Salivary duct determination in Drosophila: roles of the EGF receptor signaling pathway and the transcription factors Fork head and Trachealess

The DER signal transduction pathway, like many other signaling pathways, is used at many stages in development to change the fate of cells. It patterns the ventral epidermis and the CNS midline in the embryo (Klämbt, 1993; Raz and Shilo, 1993; Kim and Crews, 1993; Rutledge et al., 1992), specifies formation of veins in the wings and photoreceptors in the eyes of adults (Freeman, 1994; Kolodkin et al., 1994; Heberlein et al., 1993; Diaz-Benjumea and Hafen, 1994; Sturtevant et al., 1993), and is required for determination of both the anteroposterior (AP) and dorsoventral (DV) axes in the oocyte (Gonzalez-Reyes et al., 1995; Roth et al., 1995). To achieve different results in these tissues, the DER pathway must interact with different sets of regulatory genes in each tissue. To learn more about DER interactions in a particular tissue, we have studied the effects of DER signaling on the determination of salivary ducts. Near the end of germ band extension in the Drosophila embryo, a binary decision distinguishes the precursors of the salivary glands (pregland cells) from the precursors of the salivary ducts (preduct cells). We show here that this decision is mediated by at least three developmental regulators: the DER signaling pathway and transcription factors encoded by two genes, fork head (fkh) and trachealess (trh). We focus on the interactions among these regulators that distinguish the two cell types and establish a boundary between them.

The salivary primordia, one on each side of the embryo, are first detectable at about 4 hours of development (stage 10) as germ band extension is nearly finished. As shown by fate mapping experiments (Campos-Ortega and Hartenstein, 1985) and by the early expression of the dCREB-A gene (Smolik et al., 1992; S. K. B., unpublished data), these primordia, each consisting of about 100-120 cells, occupy most, if not all, of the ventral epidermis of parasegment 2. By about 6.5 hours (mid stage 11), each primordium has been divided into two distinct domains. The more dorsal 80-90 cells make up the salivary placode, the precursor of the salivary gland itself (Sonnenblick, 1939; Panzer et al., 1992). These pregland cells assume a more columnar shape than their neighbors and express several genes, including fork head (Weigel et al., 1989a; Panzer et al., 1992), huckebein (Panzer, 1994; Brönner et al., 1994) and Toll (Gerttula et al., 1988), that are not expressed in the more ventral cells. The remaining 20-30 cells ventral to the placode on each side of the embryo are the precursors of the salivary ducts (B. Zhou, unpublished observations). These cells express several genes that are specific to preduct rather than pregland cells including Serrate (Ser) and breathless (btl) (Fleming et al., 1990; Thomas et al., 1991,Klämbt et al., 1992).

After the pregland and preduct cells have been distinguished, they follow different pathways of morphogenesis. First, the cells of the salivary placodes invaginate, starting with the cells at the posterior, dorsal edge of the placode and continuing in a defined order until all placode cells have left the surface of the embryo and have formed the simple tubular glands (Panzer et al., 1992; B. Zhou, unpublished observations). The duct precursors are then internalized making a tube that has a much smaller diameter than the glands and that branches to produce the Y-shaped ducts that join the glands to the pharynx.

The first regulator of duct development we have studied, the DER signaling pathway, includes not only the Drosophila EGF receptor itself, but also several other membrane proteins and a transcription factor that seem to be uniquely associated with DER signaling. The spitz-group genes, which encode these DER-pathway-associated proteins, were initially identified because embryos mutants for each of them have similar embryonic phenotypes: ventral pattern elements were removed from the epidermis and the commisures of the ventral nerve cord were reduced or missing (Mayer and Nüsslein-Volhard, 1988). Genetic interactions between spitz-group and DER mutations then linked the spitz-group to the DER pathway (Klämbt et al., 1991; Diaz-Benjumea and Hafen, 1994; Freeman, 1994; Heberlein et al., 1993; Kolodkin et al., 1994; Sturtevant et al., 1993). Four of these genes are relevant here. The spitz (spi) gene encodes a transmembrane protein that is a member of the TGFα family, suggesting that it may act as a ligand for DER (Rutledge et al., 1992). It is likely that the active form of Spitz is the cleaved extracellular domain since constructs expressing only this domain were able to activate the DER pathway in either cultured cells or in embryos, whereas similar constructs expressing the membrane-bound form did not activate it (Schweitzer et al., 1995). Activity of the cleaved and possibly diffusible form of Spitz is consistent with evidence that signaling through the DER pathway can be spatially graded, being highest at the ventral midline in the embryo and decreasing dorsally (Raz and Shilo, 1993; Schweitzer et al., 1995). Such a graded pattern of signaling could be set up if Spitz were cleaved ventrally but could diffuse dorsally for several cell diameters.

