Heartbroken is a specific downstream mediator of FGF receptor signalling in Drosophila

Signalling by fibroblast growth factor receptors (FGFRs) has been implicated in a diversity of biological processes. These include the migration, specification, differentiation, proliferation and survival of cells during the development of numerous tissues in a wide variety of organisms (Mason, 1994; Yamaguchi and Rossant, 1995; Green et al., 1996). FGFRs are receptor tyrosine kinases (RTKs) that upon ligand binding are activated by dimerization and autophosphorylation (Schlessinger et al., 1995). This is followed by the receptor-mediated phosphorylation of additional protein substrates, leading to the stimulation of a conserved set of signal transduction molecules. The latter include the adapter protein GRB2/DRK, the guanine nucleotide exchange factor SOS, the small GTP-binding protein RAS, the serine-threonine kinase RAF and the mitogen-activated protein kinase (MAPK) that directly mediates the nuclear responses to receptor activation (Fantl et al., 1993; van der Geer et al., 1994). Many of the components of this intracellular signal transduction pathway are utilized by all RTKs, raising the question of how signal specificity is achieved for each receptor.

A number of mechanisms can potentially contribute to RTK ouput specificity (Mason, 1994; Marshall, 1995; Songyang et al., 1995; Weiss et al., 1997). Since a biological response is determined by the combination of pathways transducing a signal from a cell surface receptor, the spectrum of signalling molecules present in a given cell and capable of interacting with a particular receptor will dictate how the cell will respond to the activating ligand. Thus, specificity can be achieved by a dedicated factor acting downstream of a receptor if that factor is able to recruit or activate novel combinations of additional signalling molecules. Such a function may be served by IRS1 and FRS2 for the mammalian insulin and FGF receptors, respectively, although even these adapters are not entirely specific to a single class of receptor (White, 1994; Kouhara et al., 1997). The activation of tissue-specific effectors can also determine the unique effects of a single RTK (Clandinin et al., 1998).

The mammalian nerve growth factor receptor activates MAPK in a novel manner that is independent of the well-characterized RAS-mediated mechanism (York et al., 1998). This pathway affects the duration and magnitude of MAPK activation which, in turn, determines whether the stimulated cell will differentiate or proliferate (Marshall, 1995). A RAS-independent process also operates downstream of at least one Drosophila RTK (Hou et al., 1995). Furthermore, effects of RAS that are independent of RAF have been described (D’Arcangelo and Halegoua, 1993; Lee et al., 1996a). Modulation of RTK signals by other classes of receptors can additionally affect cellular responses (Christian et al., 1992; LaBonne et al., 1995; Pan et al., 1995). Thus, novel elements of a common signal transduction apparatus, branching pathways unique to particular receptors and interactions with other signalling systems can contribute to RTK response specificity.

Two FGFRs have been characterized in Drosophila. heartless (htl) encodes an FGFR that is expressed in the mesoderm and central nervous system where it is required for cell migration and for the determination of a subset of mesodermal cell fates (Beiman et al., 1996; Gisselbrecht et al., 1996; Michelson et al., 1998; Shishido et al., 1993, 1997). In htl mutants, the early mesoderm fails to spread into the dorsal region of the embryo. This migration defect accounts for the absence from htl mutant embryos of the heart, visceral musculature and dorsal somatic muscles since the formation of these mesodermal derivatives requires induction by DECAPENTAPLEGIC (DPP), a growth factor produced by the dorsal ectoderm (Staehling-Hampton et al., 1994; Frasch, 1995). Ectopic expression of DPP activates the expression of dorsal mesodermal markers in htl mutant embryos, indicating that htl is not required for the capacity of these cells to respond to this growth factor (Beiman et al., 1996; Gisselbrecht et al., 1996). HTL expression persists in the mesoderm at later developmental times and inhibition of HTL activity after migration is complete blocks the specification of certain mesodermal cells (Beiman et al., 1996; Michelson et al., 1998). Thus, the HTL FGFR is involved not only in cell movement, but also has a direct role in cellular commitment.

