The torso pathway in Drosophila: a model system to study receptor tyrosine kinase signal transduction

In Drosophila, the combination of classic embryological studies and genetic analyses have uncovered four different systems that establish the body coordinates of the embryo (see reviews by Nusslcin-Volhard et al., 1987; St. Johnston and Nusslein-Volhard. 1992; Fig. I). The establishment of dorsal-ventral polarity dorsal-ventral polarity utilizes a single patterning system. while cellular determination along the anteroposterior axis invoIves three independent patterning systems, the anterior. Posterior and terminal systems. Determinants for each of the four patterning systems arc synthesized during oogenesis 4 : their mRNAs or proteins are localized in the egg near their prospective sites of action (Fig. 1).

The anterior and posterior patterning systems use RNA. localization in the egg cell as a mechanism to generate two distinct morphogenetic gradients. The anterior system specifies the gnathal and thoracic regions and the posterior system specifics the abdominal region. The maternal mRNA for the bicoid (bcd) gene, the key member of the anterior system, is localized anteriorly in the egg and is responsible for a morphogenetic gradient that controls the domains of expression of early zygoticgenes (Driever and Nusslein-Volhard. 1988a.b; Struhl el 1989; Finkelstein and Perrimon. 1990). Located posteriorly in the egg. the maternal nanos (nos) mRNA generates a gradient of nos protein which defines a region in which the abdominal body pattern can be specified by other morphogens (Hulskamp et al., 1989; Struhl, 1989; Irish et al., 1989; Wang and Lehmann. 1991; Barker et al., 1992).

The terminal and dorsal-ventral systems utilize a similar strategy to organize patterns; i.e.. in both cases signal transduction pathways are triggered in the egg cell in response to ligands localized in the pcriviteline space (Stein and Nusslein-Volhard. 1991; Sprenger and Nusslcin-Volhard, 1992; Fig. 1). In each system the receptor, as well as the maternally provided downstream signal transducers, are uniformly distributed in the egg; but the signals are localized. The terminal system is required for the formation anteriorly of the acron and posteriorly of the 8th abdominal segment and the telson (as defined in Nusslein-Volhard et al., 1987). Genes of the dorsal-ventral system determine a system of positional information that results in the graded nuclear distribution of dorsal protein (Roth et al., 1989; Rushlow et al., 1989; Steward. 1989).

The establishment of terminal cell fate is under the control of a group of genes known as the terminal genes, in this paper, we review the molecules encoded by these genes and describe their respective roles in patterning. In addition, we describe the interactions of the terminal pathway with the anterior and dorsal-ventral systems that coordinately control the spatial expression of early transcription factors.

Torso (tor): Mutations in the tor gene were first isolated in screens for recessive female sterile (fs) mutations (Schupbach and Wieschaus. 1986a). Embryos derived from mothers homozygous for tor null or loss of function (lof) mutations (referred to as tor^10f embryos) do not develop the most anterior (acron) and posterior (telson and abdominal 8 segment) terminal structures (Fig. 2B). These defects have been shown to correspond to a reorganization of the blastoderm cell fates rather than a lack of the proper differentiation of the terminal structures (Nusslcin-Volhard et al., 1987: Mlodzik et al., 1987). In other words, terminal cells in tor^lof embryos adopt more central cell fates resulting in the deletion of entire developmental units, the terminal anlagen. Posteriorly, the deletion in a tor^lof embryo encompasses a region on the blastoderm fate map extending from the posterior pole (0% egg length. EL) to 20% EL. This is the region of the embryo fated to produce all structures posterior to the 7th abdominal segment, including the 8th abdominal segment, malpighian tubules, anal pads, posterior spiracles, hindgut. and posterior midgut. Anteriorly, tor^lof mutations affect the development of the clypeolabrum, part of the cephalopharyngeal skeleton, optic lobes and procephalic lobes (Nusslcin-Volhard et al., 1987; Schupbach and Wieschaus, 1986b). Thus, the mutant phenotype of tor^lof embryos indicates that it is required for the specification of embryonic terminal domains.

