A Screen for Genes That Influence Fibroblast Growth Factor Signal Transduction in Drosophila
Min Yan Zhu, Robert Wilson, Maria Leptin


The misexpression of an activated form of the FGF receptor (FGFR) Breathless in conjunction with downstream-of-FGF-receptor (Dof), an essential signaling molecule of the FGF pathway, in the Drosophila eye imaginal discs impairs eye development and results in a rough eye phenotype. We used this phenotype in a gain-of-function screen to search for modifiers of FGF signaling. We identified 50 EP stocks with insertions defining at least 35 genes that affect the rough eye phenotype. Among these genes, 4 appear to be specific for FGFR signaling, but most of the genes also influence other signaling pathways, as assessed by their effects on rough eyes induced by other activated receptor tyrosine kinases (RTKs). Analysis of loss-of-function alleles of a number of these genes in embryos indicates that in many cases the products are provided maternally and are involved in germ cell development. At least two of the genes, sar1 and robo2, show a genetic interaction with a hypomorphic dof allele, suggesting that they participate in FGF-mediated morphogenetic events during embryogenesis.

FIBROBLAST growth factor (FGF) receptors are ligand-activated transmembrane glycoproteins that transmit signals to a variety of intracellular targets. In Drosophila two FGF receptors, Breathless (Btl) and Heartless (Htl), lead to the phosphorylation of MAP kinase in the mesoderm and the tracheal system, as well as in a number of other cell types, and are required for the morphogenesis of these tissues (Klämbt et al. 1992; Beiman et al. 1996; Gisselbrecht et al. 1996; Sutherland et al. 1996; Gabay et al. 1997). A cytoplasmic molecule, downstream-of-FGF-receptor (Dof), also known as Heartbroken or Stumps, is important for the proper transduction of signals from both FGF receptors (Michelson et al. 1998; Vincent et al. 1998; Imam et al. 1999). In dof mutant embryos, mesoderm migration and tracheal branching are defective and phosphorylated MAPK fails to accumulate in these tissues. A constitutively active form of Ras can provide a partial rescue of the defects, suggesting that Dof acts upstream of Ras. Biochemical analysis and protein interaction experiments performed in yeast cells indicate that Dof binds directly to the FGF receptors via an essential protein domain, the Dof-BANK-BCAP domain, and becomes phosphorylated upon activation of the receptor (Battersby et al. 2003; Wilson et al. 2004). Functional studies with mutant forms of Dof have suggested that one protein that could bind to Dof and participate in transmission of the FGF signal is the protein phosphatase Corkscrew, but the potential binding site is neither sufficient nor necessary for all aspects of signal transmission (Petit et al. 2004; Wilson et al. 2004). It is therefore not presently fully understood how signals generated upon activation of the FGF receptors are transmitted to intracellular targets to promote cellular differentiation and morphogenesis.

Genetic screens for modifiers of mutant eye phenotypes have proved successful in identifying components of signaling pathways regulated by receptor tyrosine kinases (RTKs) (Karim et al. 1996; Huang and Rubin 2000; Rebay et al. 2000; Therrien et al. 2000). These screens were based on the observation that the expression of a gain-of-function, ligand-independent form of the RTK Sevenless in the Drosophila eye causes all ommatidial precursors to develop as neurons and leads to a rough eye phenotype (Basler et al. 1991). FGF signaling is not required for eye development (Casci et al. 1999), but we found that the expression of a constitutively active FGF receptor in the eye together with Dof (also normally absent from the eye) caused a rough eye phenotype. On the basis of this observation, we performed a gain-of-function screen to identify genes whose mis- or overexpression might influence this effect. We screened a large panel of fly stocks (referred to here as “EP lines”), which carry P-element insertions that allow the tissue-specific expression of nearby genes (rth 1996). In this article, we report the identification of 50 insertions representing at least 35 genes that can affect FGF signaling in the eye and we present further genetic and phenotypic analysis of some of these candidate genes.