In the embryo, both spi and DER are broadly expressed, so that localization of DER activity is likely to rely on other members of the spitz-group, in particular rhomboid (rho) or Star (S). Both of these encode transmembrane proteins of unknown function and the expression of both is enriched at or adjacent to the ventral midline (Bier et al., 1990; Kolodkin et al., 1994). Although both of these proteins are necessary for DER signaling in the ventral epidermis, neither is required if the secreted form of Spitz is supplied ectopically (Schweitzer et al., 1995). This epistatic effect of secreted Spitz suggests that Rho and Star act upstream of Spitz, possibly to regulate its cleavage.

The fourth member of the spitz-group is pointed (pnt), which encodes two ETS family transcription factor isoforms (Klämbt, 1993). The connection between DER-mediated signaling events at the plasma membrane and Pnt is understood, at least in outline, for one of these isoforms, Pnt P2.

Like other receptor tyrosine kinases, DER activates the Ras/MAP kinase phosphorylation cascade (Lu et al., 1993; Diaz-Benjumea and Hafen, 1994). The Pnt P2 isoform has a MAP kinase phosphorylation site in its N-terminal domain and was shown to be a target of the DER signaling cascade (Brunner et al., 1994; O’Neill et al., 1994). In transfected S2 cells, Pnt P2 was shown to promote transcription in response to activated Ras or activated MAP kinase (O’Neill et al., 1994). However, Pnt P1 was shown in these same experiments to be constitutively active and unaffected by the Ras/MAP kinase pathway. This result is relevant to the role of pnt and the DER pathway in salivary duct development because Pnt P1 is the only isoform expressed in the ventral epidermis (Klämbt, 1993). Thus, although their mutant phenotypes in the ventral epidermis are similar, the connection between pnt on the one hand and DER and the other spitz-group genes on the other is not understood.

In addition to the DER pathway, we have studied two other regulators that affect salivary duct development: the genes fork head (fkh) and trachealess (trh). The winged-helix transcription factor encoded by fkh (Kaufmann et al., 1994) is first expressed at the termini of the embryo, where it is regulated by the gap genes tailless and huckebein (Weigel et al., 1989a, 1990). fkh is required in these terminal cells for the development of the foregut and hindgut (Weigel et al., 1989b). Near the end of germ band extension (stage 10) fkh is strongly expressed in the salivary placode and is required for morphogenesis of the salivary glands (Weigel et al., 1989b). Analysis of fkh regulation in the salivary placode has shown that fkh is activated by the homeotic gene Sex combs reduced (Scr) and that the extent of its expression is limited dorsally by decapentaplegic and ventrally by the spitz-group genes (Panzer et al., 1992). Recent analysis has identified a 1 kb transcriptional enhancer element that mediates the salivary expression of fkh and has shown that this enhancer reacts directly to spitz-group signaling to prevent fkh transcription in the cells ventral to the placode (B. Zhou, unpublished observations).

trh was originally identified by the fact that trh-mutant embryos lack mature tracheae (Jürgens et al., 1984; YounossiHartenstein and Hartenstein, 1993). The gene has recently been cloned and found to encode a transcription factor with similarity to single minded (Isaac and Andrew, 1996; Wilk et al., 1996). On the basis of ectopic expression experiments, Wilk et al. (1996) have proposed that trh is a master regulator for tracheae. We began to study its role in duct development because of the observation that trh embryos completely lack salivary ducts although they have apparently normal salivary glands (Younossi-Hartenstein and Hartenstein, 1993).

