Abstract

Fibroblast growth factors (FGFs) bind to FGF receptors, transmembrane tyrosine kinases that activate mitogenic, motogenic, and differentiative responses in different tissues. While there has been substantial progress in elucidating the Ras-MAP kinase pathway that mediates the differentiative responses, the signal transduction pathways that lead to directed cell migrations are not well defined. Here we describe a Drosophila gene called stumps that is required for FGF-dependent migrations of tracheal and mesodermal cells. These migrations are controlled by different FGF ligands and receptors, and they occur by different cellular mechanisms: the tracheal migrations occur as part of an epithelium whereas the mesodermal migrations are fibroblast-like. In the stumps mutant, tracheal cells fail to move out from the epithelial sacs, and only rudimentary tracheal branches form. Mesodermal cells fail in their dorsal migrations after gastrulation. The stumps mutation does not block all FGF signaling effects in these tissues: both random cell migrations and Ras-MAP kinase-mediated induction of FGF-specific effector genes occurred upon ectopic expression of the ligand or upon expression of a constitutively activated Ras protein in the migrating cells. The results suggest that stumps function promotes FGF-directed cell migrations, either by potentiating the FGF signaling process or by coupling the signal to the cellular machinery required for directed cell movement.

DIRECTED cell migration is required during the development of many tissues and is essential for proper immune system function and wound healing. Inappropriate cell migrations contribute to spreading of cancer cells throughout the body. Cell migrations occur by different cellular mechanisms, ranging from the crawling of isolated fibroblasts in culture to the coordinated movements of entire epithelial sheets during organ morphogenesis (Bray 1992), and they are controlled by different types of signaling molecules. Fibroblast growth factors (FGFs) are a large family of secreted signaling molecules that have been implicated in the control of cell migration in a variety of contexts in vitro and in vivo (Terranovaet al. 1985; Klambtet al. 1992; DeVoreet al. 1995; Beimanet al. 1996; Gisselbrechtet al. 1996; Sutherlandet al. 1996). However, like many other signaling molecules and their receptors, FGF pathways regulate other cellular processes besides cell migration, notably mitogenic and differentiative responses (Johnson and Williams 1993; Thomas 1993; Martin 1998). While there has been substantial progress in elucidating the Ras-MAP kinase signal transduction pathway that mediates the differentiative responses of FGF receptors and other receptor tyrosine kinases, the intracellular signaling pathways that mediate directed cell migration are still poorly understood. In this article, we describe a gene required for two distinct FGF-dependent cell migrations in Drosophila melanogaster.

One FGF gene and two FGF receptor genes are known in D. melanogaster. The Branchless FGF is a ligand for the Breathless FGF receptor (Klambtet al. 1992; Sutherlandet al. 1996). A ligand for the other FGF receptor, Heartless, has not yet been identified (Beimanet al. 1996; Gisselbrechtet al. 1996). These FGF pathways are active in different tissues and play different roles in development. The Heartless FGF receptor is expressed in the developing mesoderm and plays an important role in mesodermal development. It is required for the dorsal-directed mesodermal cell migrations that are necessary to form dorsal mesodermal derivatives including the heart and dorsal somatic muscles. The Branchless-Breathless pathway plays a critical role in tracheal (respiratory) system development, as described below. Both pathways also have additional roles in the development of specific glial cells (Klambtet al. 1992; Shishidoet al. 1997).

The Drosophila tracheal system is a ramified network of epithelial tubes that delivers oxygen throughout the body (Manning and Krasnow 1993). It arises at mid-embryogenesis from 20 clusters of ectodermal cells, 10 on each side of the embryo. Each cluster of ∼80 cells invaginates and forms an epithelial sac that gives rise to one tracheal hemi-segment by a series of cell migrations and tubulogenesis events (Samakovliset al. 1996). Initially, six primary branches bud from specific positions in the sac. These branches are formed by groups of 3–20 cells that migrate out from the sac in specific directions, organizing into tubular structures as they move. Individual cells near the ends of growing primary branches then elongate and form unicellular secondary branches. Most secondary branch cells go on to extend long cytoplasmic processes that form vast arrays of fine terminal branches that transport oxygen directly to the larval tissues. Different sets of genes are expressed in or near the cells that form each type of branch, and studies of these genes have begun to elucidate the pathways that control the outgrowth and branching pattern.

The Branchless FGF controls the early tracheal branching events (Sutherlandet al. 1996). bnl is required for branching and is expressed dynamically in discrete clusters of cells surrounding the developing tracheal sacs at each position where a new branch will bud. The secreted growth factor activates the Breathless FGF receptor, which is expressed on all tracheal cells and guides the tracheal cell migrations during primary branch outgrowth (Glazer and Shilo 1991; Klambtet al. 1992; Leeet al. 1996b; Sutherlandet al. 1996). The Branchless-Breathless pathway also has another important role in branch patterning: it induces the later programs of branching in cells near the ends of the growing primary branches and thereby specifies the positions of secondary branch sprouting (Sutherlandet al. 1996; Hacohenet al. 1998). This function appears to be transduced by a Ras-MAP kinase cascade that culminates in the activation of MAP kinase and the induction of downstream effector genes including pointed, an ETS domain transcription factor (Klambt 1993), sprouty, an antagonist of FGF signaling (Hacohenet al. 1998), and blistered (pruned), which encodes the Drosophila homologue of mammalian Serum Response Factor (DSRF; Affolteret al. 1994; Guilleminet al. 1996). The signal transduction pathways that mediate the early tracheal cell migrations have not been well studied, but Ras and Raf have been implicated (Reichman-Friedet al. 1994).

