The torpedo (DER) receptor tyrosine kinase is required at multiple times during Drosophila embryogenesis

In metazoans, intercellular signalling regulates cell proliferation, differentiation and survival during devel opment. One class of molecules involved in the reception of such cell extrinsic cues are the receptor tyrosine kinases. Members of this protein family include the insulin receptor, receptors for peptide growth factors and several proto-oncogene products (reviewed in Yarden and Ullrich, 1988). The subfamily of receptor tyrosine kinases typified by the epidermal growth factor (EGF) receptor has been highly con served through evolution. Vertebrate members of this subfamily include the closely related EGF receptor, neu, HER3 and Xmrk proteins (Ullrich et al., 1984; Schechter et al., 1984; Bargmann et al., 1986; Kraus et al., 1989; Wittbrodt et al., 1989; Plowman et al., 1990), while the products of the C. elegans let-23 and D. melanogaster torpedo (DER) genes represent the invertebrate members of this class (Livneh et al., 1985;

Wadsworth et al., 1985; Schejter et al., 1986; Price et al., 1989; Schejter and Shilo, 1989; Aroian et al., 1990). The role of the EGF receptor in vertebrate develop ment has been analyzed mainly by assaying the biological effects of its primary ligand, the epidermal growth factor (Cohen, 1962; Carpenter and Cohen, 1979). Studies utilizing whole organisms, organ culture and cell culture have shown that EGF elicits a variety of cellular responses. One of the most thoroughly studied roles of the epidermal growth factor is its regulation of mitogenesis. EGF stimulates the mitotic activity of a variety of epithelial, glial and mesenchymal cells (reviewed in Carpenter and Cohen, 1979). In vivo and in organ culture, the growth factor stimulates mitogene sis in the skin, mammary duct, palate and tooth (Tyler and Pratt, 1980; Moore et al., 1983; Taketani and Oka, 1983; Grove and Pratt, 1984; Partanen et al., 1985; Topham et al., 1987; Coleman et al., 1988). Epidermal growth factor is also capable of inhibiting mitogenesis (Gill and Lazar, 1981; Imai et al., 1982; Partanen et al.,1985; Kamata et al., 1986; Coleman and Daniel, 1990). In addition, epidermal growth factor promotes the differentiation of a variety of cell types (Sun and Green, 1976; Chen et al., 1977; Schonbrunn et al., 1980; Johnson et al., 1980; Topham et al., 1987; Blecher et al., 1990; Gattone et al., 1990; Kapalanga and Blecher, 1990; der Hertog et al., 1991). Another ligand for the EGF receptor, transforming growth factor ex (Derynck et al., 1984), has been implicated in the differentiation of mouse mammary epithelial cells (Taverna et al., 1991) and the murine female hypothalamus (Ojeda et al., 1990). The EGF receptor may also transduce a signal required for cell viability. Morrison et al. (1987) have reported that epidermal growth factor promotes the survival of primary cultured neurons from rat brain. Classical genetic analysis of invertebrate homologs of the EGF receptor provides a complementary approach to investigating the developmental roles of this molecule. Through mutational analysis, it is possible to assay the developmental consequences of partial and complete elimination of receptor function on virtually every tissue of the organism. The torpedo (DER) locus of Drosophila, which encodes alternate forms of a receptor tyrosine kinase showing homology to the EGF receptor, is amenable to genetic analysis (Lev et al., 1985; Wadsworth et al., 1985; Schejter et al., 1986; Price et al., 1989; Schejter and Shilo, 1989; Wides et al., 1990; Zak et al., 1990). Genetic studies have revealed that torpedo, which was isolated in a screen for recessive female sterile mutations (Schupbach, 1987), is allelic to the faint little ball embryonic lethal locus and the Ellipse eye patterning gene. The recessive faint little ball mutations (Nusslein-Volhard et al., 1984) are severe loss-of-function alleles of torpedo (Clifford and Schup bach, 1989; Price et al., 1989); the dominant Ellipse lesions represent hypermorphic alleles of the gene (Baker and Rubin, 1989). Phenotypic characterization of torpedo mutations reveals that the gene is required not only for embryogenesis and oogenesis, but also for imaginal disc growth and differentiation (Baker and

Rubin, 1989; Clifford and Schupbach, 1989). The torpedo receptor tyrosine kinase, like the EGF receptor, may transduce signals that elicit qualitatively different responses in different cell types. Examination of the female-sterile phenotype of torpedo¹ has shown that the gene is required for the specification of cell fate in the developing egg chamber (Schupbach, 1987). During embryogenesis, the receptor may transduce a signal necessary for ectodermal cell viability (Price et al., 1989; Schejter and Shilo, 1989). In the retina, gain of-function mutations in the gene reduce the number of cells that develop as photoreceptors, possibly by maintaining cells in an undifferentiated state (Baker and Rubin, 1989).

To gain further insight into the nature of the signals transmitted by torpedo, we have extended previous characterizations of the zygotic embryonic lethal phenotype of the gene. Previous studies have revealed that loss of the gene product leads to the loss of anterior, posterior and ventral cuticle, arrested germ band retraction and reduced denticles (Nusslein-Vol-hard et al., 1984; Clifford and Schupbach, 1989; Price et al., 1989; Schejter and Shilo, 1989). torpedo is also required for the proper development of the embryonic central nervous system (Schejter and Shilo, 1989; Zak et al., 1990). In our present analysis, we show through temperature-shift experiments that torpedo gene ac tivity is required at two distinct periods during embry onic development: the gene product is first needed for the viability of the amnioserosa and certain populations of ectodermal tissues and later for the production of denticles by the ventral hypodermis. We also provide a genetic demonstration that maternally synthesized torpedo product is not supplied to the developing embryo. Therefore, both the early and late embryonic requirements for torpedo are strictly zygotic, and the mutant phenotypes accurately reflect the consequences of eliminating all gene activity from the embryo. Our analysis of torpedo’s null phenotype argues that while the gene is not necessary for mitogenesis in the embryo, it may mediate signals required for cell survival, as well as for cell determination or differentiation.