A second FGFR is encoded by the Drosophila breathless (btl) gene. btl is expressed in embryonic tracheal and midline glial cells, as well as in the border cells of the ovary (Glazer and Shilo, 1991; Klämbt et al., 1992; Murphy et al., 1995). BTL is essential for normal tracheal and glial cell migration, and also participates in the movement of border cells. Ectopic expression of a constitutively active BTL receptor suggested that proper tracheal branching depends on the spatially regulated activation of this RTK (Lee et al., 1996b). This hypothesis was confirmed with the identification of BRANCHLESS (BNL), the BTL ligand, which is expressed at sites surrounding the tracheal system where new branches form (Sutherland et al., 1996). The dynamic pattern of BNL expression guides primary tracheal branching and ectopic BNL is capable of redirecting normal tracheal migration. In addition, BNL activates BTL in the determination of the tracheal cells that give rise to secondary and terminal branches (Reichman-Fried and Shilo, 1995; Lee et al., 1996b; Sutherland et al., 1996). Thus, the Drosophila HTL and BTL FGFRs function in similar processes – the early directional migration of the cells in which they are expressed and the subsequent specification of particular mesodermal and tracheal cell fates.

The structural and functional similarities between the two Drosophila FGFRs raise the possibility that these RTKs share a common signalling mechanism. It might also be expected that some components of this pathway will confer specificity to the related responses of these two receptors. RAS1 functions downstream of both HTL and BTL, but also is involved in signalling by other RTKs in Drosophila (Simon et al., 1991; Diaz-Benjumea and Hafen, 1994; Reichman-Fried et al., 1994; Hou et al., 1995; Gisselbrecht et al., 1996). Thus, unique FGFR outputs must be achieved by more specific components. Of further relevance to the signalling specificity of HTL and BTL is the fact that another RTK, the Drosophila epidermal growth factor receptor (EGFR or DER), functions in some of the same cells as these FGFRs (Wappner et al., 1997; Buff et al., 1998). We now describe the identification of heartbroken (hbr), a novel gene that is involved in the migration of both mesodermal and tracheal cells. hbr interacts genetically with htl and btl, and reduction of hbr function is epistatic to the hyperactivation of both FGFRs. Moreover, hbr functions either upstream of RAS or on a parallel pathway leading to MAPK activation by HTL and BTL. In contrast, hbr does not participate in embryonic signalling by DER. These results indicate that hbr encodes a unique component of an RTK signalling pathway that may contribute to FGFR response specificity.

The following Drosophila mutant strains were employed – htl^AB42, htl^YY262 (Gisselbrecht et al., 1996), btl^LG19, btl^H82Δ3 (Klämbt et al., 1992), bnl^P1 (Sutherland et al., 1996) and flb^1K35 (Clifford and Schüpbach, 1994). The hbr^YY202 allele was isolated in an ethylmethane sulfonate (EMS) mutagenesis screen for lethal mutations on the third chromosome that disrupt mesoderm and nervous system development (Gisselbrecht et al., 1996). Two additional hbr alleles, hbr^ems6 and hbr^ems7, were obtained in a separate EMS mutagenesis designed to isolate mutations that fail to complement hbr^YY202. hbr also fails to complement Df(3R)ry^506-85C which was obtained from the Bloomington Drosophila Stock Center. Targeted expression of the following transgenes was accomplished using the GAL4/UAS system (Brand and Perrimon, 1993) – UAS-activated RAS1 (Gisselbrecht et al., 1996), UAS-DPP (Staehling-Hampton et al., 1994), UAS-BNL (Sutherland et al., 1996), UAS-activated DER (Queenan et al., 1997) and UAS-activated HTL (see below). Mesodermal expression was directed by twi-GAL4 (Greig and Akam, 1993) and ectodermal expression by 69B-GAL4 (Brand and Perrimon, 1993). Combinations of mutations and GAL4 or UAS insertions were generated by standard genetic crosses. Balancer chromosomes marked with lacZ expression constructs facilitated the identification of homozygous mutant embryos. Oregon R was used as a wild-type reference strain.

Standard methods were used for embryo fixation, immunohistochemical staining and in situ hybridization (Gisselbrecht et al., 1996). All antibodies and probes have been described previously (Gisselbrecht et al., 1996; Samakovlis et al., 1996; Gabay et al., 1997a,b). In the case of anti-diphospho-MAPK (Gabay et al., 1997a,b), peroxidase staining was enhanced using Tyramide Signal Amplification reagents (New England Nuclear) in combination with the Vectastain ABC Elite kit (Vector Laboratories). Embryos were sectioned and EVE expression in the dorsal mesoderm was quantitated, as described (Gisselbrecht et al., 1996; Michelson et al., 1998). The statistical significance of differences in EVE expression values was established using both one-tailed z- and t-tests.