Constructs and production of transgenic flies:

To produce a plasmid encoding an activated form of the platelet-derived growth factor (PDGF)- and vascular endothelial growth factor (VEGF)-related receptor (PVR), a 2.2-kb fragment coding for the transmembrane domain and cytoplasmic portion of the RTK was amplified in a PCR from the expressed sequence tag clone SD02957 [Berkeley Drosophila Genome Project (BDGP), obtained from Research Genetics] with the aid of the primers 5′-ACGGTAGAGATCTCGGATCTGCCCGGAATTA and 5′-GAGGCGCGCCTTAATACCTTCGTTGCTCCTTCTCGTTGACG. A BglII-AscI fragment generated from the PCR product was used to create the plasmid pUAST-λPVR (details available upon request). Standard procedures were followed to produce transgenic flies by mobilizing the P[w+, UAS-λPVR] element of pUAST-λPVR in Drosophila embryos. Stable insertions were assigned to a particular chromosome on the basis of the segregation of the transgenes from dominant markers on the second and third chromosome balancers and sex linkage for the X chromosome.


The fly stock with a rough eye phenotype used in the screen was produced by the recombination of second chromosome P[w+, GMR-Gal4], P[w+, UAS-λ-Btl], and P[w+, UAS-Flag-Dof] P-element insertions (Freeman 1996; Lee et al. 1996; Wilson et al. 2004) to synthesize a single chromosome, which was lethal when homozygous and maintained by using a CyO balancer. We refer to this as the “GMR>λ-btl, dof” chromosome. The screen was performed with 2135 EP lines obtained from the Szeged Drosophila stock center (http://gen.bio.u-szgeged.hu/servlet/jate.genetics.servlet.EPlines). The GMR>λ-btl, dof eye phenotype is mild at 18°, but much stronger at 18.5° or 19° and does not visibly change at higher temperatures. Thus, to maximize the sensitivity of the modifier screen, males of each EP line and females of the GMR>λ-btl, dof stock were crossed and maintained at 18°. The effect on the eye phenotype was scored in the progeny when at least 20 flies had emerged. To eliminate false positives due to the effects of slight fluctuations in temperature upon the GMR>λ-btl, dof phenotype, we retested the enhancers at 18° and the suppressors at 22°. In addition, we tested the candidates to see if the EP lines would produce a rough eye phenotype independently of the ectopic FGF signal by crossing the stocks to flies carrying the GMR-Gal4 transgene alone. The effects of the candidate insertions were assigned to specific genes on the basis of information provided by BDGP and FlyBase for release 3.1 of the Drosophila melanogaster genome.

Standard procedures were followed to mobilize the insertions of specific EP lines using the stock y1 w*; CyO H[w+mC Δ2-3]HoP2.1/Bc1, EgfrE1. Genomic DNA was isolated from the offspring with white eyes according to Ashburner (1989). Imprecise excisions were identified by PCR and the breakpoints of the deletions were characterized by DNA sequence analysis (see supplementary data at http://www.genetics.org/supplemental/ for details). Germline clones were generated using the FLP-DFS technique (Chou et al. 1993), and somatic clones were produced with the aid of the chromosomes 2XP[w+, Ubi-GFP] P[ry+ neoFRT]40A and P[ry+ neoFRT]82B P[w+, Ubi-GFP3-13-7]. To produce homozygous double-mutant embryos, EP insertions on the third chromosome were linked with the hbrems7 allele by recombination, and a robo28/CyO P[ry+, ftz-lacZ], hbrems7/TM6B stock was established.

