FGF8-like1 and FGF8-like2 Encode Putative Ligands of the FGF Receptor Htl and Are Required for Mesoderm Migration in the Drosophila Gastrula

Current Biology, Vol. 14, 659–667, April 20, 2004, 2004 Elsevier Ltd. All rights reserved.

DOI 10.1016/j .c ub . 20 04 . 03 .0 5 8

FGF8-like1 and FGF8-like2 Encode Putative Ligands of the FGF Receptor Htl and Are Required for Mesoderm Migration in the Drosophila Gastrula Tanja Gryzik and H.-Arno J. Mu¨ller* Institut fu¨r Genetik Heinrich-Heine-Universita¨t Du¨sseldorf Germany

Summary Background: Mesoderm migration in the Drosophila gastrula depends on the fibroblast growth factor (FGF) receptor Heartless (Htl). During gastrulation Htl is required for adhesive interactions of the mesoderm with the ectoderm and for the generation of protrusive activity of the mesoderm cells during migration. After gastrulation Htl is essential for the differentiation of dorsal mesodermal derivatives. It is not known how Htl is activated, because its ligand has not yet been identified. Results: We performed a genome-wide genetic screen for early zygotic genes and identified seven genomic regions that are required for normal migration of the mesoderm cells during gastrulation. One of these genomic intervals produces upon its deletion a phenocopy of the htl cell migration phenotype. Here we present the genetic and molecular mapping of this genomic region. We identified two genes, FGF8-like1 and FGF8-like2, that encode novel FGF homologs and were only partially annotated in the Drosophila genome. We show that FGF8-like1 and FGF8-like2 are expressed in the neuroectoderm during gastrulation and present evidence that both act in concert to direct cell shape changes during mesodermal cell migration and are required for the activation of the Htl signaling cascade during gastrulation. Conclusions: We conclude that FGF8-like1 and FGF8like2 encode two novel Drosophila FGF homologs, which are required for mesodermal cell migration during gastrulation. Our results suggest that FGF8-like1 and FGF8-like2 represent ligands of the Htl FGF receptor. Introduction Inductive events emanating from growth factors are important for controlling cell behavior in the developing and adult organism. Fibroblast growth factors (FGFs) are fundamental to many developmental processes, including mesoderm formation, neural development, cardiac and skeletal muscle development, angiogenesis, and limb development [1]. FGFs are particularly interesting for their ability to provide instructive signals during directional cell migration in a variety of organisms. In C. elegans, the FGF EGL-17 can act as an instructive guidance cue for migrating sex myoblasts [2]. In Drosophila, the FGF Branchless (Bnl) acts as a chemoattractant during development of the larval and adult tracheal system [3, 4]. *Correspondence: [email protected]

FGF signaling plays a crucial role during gastrulation in vertebrate and invertebrate development. In the early mouse gastrula, FGF4 and FGF8 are required for the migration of epiblast cells out of the primitive streak and thus for the formation of mesodermal and endodermal tissues [5]. Furthermore, targeted disruption of the FGF receptor-1 from mice results in a failure of the epithelial/ mesenchymal transition during ingression behavior in the gastrula, presumably by regulating E-cadherin and Snail expression [6]. In the chick gastrula, FGF4 and FGF8 provide chemotactic cues that coordinate cell movements during ingression of the epiblast cells through the primitive streak [7]. In Drosophila, the activity of the FGF receptor Heartless (Htl) is required for migration of the mesoderm cells in the early gastrula [8–10]. The ligand of Htl is not known, and searches in the genome database have failed so far to detect other FGF homologs besides Bnl. Htl and its adaptor protein Downstream of FGF (Dof) are specifically expressed in the mesoderm [11–13]. Dof is required for Htl signaling through activation of the Ras/ MAP kinase cassette. Based upon the spatial pattern of MAP kinase activation, it has been proposed that the ligand of Htl might form a gradient within the dorsolateral ectoderm [14]. Furthermore, as in mouse gastrulation, the synthesis of glycosaminoglycans is essential for activation of Htl, indicating that Htl activation requires proteoglycans as coreceptors. Such a requirement is a critical feature typical for FGFs [15, 16]. To better understand the molecular mechanisms that drive mesoderm migration in the Drosophila gastrula, we performed a genome-wide screen for zygotic genes required for early mesoderm morphogenesis. We found seven genomic regions, including dof and htl, that are necessary for normal mesoderm migration. Here we present the results of the genetic characterization of one genomic region, which was of particular interest because the phenotype of embryos deleted for this genomic segment is very similar to that of embryos mutant for htl or dof. Within this region, we identified two closely linked genes with homology to the FGF core domains of FGF8, 17, and 18 from vertebrates and which we therefore called FGF8-like1 and FGF8-like2. We found that these genes were only partially annotated and show that they encode two novel fly orthologs of the FGF family of growth factors. Before and during early gastrulation, both genes exhibit overlapping expression in the neuroectoderm; this overlapping expression indicates the cells that serve as a substrate for mesoderm cells during migration. Interfering with the function of both genes by using RNAi or a chromosomal deletion results in abnormal mesoderm migration during gastrulation. We further show that Htl-dependent activation of MAP kinase requires the presence of FGF8-like1 and FGF8like2. Our results suggest that Htl is activated by two different FGF ligands, which might act in a partially redundant fashion during early mesoderm development.