The dimerization domain of the bacteriophage λ cI repressor was obtained as a 0.6 kb NotI-BglII restriction fragment from the plasmid, pHSλbtl (Lee et al., 1996b). Using the polymerase chain reaction (PCR), a BglII restriction site was introduced into the htl-coding sequence at amino acids 279/280, just N-terminal to the HTL transmembrane domain (Shishido et al., 1993; Gisselbrecht et al., 1996), and a NotI site was simultaneously created in the htl 3′-untranslated region. The resulting 1.35 kb PCR product was cloned and sequenced in its entirety to verify that an intact htl coding region was obtained, flanked by the desired restriction sites. The λ repressor and htl NotI-BglII fragments were cloned into the NotI site of pUAST (Brand and Perrimon, 1993) to generate pUAS-λhtl. DNA sequencing confirmed that an in-frame fusion had been generated at the BglII site between the λ repressor and htl coding regions.

pUAS-λhtl was injected along with a helper plasmid encoding P transposase into embryos derived from a y w Drosophila strain (Spradling, 1986). w⁺ transgenic lines were selected and insertion sites were mapped by standard genetic crosses.

We previously described the isolation of mutations in the Drosophila htl gene that eliminate development of the single somatic muscle and subset of pericardial cells that express EVEN-SKIPPED (EVE) in the dorsal region of the embryonic mesoderm (Gisselbrecht et al., 1996; Fig. 1A,B). An independent complementation group with a mesodermal EVE phenotype identical to that of htl was obtained in the same genetic screen (Fig. 1C). As occurs for htl, the development of other dorsal mesodermal derivatives is severely reduced in this new mutant, including the cardial cells of the heart (Fig. 1D-I), most dorsal somatic muscles (Fig. 1D-I), pericardial cells in addition to those expressing EVE (Fig. 1J-L) and the midgut visceral mesoderm (Fig. 1M-O). On the basis of these and other phenotypic similarities to htl (see below), we have named this new gene heartbroken (hbr).

Recombination and deficiency mapping indicate that hbr is located on the right arm of the third chromosome within the 88B-E cytological interval (data not shown). One allele, hbr^YY202, was obtained in the original screen and two additional alleles, hbr^ems6 and hbr^ems7, were obtained in a subsequent mutagenesis. Of the three, hbr^YY202 has the strongest mesodermal phenotype (see below), although comparison to a deficiency that covers hbr indicates that this allele is not null but rather is a strong hypomorph.

HTL signalling is essential for normal mesoderm development (Beiman et al., 1996; Gisselbrecht et al., 1996; Shishido et al., 1997; Michelson et al., 1998). Since hbr and htl have very similar phenotypes, we were interested in whether hbr interacts genetically with htl, implying a possible role for HBR in HTL signalling. We investigated this possibility in two ways.

First, we examined whether hbr is capable of dominantly enhancing a partial loss-of-function htl allele. In contrast to a null allele of htl, the hypomorphic htl^YY262 allele (Gisselbrecht et al., 1996) does not completely eliminate mesodermal EVE expression (Fig. 2A). On average, each htl^YY262 mutant embryo has 1.5 hemisegments with at least one EVE-positive dorsal mesodermal cell. This is reduced to 0.8 EVE-positive hemisegments in the presence of one mutant copy of hbr, a highly significant difference (P<10⁻⁵). Thus, hbr not only has a htl-like mesodermal phenotype, but it also exhibits a dosage-sensitive genetic interaction with htl.

Second, we determined if hbr can dominantly suppress a hyperactivated form of HTL. The latter was generated by replacing the extracellular domain of the HTL receptor with the dimerization domain of the bacteriophage λ cI repressor. This strategy constitutively activates RTKs by causing ligand-independent receptor dimerization (Lee et al., 1996b). Such a construct was placed under GAL4 UAS control and ectopically expressed in transgenic embryos using a twi-GAL4 driver (Greig and Akam, 1993).