In a screen for mutations that affect tracheal development, a mutation called stumps was discovered that caused a striking arrest in tracheal cell migration, resulting in a tracheal phenotype resembling that of null mutations in breathless (btl) or branchless (bnl; Sutherlandet al. 1996). Here we present the genetic characterization of the gene. The gene was named stumps because no tracheal branches form in the mutant, except for an occasional short stump. stumps is also required for the Heartless-dependent migrations of mesodermal cells. The stumps mutation, however, does not prevent all FGF signaling effects, because both random cell migrations and Ras-MAP kinase-mediated induction of FGF-specific effector genes can occur upon ectopic expression of the FGF ligand. The results suggest that stumps function promotes FGF-directed cell migrations, either by potentiating the FGF signaling process or by coupling the signal to the cellular machinery required for directed cell movement.

MATERIALS AND METHODS

Fly strains and genetics: The stumps1 mutation was identified by Dr. Christos Samakovlis in enhancer trap line l(3)9904 generated by A. Spradling and colleagues (Spradlinget al. 1995). The P[ry+, lacZ] insertion at 82F8-9 and an independent lethal mutation present on the original stumps chromosome were recombined away from the stumps1 mutation at 88D using the multiply marked rucuca chromosome (Lindsley and Zimm 1992) to generate the stumps1 mutant chromosome used here, th st stumps.

btlLG18 and btlLG19 (Klambtet al. 1992), bnlP1 (Sutherlandet al. 1996), and htlAB42 (Gisselbrechtet al. 1996) are null or strong loss-of-function alleles. The lacZ enhancer trap markers used to monitor trachealess (1-eve-1), breathless (T1), pointed [l(3)7825], and sprouty [l(3)9143] expression have been described (Samakovliset al. 1996). The following chromosomal deficiencies were used: wrlnc4Rγ2 (referred to as Df(3R)γ2; M. Fuller, personal communication), Df(3R)ry506-85C [derived from Tp(3;Y)ry506-85C (Flybase; http://flybase.bio.indiana.edu)], Df(3R)red1, Df(3R)ry85, and Df(3R)red3L (Lindsley and Zimm 1992).

Genetic mapping and cytology: Meiotic recombination mapping crosses were carried out between the original stumps chromosome and the multiply marked rucuca chromosome, and between stumps1 and a cv-c P[w+, lacZ]0429 chromosome constructed by Dr. Julie Brill. Recombinants were scored for visible adult markers on the rucuca or cv-c P[w+, lacZ]0429 chromosomes and for the stumps tracheal phenotype. Recombination frequencies were analyzed in an F1 cross of w/w; stumps/cv-c P[w+, lacZ]0429 females to w; cv-c e males; the genotypes of 2617 F2 male progeny were determined.

For fine scale recombination mapping, several restriction fragment length polymorphisms (RFLPs) were identified in the 88C,D region using genomic DNA probes derived from P1 phage DS02646 (Berkeley Drosophila Genome Project; http://www.fruitfly.org) or from λ EMBL3 genomic phage (from J. Tamkun) that hybridized to P1 phage DS02646 (RFLPs 20-RT and 20-LT) and from genomic DNA flanking the P[w+, lacZ]3263 insertion (gift from J. Brill).

Preparation and orcein staining of salivary gland polytene chromosomes were carried out for stumps and Df(3R)γ2 as described (Ashburner 1989). The endpoints of the other chromosomal deficiencies shown in Figure 2C were taken from the references indicated above.

Immunohistochemistry and in situ hybridization: Embryo fixation and staining, antibodies against tracheal antigens, β-galactosidase, Engrailed, mAb 22C10, mAbBP102, and secondary antibodies have been described (Goldstein and Fyrberg 1994; Samakovliset al. 1996). Other primary antibodies used were as follows: anti-Trachealess (Wardet al. 1998) at 1:1000 dilution, anti-diphospho-ERK (Gabayet al. 1997a) at 1:500, mAb2-161 (from M. Gilman) against DSRF at 1:300, mAbFMM5 (from D. Kiehart) against muscle myosin at 1:8, anti-dMEF2 (Bouret al. 1995) at 1:2000, anti-Even-skipped (Fraschet al. 1987) at 1:3000, and anti-Tinman (Frasch 1995) at 1:400. Whole-mount in situ hybridizations using a bnl probe were carried out as described (Sutherlandet al. 1996).

Ectopic expression of bnl and ras: The Gal4/UAS expression system was used (Brand and Perrimon 1993). The following Gal4 drivers were used: btlGal4 line on chromosome II (Shigaet al. 1996), which drives expression in tracheal cells from stage 11 onward; twiGal4 (Greig and Akam 1995), which expresses in the mesoderm from stage 6 on; and hsGal4 (Brand and Perrimon 1993). The UAS lines used were UASDras1V12 on chromosome II (Leeet al. 1996a) and UASbnlB4-2 on the X chromosome (Sutherlandet al. 1996). Embryos from the following crosses were examined: (i) twiGal4 (or hsGal4)/+; stumps/+ × UASRas1/+; stumps/+, (ii) btlGal4/+; stumps/+ × UASRas1/+; stumps/+, and (iii) twiGal4 (or btlGal4)/+; stumps/+× UASbnlB4-2/UASbnlB4-2; stumps/+. The tracheal and/or mesodermal phenotypes of large numbers of embryos were scored; they fell into distinct phenotypic classes that were populated in the expected Mendelian ratios. twiGal4 and btlGal4 embryos used in the misexpression experiments were raised at 29° to maximize activity of the Gal4 driver. hsGal4 embryos were pulsed at 37° for 20 min at 4 and 5 hr after egg lay (AEL).