Current Biology 660

Figure 1. Summary of the Zygotic Requirements for Mesoderm Migration in the Drosophila Gastrula The scheme indicates the four Drosophila chromosomes (I–IV), and the numbers indicate cytological positions based on polytene chromosome bands. The black regions indicate the genomic intervals, which are essential for normal mesoderm migration. In embryos lacking these genomic regions, the mesoderm cells fail to migrate properly during gastrulation. For details on the translocation screen and phenotypes of corresponding embryos, see Supplemental Data. The gray areas represent genomic regions that we were unable to assay because these loci contain genes that are required for the specification of the mesoderm (e.g., sna [35D] or twi [59C]), genes that are required for cell formation (slam [26C]; serendipity a [99D]; bottleneck [100B]), or cell survival (thread [72D]). The interval 82F to 83C was inaccessible because of the presence of the triplolethal gene. The zygotic loci with an essential function in mesoderm migration localize to 1A–12E on the X chromosome, 33D–35B on the left arm of chromosome II, and 46E–49E (marked with an asterisk) and 50C–56D on the right arm of chromosome II. The loci on the third chromosome were mapped and identified as pebble (66A) [23], dof (88C), and htl (90E).

Results A Screen for Early Zygotic Genes Identifies Novel Genes Involved in Mesoderm Development Maternal gene products govern cleavage divisions in early embryos until mid-blastula transition, when embryonic development becomes dependent on zygotic transcription. In Drosophila, exploiting this property of early embryos can help to identify early zygotic gene functions in genetic screens [17–19]. The rationale of these screens is to generate embryos bearing large chromosomal deletions by using chromosomal translocations. An overlapping set of such synthetic deletions that together uncover the entire genome and allow the identification of gene functions, including redundant ones, required for early morphogenesis has been generated [19]. We have utilized this approach to identify genes required for mesoderm migration in the gastrula embryo. The correct migration of the mesoderm was visualized by immunolabeling of Twi to mark the presumptive mesoderm cells [20]. After invagination, the mesoderm cells undergo mitosis and subsequently migrate as an aggregate in a dorsolateral direction until a monolayer of mesoderm cells covers the basal surface of the ectoderm. In our screen, approximately 94.5% of the genome was analyzed, and seven genomic regions that contain essential genes for mesoderm migration were identified. The results of the screen are summarized in a supplemental section (Figure 1; Supplemental Data). Two loci mapped to the right arm of chromosome II. We focused on the chromosomal interval uncovered by Tp(2;3)I.707 because embryos deficient for this region

Figure 2. Embryos Lacking Chromosomal Region 46E–49E Exhibit Defects in Mesoderm Migration Embryos containing a synthetic deletion uncovering the chromosomal interval 46E–49E [from a cross of C(2)v females to Tp(3;2)I.707 males] were fixed and stained with anti Twi (brown) and anti-En (blue) antibodies. In wild-type embryos at mid-gastrulation ([A]; stage 8) or at extended germ band ([C]; stage 9), En marks 14 parasegmental borders. (B and D) In embryos carrying the synthetic deletion 46E– 49E, En stripes are absent because the deletion uncovers the en gene. (E–J) Transversal cross-sections of wild-type (E–G) or deletion embryos (H–J) at stage 7 (early gastrula; [E and H]), stage 8 (midgastrulation; [F and I]), stage 9 (extended germ band; [G,J]). Embryos were stained with antibodies against a universal membrane marker, Neurotactin (Nrt; red), and Twi (green). In the wild-type, the mesoderm spreads out on the neuroectoderm and eventually forms a coherent monolayer after gastrulation (G). In the deletion embryos, invagination occurs normally (H), but the mesoderm cells fail to migrate out on the ectoderm and remain as tight aggregates throughout mid and late gastrulation (I and J).

exhibited the strongest phenotype, and this phenotype was reminiscent of mutations in htl or dof (Figure 2). At the beginning of gastrulation, the mesoderm cells invaginated normally. However, after invagination the mesoderm cells failed to attach to the ectoderm and did not spread out but remained as a tight aggregate, which extended into the interior of the embryo (Figures 2F and 2I). At later stages some cells were attached to the ectoderm, but many cells remained aggregated and never formed a monolayer (Figures 2D, 2G, and 2J). Thus, the chromosomal region 46E to 49E contains a locus that is required for mesoderm migration and exhibits similar phenotypic features as mutations in htl or dof. Mapping of the respective chromosomal interval revealed that none of the available deletions produced a phenotype that resembled that of the translocation. Because the set of deletions used for this analysis did not encompass the entire region uncovered by the translocation, we employed the Drosophila isogenic deficiency kit to construct novel deletions in the region [21] (www.drosdel.org.uk). Of two partially overlapping deletions, Df(2R)ED2230 and Df(2R)ED2238, one produced defects in mesoderm morphogenesis. Although mesoderm migration in Df(2R)ED2230 homozygotes occurred as in the wild-type, Df(2R)ED2238 homozygotes pro-

FGF-like Genes in Drosophila Gastrulation 661

Figure 3. Mesoderm Phenotypes of Overlapping Small Deletions Df(2R)ED2230 and Df(2R)ED2238 Embryos homozygously mutant for Df(2R)ED2230 (A, C, and F) or Df(2R)ED2238 (B, D, and G) were fixed and stained with antibodies against Twi (brown [A–D]; ventral view) or Eve (E–G). (A and C) In Df(2R)ED2230 mutant embryos, the mesoderm cells are aligned along the ventral midline in early stage 8 embryos (A) and spread out laterally during stage 8 (C). (B and D) In embryos mutant for Df(2R)ED2238, the mesoderm cells failed to align along the midline in early stage 8 (B). (D) A mid-stage 8 embryo; spreading of the mesoderm failed. (E–G) Specification of Eve-positive dorsal mesodermal cells is blocked in Df(2R)ED2238 mutant embryos. (E) In the wild-type, eve is expressed in 11 clusters of dorsal mesodermal derivatives at stage 11. In addition, Eve expression is seen in the central nervous system (arrowheads in E and G). (F) In Df(2R)ED2230 embryos, eve expression is largely normal. (G) In embryos mutant for Df(2R)ED2238, eve expression in the dorsal mesoderm is absent. (H–M) Transversal cross-sections of embryos at stage 8 (I and L) or stage 9 (H, K, J, and M) are stained with anti Twi antibodies. (H–J) Df(2R)ED2230 homozygotes. (K–M) Df(2R)ED2238 homozygotes; note that mesoderm cells fail to migrate out on the neuroectoderm.