Under the influence of activated HTL, an increased number of EVE-expressing muscle and cardiac progenitors is formed in an otherwise wild-type genetic background (Fig. 2B,C). This response is equivalent to that obtained with activated forms of RAS1 or DER and reflects the involvement of both RTKs in mesodermal cell fate specification (Gisselbrecht et al., 1996; Buff et al., 1998; Michelson et al., 1998). Heterozygous hbr alone does not affect mesodermal EVE expression or mesoderm migration (Fig. 2D and data not shown). However,

one mutant copy of hbr markedly suppresses the influence of activated HTL on the formation of EVE-positive mesodermal cells (Fig. 2E). This suggests that hbr is required for the cell fate specification function of HTL, in addition to its involvement in HTL-dependent cell migration (see below). These findings establish a strong genetic interaction between hbr and htl, suggesting that hbr is involved in mesodermal FGFR signalling.

The earliest phenotype observed in htl mutant embryos is a lack of dorsolateral migration of the invaginated mesoderm (Beiman et al., 1996; Shishido et al., 1997; Gisselbrecht et al., 1996; Michelson et al., 1998; Fig. 3B,D). This abnormality accounts for the later absence of dorsal mesodermal derivatives since these structures require induction by DPP which is supplied by the dorsal ectoderm (Staehling-Hampton et al., 1994; Frasch, 1995). Since the phenotype of later stage hbr embryos is very similar to that of htl (Figs 1, 3A,C,E), we examined the effect of a hbr mutation on earlier stages of mesoderm development. As revealed by transverse sections of stage 10 TWIST (TWI)-stained embryos, there is a severe defect in the dorsolateral spreading of hbr mutant mesodermal cells (Fig. 3F). Thus, as is the case for htl, the heart, dorsal somatic muscles and visceral mesoderm fail to develop properly when hbr function is reduced because mispositioned mesodermal progenitors are not exposed to DPP.

RAS1 is a key signal transducer acting downstream of all RTKs, including HTL (Fantl et al., 1993; van der Geer et al., 1994; Gisselbrecht et al., 1996). Since htl and hbr mutants have similar mesodermal phenotypes and our genetic interaction studies suggested a functional relationship between the products of these genes, we were interested in whether hbr could also be related to RAS1 function. We have shown previously that targeted mesodermal expression of a constitutively activated form of RAS1 can partially rescue the htl mutant phenotype (Gisselbrecht et al., 1996; Michelson et al., 1998). This conclusion was reached by examining both the activated RAS1-induced migration of TWI-expressing cells and the recovery of dorsally restricted EVE-positive muscle and cardiac progenitors in htl embryos. Using these same assays, we found that activated RAS1 is capable of partially rescuing the strong hbr^YY202 mutant (Fig. 3G,H). Quantitation of the number of hemisegments in which EVE expression was recovered indicated that hbr^YY202 is even more efficiently rescued by activated RAS1 than is htl^AB42 (Fig. 3O), likely because hbr^YY202 is a hypomorph while htl^AB42 is a null allele. We also used this quantitative assay to compare the relative strengths of the three hbr alleles and found that their mesodermal phenotypes can be arranged in the following order of increasing severity – hbr^ems6<hbr^ems7<hbr^YY202 (Fig. 3O).

The above results suggest that hbr acts upstream of RAS1 or on a parallel pathway involved in either initiating or transducing the HTL signal. We next asked where hbr functions in relation to the receptor by determining if a constitutively activated form of HTL (see above) can rescue the hbr phenotype. When expressed in the mesoderm of wild-type embryos, activated HTL induced the formation of additional EVE founder cells, as previously noted, but had no effect on mesoderm migration (Fig. 3I,J). In a htl mutant background, activated HTL partially corrected the mesoderm migration defect and contributed to the specification of significant numbers of EVE progenitors (Fig. 3K,L). Quantitation of the latter effect revealed that activated HTL was significantly more efficient at rescuing loss of htl function than was activated RAS1 (Fig. 3O). In contrast, the influence of activated HTL was completely blocked by a homozygous hbr mutation (Fig. 3M-O). These results, as well as the dominant suppression of activated HTL by hbr (Fig. 2E), argue that hbr acts downstream of or parallel to this mesodermal FGF receptor.

MAPK is another important component of the RTK signalling cascade (Seger and Krebs, 1995). A monoclonal antibody specific for the dual phosphorylated, activated form of MAPK (diphospho-MAPK) has recently been shown to be highly effective for monitoring the activity of RTK pathways during Drosophila development (Gabay et al., 1997a,b). Using this reagent, high levels of activated MAPK were localized to the leading edge of the migrating mesoderm, with much lower levels present at more ventral positions (Gabay et al., 1997b; Fig. 4A). Activation of MAPK is very weakly enhanced in the ventral mesoderm by twi-GAL4-induced expression of a constitutive form of HTL, although the normal gradient of diphospho-MAPK expression does not appear to be significantly altered by this manipulation (Fig. 4B).

