Requirements of DFR1/Heartless, a mesoderm-specific Drosophila FGF-receptor, for the formation of heart, visceral and somatic muscles, and ensheathing of longitudinal axon tracts in CNS

In Drosophila, the first apparent morphogenetic event, gastrulation, takes place 3 hours after egg laying (AEL) (Campos-Ortega and Hartenstein, 1985; Bate, 1993). During gastrulation, ventral cells of the blastoderm move into the interior of the embryo along the invaginating ventral furrow. The invaginated cells form the mesoderm from which are derived many internal organs and tissues including somatic and visceral muscles, fat body and heart. All these different cells first start differentiating at the time of germ-band retraction (7-8 hours AEL; Bate, 1993).

Events decisive for mesoderm fate are thought to occur between gastrulation and germ-band retraction (Bate, 1993). Immediately after gastrulation, mesodermal cells undergo proliferation and spread out to form a monolayer beneath the overlying ectoderm (Campos-Ortega and Hartenstein, 1985). Dorsoventral and anteroposterior subdivisions of each segment have been shown to be importantly involved in partitioning of mesoderm into progenitors of different derivatives (Frasch, 1995; Azpiazu et al., 1996). Three different signals are considered to be essential for mesodermal subdivision. Decapentaplegic (Dpp) signals from the dorsal-most region of the ectoderm determine which cells of the mesoderm become competent for developing into visceral mesoderm and heart (Frasch, 1995). The opposing roles of hedgehog (hh) and wingless (wg) appear essential for what becomes of mesoderm along the anterior-posterior axis (Azpiazu et al, 1996). In addition, a recent experiment showed that Twist, a helix-loophelix transcription factor, regulates mesodermal differentiation and propels a specific subset of mesodermal cells into somatic myogenesis (Baylies and Bate, 1996).

Studies on Xenopus embryos indicate the importance of certain growth factors to mesoderm formation (reviewed in Jessell and Melton, 1992; Gurdon, 1992). Fibroblast growth factor (FGF) and transforming growth factor (TGF-β) have been shown to be associated with inducing activity of mesoderm. Exposure to FGF induces formation of mesodermal derivatives, such as muscle and mesenchyme, in the animal hemisphere isolated from Xenopus embryos (Kimmelman and Kirschner, 1987; Slack et al., 1987). On deleting an element of FGF signaling, it became evident that the FGF signaling pathway is essential for the formation of mesoderm (Amaya et al., 1991).

High-affinity receptors for FGFs (FGF-Rs) constitute a family of transmembrane proteins each possessing an intracellular tyrosine kinase domain (reviewed in Johnson and Williams, 1993). The extracellular region usually contains two to five immunoglobulin-like domains and a highly acidic region. Members of the FGF-R family have been shown present in many different mammals, amphibia and insecta. DFR1 encodes one of the two Drosophila FGF-R species (Shishido et al., 1993). In early embryos, the expression of DFR1 mRNA is specific to mesoderm. Using a small deficient chromosome from which the DFR 1 locus had been deleted, the possible involvement of DFR1 was previously shown in the patterning of muscle precursor cells (Shishido et al., 1993). But this does not provide sufficient indication of the functions of DFR1. Using the above deficiency, not only DFR1 but also several other genes with functions possibly indispensable for early mesodermal development are removed (Bellen et al., 1992).

Here, we describe the isolation and phenotype characterization of a mutant lacking the ability to produce DFR1. Dynamical changes in DFR1 expression, visualized by anti-DFR1 antibody, are also described. Mutant phenotypes indicate that DFR1 is essential not only for mesodermal cell differentiation but also for CNS formation. DFR1 mutant phenotypes are partially mimicked by targeted ectopic expression of activated Yan.

While this paper was in review, two papers (Beiman et al., 1996; Gisselbrecht et al., 1996) appeared which report requirements of DFR1 for migration and differentiation of early mesodermal cells. Our results include considerable extensions and partial amendments of their observations.