duced a phenotype similar to that of the synthetic deletion embryos identified in the original screen (Figure 3 and data not shown). In the migration phase, mesoderm cells in embryos homozygous for Df(2R)ED2238 did not spread out and remained associated with each other (Figures 3B and 3D). These defects in mesoderm migration of Df(2R)ED2238 homozygotes presumably also contribute to the failure to produce dorsal mesodermal derivatives, such as pericardial cells, which express even skipped (eve) (Figure 3E, 3F, 3G). We conclude that Df(2R)ED2238 uncovers genes zygotically required for mesoderm migration. Identification of Two Genes Encoding Novel FGF Orthologs in Drosophila In order to identify the gene that is uncovered by Df(2R)ED2238 and accounts for its defects in mesoderm morphogenesis, we performed a molecular analysis based upon the molecularly mapped chromosomal breakpoints of Df(2R)ED2238 and Df(2R)ED2230 (Figure 4). Because Df(2R)ED2230 did not affect mesoderm migration, we concluded that the gene must be localized between the distal breakpoints of Df(2R)ED2230 and Df(2R)ED2238 (Figures 4A and 4B). We identified a 179.926 bp interval that is missing in Df(2R)ED2238 but not in Df(2R)ED2230 (Figure 4B). The gene annotation release 3 of the Drosophila Genome Project predicted 14 genes within this interval (Figure 4B). Because the phenotypes of our initial screen were strictly based upon zygotic gene activity, we reasoned that prime candidates should exhibit a zygotic

expression profile. We reviewed the gene expression data of these 14 predicted genes and found that only two genes, CG12443 and CG13194, exhibited an early zygotic expression and lacked significant maternal transcripts (data not shown). We confirmed the expression profile of CG12443 and CG13194 by Northern blotting and in situ hybridization and found that both genes are expressed zygotically (Figures 4D and 5). CG12443 and CG13194 were predicted to consist of two exons each, and a BLASTp analysis exhibited a 36% amino acid identity for the primary sequences of the two genes within a stretch of 47 amino acids in the amino terminus (data not shown). Furthermore, this aminoterminal sequence of CG13194 exhibited 36% amino acid identity with FGF8 from the African clawed frog Xenopus laevis and FGF8 from the axolotl (Ambystoma mexicanum). The sequence homology of CG13194 is within the conserved FGF core domain of FGF8 and includes highly conserved amino acids such as Cys101, Phe-103 and Glu-105, which are conserved in all FGFs that can be found in the databases (Figure 4F). Because of the similarity of the predicted gene products to FGFs and because no other gene within this interval exhibited an early zygotic expression profile, we focused our further analysis on these two candidates. The alignment of the open reading frame of CG13194 with the core domain of mammalian FGFs, as well as the lack of a signal peptide sequence, suggested that the predicted exons in the genome annotation represented only part of the gene. A Northern blot analysis demonstrated that the transcripts were indeed larger than the suggested annotation of the genes. The

Current Biology 662

Figure 4. Molecular Characterization of FGF8-like1 and FGF8-like2 (A) Breakpoints of the chromosomal deletions Df(2R)ED2230 and Df(2R)ED2238 (the cytological positions are indicated by the ruler on the top). The black bars below represent the extent of the deletions. (B) Genomic organization of the interval from 48C2 to 48C5 based on the Drosophila genome annotation release 3. The distal breakpoints of Df(2R)ED2230 and Df(2R)ED2238 are indicated by vertical arrows. Horizontal arrows and arrowheads show the position and directionality of transcription units of predicted genes (for details, see www.flybase.bio.indiana. edu). The positions of CG13194 and CG12443 are indicated. The scale is 10 kb per vertical bar, and the numbering of genomic base pairs is indicated according to flybase. (C) Genomic organization of FGF8-like1. The original annotation of CG12443 is indicated on the top. The lower panel shows the organization of FGF8-like1 gene by alignment of the cDNA sequence to the genomic sequence. Four exons were detected with the following positions on the genomic sequence: exon 1, 6.855.470 to 6.855.650; exon 2, 6.857.941 to 6.858.064; exon 3, 6.872.649 to 6.872.776; and exon 4, 6.874.614 to 6.876.581. (D) Northern blot analysis with antisense RNA probes for FGF8-like1 and FGF8-like2 as indicated. 0–14: Embryos were collected for 14 hr at 18⬚C (corresponding embryonic stages 1 to 11). 14–24: embryos were collected for 10 hr and then aged at 18⬚C (embryonic stages 11–16). (E) Domain structure of the predicted proteins encoded by FGF8-like1 and FGF8-like2. (F) Amino acid alignment of the FGF core domains of FGF8-like1 and FGF8-like2 to the closest matches in Blastp analyses (Lj: Fugu rubripes). Amino acids common to all FGFs are marked with asterisks. Alignment was performed by the Clustal W method.