Activated MAPK is completely absent from the early mesoderm of htl mutants, confirming that this mesodermal expression of diphospho-MAPK is entirely HTL-dependent (Gabay et al., 1997b; Fig. 4C). Moreover, no activated MAPK is detectable at comparable stages in the mesoderm of hbr mutant embryos (Fig. 4D). Activated HTL expressed in a null htl mutant generates a low, uniform level of diphospho-MAPK throughout the mesoderm (Fig. 4E). In addition, reduction of hbr function is capable of completely blocking MAPK activation by constitutive HTL (Fig. 4F). These results suggest that hbr acts upstream of MAPK in the HTL signal transduction pathway, a hypothesis that is consistent with the findings of the above genetic epistasis experiments.

The inability of hbr mutant cells to migrate into the DPP expression domain in the dorsal region of the embryo can account for the associated absence of dorsal mesodermal structures. However, hbr might also be required in order for mesodermal cells to respond to this growth factor. We investigated this possibility by assessing the effects of ectopic DPP on the expression of the DPP target gene, bagpipe (bap; Azpiazu and Frasch, 1993; Staehling-Hampton et al., 1994; Frasch, 1995).

bap normally is expressed in a set of dorsally restricted, segmentally repeated patches of cells that give rise to the visceral mesoderm (Azpiazu and Frasch, 1993; Fig. 5A,C). This expression is markedly reduced in a hbr mutant (Fig. 5B,D), consistent with the previously documented defect in visceral mesoderm development (Fig. 1O). In an otherwise wild-type genetic background, ectopic DPP induces bap transcription in ventral and lateral mesodermal cells (Staehling-Hampton et al., 1994; Frasch, 1995; Fig. 5E). The same response to ectopic DPP is seen in hbr mutant embryos (Fig. 5F). Thus, competence to be induced by DPP does not require hbr. This mesodermal response to DPP also was found to be independent of htl (Gisselbrecht et al., 1996).

A second Drosophila FGF receptor is encoded by the btl gene (Glazer and Shilo, 1991; Klämbt et al., 1992). BTL activity is required for the migration of tracheal cells to form primary branches, and for the subsequent induction of secondary and terminal tracheal cell fates (Klämbt et al., 1992; Reichman-Fried et al., 1994; Reichman-Fried and Shilo, 1995; Lee et al., 1996b; Samakovlis et al., 1996; Sutherland et al., 1996). Mutations in btl are associated with a marked inhibition of tracheal branching (Fig. 6A-C). Given the involvement of hbr in the HTL FGF receptor signalling pathway, we examined whether HBR might also function with BTL in the tracheal system. Reduction of hbr function is indeed associated with significant defects in tracheal development. In hbr^YY202 mutant embryos, numerous primary and secondary tracheal branches are missing and the extension of those that do form is frequently stalled (Figs 6D, 7B). These results imply that hbr is necessary both for tracheal cell migration and for the acquisition of secondary tracheal cell fates. hbr^ems7 exhibits a very similar tracheal phenotype to hbr^YY202 (Fig. 6E), whereas hbr^ems6 has a more severe reduction in tracheal branching (Fig. 6F). Consistent with our earlier findings for the mesoderm, both hbr^YY202 and hbr^ems6 are hypomorphic with respect to their effects on tracheal development since more severe phenotypes occur when either allele is in trans to a deficiency (Fig. 6G,H). Homozygosity for the only available deficiency that covers the hbr locus results in severely dysmorphic embryos (Bilder and Scott, 1995; data not shown), making it impossible to use this stock to assess the tracheal phenotype associated with complete absence of hbr function. Interestingly, although hbr^ems6 has the strongest tracheal phenotype, its mesodermal defects are the least severe of the three hbr alleles (Fig. 3O; see Discussion).

