FGF ligands in Drosophila have distinct activities required to support cell migration and differentiation

Most Fibroblast growth factors (FGFs) have N-terminal signal peptides and are secreted from cells, allowing binding to and activation of their cell surface receptors (FGFRs). Studies in vertebrates, as well as simpler invertebrate model systems, have yielded valuable insights into the functions of this important gene family in multiple biological processes(Huang and Stern, 2005; Rottinger et al., 2008; Thisse and Thisse, 2005). In embryos, these functions include mesoderm induction and patterning, cell growth, migration and differentiation. Later functions include organ formation and maintenance, neuronal differentiation and survival, wound healing and malignant transformation (Chen and Deng,2005; Coumoul and Deng,2003; Eswarakumar et al.,2005). FGFs often signal directionally across epithelial-mesenchymal boundaries and exhibit diverse and dynamic patterns of expression that are both spatially and temporally restricted. Tightly regulated expression is one mechanism that helps to ensure the specificity of FGF signaling through FGFRs. An additional level of specificity is most likely imparted by their divergent amino acid sequences, resulting in differential binding affinity for the FGFR, heparan sulfate proteoglycan (HSPG)-interaction specificity (a co-factor involved in FGF-FGFR interaction), and/or FGF ligand range of action (e.g. diffusibility).

FGF and FGFR genes have been identified in organisms ranging from the nematode and fly to mouse and human. There are two FGFs and one FGFR in Caenorhabditis elegans, and three FGFs and two FGFRs in Drosophila melanogaster, as compared with 24 FGFs and four FGFRs(three of which exhibit alternative splicing) in humans and mice(Birnbaum et al., 2005; Huang and Stern, 2005). In vertebrates, over 100 potential FGF-FGFR complexes are predicted(Zhang et al., 2006). Thus,the expectation is that with so many combinations possible, tight regulation of FGF activity and receptor specificity must exist to regulate signaling. Many FGF-FGFR interactions have been studied in the vertebrate system, but the genetic redundancy can make dissection of the functional contribution of particular FGF-FGFR interactions challenging (e.g. Mariani et al., 2008). Often,more than four ligands interact with a particular FGFR isoform at any one time.

Drosophila is an excellent model system for studying FGF signaling, especially now that it appears that the full repertoire of FGF ligands and receptors has been identified (reviewed by Huang and Stern, 2005). Relatively few FGF-FGFR interactions are possible, with only three FGF ligands[Pyramus (Pyr), Thisbe (Ths) and Branchless (Bnl)] and two FGF receptors[Heartless (Htl) and Breathless (Btl)] (reviewed by Ornitz and Itoh, 2001; Szebenyi and Fallon, 1999). The FGFR Btl, and its FGF ligand Bnl, control tracheal branching in the embryo, mesoderm migration over male genital discs, and air sac formation in the larva (Ahmad and Baker,2002; Sato and Kornberg,2002; Sutherland et al.,1996). The preliminary function of the Htl FGFR is to control mesoderm migration during gastrulation(Beiman et al., 1996; Gisselbrecht et al., 1996; Shishido et al., 1997). Later in development, among other functions, Htl is also required for the differentiation of cells that form the heart and hindgut musculature(Michelson et al., 1998; San Martin and Bate, 2001). The ligands for the Htl FGFR in Drosophila had remained elusive,until pyr and ths were identified by genomic screens(Stathopoulos et al., 2004; Gryzik and Muller, 2004). Pyr and Ths share homology with the FGF8 family of vertebrate FGFs. Genetic evidence suggests that this pair of invertebrate FGFs functions through a single FGFR, Htl, to control cell migration and differentiation, as a deficiency mutant that removes both ligands phenocopies the htlmutant (Stathopoulos et al.,2004; Gryzik and Muller,2004). Thus, Drosophila melanogaster provides a unique opportunity within an invertebrate model system, amenable to genetic approaches, to gain insights into why multiple FGF ligands are utilized to activate the same receptor isoform, as is typically the case in vertebrates.