Df(3R)sr¹⁶was described previously (Shishido et al., 1993). wg^CX4is a null allele of wg. A fly line with UAS-yan^ACTwas obtained from G. Rubin (Rebay and Rubin, 1995). Four DFR1 (htl) mutants, htl⁸⁰, htl¹⁰⁸, htl¹¹⁷and htl^S1−28were identified among 112 lethals previously shown to fall within the Df(3R)P14 deficiency (Campbell et al., 1994). These htl chromosomes were lethal in trans to each other and over Df(3R)sr¹⁶. Genotypes of embryos were determined using a suitable blue-balancer. P-mediated germ-line transformation was carried out as follows. After genomic fragments derived from the DFR1 locus were cloned into a transformation vector pYCM (Fridell and Searles, 1991), resultant constructs were introduced into fly lines carrying htl⁸⁰. Several transformant lines recovered were crossed with Df(3R)sr¹⁶flies and tested for survival of htl⁸⁰/Df(3R)sr¹⁶. Two of five E12 transformants rescued htl⁸⁰/Df(3R)sr¹⁶.

Anti-DFR1 (anti-Htl) antiserum was prepared as follows. A 768bp BalI fragment of cFR4-6 (Shishido et al., 1993) was subcloned into pET3b (Rosenberg et al., 1987) to overproduce a T7-gene-10 protein fused with a part of the extracellular domain of DFR1 (from amino acid position 30 (N) to 285 (V)) in Escherichia coli cells. Partially purified fusion protein was used for immunization of rabbits and resultant antiserum was affinity-purified as described previously (Higashijima et al., 1992). The resultant antibody was specific to DFR1, there being no signals in embryos homozygous for Df(3R)sr¹⁶.

Antibody staining was carried out as described previously (Higashijima et al., 1992) with the exception of the usage of Cy3- or FITC-conjugated second antibody for confocal microscopic observation. Primary antibodies were: rabbit anti-DFR1 antibody used at 1:1000, rabbit anti-Eve antibody (Frasch et al., 1987) used at 1:50000, rabbit anti-MHC (myosin heavy chain) antibody (Kiehart and Feghali, 1986) used at 1:500, rabbit anti-Twist antibody (Thisse et al., 1988) used at 1:500, mouse anti-Fas III antibody (Snow et al., 1989) used at 1:10, mouse anti-Fas II antibody used at 1:5 (Lin et al., 1994) and mouse anti-Eve antibody used at 1:2 (Patel et al., 1992). In situ hybridization was performed as described previously (Shishido et al., 1993). tin cDNA was obtained from R. Bodmer. Araldite sections of embryos were prepared according to Azpiazu and Frasch (1993).

The mesoderm-specific expression of DFR1 RNA noted previously (Shishido et al., 1993) indicates DFR1 to possibly be required for mesodermal cell development. DFR1 protein expression must be examined in detail for clarification of its place in embryogenesis. Antibody was thus initially raised against the extracellular domain of DFR1. The DFR1 protein expression occurred initially in all mesodermal cells prior to gastrulation (stage 5; Fig. 1A,B). In contrast to uniform cell-surface distribution at later stages, DFR1 at this stage was localized exclusively on the apical cell surface (Fig. 1A, inset). Internalized mesodermal cells, spreading out below the ectoderm, persistently showed DFR1 expression (stage 6-10; Fig. 1C-E). This early mesodermal expression of DFR1 (first phase expression) is positively regulated by twist and snail (Shishido et al., 1993). As shown in Fig. 1J, it is significant that DFR1 distribution at stage 8-10 was homogeneous along the dorsoventral axis.

During stage 9, the area expressing DFR1 was divided into two domains, cephalic and postcephalic mesoderms, and the pattern of DFR1 expression in the latter began to show signs of a metameric fluctuation along the anteroposterior axis (Fig. 1E). A similar segmental expression pattern has been reported in Twist (Dunin Borkowski et al., 1995). Using ftz-lacZ stripes as internal references, we compared Twist and DFR1 expression (Fig. 1G-I), and found that, as in the case of Twist, future parasegmental borders are situated in periodic patches exhibiting relatively high DFR1 signals. During stage 10, DFR1 expression showed steady decrease, the reduction being sharp at stage 11 with residual but strong expression remaining in a fraction of somatic muscle precursor cells (Fig. 1F,K). Little or no DFR1 signals were detected in heart precursors at stage 11 (Fig. 1K). We presume the second phase expression of DFR1 in most somatic muscle precursors to start during stage 11.