The markers, rearrangements and balancer chromosomes used are described in Lindsley and Zimm (1985, 1986, 1987, 1990) unless otherwise noted. For the origin of torpedo alleles used in this study see Clifford and Schupbach (1989) and Price et al. (1989). Y. Hiromi generously provided CyO, βC1, a CyO balancer chromosome that carries a ftz-lacZ fusion gene.

Embryonic cuticles were prepared as described in Wieschaus and Nlisslein-Volhard (1986) and examined under phase contrast optics.

The zygotic embryonic lethal phenotype was studied in time-lapse video films of embryos from several mutant stocks (topJEI, topco, top^1K35) that were also homozygous for the maternal-effect mutation klarsicht on the third chromosome, as well as in embryos that were homozygous for the chromosomal deficiency Df(2R)PK1. Films were made with an Olympus compound microscope at ×250 or ×400 and a Panasonic NV 8050 video camera as described in Zusman and Wieschaus (1985) and Wieschaus and Nlisslein-Volhard (1986). Since filming of the embryos was started before the mutant defects were visible, homozygous mutant as well as phenotypically wild-type embryos were filmed. Wild-type embryos served as controls for the timing of developmental events and the extent of abnormalities seen in the mutants.

Sections of fixed, plastic embedded embryos were prepared as described in Wieschaus and Sweeton (1988). Embedded embryos were serially sectioned at two microns (for sagittal sections) or five microns (for cross sections). Analysis of older homozygous mutant embryos showed that gut differentiation proceeds normally in the mutant. For this reason, we used the length, location and morphology of the posterior midgut, as well as the location of the pole cells, to stage the sectioned animals during stages 9–11 of embryogenesis.

To examine mitotic patterns in stage 7 to 9 torpedo embryos, three to five hour old embryos were collected at 25°C from a top^CO/CyO, βC1 stock; since the CyO, βC1 chromosome contains a ftz-lacZ fusion insertion that produces detectable enzymatic activity by gastrulation, torpedo homozygotes could be identified by X-gal staining (Bellen et al., 1989). -galactosidaseanimals were dissected from the vitelline membrane by hand, then stained with Hoechst 33258 (Polysciences) as described in Wieschaus and Ni.isslein Volhard (1986). Cell division patterns were compared to the mitotic maps of Foe (1989).

To study the temporal requirement for torpedo product during embryogenesis, temperature-shift experiments were performed with the conditional allele top^1F26 In the 18°C to methanol (Wieschaus and Ntisslein-Volhard, 1986), rehy29°C shift experiment, top^1F26/CyO females and males were mated in inverted tripour beakers at l8°C and allowed to lay eggs on filter paper discs seeded with live yeast. Eggs were collected at two hour intervals and allowed to develop at 18°C for the appropriate length of time before temperature shift. Since starved Drosophila females often retain fertilized eggs, the first egg collection of each day was discarded. Temperature shifting was accomplished by transferring the yeasted filter paper to a Petri plate prewarmed to 29°C, adding a few drops of prewarmed distilled water, then immediately placing the Petri dish in a 29°C incubator. After 48 hours dead embryos were collected, dechorionated and mounted as described in Wieschaus and Ntisslein-Volhard (1986). A similar procedure was used to perform the reciprocal shifts. 95% confidence intervals for the data points in Table 1 were calculated from a statistical table in Sokal and Rohlf (1973).

Homozygous mutant germline clones were produced by gamma irradiating female larvae trans-heterozygous for the torpedo mutation of interest and the dominant female-sterile mutation Fs(2)D, as described in Schupbach and Wieschaus (1986). After eclosion irradiated females were tested in groups of 10 for three to five days on egg collection plates. Compared to clones of other female-sterile mutations located on the right arm of the second chromosome, torpedo germline clones were recovered at similar frequencies (2 – 3% of irradiated individuals) and were of equivalent size (T. S., unpublished).

To examine the expression of the decapentaplegic, single minded, snail and zerknilllt genes, single-stranded antisense DNA probes containing digoxigenin-labelled dUTP were produced by polymerase chain reaction (Saiki et al., 1988). Approximately 400 ng of template DNA and 100 ng of primer were subjected to 30 rounds of synthesis (95°C for 45 seconds, 55°C for 30 seconds, 72°C for 90 seconds) in 50 ml of reaction mixture containing 50 mM KCl, 10 mM Tris (pH 8.3), 1.5 mM MgC1₂0.1 % gelatin, 0.2 mM dATP, 0.2 mM dCTP, 0.2 mM dGTP, 0.13 mM dTTP, O.Q7 mM digoxigenin dUTP (Boehringer-Mannheim) and 2 U Taq polymerase (Perkin-Elmer Cetus).

The decapentaplegic probe was generated from the 4.5 kb cDNA insert of pBEhl (St. Johnston et al., 1990) and the zerknilllt probe from the 1.4 kb zen zl cDNA cloned into pGEM-1 (Rushlow et al., 1987). Probes to the single-minded (Crews et al., 1988) and snail (Boulay et al., 1987) transcripts were generously provided by S. Roth.