Histochemistry and microscopy:

Antibody staining, in situ hybridization, and cuticle preparation were performed according to standard protocols (Tautz and Pfeifle 1989; Roberts 1998); adult retinas were stained and sectioned as previously described (Wolff and Ready 1991). Tracheae were visualized with the monoclonal antibody 2A12 (diluted 1:20), kindly provided by N. Patel. Following the staining reaction, embryos were embedded and mounted in the Araldite (Serva) and photographed using an Axiophot photomicroscope (Zeiss) with a ProgRes 3008 (Kontron Elektronik) camera. To visualize GFP, imaginal discs and ovaries were fixed in 4% paraformaldehyde and mounted in Vectashield (Linaris). Photos were taken using an Axiophot Photomicroscope with a Quantic (Photometrics) camera. Images were processed with IPlab (Scanalytics) and Photoshop (Adobe Systems).


The effect of ectopic FGF receptor activation in the eye:

We expressed constitutively active forms of the FGF receptors Breathless or Heartless (λ-Btl or λ-Htl) (Lee et al. 1996; Michelson et al. 1998) in the eye imaginal disc using the driver line GMR-Gal4 (Freeman 1996). Neither construct had an effect on the morphology of the eyes of adult flies (Figure 1), although the activated receptors dimerize and undergo autophosphorylation in a ligand-independent fashion (unpublished data and Lee et al. 1996). However, when the activated receptors were coexpressed together with Dof, the development of the eye was perturbed. Misrotation of ommatidia as well as loss of photoreceptors was observed (Figure 1D and data not shown). We refer to this rough eye phenotype as the GMR>λ-btl, dof phenotype. This phenotype indicates that the fundamental components for at least some aspects of FGF signaling are present during eye development. We have carried out a gain-of-function genetic screen in this genetic background to identify critical components of the FGF-signaling cascade on the assumption that the misexpression of additional components in the eye could affect the strength of the ectopic FGF signal and hence the eye phenotype.

Figure 1.—

Expression of Dof and λ-Btl under the control of GMR-Gal4. (A) Expression of GMR-Gal4 alone, (B) λ-Btl under control of GMR-Gal4, or (C) Dof under control of GMR-Gal4 does not cause a rough eye phentoype. (D) Coexpression of Dof and λ-Btl produces a rough eye phenotype. The eyes shown in A–D are from animals raised at 22°. (E and F) Sections of a wild type (E) and an eye expressing Dof and λ-Btl (F) of animals raised at 18° adjacent to diagrams showing the orientations of the ommatidia seen in the sections. Arrows indicate the rotation of each ommatidium, and the red line indicates the equator. Bars, 25 μm.

Tests of potential candidates of FGF signaling:

As a preliminary step, we tested the dosage-sensitive effects of loss-of-function alleles of several different groups of known genes upon the GMR>λ-btl, dof eye phenotype (see Table 1). To our surprise, the only component of the canonical RTK-signaling pathway that affected the eye phenotype was Raf. Among the other mutations that we tested we found that RhoA, RhoGEF, and fmi enhanced the GMR>λ-btl, dof eye phenotype, while Notch and cyclin A had no effect. Hence, the GMR>λ-btl, dof phenotype can be modified by altering the dose of components of signaling pathways in the eye, but there is only a limited overlap with the components of RTK signaling identified in previous dosage-sensitive screens (Rogge et al. 1991; Simon et al. 1991; Karim et al. 1996; Huang and Rubin 2000; Rebay et al. 2000; Therrien et al. 2000).

View this table:

Dosage effects of specific mutations on the GMR>λ-btl, dof phenotype

The gain-of-function screen:

To allow the identification of components of the FGF-signaling pathway not normally expressed in the eye, in addition to common components of RTK-signaling pathways expressed in the eye, we decided to perform a gain-of-function screen. We tested the effect of 2135 EP transposon insertions, which allow the misexpression of genes that lie immediately downstream of the point of insertion, upon the eye phenotype. The eyes of the progeny containing an EP insertion and the GMR>λ-btl, dof chromosome were compared to their siblings carrying only the GMR>λ-btl, dof chromosome (see materials and methods). In the F1 generation we selected 153 EP lines showing a modified GMR>λ-btl, dof eye phenotype as potential candidates; 81 of them acted as enhancers and 72 as suppressors. Retesting confirmed 48 enhancers and 24 suppressors (see materials and methods). Of the enhancers, we excluded 22 from further consideration because their effect on the eye phenotype was independent of the FGF signal. We also found that 4 of the potential suppressors, namely EP1413, EP0355, EP1455, and EP0622, show a rough eye phenotype in the absence of the FGF signal. However, we included these EP insertions in our further analysis since they had the ability to counteract the roughening induced by the ectopic FGF signal. In summary, we found 50 EP insertions in this screen, of which 26 act as enhancers and 24 as suppressors (see Table 2).