At stage 12, the second-phase DFR1 expression occurred in heart precursors (Fig. 1L,N). Staining for Even-skipped (Eve) and DFR1 in heart primordia (Fig. 1O,P) revealed only cardial cell precursors, negative to Eve (Azpiazu and Frasch, 1993), to express DFR1. The second round of DFR1 expression in visceral muscle precursors for midgut and hindgut became discernible at stage 14 (Fig. 1Q,R), when dorsal pharyngeal muscles and foregut visceral muscles, derivatives of the cephalic mesoderm, were clearly observable as DFR1-positive tissues (Fig. 1R). From stage 12 onward, DFR1 expression was evident in a fraction of CNS cells (see below). Most DFR1 signals in somatic muscles (Fig. 1M) had faded out by stage 16 (Fig. 1S) except for signals in a few muscles including muscle 25 (ventral transverse muscle 1) in abdominal segments (A2-A7) and ventral internal segmental muscle 2, 4 and 5 in the thorax.

DFR1 has been assigned to the cytological position, 90D-E, uncovered by the deficiency Df(3R)sr¹⁶(Shishido et al., 1993). For clarification of the functions of DFR1, recessive lethal mutations were sought in the 90D-E region. Using anti-DFR1 antibody, study was made to determine whether embryos of 112 lethals, isolated by Campbell et al. (1994), are defective in DFR1 protein production. For htl⁸⁰, no protein stainable with anti-DFR1 antibody could be found throughout embryogenesis (Fig. 2D), indicating the absence of DFR1 protein production and this mutant thus to be a protein-null allele of DFR1. To further confirm this, examination was made of whether htl⁸⁰lethality is rescued by P-mediated transformation with wild-type genomic sequences. E12, a 12 kb fragment from the DFR1 locus (Fig. 3A), was found not only to rescue the lethality of htl⁸⁰/Df(3R)sr¹⁶but to recover DFR1 expression as well (Fig. 2A,G). htl⁸⁰ is thus concluded to be a protein-null allele of DFR1 all of whose essential parts are present within the E12 region. Three other DFR1 alleles, htl¹⁰⁸, htl¹¹⁷and htl^S1−28, were identified as alleles incapable of complementing htl⁸⁰. In contrast to htl⁸⁰, these alleles exhibited DFR1 expression detectable with anti-DFR1 antibody, indicating that they are not protein-null alleles of DFR1.

Using Df(3R)sr¹⁶, DFR1 was previously shown to be involved in muscle development in embryos (Shishido et al., 1993). Mutant phenotype examination was thus made using anti-MHC antibody (Kiehart and Feghali, 1986). In trans to Df(3R)sr¹⁶, all DFR1 mutants showed loss of the heart, and severely defective somatic and visceral muscles as in the case of htl⁸⁰homozygotes (Fig. 3B-I and data not shown). All defects in DFR1 mutants were eliminated by the DFR1 transgene expression (Fig. 3J). Thus, it was concluded that DFR1 is essential for the formation of the heart, midgut visceral muscles and somatic muscles.

Since all DFR1 mutants examined to date exhibit the heartless phenotype, and possible confusion in naming of Drosophila FGF-R genes (DFR1/DFR2 (Shishido et al., 1993) vs. DFGF-R2/DFGF-R1 (Klämbt et al., 1992)) should be avoided, DFR1 is hereafter referred to as heartless (htl) in accordance with the nomenclature of other workers (Beiman et al., 1996; Gisselbrecht et al., 1996). Unless otherwise stated, htl⁸⁰/Df(3R)sr¹⁶transheterozygotes are used as mutants lacking htl activity.