Embryos were prepared and hybridized according to the procedure of Tautz and Pfeifle (1989), as modified by N. Patel. RNAs were visualized with the Genius kit (Boehringer Mannheim).

For BrdU incorporation studies, embryos were labelled by a modification of the procedure of Bodmer et al. (1989). Embryos were collected in wire mesh baskets, dechorionated for 2 minutes in 50% bleach, thoroughly rinsed with distilled water, then permeabilized for 4 minutes in octane (Aldrich) saturated with D22 medium (Ashburner, 1989). Excess octane was removed by blotting, then the embryos were transferred to D22 containing 1 mg/ml BrdU (Sigma) for 30 minutes at room temperature. Labelled embryos were immediately fixed for 20 minutes in a 1:1 mixture of heptane and 4% formaldehyde in PEM. Embryos were devitellinized with methanol (Wieschaus and Ntisslein-Volhard, 1986), rehydrated, then hydrolyzed for 25 minutes with 2 N NaOH prior to antibody treatment. Hydrolyzed animals were first washed twice in PT (1 x PBS, 0.1% Tween-20), then three times for 30 minutes in 1 x PBS, 1% BSA, 0.1% Tween-20. Animals were next incubated with a 1:50 dilution of anti-BrdU (Boehringer-Mannheim) overnight at 4°C, then processed according to the instructions in the Vectastain kit (Vector Research). Embryos were mounted in Aqua-Poly/Mount (Polysciences).

Anti-engrailed staining was performed as described in DiNardo et al. (1985) using an antibody kindly provided by N. Patel.

To visualize tubulin, dechorionated embryos were incubated for 2 minutes in 1 nM taxol in PEM (0.1 Pipes, 2 mM EGTA, 1 mM MgSO₄, pH 6.95), then fixed for 20 minutes in 3.7% formaldehyde in PEM. Embryos were devitellinized with methanol. After washing three times for 30 minutes in PBT (1 x PBS, 0.1% BSA, 0.1% Tween-20), embryos were incubated with a 1:2000 dilution of antitubulin (Boehringer-Mannheim) overnight at 4°C, washed four times for 15 minutes in PGT (1 x PBS, 1% normal goat serum (Vector Research), 0.1% Tween-20), then incubated for 3 hours at room temperature with a 1:350 dilution of fluorescein-conjugated goat anti-mouse antibody (Boehringer Mannheim). Embryos were washed four times for 15 minutes in PT, mounted between two coverslips in Aqua-Poly/Mount, then viewed under fluorescence optics. Photographs were taken with Kodak T-max 100 film.

twist protein was visualized using a mouse monoclonal antibody kindly provided by S. Roth. Dechorionated embryos were fixed 20 minutes in 3.7% formaldehyde in PEM, then devitellinized with methanol. After washing three times for 30 minutes in 1 x PBS, 1% BSA, 0.1% Tween-20, embryos were incubated overnight at 4°C in a 1:1000 dilution of anti-twist antibody. Embryos were then processed according to the protocol supplied in the Vectastain kit (Vector Research). After clearing in methyl salicylate, embryos were mounted in Permount (Fisher), and photographed with Kodak T-max 100 film.

Previous analysis of torpedo has shown that the gene is required zygotically for embryonic viability (Nusslein Volhard et al., 1984; Price et al., 1989; Schejter and Shilo, 1989; Clifford and Schupbach, 1989). Animals homozygous for complete loss-of-function torpedo mutations die during embryogenesis and exhibit a characteristic cuticular phenotype. Indicative of a failure in germband retraction, mutant animals are tightly curled upon themselves, with posterior struc tures lying adjacent to anterior ones. torpedo amorphs also produce less anterior and posterior cuticle than wild type (Fig. 1B). Additionally, torpedo amorphs exhibit a loss of ventral cuticle that is similar to, but substantially more severe than, the morphological defect produced by the spitz group mutants (Mayer and Ni.isslein-Volhard, 1988). In the most severe mutant animals, a narrow stripe of naked cuticle runs along the ventral midline. In somewhat less severely affected embryos, this strip of cuticle is wider and may carry some patches of reduced denticles (Fig. 1C).

Animals homozygous for moderate alleles are similar to amorphic embryos in their overall morphology but undergo partial germband retraction, secrete greater amounts of cuticle and produce more ventral denticles (Fig. 1D). The ventral setae in these animals are organized into very narrow bands which contain at most four rows of denticles, all of which point posteriorly and are smaller than the ventral setae found in wild-type abdominal denticle belts.

Mutant animals homozygous for the weakest embry onic lethal alleles show a more intact cuticle but often exhibit reductions in the ventral plate and H piece of the anterior head skeleton. Many of these animals fail to complete germband retraction. Animals homo-zygous for weak embryonic lethal torpedo alleles occasionally suffer fusions or gaps along the dorsal or ventral midline, defects reminiscent of those produced by mutations in the spitz group loci (Mayer and Nlisslein-Volhard, 1988), and exhibit variable re ductions in the number and size of ventral setae (Fig. 1E).