A question in the FGF field is whether the specificity of receptor-ligand interactions is accomplished through the differential expression of ligands or through differences in the signaling properties of the ligand proteins themselves. An impressive analysis of all vertebrate FGF-FGFR interactions was recently completed, in which the binding specificities of ligand-receptor interactions were examined in tissue culture(Ornitz et al., 1996; Zhang et al., 2006). Specificity of receptor-ligand interactions was demonstrated in this system;however, how this relates to in vivo processes is, for the most part,undetermined. Only a limited number of in vivo studies have been conducted to analyze the function of particular FGF-FGFR interactions (e.g. Rentzsch et al., 2008; Yang et al., 2002). In this study, we examine the individual functions of the FGF ligands Pyr and Ths in Drosophila in order to define whether these ligands have distinct functions and/or function redundantly, and to gain insights into why multiple ligands are typically involved in activating a particular FGFR coordinately. We find that the regulated expression of ligands and their divergent protein sequence both contribute to the specificity of FGF-FGFR interactions.

Unless otherwise noted, fly stocks were obtained from the Bloomington Stock Center and were reared under standard conditions at 25°C. The genotype`wild type' refers to the yw genetic background. Df(2R)BSC25has been described (Stathopoulos et al.,2004). The Gal4 drivers zenVRE.Kr-Gal4 and sim-Gal4 were obtained from M. Frasch (University of Erlangen-Nuremberg, Germany) and Stephen Crews (University of North Carolina,Chapel Hill, NC, USA), respectively. Two PiggyBac insertion stocks were obtained from the Exelixis Collection Stock Center at Harvard Medical School(Thibault et al., 2004); we determined that P02915 is located in the first intron of pyr, and confirmed that P02026 is located in the second intron of ths.

Df(2R)ths238 is a deletion of the ths gene that was created in the course of this study through a genetic screen for male-specific recombination (see Preston et al.,1996) using P9.1.2, a viable P-element insertion located ∼400 bp upstream from the ths transcription start site, as mapped by RACE(see Fig. 2A). Line P9.1.2 resulted from mobilization of P10004, a viable insertion located ∼900 bp upstream from the ths transcription start site, to remove 500 bp of ths upstream sequence. Deletion breakpoints were confirmed using inverse PCR, in the case that the P-element was retained [i.e. Df(2R)BSC25 and Df(2R)ths238].

Df(2R)pyr36 is a deletion of the pyr gene that was created by mobilization of the P9.1.2 insertion and screening for excision events. The breakpoints of this deficiency were mapped by complementation using available zygotic lethal insertions (including wal^EY09961 and wal^k14026); we contend that the Df(2R)pyr36 does not extend past the walrus (wal) and ths genes. The 500 bp of ths upstream sequence that was removed in the creation of line P9.1.2 is also absent from Df(2R)pyr36. Yet, ths transcript levels and domains of expression appear similar to those of the wild type in Df(2R)pyr36/Df(2R)pyr36 mutant embryos (expression up to stage 15 was examined; see Fig. 2F and data not shown). It is formally possible that cis-regulatory sequences controlling expression at later stages were removed. However, as the phenotypes exhibited by Df(2R)pyr36 are subtle relative to Df(2R)BSC25, and as the FGF-homologous portion of Ths is located at its N-terminus, it is likely that Df(2R)pyr36 does not remove ths coding sequence.

UAS-bnl and UAS-ths have been described previously(Stathopoulos et al., 2004; Sutherland et al., 1996). UAS-pyr was generated during the course of this work. No EST is available for pyr. Therefore, the 5′ and 3′ ends of the pyr gene were determined by RACE(Stathopoulos et al., 2004). Primers based on this sequence were used to PCR amplify the intact pyr gene. PCR of the intact full-length gene from the RACE cDNA with a single primer pair was not possible, probably owing to repeats within the gene. Instead, two separate PCR reactions were performed to isolate sequence spanning the full-length gene (one to amplify the 5′ end from cDNA and the other to obtain sequence corresponding to the last exon from genomic DNA). The two products were ligated in a three-way ligation to reconstitute the full-length pyr coding sequence.