CG13194 probe detected a single band of 3.4 kb, and the CG12443 probe detected a prominent 4.9 kb band. We cloned the cDNA of CG12443 by RT PCR and obtained a 2.5 kb product, which was sequenced (Supplemental Data). Comparison of the cDNA sequence to the genomic sequence revealed that the gene, which was annotated as CG12443, contains four exons (Figure 4C). The CG12443 cDNA encodes a novel FGF with an

N-terminal signal peptide of 21 amino acids, a FGF core domain of 67 amino acids, and a long C-terminal region of 611 amino acids with no significant homologies to other proteins (Figure 4E). The FGF core domains of CG12443 and CG13194 exhibit 39% identical amino acid residues (Figure 4F). Interestingly, the homology with vertebrate FGF8, FGF17, and FGF18 is similarly high; it ranges from 32% to 35% amino acid identity. Because

FGF-like Genes in Drosophila Gastrulation 663

Figure 5. Expression of FGF8-like1 FGF8-like2 during Gastrulation

and

Embryos were fixed and hybridized with digoxygenine-labeled antisense RNA. (A and B) Before the beginning of gastrulation (stage 5; cellularization stage), FGF8-like1 transcripts are localized in a broad lateral domain of the presumptive neuroectoderm on each side of the embryo ([A], lateral view; [B], ventral view, median optical section). (C) During early gastrulation (stage 7), FGF8-like1 transcripts accumulate in the entire neuroectoderm. (H) In transversal cross-sections, the labeling is absent from the mesoderm cells and the ventralmost ectodermal cells. Transcript levels also fade out toward the dorsal ectoderm (H). (D and G) During mid-gastrulation (stage 8), continuous expression of FGF8-like1 is detected in the neuroectoderm but is absent from the amnioserosa and the ventral midline ([D], lateral view; [G], ventral view). (I and J) Cross-sections show that expression of FGF8-like1 at stage 8 is restricted to the ectodermal cell layer. (E) At stage 8, FGF8-like2 expression is downregulated in the lateral neuroectoderm but remains prominent in the dorsal-most ectoderm (arrowheads). (F) Specificity of the in situ hybridization: embryo homozygous for Df(2R)ED2238 were hybridized with the FGF8-like1 antisense probe.

of the high degree of homology of CG13194 and CG12443 to vertebrate FGF8 and the fact that FGF8 plays an essential role in vertebrate gastrulation, we propose naming these genes FGF8-like1 (CG12443) and FGF8-like2 (CG13194). FGF8-like1 and FGF8-like2 Are Expressed during Gastrulation To verify the expression data, we synthesized digoxygenine-labeled antisense RNA probes complementary to FGF8-like1 and FGF8-like2 transcripts and performed in situ hybridization experiments. No transcripts could be detected in fertilized eggs and syncytial blastoderm stage embryos (data not shown). Both genes were first expressed at mid-blastula transition in two broad lateral stripes in the prospective neuroectoderm, excluding the ventral domain of the blastoderm (Figures 5A, 5B, and 5H). At the beginning of gastrulation, expression of both genes extends to the cephalic furrow in the anterior region and to the posterior midgut invagination in the posterior part of the embryo (Figure 5C). During germ band elongation, the expression patterns of the two genes become distinct from each other. Although FGF8like1 is expressed in the entire germ band, except for a narrow ventral stripe of mesectoderm cells (Figures 5D and 5G), FGF8-like2 transcripts accumulate in the dorsal-most cells of the germ band (Figure 5E). During gastrulation, expression of both genes is confined to the ectodermal cell layer (Figures 5J; data not shown). After gastrulation, expression of FGF8-like1 and FGF8-like2 disappears from the neuroectoderm, and in later embryonic stages the two genes become differentially expressed (data not shown). These data are consistent with the Northern blot results, which showed that in early embryos FGF8-like1 is expressed at high levels in early and mid embryogenesis, whereas FGF8-like2 transcript levels are maintained and even rise during late embryogenesis (Figure 4D). These results show that FGF8-like1 and FGF8-like2 are expressed in the cells that serve as substrate for mesoderm cells during migration. Furthermore, the differential expression of FGF8like1 and FGF8-like2 during mesoderm migration suggests that the gene products might initially work in a

redundant fashion and later serve distinct functions in mesoderm morphogenesis. FGF8-like1 and FGF8-like2 Are Required for Mesodermal Migration and for the Activation of the Htl Signaling Pathway The expression pattern of FGF8-like1 and FGF8-like2 suggested that the two gene products might be required for the activation of Htl. Embryos deficient for both FGF8-like1 and FGF8-like2 exhibit defects in mesoderm migration similar to those seen in htl or dof mutants. In order to prove that FGF8-like1 and FGF8-like2 are indeed required for mesoderm migration, we performed RNA interference experiments. Injection of dsRNA targeting both genes results in a lack of Eve-positive dorsal mesodermal derivatives, whereas injection of dsRNA targeting FGF8-like1 alone as well as injection of buffer did not affect dorsal mesoderm differentiation (Figures 6A and 6B; data not shown). On the other hand, injection of dsRNA targeting FGF8-like2 alone affected the differentiation of Eve-positive mesoderm derivatives, suggesting that FGF8-like2 might have some nonredundant function for which FGF8-like1cannot compensate (data not shown). The lack of Eve-positive dorsal mesoderm cells might be due to a function of FGF8-like1 and FGF8-like2 in mesodermal patterning or to defects during the migration of the mesoderm cells. To discriminate between these two possibilities, we performed RNAi of FGF8like1 and FGF8-like2 in embryos expressing the mesoderm-specific cell surface marker twi::CD2 [22]. In the wild-type, twi::CD2 marks cell shape changes occurring during mesoderm migration [23]. The cells extend in the direction of migration and form long cellular protrusions (Figure 6C). In embryos mutant for htl, these cell shape changes do not occur [23]. Similarly, in embryos injected with dsRNA targeting FGF8-like1 and FGF8-like2, these cell shape changes were blocked (Figure 6D). We therefore conclude that FGF8-like1 and FGF8-like2 are required for cell shape changes of the mesoderm cells during migration. Because RNAi with FGF8-like1 did not affect differentiation of dorsal mesoderm derivatives, some functions of FGF8-like1 and FGF8-like2 might be redundant.