View this table:

EP insertions that modify the GMR>λ-btl, dof phenotype

In examining the effect of the EP insertions, we assumed that insertions mapping to the same regions or insertion points probably affect the same gene. On this basis at least 35 genes account for the effect of the EP insertions (see Table 2). The genes that we have identified as candidates are predicted to encode proteins with functions in diverse cellular processes and include kinases, membrane proteins, transcription factors, a mitochondrial protein, RNA-binding proteins, a ubiquitin E3 ligase, a Na+/Ca2+ exchanger, and proteins involved in vesicle transport.

The effects observed in the screen could be due to either overexpression or inactivation of the genes near the point of the EP-element insertion. In some cases, for example, CG10082, we found several insertions at the same position, but in opposite orientations, which all have the same effect upon the phenotype (Figure 2), suggesting that the insertions have disrupted a gene. While in other cases, for example CG11172 and toutatis, only EP insertions in the same orientation have an effect (Figure 2), implying that overexpression of the downstream gene is likely to be responsible for the phenotype.

Figure 2.—

Multiple EP elements with insertion sites near the same genes. (Left) EP insertions that modified the GMR>λ-btl, dof phenotype. (Right) Diagrams of the genomic regions in which the EP elements are inserted, which also include insertions that had no effect upon the GMR>λ-btl, dof phenotype. Genes transcribed from the top strand are shown as boxes above the line depicting the DNA sequence; genes transcribed from the bottom strand are shown underneath the line. Similarly, EP insertions from which the top strand is transcribed are shown as arrows above the line, and those from which the bottom strand can be transcribed are shown below the line. The type of arrowhead denotes whether the EP insertion acted as an enhancer or suppressor of the GMR>λ-btl, dof phenotype (solid and open arrows, respectively) or had no effect (solid circles). The intron/exon structure is shown only in those cases where EP elements were inserted in an intron, and only one splice form is indicated for each gene. CG32781 was formerly annotated as CG2829. The annotation is incomplete as the gene spans a large gap in the DNA sequence (cf. flybase.bio.indiana.edu/.bin/fbidq.html?FBan0032781).

Specificity of the observed effect for FGF signaling:

To determine whether the candidates specifically affected FGF signaling we tested their effect upon the rough eye phenotypes caused by constitutively active forms of the PDGF- and-VEGF-related receptor and the EGF receptor (λ-PVR and λ-EGFR) (Table 2). In this assay four genes appear to be specific for the FGF pathway, eight are common to all three pathways, and the majority are shared by the FGF- and PVR-signaling pathways, consistent with the closer evolutionary relationship between these receptors.

Analysis of the candidate genes:

If the gene products act in the FGF pathway during normal development or physiology, their products must be present in cells that respond to FGF signaling. In the case of robo2, this has been shown (Englund et al. 2002); however, the expression pattern of most of the candidates has not been examined. We therefore used in situ hybridization to examine the expression patterns of a subset of the candidate genes. These were selected because they had clear homologs in other species and were closely associated with an EP insertion. We excluded transcription factors from this group, because our primary interest was to identify components that act higher up in the signal transduction pathway. Embryos hybridized with probes derived from the cDNAs of CG2829, ago, CG3542, CG6386, sar1, CG14217, CG4266, and Tim10 all showed maternally deposited RNA (Figure 3 and data not shown). In addition to the early expression pattern, transcripts of CG4266 also accumulate in the central nervous system at stage 16. CG14217 is present at high levels at the posterior pole before and during cellularization of the blastoderm and is taken up into the pole cells, while Tim10 is expressed in the posterior and anterior midgut primordia from stage 10 onward. If the maternally provided RNAs are translated in the embryo, these protein products would be expected to be present at least in the first cells responding to FGF signaling, the presumptive mesodermal cells. Whether they are also present at later stages, for example, during tracheal development, would depend on the stability of the protein.