UAS-pyr and UAS-ths fly stocks were constructed using standard methods (Spralding and Rubin,1982). For the genetic rescue experiments, virgins of the following genetic backgrounds (#1) Df(2R)BSC25 sim-Gal4/CyO ftz-lacZ, (#2) Df(2R)BSC25/CyO wg-lacZ; zenVRE.Kr-Gal4and (#3) Df(2R)BSC25/CyO wg-lacZ; 69B-Gal4 were crossed with males from (#4) Df(2R)BSC25/CyO wg-lacZ; UAS-pyr,(#5) Df(2R)BSC25/CyO wg-lacZ; UAS-ths, (#6) Df(2R)BSC25/CyO wg-lacZ; UAS-bnl, (#7) Df(2R)BSC25/CyO ftz-lacZ; UAS-pyr, (#8) Df(2R)BSC25/CyO ftz-lacZ; UAS-ths or (#9) Df(2R)BSC25/CyO ftz-lacZ; UAS-bnl.

Embryos were manually separated and flash frozen using liquid nitrogen. Three biological replicates of each sample (i.e. 20 dechorionated embryos of stage 9/10) were obtained. Total RNA isolation was carried out using the RNeasy Micro Kit (Qiagen) and quantified using a ND-1000 spectrophotometer(NanoDrop Technologies). One microgram of total RNA was used for reverse transcription to synthesize cDNA using the Gene Racer Superscript III RT Module Kit (Invitrogen). The quantitative PCR (qPCR) reaction was performed in 96-well format using LightCycler SYBR Green 480 (Roche). The qPCR was set up as follows: 95°C for 5 minutes; 35 cycles of 95°C for 20 seconds,55°C for 20 seconds, and 72°C for 30 seconds; and a final 72°C for 10 minutes. For quantifying the transcript levels in ectopically expressing pyr and ths lines, the data were normalized to the transcript level of a housekeeping gene, Elongation factor 1(Ishimoto et al., 2005). In order to quantify the levels of `ectopic' pyr and thsexpression supported by the Gal4 driver and transgenes containing UAS, we used twist-Gal4 and UASlacZ as a control and subtracted the endogenous pyr and ths transcript levels; the domain of ectopic expression supported by twist-Gal4 is significant, as compared with expression supported by zenVRE.Kr-Gal4 and sim-Gal4, and thereby allowed for better resolution of the levels of ectopic expression.

The lines chosen [pyrline1: UAS-pyr:AMS330(III) and thsline1: UAS-ths:AMS289.22(III)] exhibit comparable levels of ectopic expression as assayed by qPCR (see Fig. S4 in the supplementary material) and support equivalent Eve⁺ expression when driven by 69B-Gal4 (see Fig. S3D,E in the supplementary material). In another set of insertions, approximately twice as much ectopic ths was supported, relative to the levels of pyr (see Fig. S4 in the supplementary material, thsline2 and pyrline2);nevertheless, the same results were obtained with this other set of lines in the zenVRE.Kr-Gal4-driven rescue experiments (data not shown). Therefore, we contend that the ability of pyr alone to support the migration of mesoderm cells at a distance is not due to differences in expression levels resulting from positional effects on transgene expression.

The following antibodies were used: guinea pig anti-Twist (1:200; from M. Levine, UC Berkeley, CA, USA), mouse anti-dpERK (1:200; Sigma), rabbit anti-β-galactosidase (1:200; Molecular Probes), rabbit anti-Eve (1:400;from M. Frasch) and monoclonal 2A12 (1:3; Developmental Studies Hybridoma Bank). In situ hybridization (Kosman et al., 2004; Lehmann and Tautz,1994), double antibody plus in situ hybridization detection of protein and transcript levels (Frasch,1995), and staining of tracheal branches using 2A12 antibody(Sutherland et al., 1996) were conducted as previously described.

For sectioning, embryos were sorted for appropriate stage and genotype(i.e. using balancer chromosomes containing reporters), and embedded in acetone-Araldite (Electron Microscopy Sciences) as described(Leptin and Grunewald, 1990). Sections (10 μm) were cut using a Historange 2218 microtome.

Because variability is present in the Df(2R)BSC25 mutant background, a system of quantitation of mutant phenotype was devised. We scored mesoderm spreading phenotypes in the deficiency and the ectopically expressing lines (see Figs S5 and S6 in the supplementary material). Clumping was used as a measure of the ability of ligands to attract cells to the ventral midline; the multilayer phenotype was scored by assessing the ability of ligands to attract cells to the dorsal-most ectoderm.