In this paper we show that the distinction between pregland and preduct cells is made by the combination of two spatially separated negative regulatory steps: the DER signaling pathway repressing fkh in the preduct cells and fkh repressing duct-specific genes in the pregland cells. Furthermore, we propose that trh is a duct-specific gene activator and that it is one of the targets of fkh repression.

A Canton-S laboratory stock was used as wild type. rho^7M43, spi^IIA14 and pnt^9j31 were obtained from the Bowling Green Stock Center. sim^H9 was obtained from K. Anderson, fkh^XT6 from H. Jäckle, trh^5D55 from V. Hartenstein and the enhancer trap line trh¹⁰⁵¹² from D. Andrew. Mutations were placed in trans to one of the lacZ-marked balancer chromosomes SM6B eve-lacZ (Panzer, 1994) or TM3Bftzlac (a gift from Y. Hiromi) so that mutant embryos could be identified by the absence of the eve or ftz pattern of β-galactosidase stripes. Transformed flies carrying one of two fkh:lacZ constructs were crossed to various mutants so salivary development could be followed by β-galactosidase expression. The fkh 1-5000:lacZ construct contains 5 kb of fkh regulatory sequence including the salivary enhancer. It is expressed in the glands and their precursors but not in the ducts (B. Zhou, unpublished observations). The fkh Δ360-505:lacZ construct contains the 1 kb fkh salivary enhancer with its ventral repression sequences removed. It is expressed in both glands and ducts and in their precursors (B. Zhou, unpublished observations).

Embryos collected on molasses/agar plates were dechorionated in 50% chlorox followed by 0.7% NaCl, 0.1% Triton X-100, and fixed in a 1:1:2 mixture of 1× PBS:10% formaldehyde (E.M. Grade, Polysciences) : heptane (HPLC grade, Sigma). The embryos were devitellinized by shaking in heptane with 90% methanol, 5 mM EGTA and were washed thoroughly in methanol. Antibody staining was performed essentially according to Patel et al. (1989) using polyclonal rabbit anti-β-galactosidase sera (5 Prime-3 Prime Inc.) in conjunction with biotinylated goat anti-rabbit secondary antibody (Jackson Immunoresearch) and Vectastain Elite ABC Kit (Vector Laboratories). Color was developed using 0.5 mg/ml diaminobenzidine and 0.06% H₂O_2. Embryos were cleared in methyl salicylate.

In situ hybridzation assays involved the use of digoxigenin-labeled antisense RNA probes (digU) as described previously by Tautz and Pfeifle (1989) with modifications (Harland, 1991). Hybridization signals were visualized via histochemical staining with alkaline phosphatase. Serrate cDNA was obtained from R. J. Fleming and breathless cDNA from B. Shilo. Embryos were staged as described by Campos-Ortega and Hartenstein (1985). Stained embryos were photographed using a Leica DMRB microscope with Nomarski DIC optics.

Serrate (Ser), which encodes a transmembrane, EGF-repeat protein that acts as a ligand for Notch, is expressed in a dynamic and complex pattern during embryogenesis (Fleming et al., 1990; Thomas et al., 1991). In the salivary primordium Ser RNA is the earliest marker for the preduct cells, appearing in germ band extended embryos (stage 11) as a pair of crescents of the most ventral cells of parasegment 2, just ventral to the salivary placode (Fig. 1A). Ser can be used to follow these cells through the rearrangements that occur during germ band retraction and head involution. During germ band retraction, the preduct cells, like epidermal cells in other parasegments, are compressed in the AP axis and expand laterally, becoming two rectangular groups of cells flanking the ventral midline (Fig. 1B-C). As retraction continues, the cells adjacent to the midline come in contact to form a continuous “lip-like” structure in which the cells are lined up in anterior and posterior rows with an intervening cleft (Fig. 1D and inset). Then, starting at the lateral ends of the “lips” and progressing toward the center, the individual ducts form, apparently by apposition of the cells on either side of the cleft and separation of the resulting tube from the surface epithelium (Fig. 1D-E). This process follows, and is aligned with, the invagination of the salivary placode to form the gland. Thus, as the ducts form, they are attached to the anterior end of the glands. While the individual ducts are forming, cells near the midline of the primordium extend forward towards the stomadeum (Fig. 1E) to form the common duct that connects the salivary glands and the individual ducts to the floor of the pharynx once head involution is complete (Fig. 1F).