Figure 3.—

The RNA expression patterns of CG4266, CG14217, and Tim10. (A and B) CG4266 is present at an elevated level in the ventral nerve chord at stage 16 (A, lateral view; B, ventral view). (C and D) CG14217 is present at high levels in the pole plasm and is taken up by the pole cells. (E and F) Tim10 is expressed in the posterior and anterior midgut primordia from stage 10 onward.

Creation of loss-of-function mutations:

Genes that are important for FGF signaling should show a mutant phenotype that recapitulates, at least in part, the loss-of-function phenotype of the FGF receptors. We detected no defects in FGF signaling when we examined the embryonic phenotypes caused by those EP insertions that were lethal when homozygous. Since the effect of the EP insertions on the downstream gene is not well defined, we also generated loss-of-function mutations for three of our candidates, namely the Drosophila homolog of sar1, which encodes a GTPase required for vesicular transport; CG3542, which encodes a homolog of human formin-binding protein 11; and CG6386, which encodes a kinase. We mobilized the P-element insertions EP3575, EP0719, and EP0863 to generate deletions by imprecise excision of the transposons (Figure 4 and supplementary data at http://www.genetics.org/supplemental/). In all three cases we found that the lethality of the EP line was associated with the P-element insertion, since the viability of the chromosome was restored upon precise excision of the P element. The deletions generated by the excisions are shown in Figure 4 and described in detail in supplementary data at http://www.genetics.org/supplemental/.

Figure 4.—

P-element-induced deletions of CG3542, CG6386, and Dmsar1. (A) The imprecise excision of EP0719 generated six internal deletions of CG3542. (B) Imprecise excisions of EP863 generated two deletions of CG6386, while (C) imprecise excisions of EP3575 generated two deletions of sar1 (CG7073). Predicted transcripts (CG3542-RA, CG6386-RA, and sar1-RA) are shown as open boxes; the predicted ORFs are shaded. Solid arrows indicate enhancers and the shaded arrow denotes a suppressor. The genomic regions of the mutant chromosomes were amplified by PCR and sequenced; for further details, see FlyBase.

As an initial test of the effect of the loss-of-function alleles of sar1, CG3542, and CG6386 upon FGF signaling, we examined the effect of the deletions upon the eye phenotype caused by the misexpression of the activated form of Breathless and Dof. We found that the deletions within sar1 and CG6386 did not modify the rough eye phenotype, suggesting that the effects of EP3575 and EP0863 upon the GMR>λ-btl, dof phenotype were due to overexpression rather than inactivation of the genes. By contrast, in the case of CG3542 we found that the deletions enhanced the phenotype, indicating that the effect of EP0719 is due to a reduction in the dose of CG3542 rather than the overexpression of this gene or a neighboring gene.

Role of the candidate genes during embryonic development:

Embryos homozygous for sar1, CG3542, and CG6386 mutations showed no defects in tracheal or mesodermal development (data not shown). To test whether maternally supplied RNA or protein masks a requirement for these zygotic gene products, we generated homozygous mutant germline clones (Perrimon 1998). For CG3542, eggs were recovered from homozygous mutant germline clones of alleles 15.5 and 18.2, but not from alleles 3.3, 9.4, 35.1, and 9.3 (see Figure 4), suggesting that alleles 15.5 and 18.2 represent weak hypomorphic alleles. In the case of CG6386, germline clones of both alleles (43 and 53) failed to produce mature eggs (see Figure 4). Thus, the function of these genes is essential for the development of the oocyte. To examine whether CG3542 and CG6386 are required for cell survival, we generated clones in imaginal discs. We recovered large mutant clones in the eye disc with the CG3542 allele 15.5 (Figure 5A), but the 3.3 allele produced only very small clones (Figure 5B). This is consistent with the CG3542 allele 3.3 being a stronger hypomorphic allele than the 15.5 allele and indicates that CG3542 has an influence upon the growth or survival of cells. For CG6386, we recovered mutant clones of both alleles in eye imaginal discs that were similar in size to the wild-type twin clones (Figure 5D), implying that this gene has no effect upon general cell growth or survival. In the case of sar1, females with a mutant germline produced a few eggs in the first few days after the induction of mitotic recombination (Figure 5E). Later, no eggs were produced, indicating that the perdurance of wild-type Sar1 protein and mRNA in the mutant germline clones is likely to account for the early production of eggs. We did not recover mutant clones of sar1 in imaginal disks (Figure 5C), consistent with an essential role of sar1 in cell survival or proliferation. Thus, in view of the requirement for these three genes in oogenesis, their potential role in FGF signaling cannot be studied in embryos lacking the maternally supplied gene products. One way of analyzing the zygotic requirement for a gene product in such a situation is to generate somatic clones during embryonic development and score the results at a stage when maternal products are likely to have been degraded. Studies using the MARCM system (Lee et al. 2000) to analyze gene function during late tracheal development showed that clones defective in FGF signaling survived until the third larval instar (A. Bilstein, M. Baer and M. Leptin, unpublished data), whereas no clones were observed in animals carrying the FRT chromosome with the sar1 loss-of-function allele, arguing that even at late stages Sar1 provides an essential cellular function in tracheal cells (not shown).

Figure 5.—

Phenotypic analysis of deletions of CG3542, CG6386, and Dmsar1. (A–D) Clonal analysis of deletions of CG3542, CG6386, and Dmsar1 in eye imaginal discs. Homozygous mutant clones are marked by the absence of GFP, while the corresponding homozygous wild-type clones exhibit a more intense fluorescent signal. (E) A white (w) mutant egg (top) and an egg from a germline mutant clone of Dmsar1 71 that yielded only a few eggs, all of which had some defects in the dorsal appendages. (F–H) Similar regions of mutant embryos stained with the antibody 2A12, which reveals the tracheal system, showing examples of the phenotypes scored in K. The images were chosen to show representative examples of the dorsal branches. The breaks in the dorsal trunk also show a large range of variation, and the number of breaks seen in these images is not representative for the whole phenotypic class. (I and J) Examples of the fraction of hbr, sar1, and hbr, robo2 double-mutant embryos with defects in germ-band retraction, stained with the 2A12 antibody. (K) Genetic interaction of sar1 and robo2 mutants with the endogenous FGF signaling pathway. The defects in dorsal branch formation were assessed in stage 14–15 embryos with the indicated genotypes. For each genotype, 160 or 100 branches (8 or 5 embryos) were evaluated. The variation in the number of the defective dorsal branches within embryos of the same genotype is indicated as “variation/side.”

robo2 and sar1 mutations enhance the tracheal defects of hbrems7 mutants:

Even though it is difficult to examine the effect of the complete loss-of-function phenotype of some of the candidate genes upon FGF signaling, an alteration in the dose of proteins that act in FGF signaling might become critical if the signal itself is partially compromised. To examine this idea, we tested whether the defects in the tracheal system of the hypomorphic dof allele hbrems7, in which FGF signaling in the tracheal system is reduced but not abolished (Michelson et al. 1998), could be exacerbated by the loss of the candidate genes. We generated double homozygous mutant embryos of hbrems7 with robo28, sar1EP3575, CG6386EP0863, dcoEP3280, and Dlc90FEP3634 and analyzed their tracheal system. We found that DcoEP3280, CG6386EP0863, and Dlc90FEP3634 had no effect, but the loss-of-function mutations robo28 and sar1EP3575 exacerbated the tracheal defects of hbrems7 mutant embryos (Figure 5G). Robo2, which we identified in the screen as a weak suppressor (EP2258), has previously been shown to be required to prevent the ganglionic branches of the embryonic tracheal system from crossing the ventral midline (Englund et al. 2002). Mutants also show slight defects in the dorsal branches (18% stalled or missing). Examination of the dorsal branches showed a significant increase in loss or stalling of branches in embryos homozygous for both robo28 and hbrems7 compared to the single mutants (Figure 5F).