It is likely that Pyr and Ths are ligands for the Htl FGF receptor, as these genes are expressed near sites of Htl expression(Stathopoulos et al., 2004). During gastrulation, pyr and ths are expressed in distinct domains within the ectoderm at the surface of the embryo (see Fig. 1A-C). Early in development, ths is expressed in ventral regions(Fig. 1F,G) and pyr is expressed in dorsal regions (Fig. 1D,E). Once invagination is complete and all mesoderm cells have entered the interior of the embryo, these cells proceed to migrate along the inner ectodermal surface, forming a monolayer. Activation of the FGFR Htl is thought to trigger the spreading of the mesoderm across the underlying ectoderm (Fig. 1A). The dynamic expression domains of the ligands suggested a possible mechanism by which directional information could be imparted to the mesoderm cells moving beneath, causing these cells to move in the dorsal direction.

To determine whether the distinct domains of ligand expression are important for proper mesoderm migration, we altered their expression domains and assayed effects on mesoderm spreading. Using the 69B-Gal4 driver,the bnl, pyr and ths FGF-encoding genes were ectopically expressed throughout the entire ectoderm(Fig. 1H-M). For pyrand ths, several transgenic lines were first compared with the control for positional effects that might influence the levels of gene expression, and a set of lines was chosen that exhibited comparable levels of ectopic expression by Gal4 as assayed by qPCR (see Fig. S4 in the supplementary material). Ectopic expression of Pyr throughout the ectoderm resulted in mesoderm migration defects(Fig. 1J,K; n=9). The pyr ectopic expression defect was specific; mesoderm spreading occurred, but a cell monolayer was not formed(Fig. 1K). By contrast, ectopic expression of Bnl or Ths in the same manner had little to no effect on mesoderm monolayer formation (Fig. 1G,I and Fig. 1L,M; n=6 and n=10, respectively).

The MAPK pathway is activated in migrating mesodermal cells as a result of FGFR receptor tyrosine kinase activation, and leads to dual phosphorylation of MAPK [also known as ERK (Gabay et al.,1997) and Rolled - FlyBase]. A monoclonal antibody that recognizes this activated form of MAPK (dpERK) can be used to monitor activation of this pathway within cells. dpERK staining is observed within gastrulating embryos at the leading edge of the migrating mesoderm cells (see Fig. S1A in the supplementary material), but is lost in the absence of Htl activation(Michelson et al., 1998; Stathopoulos et al., 2004). In embryos ectopically expressing pyr or ths via twist-Gal4, which supports their expression throughout the mesoderm,dpERK is ectopically expressed throughout all mesoderm cells (see Fig. S1C,E in the supplementary material)(Stathopoulos et al., 2004). However, when the 69B-Gal4 driver was used to ectopically express pyr or ths, only the ectopic expression of pyrsupported ectopic activation of ERK throughout the entire mesoderm (compare Fig. S1D with Fig. S1F in the supplementary material). Ectopic expression of bnl had no effect: dpERK remained localized at the leading edge of the migrating mesoderm [compare Fig. S1B with Fig. S1A (see arrowheads) in the supplementary material]. Depending on where the ligands were expressed,differences in the capacity of Pyr versus Ths to activate signaling were observed, whereas Bnl had no effect. Furthermore, the ability of these ligands to support ectopic dpERK expression was found to correlate with the inhibition of mesoderm migration (see Fig. 1; data not shown).

We then assayed the ability of Pyr and Ths to influence trachea formation,a function that is supported by Bnl-Btl. Ectopic expression of bnlvia 69B-Gal4 leads to a substantial increase in tracheal branching(compare Fig. S2B with Fig. S2A in the supplementary material)(Sutherland et al., 1996). However, no such effect on the trachea was observed upon ectopic expression of pyr or ths (see Fig. S2C,D in the supplementary material),suggesting that neither Pyr nor Ths can activate the Btl FGFR.