The complex phenotype exhibited by torpedo embryos results from a number of morphological defects. To separate the phenotype into discrete components and to assay the temporal requirements for torpedo activity during embryogenesis, we performed temperature-shift experiments with the conditional allele top^1F26. At l8°C mutant animals show a weak zygotic embryonic lethal phenotype (Fig. 2A); when raised at 29°C top^1F26 homozygotes exhibit a moderate or severe embryonic lethal phenotype (Fig. 2B). Temperature shift analysis identifies two distinct times at which torpedo is required for embryonic development. The phenocritical period for arresting germband retraction and reducing the amount of head, thoracic and ventral cuticle produced by top^1F26 animals extends from approximately three to six hours at 29°C (Table 1, Fig. 3A). This early requirement for torpedo, from the beginning of stage 7 to early stage 11 of embryogenesis, coincides with the first appearance of visible defects in the mutant during stage 10 of development: irregularities along the ventral midline and segmental grooves and increased cell death (see below). The temperature sensitive period for denticle reduction, in contrast, extends roughly from seven to ten hours of development at 29°C (Table 1, Fig. 3B), which corresponds to late stage 11 to early stage 13 of embryogenesis. Since cuticle is secreted during stage 16 of embryonic development (Campos-Ortega and Hartenstein, 1985), the critical period for ventral setae reduction appears to end several hours before denticles are actually formed. We have verified the phenocritical period for denticle reduction by performing a temperature pulse experiment in which top^1F26 embryos raised at l8°C for the first 11-13 hours of development were shifted to the restrictive temperature for three hours, then returned to l8°C to complete development. Thirty nine per cent (22/57) of the pulsed animals show an extreme reduction of all ventral setae; thus, limited exposure to the non-permissive temperature can produce severe denticle defects.

The terminal phenotypes of temperature-shifted embryos show that the requirement for torpedo in denticle formation is independent of its requirement in germband retraction and cuticle synthesis. Animals maintained at 29°C for the first seven to nine hours of development frequently show severe morphological defects and produce fewer denticles than wild type. Nonetheless, some of these animals secrete ventral setae of normal size (Fig. 2C). Conversely, animals shifted to 29°C after the temperature-sensitive period for germband retraction and cuticle formation usually show fairly normal gross morphology but often possess fewer and smaller denticles than animals raised continuously at l8°C (Fig. 2D). The fact that certain develop mental processes requiring torpedo, such as denticle formation and cuticle production, can be disrupted independently of one another shows that the defects in the mutant do not all originate from the loss of gene product at a single point in time. Rather, torpedo activity is needed at two or more periods during embryogenesis.

To determine whether the observed zygotic embryonic phenotypes reflect the complete absence of torpedo product from the embryo or whether maternal product deposited in the egg during oogenesis partially alleviates these phenotypes, we examined embryos produced from germline clones. Germline clones homozygous for the severe loss-of-function alleles top^co, as well as the weak embryonic lethal allele top¹⁰¹ were induced by irradiation, and the females were mated to their heterozygous sibs. The 25 females that carried top^IK35 and topco, top^1K35 or top^JE1 germline clones produced a total of 111 fertilized eggs. 52 of these eggs gave rise to adult animals whose phenotypes showed that they had received a wild-type torpedo allele from their father, and the remaining 59 eggs developed into severely mutant embryos that showed no difference in phenotype from amorphic embryos derived from hetero zygous females. Likewise, the 99 homozygotes derived from top¹⁰¹ germline clones and the 112 homozygous embryos derived from top^1F26 germline clones at 22°C showed the same range and distribution of intermediate to weak embryonic lethal phenotypes as did mutant embryos produced by heterozygous mothers.

The phenotype of top^CO/top^1F26 embryos derived from heterozygous females is intermediate in severity to those exhibited by topco and top^1F26 homozygotes; therefore, the development of these animals should be highly responsive to slight increases or decreases in gene activity. To test in a more sensitive way whether maternally supplied torpedo product is utilized during embryogenesis, 19 females with Jermline clones homozygous for the strong allele top^CO were mated at 22°C to males heterozygous for top^1F26. The range and distribution of the terminal phenotypes of 41 top^CO/top^1F26 and the temperature-sensitive allele top¹ embryos derived from topco clones did not differ from those of topco /top^1F26 animals derived from mothers heterozygous for top^1F26 and a wild-type chromosome. These results indicate that there is very little, if any, torpedo activity in the germ line.

To determine the course of embryogenesis in animals lacking all zygotic torpedo activity, we examined videotapes of living embryos or whole-mount prep arations of animals homoztfous for the amorphic mutations topco and top¹ and the deficiencies Df(2R)PK1 and Df(2R)top^18A. These studies show that early developmental events, such as cellularization and gastrulation, are normal in mutant embryos. When segmentation first becomes visible in the wild-type embryo (stage 10), torpedo embryos develop unusually deep grooves along the outside of the germband which may correspond to the parasegmental grooves. At this time, some mutant embryos also show a twisting of the germband and mild irregularities along the ventral midline (Fig. 4B), a defect first described by Baker and Rubin (1989).

As embryogenesis continues, torpedo animals exhibit additional developmental defects. Shortly before the onset of germband retraction (stage 11), the anterior headlobe of the mutant retracts dorsoposteriorly. Soon afterward, clumps of cells are shed from the surface of the head and thorax of torpedo embryos; most cephalic and thoracic structures crumble apart as development continues (Fig. 4D, F, H). Cells are also lost from more posterior parts of the embryonic surface, but to a less dramatic extent. Germband retraction is usually initiated at the appropriate time in development, but the germband retracts only a very short distance, and the embryo compacts itself into a tightly curled form (Fig. 4F, H). At later stages of embryogenesis mutant animals exhibit intense twitching, which indicates that they have formed some functional muscles.