tor encodes a putative transmembrane receptor tyrosine kinase (RTK) (Fig. 3). Neither tor mRNA nor protein is spatially restricted to the embryonic poles (Sprenger et al., 1989; Casanova and Struhl, 1989) suggesting that, to specify terminal development, tor is only activated at the embryonic termini. Hyperactive or gain of function (gof) tor mutations (tor^gof) have been isolated in which tor is ubiquitously activated in embryos independent of ils ligand. In such tor^gof embryos, more cells adopt the terminal fate al the expense of cells of the central cell fate (Klingler et al., 1988; Strecker et al., 1989; Sprenger and Nusslcin-Volhard. 1992; Sprenger et al., 1993). Consequently, thoracic and abdominal structures arc variably, or in extreme cases entirely deleted (Fig. 2C). The molecular nature of tor and the complementary phenotypes associated with tor^lof and tor^gof mutations suggest that tor specifies terminal cell fates through localized activation by its ligand at the embryonic poles.

D-raf (also known as l(l)pole hole): Mutations in the D- raf gene were first identified as zygotic lethal mutations associated with specific maternal effect phenotypes resembling that of tor^lof mutations (Perrimon et al., 1984, 1985). Identification of such mutations relics on the generation of homozygous mutant germlines in heterozygous viable females (Perimon and Gans, 1983). Two phenotypic classes ..re observed among D-raf embryos derived from mutant germlines (Perimon et al., 1985; Ambrosio ct al., 1989a). ‘ D-raf rescued’ embryos, corresponding to those that do not have maternal D-raf activity but have received a wild-type D-raf gene paternally, have a phenotype (.lentical to that of tor¹”! embryos (Fig. 2D). “D-raf null” embryos, corresponding to those that do not have either maternal or zygotic D-raf activity and show little cuticle differentiation (Fig. 2H). Both classes of D-raf embryos exhibit similar defects m terminal cell fate determination as shown by the expression of early segmentation genes in these mutant xickgrounds (Ambrosio et al., 1989a; Melnick et al., 1993). The poor cuticular differentiation of D-raf null embryos. decís the zygotic requirement of D-raf in other embryonic developmental processes that do not involve tor.

D-raf encodes a protein serine/threonine kinase which shows 46% identity al the amino-acid level to the human Raf1 kinase, and as much as 65% identity al the C-terminal protein kinase (CR3) domain (Mark et al., 1987; Nishida ct. 1988; Melnick et al., 1993: Fig. 3). At the N-terminal half of the molecule. D-raf shares structural features with e human Raf-I kinase: the cysteine zinc finger (CRI) and serine/threonine rich (CR2) regions. Like tor mRNA, D-raf mRNA is not spatially restricted to the embryonic poles (Ambrosio et al., 1989b).

Tatlless (tll): tll was the first gene characterized among ubryonic lethal genes that zygolically affects terminal cell fate determination (Strecker et al., 1986). Deleted in tll mutant embryos arc structures on the blastoderm fate map from approximately 7– 20% EL. including A8. the malpighian tubules, anal pads, posterior spiracles and hindgut (Fig. 2L). The posterior midgut, which originates from the most posterior region of the blastoderm fate map. is only variably affected in tll mutant embryos (Strecker cl 1988). The anterior defects associated with tll mutations are similar to those seen in tor^lofembryos (Strecker et al., 1986,1988).

The tll gene encodes a putative transcription factor of the steroid hormone receptor superfamily (Pignoni et al., 1990). The protein contains putative ligand binding as well as DNA binding domains (Fig. 3). In wild-type embryos at the Syncytial blastoderm stage, tll transcripts are expressed symmetrically from 0–20% and 80–100% EL; however, by the cellular blastoderm stage, tll expression resolves into smaller domains with an anterior “horseshoe-like” pattern extending over the dorsal side of the embryo and a posterior lain extending from 0–15% EL (Pignoni et al., 1990. 1992). The nature of the tll gene product and its restricted terminal expression pattern suggests that tll specifies terminal fate by acting as a transcription factor to turn on terminal-specific gene expression.