Figure 1.

—Tracheal phenotype of stumps. (A) Stage 12 wild-type embryo stained with TL-1 antiserum to label the tracheal lumen. Primary branches have sprouted and are migrating out from the tracheal sacs. (B) Stage 12 stumps1 homozygote. There is little sprouting or outgrowth of primary branches. (C) Stage 15/16 wild-type embryo stained with mAb2A12 to label the tracheal lumen. (D) Stage 15/16 stumps1 mutant embryo. Note the almost complete absence of branches except for an occasional stump. (E) Stage 15/16 stumps1/Df(3R)γ2 hemizygote. Note the almost complete absence of branches as in the stumps1 homozygote in D. Df(3R)γ2 homozygotes have a similar phenotype although they also show other embryonic defects that presumably result from deletion of other genes besides stumps (not shown). (F) Stage 15/16 bnlP1 homozygote. The stumps tracheal phenotype (D) is indistinguishable from that of the bnl mutant. (G) Stage 16 wild-type embryo double stained with mAb22C10 to show the peripheral nervous system (purple) and mAb2A12 to show trachea (brown). (H) Stage 16 stumps1 homozygote stained as in G. The PNS appears unaffected. Anterior is left and dorsal is up in all figures unless noted otherwise. Bar, 40 μm.

RESULTS

Genetic characterization and mapping of stumps: In a screen of P[lacZ] enhancer trap lines, Christos Samakovlis discovered a line [l(3)9904] carrying a homozygous lethal mutation that almost completely blocked tracheal branching (Figure 1D), much like null (amorphic) mutations in the Breathless FGF receptor and Branchless FGF genes (Figure 1F). The mutation complemented btlLG18 and bnlP1 mutations for viability and mapped to a different chromosomal region, indicating that it defined a separate gene (see below). The gene was named stumps (Sutherlandet al. 1996) because even at late stages in embryonic development, no tracheal branches were seen except for an occasional, rudimentary stump of a branch (Figure 1D). The stumps1 mutation is recessive, as heterozygous embryos did not display a tracheal phenotype (see below). The phenotype of hemizygous embryos [stumps1/Df(3R)γ2 or stumps/Df(3R)ry506-85C; Figure 1E)] was indistinguishable from that of the homozygote (Figure 1D), implying that stumps1 is a strong loss-of-function or null (amorphic) allele. Several other tissues examined were not grossly affected in the mutant. The overall structure of the embryo was normal, the segment polarity gene engrailed was expressed in its usual segmental stripes, and the peripheral nervous system (PNS) visualized with mAb 22C10 was unaffected (Figure 1H and data not shown). There were, however, defects in the central nervous system visualized by staining with mAbBP102 including increased separation, irregularities, and occasional breaks of the longitudinal connectives and fusions and irregularities in the transverse connectives (not shown). There were also gross defects in mesodermal development that are described in detail below.

The original stumps1 chromosome carried a P[ry+, lacZ] element at cytological position 82F8-9. However, chromosomal deficiencies that remove this region complemented the lethality and tracheal phenotype of stumps1, indicating that the stumps mutation was unlinked to the P[ry+, lacZ] insertion. The stumps1 mutation was localized to cytological interval 88D3-4 to 88D8 on the right arm of chromosome III by complementation tests with a set of chromosomal deficiencies that remove different regions of the chromosome (Figure 2, B and C). No cytological abnormalities were detected in this region of the stumps1 chromosome (Figure 2A), indicating that the mutation might be a small deletion or point mutation. Meiotic recombination mapping was used to localize stumps with respect to other genes in the region and to several RFLP markers that we identified. This analysis placed stumps 0.21 cM distal to crossveinless-c (cv-c) and 0.18 cM proximal to P[w+, lacZ]3263 (Figure 2D). For the rest of the experiments described here, we used a chromosome in which the P[ry+, lacZ] insertion and an unrelated lethal mutation had been recombined away from the stumps1 mutation (see materials and methods).

Figure 2.

—Genetic mapping of stumps. (A) Cytology of a stumps1/+ polytene salivary gland chromosome showing the relevant region of chromosome 3. No abnormalities are visible. (B) Cytology of a Df(3R)γ2/+ polytene salivary gland chromosome. Region 88A3-12 to 88D3-8 is missing in the Df(3R)γ2 chromosome. (C) Map of the chromosomal deficiencies in the 87-88 region used to localize the stumps locus. (+) and (-) indicate complementation behavior of the deficiencies with stumps1. Df(3R)ry506 (87D1-2; 88E5-6) and Df(3R)γ2 (88A3-12; 88D3-8) did not complement the stumps1 tracheal phenotype or lethality, whereas Df(3R)red3l (88B1; 88D3-4), Df(3R)ry85 (87B15-C1; 87F10-11), and Df(3R)red3l (87F12-14; 88C1-3) all complemented stumps1. This localizes stumps to the interval 88D3 to 88D8 (asterisk). (D) Recombination map showing the distances between cv-c, stumps, and P[w+, lacZ]0429. The 1.28-cM interval between stumps and P[w+, lacZ]0429 was further defined using a RFLP marker from P[w+, lacZ]3263. Six of the seven crossovers analyzed occurred distal to the RFLP marker and one occurred proximally to the marker. The other RFLP markers tested (20RT and 20LT) mapped proximal to all of the recombination breakpoints and are not indicated in the figure.