During stages 8 and 9, the mesoderm spreads with dorsolateral orientation along the underlying ectoderm to form a monolayer extending to the border between dorsal ectoderm and amnioserosa (see Fig. 1J). Heart, visceral and somatic muscles are produced through separate subdivisions of the extended mesoderm (Azpiazu and Frasch, 1993; Frasch 1995; Dunin Borkowski et al., 1995; Azpiazu et al., 1996). In these tissues or direct precursors, the second phase of Htl expression occurred from late stage 11 onward (see Fig. 1). Most, if not all, htl phenotypes observable from late stage 11 onward should thus be formed as a result of the absence of both the first and second phases of htl expression. Using Df(3R)sr¹⁶, we previously indicated htl not to be required for invagination (Shishido et al., 1993). Thus, with twist and tinman (tin; Azpiazu and Frasch, 1993), molecular markers of stages 8-10, htl (first phase expression) was studied for its being essential for early mesodermal cell migration after gastrulation.

Twist is expressed in all early mesodermal cells (Thisse et al., 1988). In wild-type embryos, the dorsal edge of extended mesoderm reached the dorsal crest at stage 9 or 4 hours AEL (Fig. 4A-D), while that of mesoderm in the htl mutant embryos failed to reach the dorsal circumference presumably because of incomplete spreading; rather, it was situated several cells away from the wild-type dorsal tip (Fig. 4E-H). Since both wild-type and mutant embryos were noted to have virtually the same number of Twist-positive nuclei (data not shown), the proliferation of mesodermal cells should be normal in htl embryos, and loss of htl activity may limit the dorsal migration of mesodermal cells.

To obtain stronger evidence of the presence of defects in htl mutant mesodermal cells in early migration, search was made for changes in the tin expression of these cells. Precursors for larval heart are derivatives of dorsal-most subdivisions of mesoderm (Bate, 1993; Bodmer, 1993; Azpiazu and Frasch, 1993; Frasch, 1995). tin encodes a home-odomain protein (Bodmer et al., 1990) essential for the formation of the heart and visceral muscles (Azpiazu and Frasch, 1993; Bodmer, 1993). It is initially expressed in all mesodermal cells (Fig. 4I; Bodmer, 1993; Azpiazu and Frasch, 1993; Frasch, 1995) and is positively regulated by twist (Bodmer et al., 1990). Soon following completion of mesodermal cell migration, twist-dependent tin expression gradually fades out to be replaced by dpp-dependent strong tin expression directly below dorsal ectodermal cells expressing dpp (Frasch, 1995). At stage 10, mesodermal cells not below dpp-expressing ectodermal cells completely lose tin RNA (Frasch, 1995). As shown in Fig. 4J, each domain of dpp-dependent tin expression splits into heart and midgut visceral mesoderm regions at stage 11 (Azpiazu and Frasch, 1993). In htl embryos, tin mRNA is initially expressed in all mesodermal cells (Fig. 4L) and later becomes restricted to dorsal part of mesoderm as with wild type (Fig. 4M). However, intensity of tin expression decreased extensively and one of the two subdomains was lost (compare Fig. 4J and M). The tin subdomain in the htl mutant at stage 11 appeared positioned virtually the same as in the wild-type subdomain for visceral mesoderm and disappeared at stage 12 as also noted for the wild-type visceral subdomain (Azpiazu and Frasch, 1993; Fig. 4N). In the wild-type heart domain, tin expression persists at later stages (Fig. 4K; Azpiazu and Frasch, 1993). Dorsal tin expression itself would thus appear independent of htl, but reduction in the number of mesodermal cells capable of receiving dpp signals from the dorsal-most ectoderm (Frasch, 1995) causes changes in the dorsoventral subdivision of mesoderm, in particular, loss of the tin domain for heart progenitors in htl mutant embryos. Consistent with this, htl mutant mesodermal cells responded to ectopic Dpp signals, with consequent induction of the expression of eve, a gene specific to heart primordia (pericardial cell precursors; Beiman et al., 1996; Gisselbrecht et al., 1996).