The phenotypic similarities between mutations in tor, D-raf and tll suggested that they may be components of the same developmental pathway that results in the differentiation of the most terminal regions of the embryo. This model was supported by the genetic episialic analyses between tor^gof and either D-raf or tll loss of function mutations (summarized in Fig. 4). tor^gof embryos exhibit repression of thoracic and abdominal segmentation with concomitant expansion of both the acron and telson regions. Embryos derived from germlines carrying both tor^gof and D-raf^lof mutations exhibit the D-raf^lof phenotype indicating that D-raf activity is necessary for tor signaling (Ambrosio et al., 1989b). Likewise, the maternal tor^gof phenotype can be suppressed by the removal of zygotic tll function (Klinger et al., 1988). These genetic interactions indicate that the effect of tor^gof mutations can be negated by mutations in cither D-raf or tll. These results led to the model that tor activates D-raf. which in turn controls the zygotic transcriptional activation of tll. Consistent with this model, the tll mRNA is not expressed posteriorly in either tor^gof or D-raf embryos (Pignoni ct al.. 1992; Melnick et al., 1993). Conversely, in strong tor^gof animals, tll is ectopically expressed along most of the anteroposterior axis of the egg (Stcingrimsson et al., 1991: Fig. 4).

It should be noted that control of tll expression is more complex in the anterior region of the egg. In both tor^lof and D-raf^lof embryos, tll is expressed al the anterior pole; however, it docs not evolve into the wild-type horseshoelike pattern (Pignoni et al., 1992: Fig. 4). Anteriorly, regulation of tll expression is under the control of the terminal, anterior and the dorsal/vcntral systems (Pignoni et al., 1992; see below for further discussion).

Genetic screens for female sterile mutations identified four additional maternal “terminal class” genes: fs( I )Nasrat (Nas:Degelmann et al., 1986). fs(l)pole hole (fs(l)ph:Perrimon et al., 1986). fs(2)trunk (trk:Schupbach and Wieschaus, 1986a). and fs(3)torso-like (tsl: Nusslein- Volhard et al., 1987). When any one of these gene activities is missing during oogenesis, embryos develop with a tor^lof phenotype. The molecular nature of the products of these genes is as yet unknown: however, in genetic epistalic analyses these mutations are unable to suppress the tor^lof phenotype, suggesting that (hey function upstream of tor. and are most likely involved in the activation of the RT K. Analyses of the tissue specificity of these mutations have provided some insights into their possible roles in tor activation. Three of the upstream genes. Nas. tor^lofh and trk. are germline dependent (Perrimon et al., 1986; Schupbach and Wieschaus. 1986b). while tsl has been shown by mosaic analysis to be required in a small number of follicle cells located at both ends of the egg (Stevens et al., 1990). Thus, tsl is likely to play a key role in restricting the activation of toral the termini (sec below for further discussion).

corkscrew (csw): Like D-raf. mutations in the csw gene were identified in screens for zygotic lethal mutations that exhibit a specific maternal effect on termini differentiation. Genetic epistasic analysis has shown that CM operates downstream of tor to positively transduce the terminal signal (Perkins et al., 1992). However, unlike D-raf. csw mutations only partially block tor signaling. This was determined by comparing tll expression in tor^lof, D-raf^lof and csw^lof mutant embryos. Posterior tll expression is completely eliminated in tor^lof and D-raf^lof mutant embryos (Pignoni et al., 1992; Melnick et al., 1993). whereas in csw^lof mutant embryos posterior tll expression is reduced (Perkinset al., 1992). Posteriorly, csw^lof embryos are missing only the posterior midgut or that part of the telson which originates from the most posterior region of the blastoderm late map (Fig. 2J).A weak D-raf allele with residual activity, D-raf^lof also ‘hows this phenotype (Melnick et al., 1993). Interestingly, embryos doubly mutant for both D-raf^lof and csw^lof arc imilar to tor^lof embryos, suggesting that csw and act concert to transduce the tor signal (Perkins et al., 1992).