Examination of serially sectioned topco embryos showed that cell death in the mutant begins during late stage 10 (time of stomodeal invagination, approxi mately 5 hours after egg deposition at 25°C), when small clusters of necrotic cells are visible in the amnioserosa, head and germband of torpedo embryos (Fig. 4J). Cell death increases during subsequent development. In stage 11 and 12 embryos, large patches of dead cells are found in the amnioserosa, the brain and the cephalic epidermis, and clumps of necrotic cells appear in what may be a segmentally repeated pattern in the epidermal tissues of the germband (Fig. 4K). Examination of cross sections of stage 12 and 13 Df(2R)top^18A homozygotes reveals an asymmetric dis tribution of epidermal cell death along the dorsoventral axis of the embryo. Necrotic cells in the ectoderm are preferentially, though not exclusively, found at two positions along the dorsoventral axis: at the ventral midline and at the boundary between dorsal and ventral epidermis (Fig. 4L, M). Within the different layers of the germband, the tissue most affected in the mutant embryos clearly corresponded to the epidermis. In many instances dead cells could be seen in the epidermal cell layer, whereas such cells were not observed within the neuropile, within the endodermal derivatives (gut, Malpighian tubules) or among the mesodermal cells. In late homozygous embryos, the gut primordia appeared to have properly fused and enclosed the yolk; gonads, salivary glands, differentiated muscles, neural tissue and nerve fibers corresponding to parts of the ventral nerve cord were also present.

Cell death appears to play a major role in the evolution of torpedo’s final embryonic lethal pheno type. In the amorph, cell death begins during the temperature-sensitive period for disrupting germband retraction and the production of anterior and ventral cuticle in top^1F26 animals and is concentrated in the amnioserosa and the epidermal tissues of the head, thorax and ventral abdomen. The late requirement for torpedo may not involve cell viability, despite the fact that extensive epidermal cell death does occur in the amorph during the phenocritical period for inducing denticle defects (late stage 11 to early stage 13). While topIF2⁶ embryos shifted to the restrictive temperature at this time frequently show a severe reduction in the size and number of denticles, they rarely suffer a noticeable loss of ventral cuticle.

The vertebrate EGF receptor has been implicated in the positive regulation of mitogenesis (reviewed in Carpenter and Cohen, 1979). From tissue culture studies, it has been proposed that this receptor tyrosine kinase controls cell division by transducing a “progression” signal that enables mitogenically competent cells to overcome a G₁ cell cycle block (Stiles et al., 1979; Pardee, 1989). Given its homology to the vertebrate EGF receptor, it is possible that the torpedo gene product plays a similar role in Drosophila embryogenesis. To address this possibility, we exam ined the pattern of cell divisions in the ectodermal tissues of mutant animals. Because torpedo is not expressed until the cellular blastoderm stage (Lev et al., 1985; Schejter et al., 1986; Kammermeyer and Wads worth, 1987; Zak et al., 1990; Katzen et al., 1991), we restricted our analysis to cell divisions occurring after cellularization.

All epidermal cells in the fly embryo undergo two rounds of cell division after the cellular blastoderm stage, while a subset of epidermal cells completes a third mitosis. The first postblastoderm round of cell division (cycle 14) occurs during stages 7 to 10 of embryogenesis, the second during stage 10, and the third during stage 11 (Campos-Ortega and Hartenstein, 1985; Foe, 1989). Cells in the peripheral and central nervous systems divide more extensively, completing at least four postblastoderm mitoses before they undergo terminal differentiation (Campos-Ortega and Harten stein, 1985; Glover, 1991). Precursors of the cuticular sense organs undergo their final rounds of division during stages 12 and 13 of embryogenesis; mitogenesis in the central nervous system continues into larval development (Campos-Ortega and Hartenstein, 1985; Bodmer et al., 1989).

We did not observe any abnormalities in the mitogenesis of ectodermal tissues in torpedo embryos, even in populations of cells that require the gene product for their survival. Clusters of cells undergoing mitosis 14 were visible in all ectodermal tissues of stage 7 to 10 top^co homozygotes (see below). Using an antibody against β -tubulin, we likewise detected territories of mitotically active cells in the cephalic, thoracic and ventral ectoderm of stage 10 and 11 mutant animals, as well as small groups of dividing cells that may correspond to PNS precursors in stage 12 Df(2R)top^18A homozygotes (Fig. 5A-F). BrdU-label ling revealed DNA synthesis in the CNS of stage 13 to 14 deficiency homozygotes (data not shown). These results argue that torpedo is not generally required in the embryonic ectoderm for progression through the cell cycle.

Although torpedo is not necessary for cell proliferation in the embryo, it might be required for the specification of cell fate. To determine whether torpedo is needed for the initial establishment of cellular identities in the embryo, we first examined the pattern of the fourteenth nuclear division. Mitosis 14, the first cell cycle under zygotic control, is asynchronous: groups of cells divide in a precise spatiotemporal sequence from stage 7 to early stage 10 of embryogenesis (Foe, 1989). Since alterations in the body plan of the embryo lead to the loss or expansion of certain mitogenic territories, these domains provide sensitive markers for detecting early defects in embryonic cell fate (Foe and Odell, 1989; Rushlow and Arora, 1990). We examined topco embryos stained with Hoechst 33258 (see Materials and methods) for the presence of 25 mitotic domains defined by Foe; each mitogenic territory (except the internal domains 10, 13 and 22, which are not clearly visible in whole-mount preparations) was observed in at least three embryos (Table 2). The two mitotic domains not scored in this experiment, N and M, were seen in antitubulin-stained topco animals (data not shown). As all visible cycle 14 domains appeared to be of normal size and location, we conclude that the initial establishment of cellular identities occurs normally in torpedo amorphs.