As was the case with htl and hbr, btl and hbr exhibit strong genetic interactions. Thus, hbr is capable of dominantly enhancing a hypomorphic btl allele (Fig. 6I) and btl can dominantly enhance the hbr tracheal phenotype (Fig. 6J). Together, these genetic interaction experiments suggest that hbr participates in both the HTL and BTL signalling pathways.

hbr completely blocks the effects of activated HTL in the mesoderm (Fig. 3M,N). We also investigated the potential requirement for hbr in mediating the effects of BTL hyperactivation. Ectopic ectodermal expression of BNL, the BTL ligand, leads to widespread BTL activation, which causes a strong inhibition of primary tracheal branching, accompanied by the induction of disorganized networks of secondary and terminal tracheal branches (Sutherland et al., 1996; Fig. 7C). A homozygous hbr mutation strongly suppresses this effect of ectopic BNL – the formation of long primary branches is recovered and the additional fine, higher order branches are markedly reduced in number (Fig. 7D). Thus, as with activated HTL in the mesoderm, a hypomorphic hbr mutation is capable of at least partially suppressing the effect of BTL hyperactivation.

Expression of activated MAPK can be used to follow RTK involvement in tracheal development. Specific tracheal cell fates are established initially under the influence of DER, whose activity is reflected in the expression of activated MAPK in the tracheal placodes at stage 10 (Gabay et al., 1997b; Wappner et al., 1997; Fig. 8A). By stage 11, BTL-dependent expression of diphospho-MAPK occurs in the tracheal pits prior to the onset of tracheal branch migration (Gabay et al., 1997b; Fig. 8B). In either btl (Gabay et al., 1997b; data not shown) or bnl (Fig. 8C,D) mutant embryos, the DER-dependent expression of activated MAPK in the tracheal placodes is not affected, while the later expression of activated MAPK in the tracheal pits is largely but not completely eliminated (Fig. 8C,D). With reduced hbr function, the DER-dependent diphospho-MAPK pattern at stage 10 is normal, while BTL-dependent expression in stage 11 is very weakly but significantly reduced (Fig. 8E,F). The extent to which tracheal pit MAPK expression is affected in the bnl and hbr mutants appears to be commensurate with the relative severities of their tracheal migration defects.

Having established that hbr is not involved in DER-dependent MAPK activation in the tracheal placodes, we next determined if hbr is required for any other DER-mediated process in embryogenesis. DER has many critical embryonic functions (Schweitzer and Shilo, 1997), as reflected in the multiple sites of DER-dependent expression of diphospho-MAPK (Gabay et al., 1997a,b). However, none of these sites of diphospho-MAPK expression is affected in hbr mutant embryos, including the head folds, cephalic furrow, dorsal folds and ventral ectoderm at stage 8 (Fig. 9A-C), the ventral midline at stage 11 (Fig. 9D-F) and the segmental epidermal pattern at stages 12/13 (Fig. 9G-I). Moreover, DER-dependent patterning of the ventral ectoderm occurs normally in hbr mutant embryos, as determined from the wild-type cuticle pattern (data not shown) and normal expression of FASCICLIN III (FAS III) in the ventral epidermis of the three thoracic segments (Fig. 9J-L). Finally, we found that a constitutively activated form of DER is able to partially rescue mesodermal EVE expression equally well in both htl and hbr mutant embryos (Fig. 9M-O), an effect that is due to the capacity of DER to activate the RAS/MAPK cascade (which also functions downstream of HTL; Gisselbrecht et al., 1996), in a HBR-independent manner. This is in contrast to the ability of the same hbr mutation to completely block the mesodermal effects of constitutively activated HTL (Fig. 3M). That is, a mutation in hbr interferes with mesodermal HTL but not DER signaling. Thus, by multiple criteria, hbr functions in the HTL and BTL but not in the DER signaling pathways.

The diversity of biological responses elicited by RTKs with common signal transduction pathways presents a challenging problem for the generation of receptor specificity. In only a few cases have unique signalling mechanisms been identified that can contribute to the generation of specific outputs (Mason, 1994; White, 1994; Marshall, 1995; Songyang et al., 1995; Kouhara et al., 1997; Weiss et al., 1997; Clandinin et al., 1998). We have isolated and characterized a Drosophila mutant, hbr, whose embryonic phenotype includes features of both htl and btl mutations. Genetic interaction and epistasis experiments, together with analysis of activated MAPK expression patterns, suggest that hbr participates in the HTL and BTL signalling pathways. In contrast, hbr does not appear to be involved in embryonic EGFR signalling. Thus, hbr may represent a novel signalling component that is specific to FGFR responses in the Drosophila embryo.