stumps is required for tracheal cell migration: Cellular and molecular markers were used to characterize the tracheal defects in the stumps1 mutant. The initial steps in tracheal development were normal in the mutant. The tracheal precursor cells invaginated to form the epithelial sacs at embryonic stage 11 (∼6 hr AEL) and expressed the early tracheal markers trachealess (trh) and breathless normally (compare Figure 3, A and B, and data not shown). The initial activation of MAP kinase in the tracheal sacs by the EGF receptor pathway at stage 10 (Gabay et al. 1997a,b) was also unaffected in the mutant, as assayed by staining with antisera specific for the diphosphorylated form of MAP kinase (data not shown). bnl was also expressed normally in clusters of cells surrounding the sac, as shown by whole-mount embryo in situ hybridizations (data not shown). However, during stages 11–14 (6–11 hr AEL), when tracheal cells in the sac normally migrate toward these clusters and form primary branches (Figures 1A and 3, A and C), tracheal cells in the stumps1 mutant moved very little (Figures 1B and 3, B and D).

During stages 13 and 14 (10–11 hr AEL), when secondary and terminal branch markers such as pointed (pnt), sprouty (spry), and DSRF have normally been induced by the Branchless FGF pathway in groups of cells at the ends of growing primary branches, only rare (pnt) or no (spry, DSRF) expressing cells were observed in stumps1 mutant embryos (Figure 3H and data not shown). Similarly, only sporadic tracheal cells showed staining with the antiserum specific for activated MAP kinase during stages 11–12, indicating that the Branchless signaling pathway was inactive or very weakly active in the trachea at this stage (Figure 3F). During stages 15–16 (12–15 hr AEL), when primary and secondary branching are complete and the first terminal branches are forming in wild-type embryos (Figure 1C), the tracheal sacs in stumps1 mutant embryos were still mostly unbranched with only an occasional rudimentary branch (Figure 1D).

stumps interacts genetically with branchless: The tracheal phenotypes described above for the stumps mutant are very similar to those of null mutations in bnl and btl (Klambtet al. 1992; Sutherlandet al. 1996), suggesting that stumps might function in this FGF pathway. Because genes in the same signaling pathway commonly display dosage-sensitive interactions, we tested for an interaction between stumps and bnl. Null mutations in bnl are haploinsufficient, as heterozygous and hemizygous embryos show sporadic defects in primary branch outgrowth. This is manifest most significantly by the ganglionic tracheal branches, 9% of which fail to reach the CNS (Sutherlandet al. 1996; Figure 4C). The stumps1 mutation is not haploinsufficient, as heterozygous embryos did not show an increase in ganglionic branch or other primary branch outgrowth defects above the low background seen in wild-type embryos (<3%; Figure 4, A and B). The stumps +/+ bnl double heterozygote, however, showed a dramatic increase in outgrowth defects, with nearly four times as many stalled ganglionic branches (33%; n = 460) as the bnl heterozygote alone (Figure 4D). In the most severe cases, all 20 ganglionic branches failed to migrate normally in the double heterozygote. The double heterozygotes also displayed increased dorsal branch outgrowth defects compared to bnl heterozygotes as well as rare dorsal trunk outgrowth defects which were never seen in bnl heterozygotes. Thus, stumps displays a dosage-sensitive genetic interaction with bnl. The interaction is very similar to the one previously described between bnl and btl (Sutherlandet al. 1996), supporting the hypothesis that stumps functions in this FGF pathway.

Figure 3.

—Tracheal cellular and gene expression defects in the stumps mutant. (A) Stage 12 wild-type embryo stained for the Trachealess transcription factor expressed in the salivary gland (SG) and all tracheal cells (Tr1 and Tr4 tracheal hemi-segments are indicated). Small groups of tracheal cells have begun migrating away from the tracheal sacs to form primary branch buds. (B) Stage 12 stumps1 homozygote. Trachealess is expressed normally but there is little movement of tracheal cells out from the sacs. (C) Stage 14 wild-type embryo. The tracheal cells have migrated out and formed mature primary branches. (D) Stage 14 stumps1 homozygote. There is still very little tracheal cell migration or branch formation. Similar results were obtained with a trachealess lacZ enhancer trap marker that labels the cytoplasm of the tracheal cells; little cell migration and few cytoplasmic projections from the cells were seen in the mutant (not shown). (E) Close-up of a hemi-segment from a Stage 12/13 wild-type embryo double stained with an antiserum against the diphosphorylated form of ERK to show MAP kinase activation (purple) and TL-1 antiserum to show trachea (brown). Activated MAP kinase is seen in tracheal cells at the ends of the DB and DTp (arrowheads) and VB (bracket) primary branches. The VB is just deep to the plane of focus. (F) stumps1 homozygote at a similar stage stained as in E. No activated MAPK is seen in the trachea, although it is expressed normally in other tissues (arrow). In 23% of stumps mutant embryos (n = 17) at stages 12 or 13, weak expression was detected in an occasional tracheal cell. (G) Hemi-segment of a stage 16 wild-type embryo double stained with mAb2-161 against DSRF (purple) and mAb2A12 to the tracheal lumen (brown). Arrowheads, DSRF-expressing cells in the lateral tracheal trunk and transverse connective. (H) stumps1 homozygote at a similar stage stained as in G. No tracheal cells express DSRF. Bar, 20 μm.

