Control of epithelial morphogenesis by cell signaling and integrin molecules in the Drosophila foregut

Pattern formation and morphogenesis are two interconnected processes in animal development (Gurdon, 1992). In Drosophila, great progress has been made on the genetic and molecular interactions that establish the body pattern in the early embryo (St Johnston and Nüsslein-Volhard, 1992; Hoch and Jäckle, 1993; Pankratz and Jäckle, 1993). Much less is known on the morphogenetic mechanisms that bring about the diverse tissues and organs of the body. In contrast to the early pattern forming processes that occur in a syncitium, these later developmental events involve interactions of cells with each other and with their extracellular environment.

One class of molecules important for such interactions are secreted molecules involved in intercellular signaling, and several conserved families of secreted growth and differentiation factors have been identified in different organisms (Jessell and Melton, 1992; Greenwald and Rubin, 1992, for reviews). The Drosophila gene hedgehog (hh), which encodes a secreted protein, and the genes wingless (wg) and decapentaplegic (dpp), which encode members of the Wnt and TGFβ families of growth factors, respectively, have been shown to act in assorted combinations in a variety of developmental contexts: hh and wg are required for epidermal segment patterning (Nüsslein-Volhard and Wieschaus, 1980); hh and dpp are required for the progression of the morphogenetic furrow in the eye (Ma et al., 1993; Heberlein et al., 1993); wg and dpp are required for the second constriction in the midgut (Immerglück et al., 1990; Panganiban et al., 1990; Reuter et al., 1990); and all three are required for limb patterning (Basler and Struhl, 1994; Diaz-Benjumea et al., 1994). Their vertebrate homologues have also been shown to be involved in patterning and induction during embryonic development (Kessler and Melton, 1994, for review).

Parallel studies have shown that cell adhesion molecules are also important mediators of cell interactions. One of the major classes of molecules modulating cell adhesion are the integrins, originally identified in vertebrates. Integrins belong to a family of cell surface adhesion receptors that mediate cell-cell and cell-extracellular matrix interactions (Hynes, 1992, for review). All integrins are αβ heterodimers, and an individual β subunit can associate with several different α subunits to form functional receptors with varying ligand and adhesive specificities. To date, 14 α and 8 β subunits are known in vertebrates. Integrins have also been identified in Drosophila, where they have been shown to be required for adhesion between different cell types and layers. These include attachments between muscles and epidermis, the visceral mesoderm and the adhering endoderm, and the dorsal and ventral parts of the wing blades (Newman and Wright, 1981; Wilcox et al., 1989; Leptin et al., 1989; Zusman et al., 1990, 1993; Brown, 1994).

In this paper, we describe a cellular system in which the roles of the cell signaling and cell adhesion molecules can be studied in a single developmental context. We show that epithelial morphogenesis during proventriculus organ development requires the activities of wg, hh and dpp, as well as the integrin class of cell surface adhesion receptors. We further provide evidence that cell signaling in the foregut operates through a distinct genetic circuitry as that in the midgut.

We used the following stocks: Oregon R, wg^IG22, wg^IL114, arm^XK, hh^IJ35, ci^D, dpp⁴⁸, ptc^IN108, en^IIB86, en^IK57 (provided by the Tübingen and Umea stock centers), HS-dpp (a gift from S. Cohen), HS-hh (a gift of P. Ingham), HS-wg (a gift of S. Cohen), HS-ptc (a gift of I. Guerrero), and mys^XG43 and if^K27e (gifts of M. Affolter). The flies were maintained and embryo collections made according to standard procedures.

BrdU (Sigma) labeling was performed, with modifications for embryos, essentially as described (Truman and Bate, 1988). The embryos were incubated for 30 minutes with BrdU prior to fixation.

Antibody staining of whole-mount embryos was carried out as described previously (Macdonald and Struhl, 1986), using the Vectastain ABC Elite-horseradish peroxidase system. NiCl₂ or Ni/CoCl₂ enhancement was used where necessary. The stained embryos were embedded in Araldite in capillaries according to the procedure of Schmidt-Ott and Technau (1992).