We also note that 60% of the double-mutant embryos show severe defects in germ-band retraction (data not shown), which do not occur in hbrems7 or robo28 mutant embryos. Thus, robo2 enhances the tracheal phenotype, and the double mutant reveals a requirement for the two genes during germ-band retraction that had not previously been recognized.

Although there was no previous evidence for Dmsar1 being involved in tracheal development, Dmsar1EP3575 also enhances the hbrems7 phenotype. Thirty-one percent of dorsal branches are missing in stage 14–15 embryos that are homozygous for both Dmsar1EP3575 and hbrems7, while 5% are stalled (Figure 5F). In addition, 24.5% of the double-mutant embryos show defects in germ-band retraction (not shown).


The hyperactivation of the Sevenless RTK-signaling pathway in the Drosophila eye causes the ommatidial precursors to differentiate as neurons and results in a rough eye phenotype (Basler et al. 1991). A number of genetic screens for identifying molecules that function in RTK signaling have been based upon the modification of this rough eye phenotype (for example, Rogge et al. 1991; Simon et al. 1991). An extension of this approach includes screens designed to identify molecules that modify eye phenotypes generated by the misexpression of genes. Screens particularly pertinent to the work described here have identified previously unknown components of the Ras-MAPK cascade (Karim et al. 1996; Huang and Rubin 2000; Rebay et al. 2000; Therrien et al. 2000). In this article we describe a screen that takes advantage of an eye phenotype caused by the misexpression of an activated form of the FGF receptor Breathless together with the signaling molecule Dof. FGF signaling is not necessary for normal eye development (Casci et al. 1999), and notably, unlike other RTKs, the activated form of the FGF receptor on its own caused no defects. However, in the presence of Dof efficient signaling occurs, indicating that other molecules required for FGF-dependent signal transduction are already present in the eye. It is likely that additional molecules involved in the regulation of FGF signaling that are not expressed in the eye could exist, but one of the advantages of the misexpression approach that we employed in this gain-of-function screen is that it should be possible to identify such genes.

In the developing Drosophila embryo, FGF receptor function can be partly replaced by other RTKs or even by activated Ras (Reichman-Fried et al. 1994; Gisselbrecht et al. 1996; Lee et al. 1996; Michelson et al. 1998; Vincent et al. 1998; Imam et al. 1999; Dossenbach et al. 2001). Thus, the ectopic FGF signal in the eye could simply mimic the function of other RTKs in the eye and lead to the overactivation of their target genes. However, most of the known components identified in related screens (Karim et al. 1996; Huang and Rubin 2000; Rebay et al. 2000; Therrien et al. 2000) did not function as modifiers of the GMR>λ-btl, dof eye phenotype. Furthermore, only Raf acted as a modifier of the phenotype caused by the ectopic FGF signal when we directly tested the components of the canonical RTK/MAPK-signaling pathway. We believe that the effect of Raf upon the GMR>λ-btl, dof eye phenotype is not due to its role in the Ras-MAPK cascade, since in a large screen to identify modifiers of RasV12 (Karim et al. 1996) only one Raf allele was isolated whereas >100 Rolled (MAPK) alleles were recovered. This implies that, in the context of the MAPK cascade within the eye, Rolled rather than Raf is limiting. It is thus more likely that the effect that we observed on FGF signaling represents an independent function of Raf. It is notable that Raf appears to function as a downstream effector of the Drosophila Sterile-20 homolog Slik, which promotes cell proliferation and cell survival via a pathway that is independent of the canonical ERK pathway (Hipfner and Cohen 2003).