Mesoderm spreading defects are observed in htl mutants and in a deficiency mutant, Df(2R)BSC25, that removes both pyr and ths (Fig. 2A)(Beiman et al., 1996; Gisselbrecht et al., 1996; Stathopoulos et al., 2004). We sought to test further the requirement for Pyr or Ths in activation of the Htl FGFR by analysis of single-mutant phenotypes. First, we identified two PiggyBac insertions located in proximity to pyr and ths (see Fig. 2A). The line containing the pyr insertion, pyr⁰²⁹¹⁵, is zygotically lethal, whereas the line containing the ths insertion, ths⁰²⁰²⁶, is only semi-lethal [i.e. homozygous flies are obtained at very low frequencies (∼0.5%) and are short-lived]. Each insertion disrupts expression of the respective gene, and we detect no effect on expression of the other ligand (see Fig. 2E,G, compare with Fig. 2A-C). As phenotypes associated with insertions can vary, we chose to identify true null alleles.

Deletions of the entire pyr or ths coding sequences were created (see Fig. 2A and Materials and methods). Because pyr and ths are linked,deletions resulting from mobilization of the insertion P10004 had the potential to remove either ths or pyr sequences. Through several rounds of male-specific recombination screening, we were able to identify a deletion that removes the ths gene in its entirety, which we named Df(2R)ths238. The entire ths coding sequence, as well as ten proximal genes, are absent, yet the entire pyr locus remains intact. Using excision-based screening with the same P9.1.2 insertion,we obtained a deletion that removes sequence upstream of ths,including the entire pyr gene and five proximal genes, which we named Df(2R)pyr36.

Complementation tests were conducted to determine genetic interactions between our allelic series of pyr and ths mutants. We believe that the pyr⁰²⁹¹⁵ allele represents a pyrnull, as no expression of pyr was detected at any time-point examined. It is unclear whether the ths⁰²⁰²⁶ allele represents a ths null, as expression could be detected in the visceral mesoderm at later stages (data not shown). Df(2R)pyr36 and Df(2R)ths238 represent null alleles of pyr and ths,respectively, because the entire coding region of each gene is absent;however, of note is the fact that other genes are also removed by these∼100 kb deletions (see Materials and methods for details). This allelic series (i.e. PiggyBac insertions: weak alleles; Df(2R)pyr36 and Df(2R)ths238 deficiencies: strong alleles; Df(2R)BSC25deficiency: double mutant; see Fig. 2A) provided the first opportunity to analyze transheterozygous combinations in order to gain insights into pyr and thssingle-mutant phenotypes. In both pyr and ths single mutants, the mesoderm spreading was aberrant(Fig. 3G-J, compare with Fig. 3A,B). The mesoderm was multilayered and a monolayer rarely formed. However, neither mutant exhibited a phenotype as severe as that of the Df(2R)BSC25 or htlmutants (Fig. 3C-F), suggesting a role for both Pyr and Ths in guiding mesoderm spreading.

To determine whether the localized expression of each ligand is important for its function, we took an ectopic expression approach using Gal4 to promote expression of a given ligand in the pyr and thsdouble-mutant background. Df(2R)BSC25 removes both pyr and ths, and 16 other genes, none of which has known embryonic phenotypes(Fig. 2A)(Stathopoulos et al., 2004). The sim-Gal4 driver (Xiao et al.,1996) delivers ligands at the ventral midline where single-minded (sim) is normally expressed (see Fig. 4F). Alternatively, the zenVRE.Kr-Gal4 driver can be used to deliver ligands to the dorsal domain of the ectoderm within the lower half of gastrulating embryos (see Fig. 4K)(Frasch, 1995).

When the sim-Gal4 driver was used to express either pyror ths in the Df(2R)BSC25 mutant background(Fig. 4F), no rescue of mesoderm spreading was observed (Fig. 4G,H). In fact, clumps of mesoderm cells were observed overlying the site of Pyr ectopic expression, suggesting that this ligand had attracted mesoderm cells to this position (Fig. 4G). This phenotype appeared to be more severe than that of Df(2R)BSC25 mutants (Fig. 4G,H, compare with Fig. 4B,C; see Fig. S5 in the supplementary material; P<0.01). Ectopic expression of ths in this manner also had a weak effect, as cells tended to pool at the ventral midline(Fig. 4H), but this effect was not supported by statistical significance as the Df(2R)BSC25 mutants themselves exhibit variability (Fig. 4B,C and see Fig. S5 in the supplementary material).