Molecular analysis of the csw gene and its products revealed that csw encodes a putative nonreceptor protein tyrosine phosphatase covalently linked to two N-terminal Src homology (SH) 2 domains (Perkins et al., 1992: Fig. 3). The csw protein is most similar to two mammalian proteins. PTPIC and SH PTP2 (also known as Syp or PTPID. Fenget al., 1993; Vogel et al., 1993) with which it shares 69 and 76 % similarity, respectively (Perkins et al., 1992; Freeman al.. 1992). While the tyrosine phosphatase domain implies enzymatic function for csw in transducing the tor signal. the presence of the SI 12 domains suggest that csw effects its role by mediating heteromeric protein interactions. Both PTPIC and SH-PTP2 have been shown to associate with cytoplasmic portions of activated mammalian growth factor receptors. PTPIC associates with the EGF receptor (Shen et al., 1991) and SH-PTP2 associates, through its SH2 domains, with the EGF and PDGF receptors, as well as chimeric receptors containing the cytoplasmic domains of the HER2-HCU, and kit-SCF receptors (Feng et al., 1993; Vogel et al., 1993; B. Neel, personal communication). Further genetic epistasis experiments arc required to precisely place <wm within the tor signaling pathway.

Ras I and Son of seven less (Sos): Recently. Lu et al. (1993) have devised a biochemical approach to demonstrate that the p21^ras GTPase, encoded by Rasl (Fig. 3). mediates tor signaling. The approach involves the injection of either activated or dominant negative forms of putative signal transducing molecules into the early, syncytial embryos. When activated p2l^ras protein, p2l^v-ras, was injected into tor^lof embryos, the embryos recovered their terminal structures suggesting that activation of p2 I ^ras/Ras I is sufficient to specify terminal identity in the absence of the tor receptor (Lu et al., 1993; Fig. 2K). This result was extended by injecting a dominant negative form of p21^ras. p2l^rasNI7. into wild-type embryos. In this experiment the injected embryos showed a terminal class phenotype similar to that of csw^lof mutant embryos. In addition. Lu et al. (1993) demonstrated that p21^ras/RasI functions upstream of D-raf in this pathway, since injection of activated p2l^v-ras into D-raf”¹ embryos did not show a rescuing activity.

Further support that p2l^ras/Rasl functions in the tor signaling pathway came from the analysis of the maternal effect phenotype of Son of sevenless (Sos), where Sos^lof embryos derived from mutant gcrmlincs show a terminal class phenotype reminiscent to tor^lof and D-raf^lof phenotypes (Lu et al., 1993; Fig. 2E.F). Sos encodes a guanine nucleotide releasing (or exchange) factor which is known to positively regulate p2l^ras/Rasl function by stimulating the accumulation of the active GTP-bound form of the protein (Bonfini et al., 1992; Simon et al., 1991; Fig. 3).

Dsorl (MAPKK): In a screen for suppressors of a reduced activity D-raf allele, a gene with 65% homology to the mammalian MAP kinase kinase (MAPKK also known as MEK. Crews et al., 1992) was isolated (Tsuda et al., 1993; Fig. 3). Although this suppressor, named Dsorl. is a gain of function mutation, it has been shown to be a normal component of the tor pathway based on the observation that embryos derived from gcrmlincs that lack wild-type Dsorl activity have a terminal class phenotype (Tsuda et al., 1993; Fig. 2G.I). In tor^gof or D-raf^lof mutant backgrounds, activity from the gain of function Dsorl mutation is sufficient to activate posterior tll expression, indicating that MAPKK operates downstream of both tor and D-raf. This is consistent with studies from mammalian cells where MAPKK has been shown to be a direct substrate of Raf-1 kinase (Kyriakis et al., 1992). The high level of homology of Dsorl to mammalian MAPKK suggests the likelihood that a MAP kinase is also implicated in tor signaling.