hbr was initially identified as a lethal mutation in which the heart and dorsal somatic muscles fail to form due to a defect in mesoderm migration. Since this phenotype is identical to that associated with loss of HTL FGFR activity (Beiman et al., 1996; Gisselbrecht et al., 1996; Shishido et al., 1997; Michelson et al., 1998), we sought and found the following functional relationships between hbr and htl. (1) hbr is capable of dominantly enhancing a hypomorphic htl allele. (2) hbr can dominantly suppress a constitutively active form of HTL. (3) Whereas constitutive HTL can partially rescue loss of htl function, it is completely unable to bypass the mesodermal requirement for hbr. (4) Activated RAS1 can partially rescue the mesoderm migration defects of both htl and hbr mutants. (5) hbr is required for the HTL-dependent activation of MAPK during mesoderm migration. (6) Like htl, hbr is not required for the competence of mesodermal cells to respond to DPP, an inducer of dorsal mesodermal identity. Collectively, these results suggest that hbr acts in the HTL signalling pathway to facilitate the migration of the embryonic mesoderm. Moreover, since heterozygous hbr modifies the capacity of activated HTL to induce the formation of additional heart and muscle progenitors under conditions where mesoderm migration is not affected, hbr must also participate in the cell fate specification function of HTL (Michelson et al., 1998).

hbr also affects branching morphogenesis of the tracheal system, another FGFR-mediated process in Drosophila. The migration of tracheal cells in the formation of primary branches is markedly reduced in hbr mutant embryos. In addition, secondary branches are frequently missing, suggesting that hbr also is required for determination of the specialized cells that give rise to these structures. Since both primary and higher order tracheal branching are BTL-dependent processes (Klämbt et al., 1992; Reichman-Fried et al., 1994; Reichman-Fried and Shilo, 1995; Lee et al., 1996b; Samakovlis et al., 1996; Sutherland et al., 1996), hbr might be involved in signalling by this FGFR. Consistent with this possibility, hbr exhibits dominant genetic interactions with btl, is capable of suppressing the effects of BTL hyperactivation and is associated with a weak but significant reduction of BTL-dependent MAPK activation in tracheal precursor cells. Thus, by multiple criteria, hbr is implicated in the functions of two FGFRs in Drosophila embryogenesis.

Our findings are most readily reconciled by a model in which hbr participates in the signalling cascade acting downstream of both HTL and BTL. Two alternative possibilities are consistent with the available data. In the first case, hbr could function downstream of the receptors but upstream of RAS in a simple linear pathway (Fig. 10A). This would explain why hbr is epistatic to constitutive receptor activity, activated RAS is epistatic to hbr, and hbr is required for HTL- and BTL-dependent MAPK activation. However, these results are equally compatible with the possibility that hbr functions downstream of the receptors in a pathway that is initially parallel to but ultimately convergent with the RAS cascade at (or above) the level of MAPK (Fig. 10B). In either case, there could be additional branching pathways emanating from hbr that contribute to the responses elicited by these FGFRs.

FGFR signalling is involved in mesodermal patterning during vertebrate embryonic development (Mason, 1994; Yamaguchi and Rossant, 1995; Green et al., 1996), and in branching morphogenesis in formation of the mammalian lung and vasculature (Hanahan and Folkman, 1996). Given the involvement of hbr in similar HTL- and BTL-dependent processes in Drosophila, a hbr-related gene(s) likely exists in these other species. In fact, candidates for both models of hbr function already have been characterized for several RTKs in mammals. For example, mammalian FGF and insulin receptors each have substrates, FRS2/SNT and IRS1, respectively, that serve as dedicated adapters to couple receptor stimulation to

activation of RAS (White, 1994; Wang et al., 1996; Kouhara et al., 1997). Such an adapter could fulfill the function of hbr illustrated in Fig. 10A. Another mammalian protein that may be a specific FGFR substrate and function upstream of RAS, p90/80K-H, might also correspond to the hbr product (Goh et al., 1996).

On the other hand, a branching pathway leading to a novel mode of MAPK activation has been characterized downstream of the mammalian nerve growth factor receptor (York et al., 1998). This pathway, which involves the GTP-binding protein RAP1, its guanine nucleotide exchange factor C3G and the adapter protein CRK-L, acts in parallel to and independently of RAS and its effectors. However, the activities of both RAS and RAP1 converge on MAPK, an arrangement that is similar to the alternative scheme of hbr function that we have considered (Fig. 10B). hbr could also be involved in a RAS1-independent mechanism of RAF activation (Hou et al., 1995). Other novel functions of hbr are, of course, not excluded. The cloning of hbr and functional analysis of its protein product will be required to resolve this issue.