Different results were obtained using the zenVRE.Kr-Gal4 driver. In this case, ectopic expression of either pyr or ths did appear sufficient to direct mesoderm cells toward dorsal regions of the ectoderm in the lower ventral half of Df(2R)BSC25 mutant embryos(Fig. 4L,M; n=17 and n=19, respectively), within the domain of ectopic expression supported by the driver (see Fig. 4K). In Df(2R)BSC25 mutant embryos, some spreading toward dorsal regions was observed in only 28% of the embryos examined (n=10 of 36 embryos scored in total) (Fig. 4B and see Fig. S5 in the supplementary material); therefore, this`rescue' in the lower half of the embryos by pyr and ths is significant (see Figs S5 and S6 in the supplementary material). Surprisingly,however, ectopic expression of pyr in this manner was also capable of directing the movement of mesoderm cells in the top half of embryos, at a distance from the source of pyr expression supported by zenVRE.Kr-Gal4. For ths ectopic expression, the effect on the top half was not as strong and exhibited more variability (see Figs S5 and S6 in the supplementary material). In Df(2R)BSC25, as well as in htl<sup>AB42</sup> mutants, such a `mixed' phenotype was not typical,as both top and bottom halves of the embryo exhibited similar phenotypes (see Fig. 4L,M, compare with Fig. 4C).

Collectively, these results suggest that localized pyr expression is important to support mesoderm spreading, as zenVRE.Kr-Gal4 is able to promote spreading, whereas sim-Gal4 is not, and also suggest the hypothesis that Pyr protein can function at a distance, whereas Ths functions more locally. To test this, embryos were assayed for dpERK activity(Fig. 4D-O). Both pyrand ths expression through sim-Gal4 was able to support dpERK activation (Fig. 4I,J). Therefore, the weaker effect of ths is unlikely to be due to an inability to activate the receptor in this domain. Moreover, pyrexpression through zenVRE.Kr-Gal4 supported strong activation of dpERK throughout the lower half of the embryos(Fig. 4N), whereas the effect of ectopic expression of ths in this same domain was more limited(Fig. 4O). Both pyrand ths expression supported dpERK activation at the leading edge of mesoderm cells in the top half of embryos(Fig. 4N,O). Mesoderm cells in the top half of the embryo are able to receive both Pyr and Ths signals, but only a subset of the cells migrates toward the dorsal ectoderm in response to Ths (see Discussion).

After mesoderm cell spreading has been completed, both Pyr and Ths are expressed in regions of the dorsal ectoderm that abut a subset of the mesoderm cells (Fig. 2D and see Fig. S3A,B in the supplementary material). pyr is present in more-dorsal regions of the ectoderm, overlying sites of dorsal mesoderm specification;limited expression of ths is also seen in this same region(Fig. 2D). The mesoderm cells that come into contact with the dorsal ectoderm receive an inductive signal from Decapentaplegic (Dpp), a Drosophila TGFβ homolog (reviewed by Frasch, 1999). Thereby,expression of genes such as even skipped (eve) and tinman, which are required for the differentiation of cardiac and pericardial tissues, depend on Dpp as well as on Htl-dependent FGF signaling.

We sought to define the role of each ligand in the specification of dorsal mesoderm cell lineages within germ-band elongated embryos. Normally, Eve is expressed within 12 hemisegments. In pyr single mutants, often only a few hemisegments exhibited Eve expression(Fig. 5D,E, compare with Fig. 5A), whereas within ths single-mutant embryos, only subtle defects in Eve⁺cell specification were observed (Fig. 5G,H, compare with Fig. 5A). The pyr mutant phenotype was much stronger than the phenotype exhibited by ths single mutants at all stages examined, and even when comparisons were made between the weakest pyr allele(pyr⁰²⁹¹⁵) and the strongest ths allele[Df(2R)ths238] (Fig. 5C,F, compare with Fig. 5I). Moreover, the phenotype of Df(2R)pyr36 was consistently stronger than that of pyr⁰²⁹¹⁵, suggesting that either pyr⁰²⁹¹⁵ is not a null allele or that ths gene function also supports Eve⁺ cell specification. Although Pyr has the more dominant role, Ths probably also plays a role in supporting Eve expression because the Df(2R)BSC25 mutant phenotype is more severe than that of Df(2R)pyr36 mutants. No Eve⁺cells are present within the dorsal mesoderm of Df(2R)BSC25 mutants,a phenotype similar to that of htl mutants(Fig. 6B)(Stathopoulos et al.,2004).