In addition to tll, hnckebein (hkb) has been shown Io be another zygotic gene that responds to tor signaling (Weigel et al., 1990). In wild-type, hkb transcripts are expressed in caps at both anterior and posterior ends of the blastoderm stage embryo. Posteriorly, this expression extends from 0 to 12% EL. or that region destined to form the posterior midgut (Bronner and Jackie. 1991). Posteriorly, in hkb mutant embryos only the posterior midgut is deleted (Weigel et al., 1990; fig. 2M), which is the only posterior structure not completely deleted in tll mutant embryos. Both the coextensive nature of the tll and hkb mutant phenotypes and the observation that tll-hkb double mutant embryos are phenolypically similar to tor^gof mutant embryos (Weigel et al., 1990) suggests that posteriorly tll and hkb arc sufficient to mediate the maternal activity of tor. The hkb gene has been shown to encode a putative transcription factor with multiple zinc lingers (H. Jackie, personal communication).

The terminal gap genes tll and hkb control the expression of more “downstream terminal” genes. Presently, these include the transcription factors; hunchback (Tautz. 1988). fork head (Weigel et al., 1990).J’ushi-larati (Mlodzik et al., 1987), the regulatory clement of the homeolic gene Abdominal II (Casanova et al., 1986). and Kruppel (Rosenberg et al., 1986). These genes are expressed at the posterior terminus of the embryo and their expression patterns are perturbed in embryos mutant for upstream terminal genes (see review by Perkins and Perrimon. 1991). These transcription factors most likely control the expression of further downstream genes involved in cellular differentiation of terminal structures.

Fig. 5 summarizes our current understanding of the tor signaling pathway. Four genes, tsl. trk. Nas and fs(l)ph, operate upstream of /or and arc probably involved in the production of an active tor ligand. Undoubtedly, the four loci encode only a subset of the proteins required for the generation, transport and secretion of the tor ligand. Three of these loci arc required in the germline, while one (tsl) is required in the surrounding follicle cells, suggesting that intercellular communication is required to activate tor.

The molecular nature of the tor ligand remains elusive. However. Sprcnger and Nusslein-Volhard (1992) have shown that it is diffusible in the perivitelline space and present in limited quantity al the egg poles. Because the tor RTK is uniformly distributed in the egg membrane, a limited amount of the diffusible ligand originating from the termini is crucial for the activation of tor within the embryonic terminal domains where terminal cell fates become specified, tsl. because of its requirement in a small number of follicle cells located at both ends of the oocyte, can either encode the tor ligand, or alternatively be involved in the activation of the tor ligand (Stevens et al., 1990). In the first .ase. an inactive form of tsl protein may be localized to the terminal regions of the vitelline membrane prior to fertllization. Following fertllization, secretion into the periv- itelline space of an activating molecule, possibly encoded by one of the three germline dependent maternal genes, will cad to the release of a diffusible activated tsl protein. In the second case, following fertllization an inactive form of the tor ligand may be secreted into the perivilclline fluid, and subsequently become activated by a molecule (possibly encoded by tsl) localized to the terminal regions of the vitelline membrane.

Activation of tor triggers a signal transduction pathway mediated by the proteins Sos and Ras I which in turn activates a phosphorylation cascade through the protein kinases D-raf and MAPKK (Dsorl). Currently, the transducers of tor and their regulatory mechanisms are similar, if not identical, to those found for other RTKs; c.g. Sevenless sev) in Drosophila (Simon et al., 1991; Dickson et al., 1992a), and PDGF and EGF growth factor receptors in mammals (review’s by Ullrich and Schlessinger. 1990; Camley et al., 1991). The guanine nucleotide releasing factor. Sos. transduces the signals generated by both tor and sev to Ras I. and activation of Ras I is sufficient to initiate either terminal development in early embryos or R7 cell differentiation in the eye imaginai disc (Fortini et al., 1992; Lu cl al., 1993). Dickson et al. (1992b) have shown that for the development of the R7 photoreceptor cell the kinase domains of either tor and sev are functionally equivalent. This indicates that the specificity of different RTKs is likely determined by the developmental potential of the cells in which they are activated. The activation of tll and hkb in early embryos may be explained by the availability of a specific transcription factor, postulated as l’eue Y (St. Johnston and Nuslein-Volhard 1992) which, once phosphorylated in response to tor signaling, activates transcription from the tll and hkb promoters.