Current Biology 664

Figure 6. dsRNA-Mediated Knockdown of FGF8-like1 and FGF8-like2 Results in Defects in Mesoderm Differentiation and Cell Shape Changes Associated with Mesoderm Migration (A and B) Postgastrulation embryos at stage 11 were either uninjected (A) or injected with dsRNA complementary to FGF8-like1 and FGF8-like2 (B). Embryos were fixed and stained with antibodies against Eve. Eve staining in the mesoderm is absent in dsRNAtreated embryos ([B]; insert shows a different focal plane of the same embryo, where Eve staining in the nervous system can be detected). (C and D) Embryos derived from the twi::CD2 stock were injected with dsRNA against FGF8-like1 and FGF8-like2, fixed at mid-gastrula stages, and stained with antibodies against CD2 protein. (C) In an uninjected control embryo, the cells at the leading edge extend dorsally and exhibit multiple cellular protrusions. (D) These cell shape changes are blocked, and only few filopodial extensions were observed in embryos treated with dsRNA.

The fact that FGF8-like1 and FGF8-like2 are expressed in the ectoderm and are required for cell shape changes of mesoderm cells indicates a non-cell-autonomous function of FGF8-like1 and FGF8-like2. On the other hand the FGF-receptor Htl is specifically expressed in the mesoderm cells. In order to test whether FGF8-like1 and FGF8-like2 are required for the activity of Htl in the mesoderm, we measured the activation of the downstream signaling component MAP kinase by using an antibody that recognizes the activated doublephosphorylated form of MAP kinase [14]. In the wildtype, activated MAP kinase can be detected in the leading-edge cells of the migrating mesoderm [24]. This early activation of MAP kinase in the mesoderm depends on the presence of Htl and its downstream signaling factor Dof [11–13, 24]. To test whether FGF8-like1 and FGF8like2 are required for activation of MAP kinase in the mesoderm cells during migration, we stained embryos homozygously mutant for Df(2R)ED2238 or Df(2R)ED2230 with the dpERK antibody. Strikingly, only embryos mutant for Df(2R)ED2238 failed to exhibit dpERK staining in the mesoderm, whereas Df(2R)ED2230 mutant embryos looked like the wild-type (Figures 7A, 7B, 7D, and 7E). The defect in MAP kinase activation in Df(2R)ED2238 mutant embryos was specific for Htl FGF receptor activation because the staining of other cells that activate

the MAP kinase pathway via the EGF receptor remained unimpaired (Figures 7C and 7F). In summary, we demonstrate that mesoderm cells in embryos that lack FGF8-like1 and FGF8-like2 fail to exhibit Htl-dependent activation of MAP kinase. These results are consistent with a model in which FGF8-like1 and FGF8-like2 represent Htl receptor ligands, which are required for the early activation of the Htl signaling cascade during gastrulation. Discussion In Drosophila two known FGF receptors, Btl and Htl, are required for cell migratory events, tracheal migration, and mesodermal cell migration, respectively [25, 26]. Despite the annotation of the Drosophila genome sequence, only Bnl, the ligand of Btl, has thus far been identified. Bnl is not expressed in the right temporalspatial pattern to serve as a ligand for Htl, and bnl mutant embryos do not display defects in early mesoderm morphogenesis [3]. In addition, expression of dominantnegative forms of Btl does not produce mesoderm migration defects, and upon ectopic expression of Bnl within the ectoderm, only Btl-expressing cells show ectopic activation of MAP kinase [12, 24, 27]. Together, these data suggest either that the ligand of Htl was Figure 7. Activation of MAP Kinase Is Blocked in Embryos Containing a Deletion for FGF8-like1 and FGF8-like2

Embryos homozygously mutant for Df(2R)ED2230 (A–C) or for Df(2R)ED2238 (D–F) were fixed and stained for the doublephosphorylated form of MAP kinase with antidpERK antibody (A, B, D, and E). Transversal cross-sections through early ([A and D]; stage 7) and mid-gastrula stage embryos ([B and E]; stage 8). In Df(2R)ED2230 mutant embryos, cells at the leading edge of the migrating mesodermal aggregate accumulate large amounts of dpERK staining (arrowheads in [A] and [B]). Additionally, ventral-ectoderm staining, which depends on the activation of the Drosophila EGF receptor (DER) is seen [14]. (D and E) In Df(2R)ED2238 mutant embryos, dpERK staining is absent in the mesoderm cells; the DER-dependent signal in the ventral ectoderm is still present. (C and F) Whole mount staining with anti-dpERK antibody of embryos at stage 10. In both cases, normal accumulation of dpERK is seen in the tracheal placodes; the activation of dpERK in this tissue also depends upon DER activation.