All three of our hbr mutants have stronger mesodermal than tracheal phenotypes when compared to null alleles of htl and btl. In addition, a hbr mutant completely blocks the effects of hyperactivated HTL in the mesoderm, whereas the same mutant only partially suppresses hyperactivated BTL in the trachea. These phenotypic differences may reflect allele-specific effects of hbr on each of the FGFR pathways, or may indicate a true differential involvement of hbr in signalling by these two RTKs. In this regard, it may be relevant that the hbr^ems6 allele has a more severe tracheal but less severe mesodermal phenotype than the other two hbr alleles, raising the possibility that HBR has independently mutable domains that function differentially in HTL and BTL signaling. Additional genetic and molecular analyses will be required to assess these various hypotheses.

BTL provides an instructive signal for the migration of primary tracheal branches through the localized expression of its activating ligand (Sutherland et al., 1996). Does HTL have a similar function in mesoderm migration? Although the HTL ligand has not yet been identified, the HTL-dependent, graded activation of MAPK in the migrating mesoderm suggests that HTL is indeed instructive for this process (Gabay et al., 1997b). Consistent with this hypothesis, we found that a constitutively active form of HTL, which generates uniform, albeit low, MAPK activation throughout the mesoderm, induces only partial rescue of a null htl mutant. This is in marked contrast to the complete rescue obtained by wild-type HTL expressed in an identical temporal and spatial manner (Michelson et al., 1998). Thus, graded activation of HTL is likely to be essential for the proper directional migration of mesodermal cells. In contrast, constitutive receptor activity may facilitate random cell movement, increasing the probability that some cells reach the dorsal ectoderm, but not reconstituting a normal pattern of cell migration. This could explain why activated HTL did not perturb migration in an otherwise wild-type background since, under these conditions, we still observed graded diphospho-MAPK expression. To further test the importance of graded signalling in mesoderm migration, it will be necessary to achieve uniform MAPK activation at a level comparable to that normally found in cells at the leading edge.

Although hbr is important for signalling by two FGFRs, it is not required for EGFR responses in the Drosophila embryo. This conclusion is based on the normal ventral ectodermal patterning and the completely wild-type DER-dependent expression of diphospho-MAPK seen in hbr mutant embryos. In addition, a hbr mutant fails to block the mesodermal effects of constitutive DER, which is able to partially substitute for HTL due to its ability to activate the common RAS/MAPK cascade in a HBR-independent manner. This is in contrast to the markedly reduced effects of constitutive HTL and BTL in the same hbr mutant background. Thus, hbr may be a key component that confers specificity to FGFR signalling. This is of particular interest since EGF and FGF receptors both function in the mesoderm, but with different biological outcomes (Beiman et al., 1996; Gisselbrecht et al., 1996; Shishido et al., 1997; Buff et al., 1998; Michelson et al., 1998). Both types of receptor also function in tracheal cells, but again each elicits a unique response (Klämbt et al., 1992; Reichman-Fried and Shilo, 1995; Lee et al., 1996b; Wappner et al., 1997). The availability of a dedicated RTK signal transducer like HBR, in addition to shared components such as RAS1, could insure that the appropriate output is generated by each receptor.

We would like to thank the following for generously providing fly stocks, antibodies and DNA clones – B. Shilo, M. Frasch, D. Montell, D. Kiehart, B. Patterson, L. Perkins, N. Perrimon, S. Roth, M. Akam, M. Krasnow, T. Schüpbach, T. Volk and the Bloomington Drosophila Stock Center. We are grateful to L. Perkins, M. Halfon and N. Perrimon for valuable discussions and comments on the manuscript, and to R. Gupta and L. Sauter for technical assistance. J. B. S. thanks C. Doe for his support during the early stages of this work. J. B. S. also was supported by a grant from the NINDS (R01-NS36570), the Cancer Research Fund of the Damon Runyon-Walter Winchell Foundation (DRS-9) and by a Howard Hughes Medical Institute Research Resources Program for Medical Schools Junior Faculty Award (#76296-538202). A. M. M. is an Assistant Investigator of the Howard Hughes Medical Institute.