Some downstream effects of Branchless FGF signaling can be activated in the stumps mutant: Like bnl and btl mutations, the stumps mutation blocks tracheal cell migration and causes failure of MAP kinase activation, loss of induction of secondary and terminal branch markers, and failure to form secondary and terminal branches. This could mean that the Branchless signaling pathway is completely inactivated by the stumps mutation. Alternatively, the secreted signal might fail to diffuse properly and reach the tracheal cells, or the tracheal cells might receive the initial signal but fail to migrate toward the signaling source and thus never receive enough signal to trigger the later gene inductive responses. To test whether any downstream signaling effects of the Branchless pathway were intact in the stumps mutant, we ectopically expressed the Branchless FGF near the tracheal sacs and monitored the tracheal response.

Figure 4.

—Dosage-sensitive genetic interaction between bnl and stumps. (A) Ventral view of a stage 16 wild-type embryo stained with mAb2A12. The 10 pairs of ganglionic branches have grown out onto the ventral nerve cord. (B) stumps1/+ heterozygote. The ganglionic branches have grown out normally. (C) bnlP1/+ heterozygote. Outgrowth of 4 of the 20 ganglionic branches has stalled (asterisks). (D) A stumps1/bnlP1 double heterozygote. Outgrowth of 14 ganglionic branches has stalled. btlLG18/+ heterozygote (Sutherlandet al. 1996) and stumps1 +/+ btlLG18 double heterozygotes did not show any tracheal defects (data not shown).

The Gal4/UAS system was used to express Branchless in either the mesoderm of stumps1 mutants beginning at stage 6 or the tracheal cells themselves beginning at stage 11. Both conditions produced similar results. All of the later FGF signaling functions assayed were strongly activated in the portions of the tracheal sacs near the ectopic Branchless-expressing cells. MAP kinase was phosphorylated in these cells (Figure 5B), DSRF expression was induced (Figure 5E), and there was significant cytoplasmic outgrowth and ramification of terminal branches (Figure 5G). Some erratic outgrowth of primary branches also occurred (Figure 5G). Similar effects of ectopic Branchless expression were observed in Df(3R)γ2 homozygotes (not shown). We also obtained very similar effects when a constitutively activated Ras protein (Dras1V12) was expressed in the developing tracheal system of stumps1 mutants using a btlGal4 or a hsGal4 driver. Under these conditions too, downstream FGF signaling functions were activated in many or all tracheal cells (Figure 5, C, F, and H), and there was erratic outgrowth of primary branches (Figure 5H). Thus, many downstream signaling events in the Branchless FGF signaling pathway can be activated in the stumps mutant upon ectopic expression of the ligand or by expression of an activated Ras. Interestingly, expression of the activated Ras in wild-type embryos did not disrupt the primary branching pattern, resulting only in ectopic induction of downstream genes and in extra terminal branching (Figure 5I and data not shown).

stumps is required for mesodermal cell migration mediated by the Heartless FGF receptor: In addition to the tracheal defects, the stumps1 mutant displayed a severe defect in mesodermal development. In wild-type embryos, mesodermal cells invaginate at the ventral midline at stage 6 (3 hr AEL) and migrate along the inner surface of the epidermis until reaching the dorsal margin of the epidermis two hr later at stage 10 (Michelsonet al. 1998). The migrating cells turn on different muscle markers and differentiate into visceral, somatic, and cardiac muscles and other mesodermal derivatives. In the stumps1 mutant, mesodermal cells appeared to invaginate normally at the ventral midline but failed in their dorsal migrations. The migration defect was evident in stage 9 embryos where mesodermal cells stained for Tinman remained clustered near the ventral midline (Figure 6B). In wild-type embryos of the same age many of the cells have already moved out of the region (Figure 6A). In older embryos, the migration defect was manifest by the absence of the heart and most other dorsal mesoderm derivatives (Figure 6, D, F, and I). The mesodermal phenotype decreased in severity from dorsal to ventral positions (compare Figure 6I to 6H), implying that the mesodermal cells that fail to migrate can still undergo many aspects of the normal ventral muscle differentiation program.

Although the tracheal phenotype of stumps mutant embryos is nearly identical to those shown by bnl and btl mutants, null mutants in these genes do not display any obvious mesodermal defects. The other known FGF receptor in Drosophila, however, encoded by heartless, is known to be expressed in and function in the developing mesoderm. Null htl mutations cause a mesodermal phenotype that is indistinguishable from that of the stumps1 mutant in every aspect in which they have been compared. This includes all of the morphological effects described above (Beimanet al. 1996; Gisselbrechtet al. 1996; see also Figure 6, G and J) and all of the effects on marker expression that have been tested (see below). This suggests that stumps functions in the Heartless FGF receptor pathway as well as the Branchless-Breathless pathway.

In the stumps1 mutant, as in null htl mutants, mesodermal cells failed to activate MAP kinase (data not shown), the dorsal mesodermal marker Tinman was only sporadically induced (Figure 6D), and the Even-skipped marker was never induced (Figure 7B). In htl mutants, the failure to induce the dorsal markers was shown to be a secondary consequence of the failure of most mesodermal cells to reach dorsal positions near the DPP inductive signal secreted by the dorsal-most ectodermal cells (Frasch 1995; Beimanet al. 1996; Gisselbrechtet al. 1996). The results described below imply a similar cause in the stumps mutant.

Figure 5.