FGF-like Genes in Drosophila Gastrulation 665

missed in the genome annotation or that Htl might not be activated by FGF-like ligands. In the present work, we demonstrate the identification of two genes encoding for novel fly FGF homologs, which exhibit features consistent with being ligands for Htl. The early expression of FGF8-like1 and FGF8-like2 is restricted to the neuroectoderm and thus corresponds to the predicted source of ligands required for Htl activation [24]. The knockdown of the function of these genes by RNAi leads to a mesoderm cell phenotype that is very similar to that produced by mutations in the Htl receptor. Thus, FGF8-like1 and FGF8-like2 act non-cellautonomously in the early embryo and are required for mesodermal cell shape changes during gastrulation. In addition, deletion of the two genes blocks activation of MAP kinase in early mesoderm cells. Our genetic mapping using isogenic deletions showed that an interval containing 14 genes is responsible for the observed phenotypes. Because the RNAi experiments are able to reproduce the migration phenotype of the deletion embryos, we propose that FGF8-like1 and FGF8-like2 together are responsible for the mesoderm defects observed in the deletion. In conclusion, these results strongly suggest that FGF8-like1 and FGF8-like2 represent ligands for Htl. The similarity of the early expression patterns of FGF8-like1 and FGF8-like2 suggests that their role during early gastrulation might be partially redundant. This idea is consistent with our result that RNAi knockdown of FGF8-like1 alone is not sufficient to produce defects in dorsal mesodermal derivatives. In contrast, during late gastrula stages the expression patterns of FGF8like1 and FGF8-like2 differ, suggesting that the two genes might have distinct functions during later morphogenesis. This idea is supported by our observation that knockdown of FGF8-like2 alone does produce defects in mesoderm differentiation. It has been shown that the Htl receptor has dual functions in mesoderm morphogenesis [28, 29]. During gastrulation, Htl is required early for adhesive interactions of the mesoderm to the ectoderm and for cell shape changes associated with migration of the mesoderm cells [23]. After gastrulation, Htl is required for specification of dorsal mesodermal derivatives that later will give rise to pericardial cells [30]. The differential expression of FGF8-like1 and FGF8-like2 in later development suggests that the two ligands might act in a nonredundant fashion during mesoderm differentiation. Several pieces of evidence suggest that FGF signaling via the Htl receptor is required for setting the correct timing for the interaction of mesoderm to ectoderm in early stages of gastrulation. The most robust migration phenotype of htl loss-of-function mutants occurs during early stages of mesoderm migration, at a time when the cells contact the ectoderm and migrate in dorsolateral direction. In late gastrula embryos, mesoderm cells exhibit directional protrusive activity in htl mutant embryos, indicating that htl is not essential for the migratory properties of the cells in a more general way [23]. Ligandindependent activation of Htl in a htl mutant background is able to rescue the early defects in cell shape changes but fails to completely rescue the late defects in mesoderm migration and differentiation [23, 28]. At the begin-

ning of gastrulation, FGF8-like1 and FGF8-like2 are uniformly expressed in the neuroectoderm, consistent with a permissive function for FGF signaling during early stages of gastrulation. This early expression pattern is likely to depend upon the Dorsal transcription factor because CG12443 has been described as a target of Dorsal [31]. The local activation pattern of MAP kinase suggests that during early mesoderm migration Htl is specifically activated in the leading-edge cells of the migrating mesoderm [24]. Because htl is expressed in all mesodermal cells, it has been proposed that the potential ligands might be present in a graded fashion along the dorsoventral axis. Although FGF8-like1 is expressed in a uniform pattern throughout the germ band during gastrulation, FGF8-like2 expression is downregulated in the ventral-lateral ectoderm and only remains expressed in the dorsal-most ectodermal cells. Thus, FGF8-like2 might act as an instructive cue that guides mesoderm cells during the migration to their dorsal targets. The results of the single knockdown of FGF8-like2 by RNAi supports this model. The FGF core domains of FGF8-like1 and FGF8-like2 exhibit a high degree of identity with vertebrate FGFs, in particular with mammalian FGF8. During mouse gastrulation, FGF8 is required for progenitor cells to migrate away from the primitive streak [5]. At the primitive streak, epiblast cells undergo an epithelial/mesenchymal transition followed by ingression movement of the endodermal and mesodermal progenitor cells [32]. Interestingly, in FGF8⫺/⫺ embryos, the epithelial/mesenchymal transition in the streak occurs normally, but the cells fail to migrate and form an aggregate in the streak region [5]. This cellular phenotype is reminiscent of the phenotype of Drosophila embryos mutant for htl. The mesoderm cells of htl mutants invaginate normally and undergo epithelial/mesenchymal transition, but fail to migrate out on the underlying ectoderm [23]. Thus, the cellular functions of FGF8 signaling during gastrulation movements of mesodermal precursor cells in species as different as mouse and Drosophila share similar features. Conclusions Two FGF receptor homologs, Htl and Btl, are present in the Drosophila genome. Although the ligand of Btl is represented by Bnl, the ligand of the Htl receptor remained unknown. We have identified in Drosophila two novel FGF family members, FGF8-like1 and FGF8like2, that are expressed at the right time and in the right place to serve as ligands for Htl. We show that FGF8-like1 and FGF8-like2 are required for Htl-dependent cell shape changes during mesoderm migration and for signaling events emanating from the Htl receptor but are dispensable for signaling events emanating from other RTKs. We conclude that FGF8-like1 and FGF8like2 are required for promotion of mesoderm migration during Drosophila gastrulation and thus represent likely ligands of the FGF receptor Htl. Experimental Procedures Genetics Fly stocks were kept under standard conditions. The chromosomes utilized in this study are described in the Drosophila genome data-