We used the following antibodies at the dilutions indicated in parenthesis: mAb22C10 (Zipursky et al., 1984; 1:20), anti-β-galactosidase (Cappel; 1:10000), anti-armadillo (Riggleman et al., 1990; 1:100), anti-MHC (Kiehart and Feghali, 1986; 1:1000), anti-forkhead (Weigel et al., 1989a; 1:150), and anti-crumbs (Tepass et al., 1990; 1:50). All antibodies were preabsorbed against wild-type embryos before use.

In situ hybridization was performed essentially as described in Tautz and Pfeifle (1989). The probes used were: dpp (a gift of S. Cohen), hh (a gift of P. Ingham), wg (a gift of S. Cohen), ptc (a gift of I. Guerrero), en (a gift of S. Cohen), ci (a gift of R. Holmgren), α1, α2 and β-integrins (gifts of T. Bunch and D. Brower).

HS-dpp: 0-20 hour embryo collections at 18 °C were placed at 37 °C for 45 minutes two times with 3 hours at 18°C between each heat shock, allowed to recover for 3 more hours at 25 °C, then fixed as described for immunohistochemical staining. The same protocol with wild type, HS-ptc, HS-hh or HS-wg harboring transgenic strains did not produce proventricular defects.

wg^IL114: 0 –20 hour collections were taken at 18 °C and then the embryos were transferred to 29 °C for 12 –16 hours and fixed as above. For the feeding assay, 0 –24 hour embryos were placed at 29 °C for 1 hour and then returned to 18 °C until hatching.

The larvae were allowed to grow on applejuice plates containing yeast that had been dyed with Carmine red (Sigma). Mutant feeding phenotypes were scored at various times under the dissecting microscope.

The foregut of the Drosophila larva is functionally and structurally subdivided into the pharynx, the esophagus and the proventriculus (Fig. 1A,B). The proventriculus is located at the caudal end of the esophagus and serves as a valve in regulating food passage into the midgut (Strasburger, 1932; Graham-Smith, 1934; Rizki, 1956). It is composed of two tissue layers, the ectodermal epithelial layer and the ensheathing visceral mesoderm (Fig. 1C,D; Skaer, 1993, for review). An exception is the area that will form the inside portion of the proventriculus, which is completely free of mesodermal tissues (Fig. 1D; Tepass and Hartenstein, 1994). This internal portion, called the cardiac valve (the proventriculus is also referred to as the cardia, Snodgrass, 1935; King, 1988), is innervated by three axons from the proventricular ganglion (Fig. 1F), one of four major interconnected ganglia that constitute the stomatogastric nervous system (Poulson, 1950; Willey, 1961; Schoeller, 1964; Penzlin, 1985; Campos-Ortega and Hartenstein, 1985; Hartenstein et al., 1994).

The proventriculus develops at the junction of the foregut and the midgut (Fig. 2A; Poulson, 1950; Campos-Ortega and Hartenstein, 1985). There is initially an outward buckling of the foregut tube, in a region that is free of visceral mesoderm, to form what we refer to as the ‘keyhole’ structure (Fig. 2B). This area will undergo further outward movement, then fold back on itself and move inwards to form the mature, multilayered proventriculus (Fig. 2C). The cells moving inwards assume a stretched appearance with long cytoplasmic extensions (Fig. 1E). These late steps in proventriculus morphogenesis are due to migration of cells and are not accompanied by cell proliferation, as assayed by BrdU incorporation experiments (data not shown; Hartenstein and Campos-Ortega, 1985). A major advantage of the proventriculus for studying morphogenesis is the relative ease with which one can monitor the movement of the epithelial cells at all stages of development.

Both hh and wg are expressed in spatially restricted domains in the developing proventricular epithelium. hh is expressed in the keyhole region (Fig. 3A,C) and persists until the late stages of proventriculus formation (Fig. 3B,D). The expression pattern of wg is more dynamic. It is initially expressed as a contiguous band spanning the keyhole (Fig. 3E); this domain then splits in the middle to form two narrow bands that now flank the keyhole (Fig. 3G). The two expression bands remain until the final stages of proventriculus development (Fig. 3F,H). The hh and wg expression domains show striking spatial correspondence with the mesoderm-free area: the posterior border of hh corresponds with one of the mesoderm borders while the two wg domains flank both mesodermal borders (Fig. 3C,G).