—Effects of ectopic expression of Branchless and an activated Ras on the trachea in the stumps mutant. (A) Stage 12/13 stumps1/stumps1 embryo double stained with TL-1 for the tracheal lumen (brown) and the anti-dpERK antiserum to show MAP kinase activation (purple). No expression of activated MAP kinase is seen in the two tracheal hemi-segments shown although some can be seen in the tissues nearby. (B) UASbnlB4-2/+; twiGal4/+; stumps1/stumps1 mutant embryo stained as in A. twiGal4 drives UASbnl expression throughout the mesoderm, which lies just deep to the developing tracheal system. Ectopic Branchless expression leads to MAP kinase activation in virtually every tracheal cell. (C) Similarly staged UAS-Dras1V12/btlGal4; stumps1/stumps1 mutant embryo stained as in A. btlGal4 drives activated Ras expression throughout the developing tracheal system. Expression of the activated Ras leads to activation of MAP kinase in every tracheal cell. (D) Stage 16 stumps1/stumps1 embryo double stained for DSRF (purple) and the mAb2A12 tracheal antigen (brown). No tracheal expression of DSRF is seen. (E) Similarly staged UASbnlB4-2/+; twiGal4/+; stumps1/stumps1 embryo stained as in D. Ectopic bnl expression induces DSRF expression in most tracheal cells. (F) Similarly staged UAS-Dras1V12/btlGal4; stumps1/stumps1 embryo stained as in D. Expression of the activated Ras induces DSRF expression in most tracheal cells. (G) Stage 16 UASbnlB4-2/+; twiGal4/+; stumps1/stumps1 embryo stained for the mAb2A12 tracheal antigen. The ectopic bnl expression causes erratic outgrowth of primary branches (arrowheads), occasional dorsal trunk fusion (asterisk), and sprouting of fine terminal branches (bracket). (H) Stage 16 UAS-Dras1V12/btlGal4; stumps1/stumps1 mutant embryo stained for mAb2A12 tracheal antigen. Expression of the activated Ras causes a similar constellation of effects to those caused by ectopic Branchless in G. (I) Stage 16 UAS-Dras1V12/btlGal4 embryo. Expression of the activated Ras causes excessive sprouting of fine terminal branches throughout the tracheal system but does not disrupt the primary branch pattern. Bar, 20 μm.

We tested whether the mesodermal defects in stumps1 mutants could be ameliorated by expression of the constitutively activated Ras protein in the developing mesodermal cells using the same approach used to test for rescue of the tracheal defects. As in the tracheal system, expression of the activated Ras caused activation of MAP kinase and a general increase in motility, as well as induction of the dorsal muscle markers Even-skipped and Tinman (Figure 7D and data not shown). The effects of activated Ras in the stumps mutant were very similar to the reported effects of activated Ras in htl mutants (Gisselbrechtet al. 1996). We conclude that the stumps mutation causes severe defects in the Heartless-dependent migrations of mesodermal cells, and that some of these defects can be overcome by expression of an activated Ras.

Figure 6.

—Mesodermal phenotype of stumps1 mutant. A, C, E, and H are wild-type embryos; B, D, F, and I are stumps1 mutant embryos; G and J are htlAB42 mutant embryos. (A) Stage 9 wild-type embryo stained for Tinman, a mesodermal marker (ventral view). (B) Stage 9 stumps1 mutant stained as in A. Note the mesodermal cells are defective in their migrations out from the ventral midline. The same phenotype is seen for a null heartless mutation (Gisselbrechtet al. 1996). (C) Stage 16 wild-type embryo stained with Tinman (dorsal view). At this stage, Tinman expression is restricted to just the cardial cells of the heart at the dorsal midline. (D) Stage 16 stumps1 mutant stained as in C. Only very few Tinman-expressing cells are present at the dorsal midline. The same phenotype is seen for a null htl mutation (Beimanet al. 1996; Gisselbrechtet al. 1996). (E) Stage 16 wild-type embryo (dorsal view) stained with anti-DMEF2, a muscle cell transcription factor. The cardial cells are seen at the dorsal midline flanked by myocytes of the dorsal somatic muscles. (F) Stage 16 stumps1 mutant stained as in E. Most of the cardial and dorsal somatic muscle cells fail in their dorsal migrations, and only a small number of scattered DMEF2-expressing cells are seen. (G) Stage 16 htlAB42 mutant stained as in E. The null htl phenotype is the same as that of stumps1 (shown in F). (H) Stage 16 wild-type embryo (lateral view) stained for muscle myosin (mAbFMM5) to show the pattern of the muscles. (I) Stage 16 stumps1 mutant stained as in H. Note the almost complete absence of dorsal muscles but the presence of some of the ventral muscles. (J) Stage 16 htlAB42 mutant stained as in H. The null htl phenotype is virtually the same as that of stumps1 (see I). The stumps1 mutation also causes a visceral muscle phenotype that includes the absence of the second gut constriction (not shown) as does the null htl mutation (Gisselbrechtet al. 1996).

DISCUSSION

stumps is required for two distinct sets of embryonic cell migrations: We have characterized the stumps1 mutant and shown that the gene is required for tracheal and mesodermal cell migrations. In the stumps mutant, tracheal precursor cells execute the initial steps of their morphogenetic program normally and invaginate to form epithelial sacs. However, the migrations of tracheal cells out from the sacs to form primary branches fail, and only an occasional stump ever develops. Similarly, the mesodermal precursor cells undergo the initial steps of their morphogenetic program normally, invaginating at the ventral midline to form the mesodermal primordium. The subsequent dorsal migrations of mesodermal cells fail, and the heart and other dorsal mesodermal derivatives are almost completely absent. Although many late tracheal and mesodermal processes fail in the stumps mutant, these may all be secondary to the migration failure because many of the later differentiative processes can be restored by ectopic expression of signaling pathway genes, as discussed further below. stumps does not appear to be a general factor required for cell migration because many other cell migrations, such as the endodermal migrations required to establish the gut, the neural migrations that form the PNS, and the spreading of the epidermal sheet during dorsal closure are all unaffected in the mutant. The two types of migrations that are affected do not have any notable properties in common with respect to migration mechanism, as the tracheal cell movements occur as part of an epithelium whereas the mesodermal migrations apparently occur by fibroblast-like movements. The only obvious shared feature of the two types of migrations affected by stumps is that both depend on FGF signaling pathways.