Current Biology 666

base (www.flybase.bio.indiana.edu) unless otherwise indicated. Df(2R)ED2230 and Df(2R)ED2238 were generated with the DROSDEL isogenic deficiency kit (www.131.111.146.35/ⵑpseq/ drosdel/ddindex.html) [21]. Df(2R)ED2230 was constructed with the RS elements CB-0208-3 and 5-SZ-3066 and represents a 341.051 bp deletion at cytological position 47F7 to 48C2 (starting at 6.435.843 to 6.776.894 genomic base pairs; 36 annotated genes were uncovered). Df(2R)ED2238 is based upon the RS elements 5-HA-1544 and CB-6286-3 and uncovers 439.100 bp at the cytological position 47F13 to 48D1 (from 6.517.720 to 6.956.820; 33 annotated genes were uncovered). Homozygous embryos were selected with balancer chromosomes marked with ftz::lacZ transgenes. Oregon R flies were used as a wild-type control. Molecular Biology Molecular cloning was performed after standard procedures. For cloning the cDNA of CG12443, RT-PCR was performed with embryonic poly A⫹ RNA or cDNA libraries as templates. 5⬘ prime oligonucleotide primers were designed based upon sequence information of the partially sequenced EST clone RE28585, which mapped 19.437 bp upstream of the annotated CG12443 exon. A similar strategy was followed for the CG13194 exons but has failed so far. Primer sequences are available upon request. Products from RT-PCR were subcloned with the TOPO-cloning kit (Invitrogen) and sequenced (Seqlab, Go¨ttingen, Germany). Antisense RNA probes were produced with either the FGF8-like1 cDNA or genomic DNA from the predicted CG13194 exons. Northern blot experiments were performed according to standard procedures with digoxygeninelabeled antisense RNA probes. Processing of sequences and alignments were conducted with the Lasergene Software via the Clustal W method and with the help of the EMBL computational services (www.bork.embl-heidelberg.de/). Immunohistology and In Situ Hybridization Embryos were fixed and stained essentially as described elsewhere [19, 23]. For genotyping, embryos were stained with anti␤-galactosidase antibodies, and homozygous embryos were selected under the dissecting microscope. Plastic and araldite sections were produced as described [23]. In situ hybridization was conducted with digoxygenin-labeled antisense RNA probes after standard protocols [33]. The following antibodies were used: mouse anti-dpERK (Sigma); rabbit anti-Twist (Siegfried Roth, Cologne); rabbit anti-␤Gal (Cappel); mouse anti-␤Gal (Promega); mouse anti-Eve; mouse anti-Engrailed; mouse anti-Neurotactin (DSHB, Iowa); mouse anti-CD2 (Serotec, Germany); and mouse anti-digoxygenine conjugated with alkaline phosphatase (Roche, Germany). Fluorescenceconjugated secondary antibodies and antibodies conjugated with alkaline phosphatase were from Dianova (Germany). Detection with horseradish peroxidase was performed with the Vecastain kit (Vector, USA). Micrographs were taken on an Axiophot compound microscope with Nomarski optics or on a Leica TCS NT confocal microscope. Images were processed with Adobe Photoshop on an Apple computer. RNA Interference dsRNA-mediated interference was performed essentially as described [34]. dsRNA was produced with primers containing flanking T7 promotors. For FGF8-like1, bp 1605–2077 of the cDNA were used as a template. For CG13194 genomic sequence (bp 6.789.377– 6.789.815), was used as a template. For microinjection of dsRNA, 0- to 30-min-old embryos were dechorionated, mounted onto cover slips with glue, and covered with 3S Halocarbon oil. Embryos were injected with a constant volume into the posterior third with a Z820 pneumatic micropump (WPI, USA) on a Zeiss inverted microscope. Injected embryos were grown at 18⬚C, fixed at the appropriate stages with 8% formaldehyde, and manually devitellinized. Immunostaining was performed as described. Supplemental Data Supplemental Data are available with this article online at http:// www.current-biology.com/cgi/content/full/14/8/659/DC1/.

Acknowledgments We thank the Bloomington Drosophila Stock Center, the Developmental Studies Hybridoma Bank, and Kevin Cook, Gunnar Ising, Gunter Reuter, Siegfried Roth, John Roote, Ed Ryder, and Janos Szidonya for fly stocks and advice. We are indebted to Andre´ Bachmann for advice, discussion, and motivation during the work. We thank Eric Wieschaus for support and advice during the translocation screen and Sylvia Tannebaum and Nora Hinssen for expert technical assistance. The authors declare that no conflicts of financial interests exist. This project was funded by a grant from the Deutsche Forschungsgemeinschaft (SFB590-TP B1). Received: March 12, 2004 Accepted: March 24, 2004 Published online: April 1, 2004 References 1. Szebenyi, G., and Fallon, J.F. (1999). Fibroblast growth factors as multifunctional signaling factors. Int. Rev. Cytol. 185, 45–106. 2. Burdine, R.D., Branda, C.S., and Stern, M.J. (1998). EGL-17(FGF) expression coordinates the attraction of the migrating sex myoblasts with vulval induction in C. elegans. Development 125, 1083–1093. 3. Sutherland, D., Samakovlis, C., and Krasnow, M.A. (1996). branchless encodes a Drosophila FGF homolog that controls tracheal cell migration and the pattern of branching. Cell 87, 1091–1101. 4. Sato, M., and Kornberg, T.B. (2002). FGF is an essential mitogen and chemoattractant for the air sacs of the Drosophila tracheal system. Dev. Cell 3, 195–207. 5. Sun, X., Meyers, E.N., Lewandoski, M., and Martin, G.R. (1999). Targeted disruption of Fgf8 causes failure of cell migration in the gastrulating mouse embryo. Genes Dev. 13, 1834–1846. 6. Ciruna, B., and Rossant, J. (2001). FGF signaling regulates mesoderm cell fate specification and morphogenetic movement at the primitive streak. Dev. Cell 1, 37–49. 7. Yang, X., Dormann, D., Munsterberg, A.E., and Weijer, C.J. (2002). Cell movement patterns during gastrulation in the chick are controlled by positive and negative chemotaxis mediated by FGF4 and FGF8. Dev. Cell 3, 425–437. 8. Beiman, M., Shilo, B.Z., and Volk, T. (1996). Heartless, a Drosophila FGF receptor homolog, is essential for cell migration and establishment of several mesodermal lineages. Genes Dev. 10, 2993–3002. 9. Shishido, E., Ono, N., Kojima, T., and Saigo, K. (1997). Requirements of DFR1/Heartless, a mesoderm-specific Drosophila FGF-receptor, for the formation of heart, visceral and somatic muscles, and ensheathing of longitudinal axon tracts in CNS. Development 124, 2119–2128. 10. Gisselbrecht, S., Skeath, J.B., Doe, C.Q., and Michelson, A.M. (1996). heartless encodes a fibroblast growth factor receptor (DFR1/DFGF-R2) involved in the directional migration of early mesodermal cells in the Drosophila embryo. Genes Dev. 10, 3003–3017. 11. Vincent, S., Wilson, R., Coelho, C., Affolter, M., and Leptin, M. (1998). The Drosophila protein Dof is specifically required for FGF signaling. Mol. Cell 2, 515–525. 12. Michelson, A.M., Gisselbrecht, S., Buff, E., and Skeath, J.B. (1998). Heartbroken is a specific downstream mediator of FGF receptor signalling in Drosophila. Development 125, 4379–4389. 13. Imam, F., Sutherland, D., Huang, W., and Krasnow, M.A. (1999). stumps, a Drosophila gene required for fibroblast growth factor (FGF)-directed migrations of tracheal and mesodermal cells. Genetics 152, 307–318. 14. Gabay, L., Seger, R., and Shilo, B.Z. (1997). In situ activation pattern of Drosophila EGF receptor pathway during development. Science 277, 1103–1106. 15. Garcia-Garcia, M.J., and Anderson, K.V. (2003). Essential role of glycosaminoglycans in Fgf signaling during mouse gastrulation. Cell 114, 727–737. 16. Lin, X., Buff, E.M., Perrimon, N., and Michelson, A.M. (1999).