We also examined early patterning events in the mutant by analyzing the expression of a variety of genes whose products are spatially and temporally restricted during embryogenesis. To examine patterning along the dorsal-ventral axis, we examined the early expression of the dorsally restricted decapentaplegic (dpp) and zerk niillt (zen) genes (St. Johnston and Gelbart, 1987; Padgett et al., 1987; Rushlow et al., 1987) and the ventrolaterally expressed single-minded (sim) gene (Crews et al., 1988) by whole-mount in situ analysis of RNA transcription. Using an antibody against its protein product, we examined the expression of twist, whose early expression is restricted to ventral cells (Thisse et al., 1987, 1988). All four genes appear to be activated normally in stage 6 to 10 amorphic or null embryos (data not shown).

To assay patterning along the anterior-posterior axis, we stained topco animals with an antibody against the product of the segment polarity gene engrailed which is expressed in the cells comprising the posterior half of each developing segment (DiNardo et al., 1985; Fjose et al., 1985; Kornberg et al., 1985). As was observed by Schejter and Shilo (1989), we detected no abnormalities in the expression of this protein (data not shown). Likewise, the segmentally repeated pattern of larval muscle precursors (Bate et al., 1991) appears normal in stage 11 to 14 embryos homozygous for the null allele Df(2R)top^18A (Fig. 6B, D).

Based on these experiments, we conclude that the first visible defect in the development of torpedo embryos, cell death, is not preceded by any obvious defect in the establishment of the embryonic body plan. The expression of several anteroposteriorly and dorso ventrally restricted genes is initiated properly in the absence of functional torpedo protein, and the sequence of cell division during mitosis 14 appears normal.

To investigate whether torpedo is required for establishment or maintenance of secondary cell fates after the initial body plan is established, we examined the late expression of dpp in mutant animals. dpp shows a complex pattern of transcription during embryogenesis (St. Johnston and Gelbart, 1987; Immergli.ick et al., 1990; Panganiban et al., 1990; Blackman et al., 1991). At the blastoderm stage dpp RNA is found in all dorsal cells. At stage 9, transcription in the head is restricted to three broad domains and message in the thorax and abdomen is limited to dorsal epidermal cells (Fig. 7A). Between stages 9 and 11 of development, dpp ex pression in the ectodermal cells of the head resolves from three broad domains into seven discrete spots (Fig. 7B). At the same time, the dorsal band of dpp expression in the trunk of the animal resolves into two lateral stripes (Fig. 7B); the dorsal stripe marks the boundary between dorsal epidermis and amnioserosa, while the ventral stripe may delimit the boundary between dorsal and ventral epidermis (St. Johnston and Gelbart, 1987). During germband-extended stages clusters of mesodermal cells also express dpp (Fig. 7E). These refined domains of expression persist through stage 13 of development (Fig. 7C, E, G).

At the gross level, the refined transcriptional pattern of dpp in torpedo amorphs is similar to that seen in wild type animals. The seven cephalic ectodermal territories of dpp expression, while often reduced in size relative to wild type, usually can be found in the severely degenerated heads of stage 13 topco animals (Fig. 7D, F). In stage 13 topco animals, both lateral domains of expression are present in the thorax and abdomen (Fig. 7D), as are the mesodermal territories of expression (Fig. 7F).

The distribution of cells expressing dpp in the mutant is not completely normal, however. In the head, domains of dpp expression are shifted relative to one another (Fig. 7D); we believe this is due to the extensive loss of cephalic tissue observed in torpedo amorphs. In the trunk, the separation between the left and right ventral stripes is often drastically reduced (Fig. 7H). By measuring the thoracic area bordered by the more ventral stripes of dpp expression in wild-type and topco animals, we found that mutant ventral thoracic area (which should reflect the number of ventral thoracic cells) is 91% of the wild-type area at stage 11, 87% at stage 12, and 71% at stage 13 (Fig. 7I).

Amorphic animals, therefore, suffer a progressive reduction in the number of ventral epidermal cells during these stages of embryogenesis.

An intriguing alteration in dpp expression is seen in stage 12 to 14 torpedo embryos: ectopic expression of the gene appears in the segmental grooves (Fig. 7D). We consider it unlikely that this aberrant hybridization reflects the nonspecific binding of dpp probe to necrotic cells in this region because ectopic hybridization is not seen in the degenerating cephalic tissues of mutant animals. The abnormal accumulation of transcript may result from the migration of dpp-expressing cells from the dorsal and lateral epidermis into the intersegmental grooves to replace dying cells. Alternatively, ectopic dpp accumulation may reflect a transformation of cellular identities in the vicinity of the segmental grooves.

After the onset of cell death in torpedo embryos, the expression of the sim gene is also slightly altered. In early germband-extended wild-type embryos, sim mess age is present in a continuous row of mesectodermal cells positioned along the ventral rnidline of the embryo; during stage 11 short gaps appear in this stripe of gene expression (Crews et al., 1988; Thomas et al., 1988; Fig. 8A). Some stage 11 topco and Df(2R)top¹⁸A animals show a subtle abnormality in sim expression, larger discontinuities in the midline stripe (Fig. 8B).

Nonetheless, sim expression is maintained in torpedo amorphs, as in wild type, at least until stage 14 of embryogenesis (data not shown).

We have observed one other defect in gene ex pression in torpedo embryos. In late stage 14 wild-type animals, the snail gene is expressed in a posterior, subepidermal group of cells (Boulay et al., 1987; Alberga et al., 1991). This posterior domain of snail expression is never seen in late torpedo amorphs, even though snail message is abundant in thoracic tissues (Fig. 8D).