To examine the effects of spitz-group and trh mutations on salivary gland and duct development, we used lacZ reporter genes that are expressed either in just gland cells or in gland and duct cells. We previously showed that in spitz-groupmutant embryos the expression of the placode-specific gene fkh is expanded ventrally to include the preduct cells (Panzer et al., l992). We have repeated this result (Fig. 2A,C), using a reporter construct, fkh 1-5000:lacZ rather than fkh itself to mark placode cells. Next, to determine whether the inappropriate ventral expression of fkh (and perhaps other placode-specific genes; Panzer et al., 1992) affects the subsequent development of the salivary ducts, we have used another reporter, fkh Δ360-505:lacZ, which marks both pregland and preduct cells and later the mature glands and ducts (Fig. 2B). In embryos carrying this reporter but mutant for pnt or spi, ducts do not appear. In most of these embryos all of the labeled cells invaginate to form glands that have no connection to the surface of the embryo (Fig. 2D). In some embryos a few cells remain as a cluster on the surface at the anterior end of the embryo (data not shown). These surface cells remain connected to the internalized cells by long, slender processes. We cannot tell whether this is a terminal phenotype or whether these surface cells would eventually have joined the internal majority. In any case, in most spitz-group embryos the presumptive salivary duct cells appear to be converted to salivary gland cells.

Embryos mutant for trh also lack ducts but for different reasons. Using the placode-specific fkh:lacZ marker in trh embryos, we do not see expression in the preduct cells (Fig. 2E). Instead the expression is confined to the placodes as it is in wild-type embryos, indicating that trh has no role in preventing fkh expression in the preduct cells. At later times we see, using the fkh:lacZ marker for all salivary cells, that the preduct cells do not invaginate. They remain on the surface and move to the front of the embryo (Fig. 2F). The cells of the placode do invaginate but do not retain any connection to the surface of the embryo. Instead they form internal glands whose anterior and posterior ends are closed. Thus in trh embryos the preduct cells are not transformed to gland precursors. Rather, they appear to lose their salivary identity.

Because the development of the salivary ducts is abnormal in embryos mutant for either spitz-group genes or trh, we wanted to know whether the expression of duct-specific markers was altered in these mutant embryos. We looked at the expression of two marker RNAs, Serrate, whose developmental expression is shown in Fig. 1, and breathless (btl), the Drosophila FGF-receptor homolog (Klämbt et al., 1992). Like Ser, btl is expressed in all the preduct cells, though it starts a little later (early stage 12) and is not as strongly expressed (Fig. 3B and data not shown). In embryos mutant for spi, rho or pnt or in trh embryos, neither Ser nor btl is expressed in the presumptive preduct cells (Fig. 3C-F and data not shown). Since Ser is expressed early in the preduct cells of wild-type embryos (Fig. 1A), its absence in these mutants shows that both trh activity and spitz-group signaling act near the upstream end of the pathway that defines preduct cells and distinguishes them from pregland cells.

One possible interpretation of the results so far is that the spitz-group genes mediate a choice between salivary gland and salivary duct identities and that trh acts downstream of the spitz-group signaling pathway to carry out a necessary step in duct development. If that were the case, trh expression or trh activity might depend on the spitz-group genes. We used a trh enhancer trap to test the first possibility, that trh expression is spitz-group-dependent. As shown in Fig. 4A, this enhancer trap drives β-galactosidase expression throughout wild-type salivary glands and ducts, a pattern that matches early trh RNA expression (Isaac and Andrew, 1996). In pnt embryos the duct cells have been converted to gland cells and have been invaginated with the rest of the gland. These cells do express the trh enhancer trap, showing that trh expression is independent of the spitz group. There is an important limitation to this experiment. The stability of the β-gal protein precludes analysis of changes in the expression pattern after its initiation. Thus, it is possible that continued expression of trh would depend on pnt or other spitz-group genes. This issue is considered further in the discussion.