Only a small subset of our candidates modified a rough eye phenotype caused by ectopic EGFR signaling. This observation suggests that the overlap in the components that have a major effect on most RTK-signaling pathways is small and, specifically, that the FGF receptor/Dof pathway triggered in the eye differs from that triggered by the EGF receptor. Many more of our candidates were able to modify the defects in eye development associated with misexpression of an activated form of PVR, indicating that the targets of these receptors may overlap to a much greater extent.

The predicted proteins encoded by the genes identified in this screen act in many distinct cellular processes, suggesting that the strength of the ectopic FGF signal responsible for the eye phenotype could be regulated at a number of different levels in vivo. Among the candidates that we have characterized in detail, we found that Sar1 and Robo2 interact with endogenous FGF signaling.

Sar1 is a small GTPase essential for the formation of COPII transport vesicles, as well as for the selection of the cargo. Thus, the strength of FGF signaling could be influenced by the rate of COPII-mediated transport. For instance, the efficient transport of the FGF receptor to the surface of the cell may depend upon selective enrichment of the protein in COPII vesicles, as shown for glycosylated pro-α-factor and Gap1p in Saccharomyces cerevisiae (Malkus et al. 2002). The Drosophila genes encoding putative homologs of other COPII components, namely the protein complexes Sec13 (CG6773)/Sec31 (CG8266) and Sec23 (CG1250)/Sec24 (CG10882 and CG1472), as well as the Sar1 GTP exchange factor Sec12 (CG9175), were not found. However, most of these genes are not associated with an EP element and therefore could not have been identified by this screen.

A role for the cell surface receptors Robo and Robo2 and their ligand Slit in the development of the trachea has been described previously (Englund et al. 2002). Interestingly, the function of Robo2 during outgrowth of tracheal cells toward sources of Slit appears to require FGF signaling, since the misexpression of Slit was able to induce ectopic branch outgrowth only in wild-type embryos but not in embryos mutant for the Drosophila FGF Branchless (Englund et al. 2002). However, the nature of the interaction between the FGF-signaling pathway and Slit/Robo2-signaling pathway has not been established. robo2 is known to interact genetically with Abl, which encodes a kinase involved in the regulation of actin dynamics (Wills et al. 2002). Notably, the activation of Abl correlates with tyrosine phosphorylation (see Hernandez et al. 2004), and phosphorylation of Abl is observed following the activation of Src family kinases by several RTKs (Plattner et al. 1999). Thus, Abl could represent one point at which the FGF- and Slit/Robo2-signaling pathways interact in vivo, and this will be an interesting aspect to explore.


We are very grateful to J. Curtiss and M. Mlodzik for their comments and help with the analysis of the GMR>λ-btl, dof eye phenotype. We thank I. Stüttem for allowing us to screen her copy of the EP collection; S. Roth, D. Montell, and B. Dickson for fly stocks; N. Patel for the 2A12 antibody; and the Bloomington and Szeged Drosophila Stock Centers and FlyBase for providing excellent resources. We are also very grateful to E. Vogelsang for assistance with the generation of the P[w+, UAS-λ-PVR] insertions. T. Hummel, C. Klämbt, and K. Johnson were kind enough to read the manuscript critically prior to its submission. This work was supported by the Deutsche Forschungsgemeinschaft Research Training Group 296 Genetics of Cellular Systems program, the Deutsche Forschungsgemeinschaft grant LE-546/3, and the Sonderforschungsbereich 572.


  • 1 Present address: Institut für Neurobiologie, Universität zu Münster, Badestr. 9, D-48149 Münster, Germany.

  • Communicating editor: K. V. Anderson

  • Received December 13, 2004.
  • Accepted March 11, 2005.


View Abstract