FGF-like Genes in Drosophila Gastrulation 667

17.

18.

19.

20. 21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32. 33.

34.

Heparan sulfate proteoglycans are essential for FGF receptor signaling during Drosophila embryonic development. Development 126, 3715–3723. Wieschaus, E., and Sweeton, D. (1988). Requirements for X-linked zygotic gene activity during cellularization of early Drosophila embryos. Development 104, 483–493. Merrill, P.T., Sweeton, D., and Wieschaus, E. (1988). Requirements for autosomal gene activity during precellular stages of Drosophila melanogaster. Development 104, 495–509. Mu¨ller, H.-A.J., Samanta, R., and Wieschaus, E. (1999). Wingless signaling in the Drosophila embryo: zygotic requirements and the role of the frizzled genes. Development 126, 577–586. Leptin, M., and Grunewald, B. (1990). Cell shape changes during gastrulation in Drosophila. Development 110, 73–84. Golic, K.G., and Golic, M.M. (1996). Engineering the Drosophila genome: chromosome rearrangements by design. Genetics 144, 1693–1711. Dunin-Borkowski, O.M., and Brown, N.H. (1995). Mammalian CD2 is an effective heterologous marker of the cell surface in Drosophila. Dev. Biol. 168, 689–693. Schumacher, S., Gryzik, T., Tannebaum, S., and Mu¨ller, H.A.J. (2004). The Rho-GEF Pebble is required for cell shape changes during cell migration triggered by the Drosophila FGF receptor Heartless. Development, in press. Gabay, L., Seger, R., and Shilo, B.Z. (1997). MAP kinase in situ activation atlas during Drosophila embryogenesis. Development 124, 3535–3541. Ghabrial, A., Luschnig, S., Metzstein, M.M., and Krasnow, M.A. (2003). Branching morphogenesis of the Drosophila tracheal system. Annu. Rev. Cell Dev. Biol. 19, 623–647. Wilson, R., and Leptin, M. (2000). Fibroblast growth factor receptor-dependent morphogenesis of the Drosophila mesoderm. Philos. Trans. R. Soc. Lond. B Biol. Sci. 355, 891–895. Reichman-Fried, M., and Shilo, B.Z. (1995). Breathless, a Drosophila FGF receptor homolog, is required for the onset of tracheal cell migration and tracheole formation. Mech. Dev. 52, 265–273. Michelson, A.M., Gisselbrecht, S., Zhou, Y., Baek, K.H., and Buff, E.M. (1998). Dual functions of the heartless fibroblast growth factor receptor in development of the Drosophila embryonic mesoderm. Dev. Genet. 22, 212–229. Carmena, A., Buff, E., Halfon, M.S., Gisselbrecht, S., Jimenez, F., Baylies, M.K., and Michelson, A.M. (2002). Reciprocal regulatory interactions between the Notch and Ras signaling pathways in the Drosophila embryonic mesoderm. Dev. Biol. 244, 226–242. Carmena, A., Gisselbrecht, S., Harrison, J., Jimenez, F., and Michelson, A.M. (1998). Combinatorial signaling codes for the progressive determination of cell fates in the Drosophila embryonic mesoderm. Genes Dev. 12, 3910–3922. Stathopoulos, A., Van Drenth, M., Erives, A., Markstein, M., and Levine, M. (2002). Whole-genome analysis of dorsal-ventral patterning in the Drosophila embryo. Cell 111, 687–701. Tam, P.P., and Behringer, R.R. (1997). Mouse gastrulation: the formation of a mammalian body plan. Mech. Dev. 68, 3–25. Tautz, D., and Pfeifle, C. (1989). A non-radioactive in situ hybridization method for the localization of specific RNAs in Drosophila embryos reveals translational control of the segmentation gene hunchback. Chromosoma 98, 81–85. Kennerdell, J.R., and Carthew, R.W. (1998). Use of dsRNA-mediated genetic interference to demonstrate that frizzled and frizzled 2 act in the wingless pathway. Cell 95, 1017–1026.