We have shown that in embryos mutant for spitz-group genes, fkh is expressed ectopically in the presumptive preduct cells (Panzer et al., 1992), and in those same cells we find that the duct markers Ser and btl are not expressed (Fig. 3C-D). These results suggested that the absence of Ser and btl might be caused by the inappropriate expression of fkh in these mutants. If so, fkh expression in its normal location, the salivary placode, might also negatively regulate Ser and btl. To test this possibility, we examined expression of Ser and btl RNAs in fkh-mutant embryos (Fig. 3G-H). As in wild-type embryos Ser is expressed in the preduct cells on either side of the ventral midline, but is now also expressed in the posterior cells of the placode (Fig. 3G). Thus, fkh expression in the placode normally blocks the expression of Ser. In contrast, btl expression seems to be little affected by mutations in fkh. btl RNA is still confined to a small number of preduct cells near the ventral midline.

To test more directly whether it is the ectopic fkh expression in spitz-group mutants that prevents Ser and btl expression, we looked for expression of these RNAs in rho fkh and pnt fkh embryos. In both cases the removal of fkh along with one of the spitz-group genes restores Ser and btl expression (Fig. 5A-D). This result shows that loss of the duct markers in spitz-group mutants is not just due to a general fate map shift that lateralizes the ventral epidermis. Instead, it is the expanded expression of a particular ventrolateral gene, fkh, that blocks expression of the two duct-specific genes.

These doubly mutant embryos also show that the spitz-group genes rho and pnt have slightly different effects on duct development. When the rho fkh and pnt fkh embryos are compared, differences in the expression of both Ser and btl are seen (Fig. 5A-D). In rho fkh embryos Ser RNA is expressed ventrally between the two placodes and posteriorly in the placodes themselves, much as it is in fkh single mutants (compare Figs 5A and 3G). In contrast, in pnt fkh embryos, Ser is only expressed in a posterior stripe that includes the same posterior placode cells, but only the most posterior of the ventral cells (Fig. 5C).

There also appear to be differences in btl expression between rho fkh and pnt fkh embryos (Fig. 5B,D). The differences in expression of these two duct markers suggest that pnt and rho affect overlapping but not identical groups of cells. Consistent with these results, Bier et al. (1990) found additive effects of rho and pnt mutations on the formation of the ventral nerve cord.

We also wanted to look for interactions between trh and either fkh or spitz-group mutations. Single trh or fkh mutations have opposite effects on Ser expression, trh eliminating it and fkh allowing it to expand (Fig. 3E,G). btl expression is also eliminated by a trh mutation though it is little affected by loss of fkh. In trh fkh mutants trh is epistatic. Neither Ser nor btl is expressed in the salivary region of the embryo (Fig. 5E,F). This result suggests that, in the pathway to Ser or btl expression, trh and fkh act independently or trh acts downstream of fkh.

In contrast to spitz-group mutations, trh has no effect on the expression of fkh. Thus, we expected that spitz-group trh double mutants would allow ventral derepression of a fkh-lacZ construct, the result that is shown for pnt trh in Fig. 5G. As in the previous experiment, this result is consistent with a model that places trh in an independent or downstream position.

spitz-group and trh mutations both prevent salivary duct development, but they do it by profoundly different mechanisms. spitz-group signaling is necessary to convert pregland cells to preduct cells. In the absence of this signal, lateral fates expand ventrally and all of the salivary primordium differentiates as salivary gland. This result is similar to results showing that markers for two other lateral structures, the Keilin’s organs and tracheal pits, expand ventrally to the midline in DER-mutant embryos (Raz and Shilo, 1993). One difference here is that we have followed the development of the affected cells beyond the time of early expression patterns. We see that the potential preduct cells not only express markers for more lateral cells, they also behave as though they have been converted to a different tissue type. In many embryos all of the ventral cells in PS2 invaginate with the glands and form a tube that is the same diameter as the glands, not the much smaller diameter of salivary ducts. In contrast, the preduct cells in trh mutant embryos simply fail to differentiate as duct cells rather than being converted to some other fate. They remain as a disorganized clump on the ventral surface of the head. We conclude that in spitz-group mutants there is a switch in determination of the ventral cells from preduct to pregland, while in trh mutants duct determination does not occur.

