstream components of the HTL pathway by measuring MAP kinase activation in pbl mutants. Strikingly, MAP kinase is still activated in pbl mutants in a pattern similar to that of wild-type embryos(Fig. 7H,I). This result indicates that PBL function is dispensable for HTL-dependent activation of MAP kinase. We therefore conclude that PBL is unlikely to act upstream of the HTL receptor and its signaling cascade.

The mesoderm defects of htl homozygous embryos can be completely rescued by expression of UAS::htl using twi::Gal4(Fig. 7A,B)(Michelson et al., 1998b). When, instead, the constitutively active form of HTL, λHTL, is expressed under the same conditions, the cellular defects in early mesoderm migration were rescued, but mesoderm differentiation defects were only partially rescued (Fig. 7C,D; Table 1)(Michelson et al., 1998b). These results indicate that expression of λHTL with the Gal4/UAS system is sufficient to trigger the early cell shape changes in htlhomozygously mutant embryos.

To test whether pbl is required for HTL triggered cell shape changes in mesoderm migration, we overexpressed wild-type or constitutively active forms of HTL in a pbl mutant background. Both, expression of HTL or λHTL, were unable to trigger early cell shape changes in pbl homozygous embryos (Fig. 7E and data not shown). Although HTL or λHTL expression failed to rescue phase 1, some mesoderm spreading was observed in phase 3 and later (Fig. 7G). Because the rescue in phase 3 was rather moderate and variable, we examined mesoderm differentiation by analyzing the number of EVE-positive hemisegments. Mesoderm differentiation defects in pbl homozygotes are rescued significantly by expression of λHTL but not by expression of HTL(Fig. 7F; Table 1). This result suggests that mesoderm migration might be delayed in these embryos, in a way that might render signaling events leading to EVE expression ineffective. In addition,the partial rescue of pbl mutants by λHTL might reflect a second function of the receptor in differentiation of dorsal mesodermal derivatives (Michelson et al.,1998b). These results are consistent with a role of PBL for early cell shape changes triggered by HTL. The fact that λHTL rescues some late defects in pbl mutants suggests the presence of additional mechanisms required for mesoderm migration, which might be independent of pbl.

HTL exerts different functions in mesoderm morphogenesis(Michelson et al., 1998b). We analyzed the requirements of HTL for cell shape changes during mesoderm migration. Our analysis of cell shape changes revealed that the cells extend in the direction of their migration and form long cellular protrusions. These observations provide evidence that the leading cell rows of the mesodermal aggregate migrate directionally in dorsolateral direction.

Although the ligand of the HTL receptor is unknown, it seems unlikely that signals mediated by HTL provide the only directional cues for mesoderm migration. Localization of activated forms of MAP kinase in the leading edge cells during mesoderm migration suggests that local activation of the receptor might be important for proper migration(Gabay et al., 1997). In addition, activation of HTL in a ligand-independent fashion throughout the mesoderm is unable to completely rescue the htl mutant phenotype(Michelson et al., 1998b)(this paper). Thus local activation of the receptor occurs in vivo and is essential for proper mesoderm migration. We show that HTL is required for protrusive activity only during phase 1 and 2 of mesoderm migration. However,HTL activity is not essential for the protrusive activity of the cells per se,because cells do extend dorsolaterally during phase 3 in htl mutant embryos. These data demonstrate that HTL activation is unlikely to provide the only directional cue in mesoderm migration. The results presented in this paper suggest that HTL signaling provides temporal information for protrusion formation during phases 1 and 2, and might be therefore acting as a permissive factor during mesoderm migration.

PBL is required for protrusive activity of mesoderm cells also in phase 3 and later. It is therefore possible that PBL function might be required in a more general way for the cell to extend protrusions. The specificity of PBL for protrusive activity is also supported by the fact that loss of epithelial characteristics is unaffected in pbl mutant embryos. Although the specific mechanism of PBL function in cell migration is currently unknown, it is important to note that not all morphogenetic movements are compromised in pbl mutants. For example, cephalic furrow formation, invagination of the ventral furrow and germband extension movements, which all depend on a functional cytoskeleton are normal in pbl mutant embryos (data not shown). We therefore propose that PBL might constitute an important component for cytoskeletal changes, which are triggered by FGFR signaling events.

Of the multiple responses generated downstream of FGFR activation, only little is known of the molecular pathways by which FGFRs trigger cell shape changes in vivo. The Rho GEF PBL represents a good candidate for mediating cell shape changes triggered by HTL signaling. Importantly, the early phenotypes of htl and pbl mutants are almost identical,indicating that both gene products are required in a narrow time window for early cell shape changes after invagination of the mesoderm. Furthermore, in both mutants this phenotype is completely penetrant, indicating that the gene products do not act in a redundant fashion.

The function of PBL for mesoderm migration is specific for mesoderm cells. Because htl is expressed only in the mesoderm at this stage of development, PBL might be involved in the presentation of the receptor or its unknown ligand and thus acting upstream, or PBL might be involved in downstream events triggered by the HTL signaling cascade. If PBL was acting upstream of HTL, signaling events downstream of HTL should be blocked in such mutants. By contrast, here we show that PBL is dispensable for activation of MAP kinase in the early mesoderm cells. These results suggest that PBL does not act upstream of HTL and favor a model in which PBL acts downstream of the HTL signaling cascade.