stumps functions in FGF signaling pathways: Three lines of evidence support the hypothesis that stumps functions in FGF signaling pathways. First, the only two processes that are known to be affected in stumps mutant embryos are both dependent on FGF signaling pathways. The tracheal cell migrations require the Branchless FGF, which binds and activates the Breathless FGF receptor on tracheal cells and guides their migrations. Although the ligand for the mesodermal migrations is not known, these migrations require the Heartless FGF receptor, which is expressed specifically on the migrating cells. The tracheal and mesodermal migration defects observed in the stumps mutant are a virtual composite of those seen in null mutants of bnl or btl, and htl. No mutations in any other known genes cause a phenotype that so closely mimics either the bnl and btl phenotype or that of htl. Second, we observed dosage-dependent genetic interactions between bnl and stumps, as are often seen for genes in the same signaling pathway. The bnl FGF gene is haploinsufficient as bnlP1 heterozygotes show mild tracheal outgrowth defects (Sutherlandet al. 1996). Although stumps1 embryos did not show any defects when heterozygous, the bnlP1/stumps1 double heterozygote showed a substantially increased severity of outgrowth defects compared to the bnlP1 heterozygote. Indeed, the genetic interaction between bnl and stumps was similar in strength to the interaction observed between bnl and the btl FGF receptor (Sutherlandet al. 1996). Finally, we were able to obtain partial rescue of the stumps mutant phenotypes by expression of FGF pathway components. Some of the tracheal migration defects in the stumps mutant were ameliorated by ectopic expression of the Branchless FGF or by expression of a constitutively activated Ras in the tracheal cells. Similarly, expression of the activated Ras in mesodermal cells of the stumps mutant caused increased migration and rescue of dorsal mesodermal cell fates, just as the activated Ras partially rescues htl mutants (Gisselbrechtet al. 1996). Taken together, the results suggest that stumps functions as a critical component of both the Branchless-Breathless and Heartless FGF signaling pathways.

Figure 7.

—Effect of mesodermal expression of an activated Ras in the stumps1 mutant. (A) Stage 11 wild-type embryo stained for the Even-skipped homeodomain (EVE) transcription factor. EVE is expressed in segmentally repeated clusters of three cells. (B) stumps1 homozygote stained as in A. The EVE-staining mesodermal cells are absent, although normal EVE expression is detected in the CNS (inset shows CNS from same embryo). (C) UAS-Dras1V12/twiGal4 embryo stained as in A. Expression of activated Ras throughout the mesoderm increases the number of EVE-expressing cells in each mesodermal cluster (see also Gisselbrechtet al. 1996). (D) UAS-Dras1V12/twiGal4; stumps1/stumps1 mutant stained as in A. Expression of activated Ras throughout the mesoderm restores some random mesodermal migrations and expression of EVE. A similar restoration in the mesoderm is seen when activated Ras is expressed in htl mutants (Gisselbrechtet al. 1996).

Other cell migrations in Drosophila, such as glial cell migrations and border cell migrations in the ovary, also involve these two FGF signaling pathways (Klambtet al. 1992; Montell 1994; Murphyet al. 1995; Shishidoet al. 1997). Aside from the involvement of FGF signaling, these other migration processes appear to share little in common with the tracheal or mesodermal migrations that require stumps function. The glial cell migrations are a complex set of cellular rearrangements that take place on developing CNS neurons, and the border cell migrations occur as small groups of cells that weave their way through a matrix of nurse cells to reach the oocyte. It will be interesting to determine whether stumps is required for all FGF-dependent cell migrations. Indeed, the CNS defects we observed in the stumps mutant are consistent with a role for stumps in both the breathless- and heartless-dependent glial cell migrations.

Models of stumps function in FGF signaling and cell migration: An important consideration with regard to the role of stumps in FGF-mediated tracheal migrations is that the gene does not appear to be required for all aspects of the tracheal FGF signaling pathway. All downstream effects of the Branchless FGF pathway we examined in addition to cell migration failed in the stumps mutant, including the activation of MAP kinase, the induction of secondary and terminal branching genes, and the formation of secondary and terminal branches. However, all of the gene inductive and differentiative effects could be restored by ectopic expression of the Branchless FGF or expression of the activated Ras. These treatments also induced some tracheal cell migration, although the migrations were limited in extent and did not follow the normal outgrowth pathways. Thus, the stumps1 mutant has the ability to generate an active FGF signal, receive the signal, and transduce it through MAP kinase and on to downstream target genes. The residual signaling capability is unlikely to result from leakiness of the stumps1 allele because stumps1 behaved as a null allele in genetic tests, and the same residual signaling capability was found in Df(3R)γ2 homozygotes that presumably lack the stumps locus.

Figure 8.

—FGF pathway. (A) Potentiator models. stumps function is not absolutely required for any step in the pathway, but it serves to enhance the efficiency of one or more of the steps required for general motility, directional movement (guidance), and gene induction. (B) Bifurcating pathway models. The Branchless signaling pathway is a bifurcating pathway, and stumps function is required only for the branch required for directional movement. Gene induction and general motility can occur in the absence of stumps, but the cell movements would not be guided toward the signaling source and thus would secondarily diminish all downstream signaling events. The data do not distinguish between the two pathways shown which differ in the point of bifurcation relative to Ras.