Thus, although the initial establishment of the embryonic body plan is not perturbed in the complete absence of torpedo activity, defects in the accumulation of several gene products are seen after the onset of ectodermal cell death in torpedo animals. Some alter ations in gene expression, such as the abnormal distribution of thoracic and abdominal dpp-expressing cells and the loss of sim+ and snail+ cells, may arise as secondary consequences of cell loss. Another defect in gene expression, the ectopic accumulation of dpp in cells along the segmental grooves, may occur in response to a cell fate change.

Despite the torpedo product’s structural similarity to the mammalian EGF receptor, it does not appear to be generally required for cell cycle progression in the Drosophila embryo. In animals homozygous for complete loss-of-function torpedo alleles, all three epider mal postblastoderm mitoses, as well as cell division in the central and peripheral nervous systems, can be observed. torpedo is, however, required for the viability of certain populations of embryonic cells; increased cell death, which begins late in stage 10 of embryogenesis and continues throughout embryonic development, is one of the earliest visible defects in torpedo mutants. Cell death is largely restricted to the amnioserosa, the epidermis and neural tissues inside the head. Within the ectoderm, it seems that the head and thoracic segments are more severely affected than the abdominal segments, and the ventral hypoderm is more reduced than the dorsal hypoderm. Most, if not all, of the affected cells are mitogenically quiescent: the amnioserosa is a nondividing tissue, one population of epidermal cells completes its final round of division during stage 10 of embryogenesis, and the remaining epidermal cells undergo their last mitosis during stage 11 (Campos Ortega and Hartenstein, 1985).

Reduced cell viability observed in torpedo animals may occur in response to a variety of physiological defects. One possibility is that cell death in the mutant arises as the secondary consequence of a defect in the establishment or maintenance of cellular identity. Alterations in cell fate resulting from mutations in the segment polarity genes, for example, lead to increased embryonic cell death (Martinez Arias et al., 1988). However, we do not observe any obvious defects in the establishment of the segmented body plan in torpedo amorphs; the pattern of spatially and temporally asynchronous mitosis that occurs during division four teen is normal in the mutant, as are the expression patterns of the early zygotic genes dpp, zen, sim, twist and engrailed. The late expression patterns of the majority of these genes also appear normal in the mutant. Alterations in the distribution of some cells expressing the dpp, sim and snail genes in torpedo embryos occur only after the onset of cell death, and therefore may be byproducts of this defect. While our observations do not support the idea that torpedo is required for the specification or maintenance of cell fate in the early embryo prior to the onset of cell death, it is conceivable that the receptor mediates intercellular signalling required for fine scale patterning within segments that cannot be assayed with the molecular probes we have used.

A possibility that we find attractive is that the torpedo gene product transduces a maintenance signal in the embryonic ectoderm. Studies employing mammalian cell and tissue culture suggest that receptor tyrosine kinases, including the EGF receptor, may play a role in the transmission of signals that promote the survival of neurons (Morrison et al., 1986, 1987; Warlicke et al., 1986; Hempstead et al., 1991; Kaplan et al., 1991; Klein et al., 1991; Squinto et al., 1991). Likewise, the viability of nonneural tissues may depend upon intercellular signals mediated by receptor tyrosine kinases. NIH 3T3 murine fibroblasts, under certain culture conditions, require growth factors for their survival (Zhan and Goldfarb, 1986; Glass et al., 1991).

In addition to increased cell death, animals homo zygous for total loss-of-function torpedo alleles produce fewer and smaller ventral setae than wild type. Analysis of the conditional allele top^{1 F26} strongly argues that the setal defects seen in the mutant are not the secondary consequences of cell death in the ventral hypodermis. Temperature-shift experiments employing top^1F26 further show that the requirement for torpedo in denticle production is postmitotic.

The loss and abnormal morphogenesis of denticles in torpedo amorphs could reflect a reduced ability of terminally differentiating ventral hypodermal cells to secrete denticle material. As the result of a defect in cytoskeletal reorganization or cellular metabolism, large setae could be reduced in size and small denticles could disappear altogether in torpedo embryos. Like wise, the delayed initiation or premature termination of denticle synthesis could result in a global reduction in denticle size. Alternatively, the denticle abnormalities seen in torpedo mutants may result from a transformation of hypodermal cell fate. If cellular identities within the segments were shifted toward more anterior values, the amount of naked cuticle would expand and only small denticles characteristic of the most anterior rows of setae would be produced.

Temperature-shift experiments demonstrate that torpedo acts at least twice during embryogenesis. The phenocritical period for inducing the loss of anterior and ventral cuticle and arresting germband retraction in the conditional allele top^1F26 extends roughly from stages 8 to 11 of embryonic development. Given that the onset of cell death in torpedo mutants correlates well with the temperature-sensitive period for these morphological abnormalities, it is likely that cell death is responsible not only for the loss of cephalic, thoracic and ventral cuticle but also for the twisted germband extension and arrested germband shortening seen in total loss-of-function embryos.

Not every aspect of the torpedo embryonic lethal phenotype results from the early loss of gene product. The temperature-sensitive period for generating den tide abnormalities, late stage 11 to early stage 13 of embryogenesis, is temporally distinct from that for germband shortening and cuticle deposition. Moreover, since embryos downshifted from the nonperm1ss1ve temperature before this phenocritical period show severe reductions in the amount of ventral cuticle but still produce some denticles of wild-type size, the requirements for cell viability and denticle production are functionally separable.

Molecular studies have shown that torpedo is expressed in a dynamic pattern during embryogenesis (Schejter et al., 1986; Kammermeyer and Wadsworth, 1987; Zak et al., 1990; Katzen et al., 1991). Defects observed in mutant animals generally correlate well with the expression pattern of the gene.