This difference in mutant phenotype suggested that spitz-group activity might be near the top of the regulatory pathway that distinguishes preduct cells from the more dorsal placode or pregland cells. Then trh might be a downstream gene necessary to carry out the differentiation of ducts once the duct-gland distinction had been made. If so, we expected that trh expression or activity would be dependent on the spitz-group. However, since we have shown that the trh enhancer trap is expressed normally in pnt embryos, we conclude that the initiation of trh transcription is independent of the spitz group. It is still possible that the activity of the trh protein is regulated by spitz-group signaling. For example, since the spitz-group signaling pathway operates via a protein kinase cascade, trh might be a phosphorylation target downstream of the protein kinases. Although trh is necessary for expression of duct markers, it does not appear to be involved in the spitzgroup-mediated repression of fkh. In fact, we argue below that fkh probably represses trh expression.

The spitz-group and fkh seem to oppose each other in establishing the boundary between pregland and preduct cells. In spitz-group mutants, fkh expands ventrally, preventing duct formation and forcing the cells to assume the gland identity. If fkh is removed, duct markers typified by Ser expand dorsally into the placode. These results suggest a double negative pathway for determining and positioning the preduct cells:

Because the homeotic gene Scr either directly or indirectly activates them, both fkh and duct markers like Ser have the potential to be expressed throughout the DV extent of the salivary primordium. It is the ventrally localized activity of the spitz-group pathway that leads to the spatial separation of pregland and preduct cells. spitz-group activity restricts fkh expression to the more dorsal cells that become the salivary placode. fkh in turn restricts duct-specific genes to the more ventral, preduct cells. As a result of this pathway, spitz-group activity and the expression of duct-specific genes are found in the same ventral epidermal cells. This colocalization leaves open the possibility that the spitz-group genes act not only negatively to block ventral fkh expression, but also positively to activate at least some of the duct genes. This possibility is discussed further below.

Because the preduct cells in spitz-group mutant embryos seem to be transformed to pregland cells, we expected that many of the gland-specific genes would be ectopically expressed as fkh is. Indeed, the enhancer trap line N33, which is expressed early in the placode and is independent of fkh, is expressed ventrally in spi embryos (Panzer, 1994; data not shown). Then the block that prevents preduct expression of Ser and btl in these embryos might be caused by one of these other gene products and be unrelated to ectopic fkh expression. However, when a fkh mutation was added to rho or to pnt, expression of Ser and btl was restored. Thus, fkh is necessary to block expression of the duct markers. Whether fkh is the only gland-specific gene required to prevent their expression is currently being tested with a hsfkh construct.

This pathway accounts for the limitation of the duct markers at their dorsal edge, but we must also ask which genes are responsible for activating the duct markers and how they might be negatively affected by fkh expression. Although both Ser and btl expression in the preduct cells is likely to be dependent on the homeotic gene Scr, we do not favor the idea that Scr activates their expression directly. At about the same time as expression of these genes is initiated, Scr transcript and protein begin to disappear from the ventral part of parasegment 2 and are not detectable in invaginated glands or ducts (LeMotte et al., 1989; Panzer, 1994). In addition, several genes, including fkh, dCREB-A and trh, are activated in the salivary placodes or in the entire salivary primordium earlier than either Ser or btl is (Panzer et al., 1992; Smolik et al., 1992; Isaac and Andrew, 1996). It seems likely that one of these early genes, rather than Scr itself, activates Ser and btl.