Given that aspects of the FGF pathway are intact in the stumps mutant, including expression of the Branchless FGF and Breathless FGF receptor genes, we suggest two general models for how stumps may function in the tracheal FGF pathway. The first is that stumps acts as a potentiator in the pathway (Figure 8A). In this model, stumps is not absolutely required for any step in the process, but it is necessary to amplify or focus one or more steps. For example, it might facilitate transport or reception of the ligand, as heparan sulfate proteoglycans are believed to function in presenting the FGF ligand to FGF receptors in cultured mammalian cells (Spivaket al. 1994). Or, it might function like the LIN-2, LIN-7, and LIN-10 proteins in C. elegans to concentrate the receptor tyrosine kinase at the basal (signal receiving) side of the tracheal cells (Kaechet al. 1998). It could also function downstream as a scaffold that concentrates components of the signal transduction machinery, thereby amplifying the signal (Pawson and Scott 1997). If the potentiator model is correct, then our data do not allow us to more precisely place stumps on the FGF pathway, because overexpression of the ligand or introduction of the activated Ras could either force signaling through or bypass the signaling bottle-neck caused by the absence of stumps.

The second general class of models for stumps function requires a bifurcation in the FGF signaling pathway (Figure 8B). In these models, stumps is required only for one fork in the pathway—the fork needed for directed cell migration (guidance). The other fork would remain intact in the stumps mutant. Because MAP kinase activation and downstream effector genes were able to be induced in the stumps mutant, FGF signal production, reception, and this fork of the signal transduction pathway would be independent of stumps. General cell motility would also not require stumps, as tracheal cells in the stumps mutant could be induced to move, albeit erratically, upon ectopic expression of the Branchless FGF or expression of an activated Ras. Ectopic expression of activated Ras in the wild type did not perturb directed cell migration, demonstrating that the cells continued to receive the normal FGF guidance cues under these conditions. According to the bifurcating pathway models, all of the other FGF signaling failures observed in the stumps mutant would be secondary consequences of the tracheal cells’ inability to follow the FGF signal toward its source and to receive continuous high levels of signal. Thus, if the signal were artificially brought close to the tracheal cells, as in our ectopic expression experiments, the mutant cells could respond normally in many respects, except for proper directional migration, just as was observed.

In the bifurcating pathway models, stumps functions downstream of the receptor. It might serve, for example, to mark the highest point of signaling activity on the membrane of the receiving cell and couple the activated receptor to the cytoskeletal machinery required for directed cell movement. Our data, however, do not allow us to place the split in the pathway with respect to Ras (or MAP kinase), as indicated by the alternative models diagramed in Figure 8B.

Although our analysis of the mesodermal migration defects in the stumps mutant was more limited because the ligand for the mesodermal FGF pathway is not known, the available data also fit well with a bifurcating signal transduction pathway. Mesodermal cells in the stumps mutant failed in their directed migrations, although they were still able to undergo limited, apparently random migrations. Expression of activated Ras stimulated motility and restored some dorsal mesodermal development. If the bifurcation model is correct (Figure 8B), it would suggest that the Heartless and Breathless FGF receptors use related signal transduction machineries to mediate directed cell migrations. Indeed, the intracellular portion of the Heartless FGF receptor can partially rescue defective tracheal outgrowth in a Breathless FGF receptor mutant, implying common downstream effectors in migration (Reichman-Friedet al. 1994).

Many FGF receptors and other receptor tyrosine kinase pathways stimulate a variety of cellular effects including mitogenic, motogenic, and differentiative responses (Johnson and Williams 1993; Thomas 1993; Martin 1998). There is growing evidence that directed cell migrations are mediated by a special branch in RTK signal transduction pathways that is not shared by the mitogenic and differentiative responses (Leeet al. 1996a; Sundaramet al. 1996). Various intracellular signaling molecules including Ras, Rac, phospholipase C-γ, phosphoinositide-3′ kinase, and Src have been implicated in signal transduction processes associated with migration (Anand and Zetter 1997). Molecular characterization of the stumps gene and biochemical analysis of its products should help elucidate the FGF signal transduction pathways that control tracheal and mesodermal cell migrations and may provide insight into how these and other RTK pathways guide cell movements.

Acknowledgments

We are indebted to Dr. Christos Samakovlis who identified the original stumps mutation. We are also grateful to Dr. Julie Brill for valuable discussions and for strains and reagents in the 88D region and to Matt Fish for technical assistance. We thank Drs. S. Crews, D. Kiehart, E. Furlong, S. Roth, and A. Michelson for strains and antibodies, and the members of our lab for comments on the manuscript. This work was supported by a National Institutes of Health (NIH) grant (to M.A.K.) and NIH Medical Scientist Training Program fellowships to F.I. and D.S. M.A.K. is an investigator at the Howard Hughes Medical Institute.

Footnotes

  • Communicating editor: R. S. Hawley

  • Note added in proof: Complementation tests demonstrate that two recently described genes, Dof (Vincent, S., R. Wilson, C. Coelho, M. Affolter and M. Leptin, 1998, The Drosophila protein Dof is specifically required for FGF signaling. Mol. Cell 2: 515–525) and heartbroken (Michelson, A. M., S. Gisselbrecht, E. Buff and J. B. Skeath, 1998, Heartbroken is a specific downstream mediator of FGF receptor signalling in Drosophila. Development 125: 4379–4389), are allelic to stumps.

  • Received October 22, 1998.
  • Accepted February 8, 1999.

LITERATURE CITED

View Abstract