The larval heart is a pulsating vessel situated quite close to the ectoderm in the dorsal midline region (Bate, 1993; Zaffran et al., 1995) and comprises of two apposed layers of myoendothelial cells (cardial cells) surrounded by pericardial cells. Pericardial cell precursors are detectable as cells expressing a pair-rule gene, eve (Fig. 2B; Frasch et al., 1987). In htl embryos, no Eve-expressing cell clusters could be found in the region for prospective heart precursors (Fig. 2E). At the start of the germ-band retraction, tin expression is evident only in a continuous row of cells on the dorsal rim in wild-type embryos, which later form the heart (Fig. 4K; Bodmer et al., 1990; Bodmer, 1993): tin expression in the prospective midgut visceral mesoderm has faded away by the end of stage 11 (Azpiazu and Frasch, 1993). In htl embryos, tin expression could be no longer detected from stage 12 onward (Fig. 4N), indicating that heart precursors are not specified or formed in the htl mutant. Defects in the expression of eve and tin in the htl mutant was eliminated completely by the expression of the htl transgene (Fig. 2H and data not shown). eve expression was partially rescued by ectopic dpp expression driven by twist-Gal4 (Beiman et al., 1996; Gisselbrecht et al., 1996). Thus, it is concluded that, in htl embryos, presumptive heart primordia fail to come into contact with dorsal ectoderm from which Dpp signals are emitted and consequently never take on the status of heart primordia.

The larval gut is coated with a layer of visceral muscles as the source of power for peristaltic movement of digestion (Bate, 1993). Fasciclin III (Fas III) may serve as a marker for progenitors of midgut visceral mesoderm. At stage 11, Fas III is initially expressed in segmental patches. The expression becomes strong in cells situated along the ventral periphery of the patches and these cells form a band of visceral mesoderm in which Fas III expression continues up to stage 13 (Fig. 2C; Bate, 1993; Azpiazu and Frasch, 1995). In htl embryos, Fas III expression was frequently interrupted (Fig. 2F).

Connectin (Con) is a cell surface molecule, which marks a subset of visceral midgut muscle precursors (Nose et al., 1992; Bate, 1993). Our unpublished data showed Con expression cells to decrease considerably in mutant embryos. The defects in Fas III and Con expression in the mutant were completely rescued by the E12 transgene (Fig. 2I and data not shown).

The presence of Fas-III-positive and Con-positive cells indicates that specification of visceral muscle precursors takes place to some extent in htl mutant embryos, although Fas-III- or Con-positive cells are greatly reduced in number, consistent with the failure of spreading mesodermal cells in htl mutants.

Larval abdominal hemisegments (A2-A7) normally each contain a set of 30 syncytial muscle cells all of which can be distinguished in size, shape and point of attachment from late embryogenesis (Bate, 1993). Larval muscle formation is initiated with start of germ-band retraction (7.5 hour AEL). Somatic muscles in 111 hemisegments from late htl embryos were examined using anti-MHC antibody and dorsal oblique muscles were observed to have decreased to 26% of those of the wild-type and ventral longitudinal and oblique muscles, to 54%. Pleural and dorsal transverse muscles in the lateral body wall showed moderate decrease (86%) and, as expected, no Eve-positive dorsal-muscle cells corresponding to muscle 1 could be found at stage 14 in the htl mutant (Fig. 5A,B). Many unfused myoblasts were seen scattered in mutant embryos (Fig. 3D), possibly indication of fewer dorsal muscle precursor cells (putative muscle founder cells; Bate, 1993; Rushton et al., 1995) to which undifferentiated myoblasts become fused.

Anti-Htl antibody staining showed Htl expression in abdominal ventral transverse muscle 1 (muscle 25; Fig. 5C,D) and thoracic ventral internal segmental muscles (2, 4 and 5; Fig. 5E,F), strongly expressing Htl from stages 13 and 15, respectively, in the wild-type, to be absent from or reduced in most hemisegments of htl^S1−28, a producer of an inactive form of Htl. Although Htl expression in cephalic mesoderm was evident at stage 10, Htl expression disappeared in the derivatives of the cephalic mesoderm, dorsal pharyngeal muscles and foregut visceral muscles, at later stages (Figs 1L, 5G).

It is quite feasible that late defects in somatic musculature are partly due to the absence of the second phase or somatic-muscle-specific expression of htl. Intricate somatic muscles are considered to form in accordance with positional information in the ectoderm (Bate, 1993; Staehling-Hampton et al., 1994). Formation of the mutant cuticle was followed but nothing abnormal was found. Muscle defects in mutant embryos thus appeared to be intrinsic to the mesoderm.