One likely possibility is that trh, which is activated by Scr (Isaac and Andrew, 1996), in turn activates the duct-specific genes. As we have described, trh is necessary for both ductspecific genes and for duct development. Since trh is initially expressed throughout the salivary primordium, its effect on duct but not gland development might result from the restriction of a coactivator to the preduct cells. There is an interesting alternate possibility, that trh activity is limited to the preduct cells via the spitz-group/fkh pathway. Since trh is required for both duct markers the pathway would be:

This is an attractive alternative because it doesn’t require fkh to restrict the expression of multiple duct-specific genes. Instead fkh could limit the expression of just one gene, trh, which would then be required for the activation of other ductspecific genes. One prediction of this pathway has recently been confirmed. Isaac and Andrew (1996) showed that trh RNA initially appears throughout the salivary primordium as does the β-gal expression from the trh enhancer trap that we used. But in contrast to the β-gal, they found that trh RNA disappears from the placode, while remaining in the preduct cells. As predicted by the model, the disappearance of trh from the placode is dependent on fkh. In fkh-mutant embryos trh RNA remains in the placodes as well as in the preduct cells (Isaac and Andrew, 1996). This dorsal persistence of trh RNA may explain why Ser RNA is expressed in the placodes of fkh⁻ embryos.

trh is the third salivary gene that we know is directly or indirectly repressed by fkh. In addition to Ser and trh, the transcript of the lune gene, which encodes a protein of the paired-homeodomain family (Wilson et al., 1993), is repressed in the placode by fkh. Like trh, lune is initially expressed in most if not all of the salivary primordium, but then disappears from the pregland cells in a fkh-dependent process (Panzer, 1994).

As mentioned above, the spitz-group genes might activate duct markers directly in addition to preventing fkh from blocking their expression. This is an attractive possibility because spitz-group signaling is required for ventral epidermal gene expression in many segments other than PS2, and fkh is not expressed in these other segments (Wieschaus et al., 1992; Kim and Crews, 1993; Raz and Shilo, 1993). For example, orthodenticle (otd) is expressed throughout the trunk of the embryo in a narrow row of cells on either side of the midline (Finkelstein et al., 1990; Kim and Crews, 1993). Its expression was shown to be dependent on several spitz-group genes that were tested (spi, rho, S, pnt), although pnt embryos showed some residual otd expression (Kim and Crews, 1993).

Initially the results for the pnt fkh and rho fkh mutant embryos do not seem to favor the idea that spitz-group genes activate Ser and btl. Each of these genotypes includes a spitz-group mutation yet Ser and btl are expressed. However, these markers are not expressed in the same cells in the two double mutant embryos. This difference in expression pattern would be understandable if Pnt and the DER-signaling pathway act independently and in overlapping but distinct groups of cells to activate Ser and btl. Either one would be sufficient to activate Ser/btl expression in the ventral epidermis although their partially redundant activities are masked in single mutants by the ventral expression of fkh. Independent activities for Pnt and the DER signaling pathway are consistent with the additive effects of rho and pnt mutations (Bier et al., 1990) and with the fact that only the Pnt P1 isoform is expressed in the ventral epidermis (Klämbt, 1993). In cultured cells Pnt P1 is constitutively active and is insensitive to the activity of Ras1 or MAPK, members of the protein kinase cascade that transmits the DER signal (O’Neill et al., 1994). Presumably there is another transcription factor in the ventral epidermis that does respond to DER signaling. Both Pnt P1 and this other transcription factor would then act directly or indirectly to activate Ser/btl transcription and to repress fkh.

Our analysis of boundary formation in the salivary primordium can be summarized by the model in Fig. 6 which shows two cells, one on either side of the pregland-preduct boundary. DER signaling is stronger in the preduct cell and suffices to prevent fkh expression. Therefore, trh is active and it activates Ser and other preduct markers. In the pregland cell, DER signaling is not strong enough to block fkh expression. fkh then blocks trh expression and duct markers cannot be activated. Also indicated is the possibility that Pnt or DER signaling or both may activate duct markers directly rather than just indirectly by repressing fkh.

We would like to thank Volker Hartenstein for pointing out that salivary glands are abnormal in trh mutant embryos. We are also grateful for fly stocks from K. Anderson, H. Jäckle, V. Hartenstein, D. Andrew and Y. Hiromi. We thank B. Shilo for the btl cDNA and R. Fleming for the Ser cDNA. Finally we wish to thank Kim Bland and Chris Potter for technical assistance. This work was supported by a grant to S. K. B. from the National Institutes of Health.