To correlate these gene expression patterns with morphogenetic function, we examined the proventriculus phenotypes of the corresponding mutant embryos using anti-forkhead (nuclear marker for the foregut epithelial cells), anti-myosin heavy chain (MHC) (mesodermal muscle marker) or anti-crumbs (epithelial apical membrane marker) antibodies (see Materials and Methods). In hh mutants, the keyhole structure develops normally, but the internal cardiac valve is not formed; the outer wall of the proventriculus is still present but is narrow and hollow as compared to the wild-type morphology (Fig. 4A,B). There are clearly foregut cells on top of this hollow proventriculus, but they are clustered and crowded, suggesting that the defect observed in hh mutants is due to the failure of the cells to move into the internal region of the proventriculus. For descriptive purposes, we use the term ‘cardiac arrest’ phenotype to describe this specific type of proventriculus defect. In some mutant embryos, the foregut tube succeeds in inserting into the proventriculus but does not complete the full range of movement into the midgut (data not shown).

Interpreting the phenotype of wg mutants through analysis of null alleles is problematic since most of the esophagus and the proventriculus are missing due to failure of the stomodeum to invaginate properly (Skaer, 1993; see also Fig. 4C). However, the analysis of a temperature-sensitive allele (wg^ts) demonstrates that wg activity is required for morphogenetic movements during proventriculus development (Fig. 4D). This phenotype is very similar to what is observed in hh mutants. There is a wide range of severity with regard to esophageal and proventricular defects, which most likely reflects the precise time at which the embryo has experienced the temperature shift. To see whether we could obtain proventricular defects in the absence of any other morphological foregut defects, we performed pulsed temperature-shift experiments and tested the larvae for their ability to feed (see Materials and Methods), the reasoning being that if the proventriculus does not function properly due to specific structural defects, the larvae should also not be able to feed normally. Using this assay we could, remarkably, recover larvae that move about normally but which cannot feed due to a specific defect in the inward movement of the proventricular cells (Fig. 4E,F); consequently, the food cannot be efficiently transported into the midgut, resulting in an engorged esophagus (Fig. 4E).

The above results indicate that both hh and wg activities are required for proper epithelial morphogenesis of the proventriculus. hh and wg are members of the segment polarity class of genes, which constitute a cell signaling network in epidermal patterning (Martinez Arias, 1993; Perrimon, 1994, for reviews). The gene cubitus interruptus (ci), which encodes a zinc-finger transcription factor, is another member of the segment polarity genes and is expressed in the keyhole region (Fig. 4G); unlike hh and wg, this expression domain disappears soon afterwards (data not shown). In ci mutant embryos, the keyhole structure does not form (Fig. 4H). Thus, ci could be involved in the specification and/or outgrowth of the foregut region to form the keyhole structure. However, not all of the segment polarity genes are involved in proventriculus morphogenesis. For example, the gene patched (ptc), which encodes a transmembrane protein (Hooper and Scott, 1989; Ingham et al., 1991), is expressed in the foregut but ptc mutants do not show the cardiac arrest proventricular phenotype (data not shown). Most notably, engrailed (en), which plays a key role in epidermal as well as limb patterning, is not expressed in the foregut, and en mutants show no proventricular defects (not shown).

The range of movements that the cells of the keyhole region undergo in forming the proventriculus suggested that factors involved in cell-cell or cell-extracellular matrix adhesions may also be important for this process. In addition, it has been documented that the region inside the proventriculus is rich in extracellular matrix components (King, 1988). We therefore investigated the role of genes involved in cell adhesion: armadillo (arm), encoding a homolog of the vertebrate β-catenin which is a component of the adherens junctions (Peifer and Wieschaus, 1990; Peifer, 1993; Peifer et al., 1993), and genes encoding the integrin subunits, which form heterodimeric cell surface receptors involved in cell-extracellular matrix interactions (Hynes, 1992; Brown, 1993, for reviews). As arm has been shown to be post-transcriptionally regulated (Riggleman et al., 1990), we used, in the case of arm, antibodies to monitor the distribution of the gene product. Armadillo is distributed throughout the foregut but is concentrated in specific areas of the developing proventriculus. At early stages, it is concentrated near the keyhole region (Fig. 5A); at later stages, Armadillo becomes highly concentrated in areas undergoing the most extensive cell movements (Fig. 5B). In arm mutant embryos, the proventriculus phenotype is similar to that of wg null embryos in that the entire foregut region is affected (Fig. 6A,B). The phenotype can vary, and we observe embryos in which the foregut tube has formed to a greater degree; a comparable temperature-sensitive allele as that of wg is not available for arm at this point, so we do not have unequivocal evidence as to how arm acts at later stages of proventriculus development. However, the overlap of the expression patterns between arm and wg through the final stages of proventriculus formation, the similarities of the foregut phenotypes in the mutants of the two genes, and the demonstration that wg activity affects the level of Armadillo in tissue culture cells (van Leeuwen et al., 1994), suggest the requirement of arm during proventriculus morphogenesis.