The most striking abnormality seen in torpedo mutants is increased cell death in the amnioserosa and epidermis. Of epidermal tissues, those of the head, thorax and ventral abdomen, are most severely affec ted. torpedo is highly expressed in both the amnioserosa and ectoderm until the germband shortening stages, with mRNA and protein being particularly abundant in cephalic structures. After germband retraction torpedo protein persists at low levels throughout the epidermis and remains abundant in the clypeolabrum, terminal hindgut and in the vicinity of the segmental grooves. Although ventral cuticle is more sensitive to a reduction in torpedo activity than is dorsal cuticle, there is no obvious difference in the level of gene product in these two populations of cells. (Schejter et al., 1986; Kammermeyer and Wadsworth, 1987; Zak et al., 1990; Katzen et al., 1991). torpedo RNA and protein are present at high levels in the telson in germband shortened embryos (Zak et al., 1990; Katzen et al., 1991). Interestingly, mutant embryos lack the posterior domain of snail gene expression that normally appears during stage 14. We do not known whether this defect reflects the death or reprogramming of cells.

Around the time of germband shortening, torpedo is necessary for the production of ventral setae. torpedo protein is found throughout the epidermis during stages 12 and 13, with highest levels being in the segmental grooves (Zak et al., 1990). In torpedo amorphs, during the period of germband retraction, we have observed an intriguing ectopic accumulation of dpp message along the segmental grooves.

Shilo and coworkers have defined a requirement for torpedo in the embryonic central nervous system. As neurons do not express torpedo protein, the nervous system abnormalities seen in torpedo mutants might arise as a secondary consequence of a defect in glial cell viability or differentiation (Schejter and Shilo, 1989; Zak et al., 1990).

torpedo is expressed in some tissues of the embryo that do not show an obvious requirement for the gene product. torpedo mRNA and protein are found in the visceral and splanchnic mesoderm throughout embryo genesis (Kammermeyer and Wadsworth, 1987; Zak et al., 1990; Katzen et al., 1991). Using an antibody against the twist protein, we could detect no abnormali ties in mesodermal development in amorphic animals.

Likewise, torpedo is expressed in dorsal epidermis, yet complete loss-of-function animals produce seemingly normal dorsal cuticle. The apparent lack of a requirement for torpedo in these tissues can be explained in several ways. One possibility is that the gene product is functionally redundant in certain tissues. Alternatively, the ligand for torpedo may be spatially restricted within the embryo, so that the protein is activated only in a subset of cells that express it. Another possibility is that torpedo may be required in one tissue for the proper development of another cell population. For example, activation of the torpedo receptor tyrosine kinase in mesodermal cells may modulate the expression of a signalling molecule that regulates the development of adjacent epidermal cells.

Phenotypic analysis of mutations at the torpedo locus reveals that the gene performs multiple functions during development. In the ovary, torpedo serves to establish or maintain dorsal follicle cell identities (Schupbach, 1987). During larval development, the gene product regulates the proliferation or viability of imaginal disc cells (Clifford and Schupbach, 1989; Dfaz Benjumea and Garcia-Bellido, 1990; our unpublished observations). In the pupa, torpedo acts to specify cellular identities in the imaginal disc derivatives. Animals of many heteroallelic mutant genotypes exhibit bristle duplications and the loss of specific wing veins (Clifford and Schupbach, 1989), and the pheno type exhibited by Ellipse1, a dominant gain-of-function mutation at the torpedo locus, suggests that the gene product could function in the developing compound eye to mediate the switch from an undifferentiated, pro liferative cell state to a differentiated, nondividing state (Baker and Rubin, 1989). Our present work shows that torpedo even can act at multiple times in the development of a single tissue.

The range of cellular responses elicited by the torpedo gene product argues that the receptor triggers different downstream signalling pathways in the various tissues. Vertebrate receptor tyrosine kinases interact with multiple cellular substrates (reviewed in Ullrich and Schlessinger, 1990; Cantley et al., 1991); therefore, it is likely that the torpedo proteins also phosphorylate a number of substrate molecules. How a cell responds to the activation of torpedo might depend upon which primary and/or secondary substrates it expresses, and its complement of substrates may change as cellular differentiation proceeds. Genetic screens for suppres sors and enhancers of torpedo mutations provide a powerful means of identifying general and tissue specific downstream elements of the torpedo signalling pathways. This experimental approach has recently implicated the products of the Son of sevenless gene, which encodes a CDC25 homolog, and the Rasl gene as components of the torpedo signal transduction pathway in the compound eye (Rogge et al., 1991; Simon et al., 1991; Bonfini et al., 1992). The identification and characterization of additional interacting loci should prove useful for dissecting the molecular mechanisms by which torpedo and other receptor tyrosine kinases mediate cell proliferation, survival and differentiation.

We thank Yash Hiromi for the CyO, Cl chromosome, Nipam Patel for the engrailed antibody and Rob Ray for the plasmid containing dpp sequences. We thank Siegfried Roth for providing the twist monoclonal antibody, single-minded and snail probes and technical advice. We are grateful to Dari Sweeton for her generous help in preparing sections. Our appreciation is extended to past and present members of the Princeton fly groups for stimulating discussions and helpful advice on various aspects of this work, and we especially thank Valerie Lantz, Siegfried Roth, Eyal Schejter and Eric Wieschaus for their critical reading of this manuscript. We also thank Gordon Gray and Cubit Case for preparing fly food. This work was supported by grants from the National Institutes of Health (GM 40558) and the National Science Foundation (BNS 8616928). R. C. was supported, in part, by an NIH Genetics Predoctoral Training Grant to the Princeton University Department of Molecular Biology.