Little is known about how the overlapping spatial domains of ill and hkb expression are achieved. Previous studies (Casanova and Slruhl, 1989) have proposed that graded levels of tor signaling are required to direct pattern formation within the posterior unsegmented terminal anlagen (O-2% EL). This gradient of tor activation is thought to be generated by diffusion of localized tor ligand originating from the poles. However, we have observed that posterior terminal structures were rescued at their normal posterior position even when activated p21^ras/RasI was injected anteriorly or in the central region of tor^1of embryos. These results suggest that embryos devoid of the tor RTK still possess an intrinsic polarity for tail formation. This polarity arises independently of tor activation somewhere downstream of p21^ras /Ras I. These observations are consistent with the recent postulation by Sprenger and Nusslein- Volhard (1992) that “activated tor might act more like a switch, triggering terminal development after a threshold level of tor activity is achieved”, and “gradient(s) of cytoplasmic molecules” would then be responsible for subdividing the terminal anlagen.

Even though the four maternal patterning systems are separable genetically, they interact al the molecular level. For cells at the termini, three systems (anterior or posterior, terminal and dorsal-ventral) converge to establish the positional values of individual nuclei. These interactions have been especially studied at the anterior end. Many genes activated by bicoid (bed), such as orthodenticle and hunchback, arc repressed in the anterior tip by the tor signaling pathway (Taut/.. 1988. Finkelstein and Perrimon. 1990). This is also the case for snail, a gene activated by the dorsal (dl) morphogen gradient on the ventral side of the embryo (Casanova. 1990; Ray et al., 1991). Similarly, the anterior tll expression pattern is also influenced by the anterior (bcd) and dorsal-ventral (dl) systems. Pignoni et al. (1992) have shown that the anterior “horseshoe-like” pattern of til expression al the late blastoderm stage is the result of positive activation by intermediate levels of bed and tor activities phis repression at a high level of tor. bed and dl concentrations.

We have described the molecular nature and roles of each of the known components involved in the tor signaling pathway. There are certainly more terminal genes that remain to be identified and one approach to identify these genes is to pursue the isolation of mutations associated with terminal defects. Although genetic screens for recessive female sterile and embryonic lethal mutations arc near saturation. previous searches for zygotic lethal mutations associated with specific maternal effect phenotypes have so far only been conducted for the X-chromosome (Perrimon et al., 1989; Chou and Perrimon. 1992). We expect that many other genes will be identified, as were D- raf and csw. once screens to detect (he maternal effects of essential genes are extended to the remaining of the genome. Another powerful approach involves the isolation of mutations that modify (by suppression or enhancement) the phenotypes of a preexisting mutations, l or example. Dsor-1 was identified as a supressor of a weak D-raf allele (Tsuda et al., 1993). Such interaction screens allow one Io isolate essential genes with multiple functions throughout development, because the inactivation of one copy of the gene is often enough to show interactions in a sensitized genetic backgound (Simon et al., 1991). Finally, complementary approaches such as the injection into embryos of various molecules can be used to est their putative roles in tor signaling. Good candidates for this approach include previously identified components of other RTKs signaling pathways.

Once new components are identified, it will be important :o address their necessity and sufficiency in signaling. Such analysis, in combination with detailed biochemical analysis, will allow a full understanding of the role each component plays in tor RTK signaling. The tor signaling pathway, with its unique opportunity to combine genetic, cytoplasmic injection of molecules and biochemical approaches, provides an excellent model system not only to identify the components involved in RTK signaling but also to understand the epistatic and biochemical relationships between the various signaling molecules.