The different subunits of the integrins (PS1α, PS2α and PSβ, collectively known as the position-specific integrins) have been shown to be distributed differentially in the developing embryo (Bogaert et al., 1987; Leptin et al., 1989; Zusman et al., 1990; Wehrli et al., 1993). In the foregut region, the α2 subunit is found in the mesodermal layer and as a narrow band in the ectoderm in the keyhole region (Fig. 5C); the expression pattern persists into the late stages of proventriculus development (Fig. 5D). The α1 subunit is found in the foregut ectoderm (Fig. 5E) and remains at low levels at later stages (not shown). The β subunit is uniformly distributed at low levels throughout most of the foregut (Fig. 5F).

The β integrin subunit is encoded by the myospheroid (mys) gene (MacKrell et al., 1988; Leptin et al., 1989). In mys mutant embryos, we observe a proventricular phenotype which is very similar to that of hh and wg^ts mutant embryos: the cells are clustered at the top of the esophagus, but cannot migrate inside to form the cardiac valve (Fig. 6C,D). The α 2 subunit is encoded by the inflated gene (Brower and Jaffe, 1989; Wilcox et al., 1989; Brown, 1994); we did not observe proventricular cell migration defects in inflated mutant embryos (data not shown). For the α1 subunit, a mutant for the corresponding gene does not yet exist. Taken together, these results indicate that the integrin β subunit molecules, which can heterodimerize with either of the two α subunits (Brower et al., 1984; Wilcox et al., 1984; Leptin et al., 1987), are required for epithelial morphogenesis in proventriculus development.

dpp is expressed only in the anterior region of the foregut tube and not in the keyhole region (Fig. 7A,B). In dpp mutants, despite drastic effects on the global morphology of the embryo, much of the proventriculus still forms. However, more cells move into the proventriculus as compared to wild-type embryos (Fig. 7C). In certain embryos, one observes an extra outbudding zone in the foregut epithelium near the proventriculus (Fig. 7D). Although it is not clear whether this is a primary effect of dpp mutations or a secondary consequence of the torsional stress brought about by the twisting of the esophagus, it is possible that dpp functions to suppress cell movements in more anterior regions of the foregut, thereby allowing only the keyhole region to move outwards. In this case, the extra outbudding zone in the esophagus could represent attempts to form an ectopic proventriculus due to the lack of repressing activity by dpp. To explore this possibility, we examined embryos in which the dpp gene was ectopically expressed under heat-shock control (see Materials and Methods). In these embryos the development of the proventriculus can be suppressed, resulting in the cardiac arrest phenotype (Fig. 7E,F). Thus, dpp activity imposes an opposite effect on proventriculus morphogenesis as wg, hh and the integrins.

The foregut arises from an anterior region of the blastoderm embryo; the anterior portion of the invaginating foregut will form the esophagus while the posterior portion (the ‘keyhole’) will give rise to the proventriculus. ci, hh and wg are all expressed in the keyhole region, whereas dpp is expressed only in the anterior part of the foregut. As a first step towards examining the potential genetic interactions between the genes involved in proventriculus development, we monitored the expression of hh, wg and dpp in various mutant backgrounds. In ci mutants, both hh and wg expression are normal at an earlier stage of proventriculus development, but decrease at a later stage (Fig. 8A,B), suggesting that ci is required to maintain the expression of hh and wg. In hh mutants, wg is expressed normally at early stages but is decreased at later stage, indicating that maintenance of wg expression also requires hh activity (Fig. 8C). We do not observe a significant change in dpp expression in the foregut of hh mutants, although the dpp expression in the gastric caecae of the midgut is completely lost (Fig. 8D). Taken together, these studies indicate that the genetic interactions are mostly modulatory in nature, rather than one gene being strictly dependent on another for expression. The initial activation of wg and hh is most likely carried out by other genes controlling foregut development, e.g., forkhead (fkh) is required for the expression of both wg and hh (Pankratz and Hoch, unpublished; Mohler et al., 1995).