Our previous experiment (Shishido et al., 1993) showed htl to be expressed in CNS cells. To determine whether these cells are neuronal or glial, wild-type embryos at late stage 11/early stage 12 were subjected to double-staining with anti-Htl antibody and anti-Reverse potential (Repo), which specifically labels virtually all CNS glial cells (Xiong et al., 1994; Campbell et al., 1994; Halter et al., 1995). Most, if not all, Htl-positive CNS cells (several cells per hemisegment (see Fig. 6K)), occupying a position for future longitudinal tracts (Jacobs and Goodman, 1989), were found to express Repo (Fig. 6L), indicating Htl-positive CNS cells to be interface glia (IG), which consist of longitudinal glia (LG) and their immediate glial neighbours (Fig. 4 in Halter et al., 1995). Consistent with this, two rows of Htl-positive non-neuronal cells, running parallel with each other along the dorsal face of devel-oping longitudinal connectives, were visualized by anti-Htl antibody at late stage 12, early stage 13 (Fig. 6G). At stage 16, nuclei of LG cells are arranged in two rows on each side of the midline (Campbell et al., 1994) and LG cells ensheath longitudinal axon tracts (Jacobs and Goodman, 1989). Anti-Htl antibody staining of the wild-type CNS showed Htl to be expressed in the extended cytoplasmic sheets growing out of the IG/LG cell bodies, which encircle the neuropil from lateral and medial sides (Fig. 6A,C).

It is possible to identify htl mutant IG/LG cells in htl^S1−28/Df(3R)sr¹⁶embryos by staining with anti-Htl antibody. As shown in Fig. 6H, in mutant embryos at stage 12/13, IG/LG cells were not only spatially disorganized but also exhibited a significant reduction in migration rate (compare Fig. 6H with G). Unlike wild-type cells, mutant IG/LG cells were rounded and appeared to be incapable of increasing their surface area. In contrast to wild type, mutant IG/LG cells at stage 16 could not line up in two files and appeared to fail to enwrap longitudinal axon tracts (Fig. 6B,F). Fig. 6D,E shows the interval between bilateral rows of IG/LG cells in the mutant to be 2-3 times larger than that of the wild-type. Mutant longitudinal axon fascicles were virtually normal in appearance (Fig. 6I,J). Thus, we conclude that htl is essential for IG/LG cell migration and ensheathing longitudinal connectives.

The MAP-kinase pathway has been shown to be involved in signaling by Breathless, another Drosophila FGF-R (Reichman-Fried et al., 1994). Thus, reduction in MAP-kinase signaling activity in mesodermal cells may result in htl mutant phenotype. To test this possibility, flies with a construct that drives spatially restricted Gal4 expression under the control of htl regulatory elements (Ono and Saigo, unpublished data) and flies expressing an activated form of Yan, a downstream repressor of the MAP-kinase pathway, under the control of Gal4 upstream sequences (Rebay and Rubin, 1995) were crossed. Embryos derived from this cross express activated Yan in all mesodermal cells from stage 10 (data not shown). The effects of activated Yan on the expression of Eve, Fas III, and Htl in the wild-type background were studied.

The ectopic expression of activated Yan apparently caused reduction in Eve-expressing cells (Fig. 7A,B); a much less extensive effect on Fas III expression was detected (Fig. 7C,D). Anti-MHC antibody staining indicated normal heart and somatic muscle formation in part not to occur (Fig. 7E,F), though that of midgut visceral muscles appeared to proceed normally (data not shown). These findings, and the fact that the htl mutant phenotype is partially rescued by the targeted mesodermal expression of an activated form of Ras1 (Gisselbrecht et al., 1996), may indicate that the MAP-kinase pathway acting downstream of Htl or parallel to its pathway or both is essential for the formation of heart and somatic muscles.