To investigate the roles of Pyr and Ths in controlling cell differentiation, which might differ from their roles during cell migration, we focused on how ectopic expression of these ligands affects specification of Eve⁺ cells within the dorsal mesoderm of the embryo. We have demonstrated previously that when Ths is ectopically expressed in the ectoderm using the 69B-Gal4 driver, ectopic Eve⁺ cells result; this is similar to the phenotype observed when constitutively activated Htl is ectopically expressed in the same manner (see Fig. S3E in the supplementary material) (Stathopoulos et al.,2004). We found that ectopic expression of Pyr driven by 69B-Gal4 also increases the number of Eve⁺ cells (see Fig. S3D in the supplementary material). Normally, three to four Eve⁺cells are specified in 12 hemisegments (see Fig. S3C in the supplementary material). In the presence of excess levels of Pyr or Ths, ∼6-15 cells were specified and Eve expression was expanded to 14 hemisegments (compare Fig. S3C with Fig. S3D in the supplementary material). When activated Ras is ectopically expressed in the ectoderm, a similar increase in the number of Eve⁺ cells is observed (Carmena et al., 1998).

Next, we compared the ability of pyr, ths or bnl to rescue Eve⁺ expression in the Df(2R)BSC25 mutant background under conditions of equivalent expression. Either Pyr or Ths was able to support Eve⁺ expression within dorsal mesoderm cells within Df(2R)BSC25 mutants when expressed using 69B-Gal4 (`full rescue'; Fig. 6D,E, compare with Fig. 6B) or zenVRE.Kr-Gal4 (`partial rescue'; Fig. 6G,H, compare with Fig. 6B) drivers. Similar expression of Bnl failed to support the specification of any Eve⁺cells (Fig. 6F,I). In addition,when expressed at the ventral midline using sim-Gal4, Bnl, Pyr and Ths all failed to support Eve expression (see Fig. 6C; data not shown). In summary, these results suggest that FGF receptors exhibit ligand-binding preferences, and that the ligands must be expressed in proximity to the Htl FGFR for activation to occur.

The experiments outlined above demonstrate that the DrosophilaFGFs Pyr, Ths and Bnl have different functions and that the activation of FGF receptors by specific ligands affects particular biological processes. Examination of an allelic series of pyr and ths mutants suggests that pyr and ths are not redundant in function:both influence mesoderm spreading, whereas pyr is the dominant player controlling Eve⁺ cell specification within the dorsal mesoderm(Figs 3 and 5). We have demonstrated previously that ectopic expression of ths by twist-Gal4 and 69B-Gal4 in the Df(2R)BSC25 mutant background can support Htl FGFR activation (Stathopoulos et al.,2004). However, in this study, we assayed whether the expression supported in distinct domains would support Htl activation. By a series of`rescue' experiments, through ectopic expression of one ligand in the Df(2R)BSC25 mutant background, we obtained evidence that localized expression of the ligands is important for proper mesoderm spreading. We find,surprisingly, that the ligands exhibit differences in their functional range of action (Fig. 4). In addition, using this same approach, we find that either Pyr or Ths can support Eve⁺ cell specification within the dorsal mesoderm, but that Bnl cannot (Fig. 6). Collectively,these data suggest that the Pyr and Ths FGFs function as ligands for the Htl FGFR and that specificity of FGF-FGFR interactions exists in Drosophila.