The present results render it unlikely that PBL is directly involved in a signaling pathway downstream of HTL FGFR. In contrast to htl mutants,no cell shape changes and no protrusive activity was observed in pblmutant mesoderm cells in phase 3. In addition, the pbl null mutant phenotype still allows a few cells to undergo eve expression,probably owing to the larger cells and abnormal cytoarchitecture in the division defective embryos. This is in contrast to htl loss of function mutants where EVE-positive mesoderm cells are never observed. If pbl was essential for signaling downstream of the HTL receptor, the phenotype of htl and pbl mutants should be more similar with respect to mesoderm differentiation; for example, the phenotypes of htl and dof mutant embryos are identical(Vincent et al., 1998; Michelson et al., 1998a). We therefore propose that PBL might represent a regulator of the cytoskeleton or adhesive mechanisms of the cell, which provide targets of the HTL signaling cascade to trigger cell shape changes.

Although the activation of MAP kinase in the mesoderm depends on HTL, it is not known whether this is a direct response or whether it is indirect, i.e. MAP kinase may not be directly activated by HTL itself, but through interactions of the mesoderm with the ectoderm. In this case, activation of MAP kinase would be a response rather than a cause of the cell shape changes. The phenotype of pbl mutants, however, argues against the latter possibility, because it shows that in the absence of cell shape changes, MAP kinase can still be activated. This result also suggests that activation of MAP kinase alone cannot account for the cell shape changes that occur. This idea is supported by the fact that activated forms of RAS1 are unable to completely rescue the defects in mesoderm migration of htl or dof mutant embryos, including the defects in cell shape changes in phase 1 and 2 (data not shown) (Vincent et al., 1998; Michelson et al.,1998a; Michelson et al.,1998b; Iman et al., 1999). These results suggest the presence of a signaling pathway acting in parallel to the Ras/Raf MAP kinase pathway to be involved in mesoderm migration.

The pbl gene was originally identified and characterized as an essential factor for cytokinesis (Hime and Saint, 1992; Lehner,1992; Prokopenko et al.,1999). We describe a function of PBL in interphase cells that can be genetically separated from its requirement for cytokinesis.

Two lines of evidence indicate that the function of PBL in cell migration is mediated through a pathway different from the cytokinesis pathway. First,expression of a dominant-negative form of RHO1, which blocks cytokinesis in the mesoderm has no effect on mesoderm migration, cell shape changes associated with it or expression of differentiation markers specific for dorsal mesoderm derivatives. Second, a mutation in pbl,pbl^11D exhibits significantly weaker defects in mesoderm differentiation compared to the strong loss of function mutation pbl³. These allele-specific differences indicate distinct requirements of the PBL protein for its two functions, because both alleles exhibit identical cytokinesis defects and only differ in mesoderm differentiation defects significantly. We therefore propose that the function of PBL for cell migration might not involve RHO1 and might therefore be using another mechanism.

How does PBL act in cell migration and which GTPase represents its substrate? Both PBL and its mammalian orthologs belong to the Dbl family of Rho-GEFs, which promote activation of Rho GTPases through a conserved Dbl-homology (DH) domain (Prokopenko et al., 2000b). The DH domain is required for both functions of PBL,because a missense mutation in pbl, called pbl⁵,in which an amino acid exchange renders the DH domain inactive, exhibits equally strong defects in cell migration and cytokinesis (S.S. and H.A.J.M.,unpublished) (M. Smallhorne, M. Murray and R. Saint, personal communication). Data from yeast two-hybrid assays, as well as genetic interactions indicate that PBL binds to and interacts with RHO1(Prokopenko et al., 1999). During cytokinesis, PBL is proposed to locally activate RHO1, which then interacts with its effector Diaphanous, a Drosophila homologue of the Formin family of actin regulators (O'Keefe et al., 2001; Prokopenko et al., 2000b; Somers and Saint,2003). Although PBL appears to interact with RHO1, but not with CDC42 or RAC1, mammalian homologs of PBL promote GTP/GDP exchange of the GTPases RHO1, RAC1 and CDC42 (Tatsumoto et al., 1999). Because these discrepancies might reflect differences in the sensitivity of the assays applied, it remains to be determined which substrate PBL uses for its function in cell migration.

Although we have detected a role of PBL in FGFR triggered cell migration,it is currently unclear how general the requirement of PBL is for the protrusive activity of migrating cells. Interestingly, mutations in pbl have been discovered in a screen for genes required for the development of the peripheral nervous system(Salzberg et al., 1994). These mutants affected the correct migration of the axons in the PNS without obvious defects in cytokinesis. It will therefore be interesting to assess the function of PBL in a variety of migrating cells to further characterize its potential role as a mediator of cell shape changes triggered by extracellular signals.

We thank the Bloomington Stock Center, the Developmental Studies Hybridoma Bank (Iowa, USA), Nick Brown, Christian Lehner, Alan Michelson, Marek Mlodzik,Hiroki Oda, Siegfried Roth, Rob Saint, David Strutt and Eric Wieschaus for fly stocks and antibodies. We thank Rob Saint and Masha Smallhorne for communicating results prior to publication and Jörg Großhans, Kevin Johnson, Eli Knust, Helen McNeill and Andreas Wodarz for useful comments on the manuscript. We thank Nora Hinssen for excellent technical assistance,Meike Löbbecke for initial studies on cell shape changes and Monika Meyer for preparing fly food. This work was funded by the Deutsche Forschungsgemeinschaft (SFB 590TP-B1 and MU1168/3-1).