The second phase of Htl expression extends from stage 11 onward in putative precursors for heart, somatic and visceral muscles. Heart formation was previously shown to require both wg and hh, with only hh being required for normal visceral mesoderm formation (Wu et al., 1995; Park et al., 1996; Baylies et al., 1995). Htl expression in somatic muscles is segmental (see Fig. 1E). Study was thus made to see whether Htl expression in somatic muscles requires Wg and/or Hh signals. Anti-Htl staining showed many muscle precursors in hh mutants to be capable of producing Htl (data not shown). In contrast, Htl expression in a large fraction of presumptive somatic muscle precursors was wg-dependent (Fig. 7G,H). wg is expressed in the mesoderm only for a short period after gastrulation (Lawrence et al., 1994); the results thus suggest that the second phase of Htl expression may be controlled by Wg signals from ectodermal cells.

The mode of Htl expression suggest that signaling relevant to Htl occurs in at least three stages of mesodermal development:(1) an early stage for spreading of dorsolaterally orientated mesodermal cells, and (2) that for segregation of the mesoderm subdivided by inductive signals. The third stage of signaling is induced by the second phase of Htl expression in precursors of each mesodermal tissue or organ. This study, and those of Beiman et al. (1996) and Gisselbrecht et al. (1996), established the functions of Htl at the earliest stage of mesodermal development by characterizing mutant phenotypes.

The dorsal subdivision of the mesoderm has been shown elicited by signals derived from the ectoderm (Frasch, 1995; Staehling-Hampton et al., 1994). On preventing the mesoderm from interacting with the dorsal part of the ectoderm, induction of the formation of the heart and visceral muscles was not possible (Baker and Schubiger, 1995). Genetic analysis showed Dpp to serve as central signaling molecules from the dorsal ectoderm to underlying mesoderm and these Dpp-dependent signals to be essential for tin activation in dorsal subdivision of mesoderm (Frasch, 1995). Since tin is a homeobox gene required for the formation of the heart and visceral muscles (Bodmer et al., 1990), restriction of tin expression to dorsal mesodermal cells is absolutely required for early regional subdivision of the mesoderm (Frasch, 1995).

Maggert et al. (1995) noted that narrowing of the ventral furrow (prospective mesoderm) caused loss or reduction of heart and visceral muscles. In their mutants, there was no spreading of mesoderm to the dorsal region; therefore, fewer cells could come into contact with dpp-expressing ectodermal cells and, as a result, dorsal tin-expression region failed to be formed. These phenotypes appear to be the same as those observed here for htl mutants. In the absence of Htl, not only does properly directed migration of mesodermal cells fail to occur (see Fig. 4A-H), but there is also no normal formation of heart and visceral muscles (see Figs 2, 3). tin domains for heart precursors and visceral muscles, respectively, did not form and was reduced in size (see Fig. 4J,M). Since htl mutant mesodermal cells are capable of expressing eve, a homeobox gene specific to pericardial cells (Frasch et al., 1987), in response to exogenous dpp signaling (Beiman et al., 1996; Gisselbrecht et al., 1996), the first phase of Htl expression in primitive mesoderm may be essential for the dorsolateral migration of mesodermal cells and the loss of heart precursors and reduction of visceral mesoderm in htl mutants may be due to the failure of mutant mesodermal cells to reach the dorsalmost region.

A recent model of trachea formation suggests that the pathway involving Branchless FGF and Breathless (Btl) FGF-R specifies the mode of trachea branching by guiding tracheal cell migration during primary branch formation (Sutherland et al., 1996). However, whether Htl FGF-R responds to putative chemotactic signals for primitive mesodermal cell migration is not known. Anti-Htl antibody staining of transverse sections of wild-type embryos could indicate no preferential distribution of Htl (see Fig. 1J). Htl appeared distributed evenly on the surface of all primitive mesodermal cells except for those at stage 5, a stage before gastrulation, in which Htl is localized in the apical cell surface (see the inset of Fig. 1A). It is possible that mesodermal cells are activated by putative Htl ligands possibly present in perivitelline fluid at stage 5 and, consequently, acquire the ability for chemotaxis at later stages.