Our results demonstrate that both Pyr and Ths FGF ligands can activate the Htl FGFR, whereas only the Bnl FGF ligand can activate the Btl FGFR(Fig. 6 and see Fig. S2 in the supplementary material). Specificity of FGFR activation was observed: pyr or ths, but not bnl, expression is able to activate Htl to affect expression of Eve, and bnl, but neither pyr nor ths, is able to support tracheal specification. No evidence was obtained that other cross-interactions occur (i.e. Pyr-Btl,Ths-Btl or Bnl-Htl), which demonstrates that Gal4-mediated ectopic expression does not simply `swamp the system'. This experimental approach also `levels the playing field', as expression of each ligand is driven at the same time and place and presumably at similar levels. We conclude that only three FGF-FGFR combinations function in Drosophila (i.e. Pyr-Htl, Ths-Htl and Bnl-Btl), which supports the idea that FGFRs exhibit ligand-binding preferences. Previous studies have investigated FGF signaling specificity by analyzing the ability of other receptor tyrosine kinases to support cell migration or by activating particular intracellular signaling pathways to examine which are required to effect FGFR-dependent cell migration versus cell differentiation (Dossenbach et al.,2001; Wilson et al.,2005). In this work, we analyzed the specificity of FGF ligand-receptor interactions and how they contribute to particular developmental processes.

When ligand expression is supported by twist-Gal4, Htl FGFRs presumably become saturated because dpERK is ectopically activated in all cells and spreading is negatively affected (see Fig. S1 in the supplementary material). One explanation for why this might affect mesoderm cell spreading is that these FGF-saturated mesoderm cells may no longer be competent to respond to endogenous ligands that provide directional cues. Recently, we have shown that movement of the mesoderm cells during gastrulation is in fact directional (McMahon et al.,2008). Pyr and Ths ligands are differentially expressed during gastrulation and this might provide the necessary positional information required to direct migration of the mesoderm (see Fig. 7A).

We propose that Pyr and Ths have different activities that fulfil aspects of FGFR activation required to support cell migration(Fig. 7A). Ectopic expression of Pyr within the ectoderm negatively affects mesoderm spreading(Fig. 1H-M), which suggests that the refined expression domain of pyr within cells of the dorsal ectoderm is normally required to guide the mesoderm cells toward dorsal regions. However, even though ectopic expression of ths in the ectoderm has no effect on mesoderm spreading, ths mutants also exhibit defects in mesoderm spreading, demonstrating that both genes are required, perhaps to control different aspects of the migration. Our `rescue'experiments using the zenVRE.Kr-Gal4 driver support the view that Pyr has a longer functional range than Ths(Fig. 4N,O). These differences in range of function might correlate with different diffusion capabilities,but an alternative explanation is that the ligands activate the receptor with different affinities. Additional experiments will be necessary to distinguish their exact functions and to uncover the molecular basis for the differential functions of Pyr and Ths; we suggest that in vivo imaging and quantitative analysis (McMahon et al.,2008) of single-mutant phenotypes will provide insights.

With regard to the FGF-dependent cell differentiation, our `rescue'experiments suggest that ectopic expression of either Pyr or Ths is sufficient to support Eve⁺ cell specification(Fig. 6D,E). The reason why loss of ths has less of an effect on Eve⁺ cell specification is most likely because pyr is prominently expressed in the vicinity of the future Eve⁺ cells; normally, Pyr supports this function, but Ths can support this activity if presented at sufficient levels within the correct domain. Furthermore, we propose that FGF signaling might not play an instructive role in supporting eve expression(Fig. 7B). Other signaling pathways already provide positional information required for the specification of Eve⁺ cells; FGF signaling pathway activation might simply serve a permissive role, and in this context either ligand would suffice.

We have used Drosophila to study FGF signaling and determine why multiple ligands are utilized to activate the same receptor. It will be informative to obtain additional insights into how these ligands differ in their activities. The expression domains of the ligands do confer information that is important for controlling developmental processes, but their individual protein sequences also impart differential functionality. For instance, the ligands might exhibit different affinities for specific HSPGs that influence their range of diffusion, or the proteins themselves may have different stabilities. Future experiments will also define the FGF ligand preferences that exist to support FGF signaling at later stages of development.

At least 15 human genetic diseases result from mutations within FGFR genes and each disease is caused by a different mutation shown to affect receptor activation (Chen and Deng,2005). Several mutations in FGFR lead to an expansion of FGF ligand-binding preference (Ornitz,2005); however, it is still not clear why different mutations yield different syndromes. Continuing this work in order to understand how different FGF ligands activate the same receptor to effect different outcomes is an important goal, as this may provide insights into why different mutations in the same FGFR lead to various dysplasias and diseases(Wilkie, 2005).