Twist expression is known to be modulated in a metameric fashion at stage 9/10 of development (Dunin Borkowski et al., 1995). Baylies and Bate (1996) recently showed that levels of Twist expression are critical determinants of mesodermal differentiation in the early embryo. The absence of Twist causes defects in somatic muscle formation, while high levels of Twist switch cells into somatic myogenesis during subdivision of the mesoderm, resulting in various derangements in heart and visceral muscle formation. Thus, similarity in modulation of expression between Twist and Htl is an intriguing finding (see Fig. 1G-I). One might imagine that Htl serves as a downstream element of Twist to segregate the subdivided mesoderm.

The second phase of Htl expression is wg-dependent and most, if not all, positive signals emerged in the regions of each parasegment with relatively high levels of the first phase of Htl expression (see Fig. 1F,L), possibly suggesting that the metameric modulation of the first phase of Htl expression is also wg-dependent. Since gene expression and/or differentiation of a fraction of somatic muscles, pharyngeal muscles and foregut visceral muscles were affected at later stages by the absence of htl (see Fig. 5), it is quite feasible that htl is also implicated in late differentiation of mesodermal cells.

As with mesodermal cells, Htl is required for migration and morphogenesis of interface glia including LG cells (see Fig. 6). Klämbt et al. (1992) showed another Drosophila FGF-R, Btl, to play an important role in migration of midline glia. Together, these findings indicate the concerted expression of two FGF-Rs in glial cells to be essential for CNS formation.

In chicken embryos, cardiac myocytes are derived from anterolateral mesoderm called the heart-forming region (HFR) (Mima et al., 1995; Lough et al., 1996). When cells in HFR become committed to a cardial lineage, FGF and FGF-R are expressed at high levels (Mima et al., 1995). Introducing dominant-negative FGF-R showed that FGF-R is required for cardiac myocyte proliferation in early heart development (Mima et al., 1995). FGF-R has been shown necessary for the proper cell proliferation of mesoderm in mice (Deng et al., 1994). Abnormal axial formation in FGF-R-deficient mice does not arise from intrinsic blocks in mesodermal differentiation since cells from FGF-R-deficient mice can produce many mesodermal cell types in vitro.

Lough et al. (1996) showed BMP-2 of the TGF-β growth factor family and FGF-4 to be involved in the early heart development in vertebrates. Although neither FGF-4 nor BMP-2 alone induces the formation of cardial cells, together, they induce cardiogenesis. BMP-2 mimics cardiogenic effects on precardiac mesoderm but cannot support the viability of mesoderm. FGF-4 does not induce cardiogenesis in non-pre-cardiac mesoderm. In Drosophila, heart formation requires FGF-R (Htl) and TGF-β (Dpp). As with Drosophila tin, Nkx2-5, a mammalian tin homolog, is expressed in early mesodermal development (reviewed in Harvey, 1996). A functional relationship would thus appear to exist between tin and Nkx2-5. It is of interest that vertebrates and invertebrates (Drosophila) both require essentially the same molecules for establishing the cardial system during embryogenesis.

Examination of the targeted expression of activated Yan (see Fig. 7A-F) indicated the involvement of the MAP kinase pathway in the heart formation and somatic muscles. Since no activated Yan expression could be detected prior to stage 10 by antibody staining (Ono and Saigo, unpublished data) and no appreciable effect of activated Yan on Fas III expression was detected (Fig. 7D), it is quite feasible that, under the present conditions, activated Yan can affect only late mesodermal differentiation starting from stage 11/12. Gisselbrecht et al. (1996) showed that the targeted mesodermal expression of an activated form of Ras1 can partially rescue hypomorphic htl mutant phenotypes. They also showed that activated Ras1 is capable of partially eliminating defects from oriented migration of mesodermal cells. It may thus follow that the MAP kinase pathway is required for normal differentiation of the heart throughout development.

We thank Andrew Tomlinson for his generous gift of mutant flies. We are also grateful to Tadashi Uemura, Bernard Thisse, Christine Thisse, Daniel Kiehart, Manfred Frasch, Siegfried Roth, Shigeo Hayashi, Akinao Nose, Rolf Bodmer and Gerald Rubin for antibodies, cDNA or fly strains. We thank Akinao Nose and Masatoshi Takeichi for discussion. E.S. was supported by JSPS Research Fellowships for Young Scientists. This work was supported in part by grants from the Ministry of Education, Science and Culture of Japan to K. S.