In dpp mutants, wg expression remains quite strong in the keyhole region; in fact, the level of expression is consistently stronger than in wild-type embryos (Fig. 8E). Thus, dpp could be involved in repressing wg activity in the proventriculus. In the absence of internal staining control, we cannot estimate to what extent wg expression is derepressed in dpp mutants. However, the genetic interaction that we observe is clearly different from what is found during the formation of the second midgut constriction, where dpp activates wg expression (Immerglück et al., 1990). We did not observe any major change of hh expression in dpp mutants (Fig. 8F).

We have described a model system where the process of epithelial morphogenesis is accessible to genetic analysis. We discuss below how signaling molecules and integrins may function to coordinate this morphogenetic process in proventriculus development (Fig. 9).

During late stages of embryogenesis, a specific region of the foregut epithelium moves outwards, folds back on itself and moves inwards, thereby transforming an epithelial tube into a multiply folded organ, the proventriculus. The genes wg, hh and dpp are involved in controlling these morphogenetic events. hh and wg are both expressed in the foregut epithelium that moves inside during proventriculus formation, and the corresponding mutant embryos display a very similar mutant phenotype, namely, the failure of the final inward movement to occur. This results in a proventriculus that is hollow inside and where cells are clustered on top (the ‘cardiac arrest’ phenotype), rather than the normal multiply folded structure.

Several cellular processes could underlie this failure in epithelial morphogenesis. One is a defect in cell proliferation which results in the absence of the cells that normally move inwards. Another is a defect in coordinating cell movements or shape changes, whereby the cells cannot properly migrate into the proventriculus. It has been shown that wg is involved in cell proliferation at early stages of gut development (Skaer and Martinez Arias, 1992; Skaer, 1993). In hh mutants, the esophagus is shortened, suggesting that hh may also be involved in cell proliferation at the early stages of foregut development. For the proventricular ‘cardiac arrest’ phenotype at the later stages of embryogenesis, however, we favor the view that a failure to coordinate cell movements is responsible. This is based on several considerations. First, in both hh and wg^ts mutant embryos, many cells are clustered on top of the proventriculus, indicating that cells are in fact available for moving inwards. Second, we can obtain, from experiments with the wg^ts allele, crawling larvae with a specific proventricular defect in which cells have not moved inwards but remain looped out. Third, embryos mutant for the gene encoding the integrin β subunit, a molecule involved in cell adhesion, show a very similar proventricular defect as wg^ts and hh mutants (see below).

dpp mutants show an opposite phenotype as hh and wg^ts mutants, namely, the uncontrolled movement of cells in the proventriculus and in parts of the esophagus. As dpp expression is restricted to the esophagus, and ectopic dpp can shut down proventriculus morphogenesis, dpp formally suppresses proventriculus development, perhaps through delimiting the area within the foregut tube where the proventriculus can form. The mechanism by which dpp performs this function is unclear. wg expression appears to be negatively regulated by dpp as assayed in dpp mutant embryos, but we do not observe a significant repression of wg in heat-shock dpp transgenic embryos (data not shown). However, this may be due to the lack of sensitivity of our assay system. An alternative possibility is that dpp acts on the downstream effectors of wg and hh activities through a parallel pathway.

hh is required to maintain wg expression in the proventriculus but not for the initial activation. hh and wg are in turn dependent on ci for maintenance of their expression pattern, but again, not for initial activation. A potential activator of both wg and hh is fkh, since both hh and wg expression in the foregut is gone in fkh mutants (Pankratz and Hoch, unpublished; Mohler et al., 1995). fkh is required for the development of the esophagus and the proventriculus, and is expressed throughout the foregut region at high levels from the blastoderm stage to the end of embryogenesis (Weigel et al., 1989a,b). As fkh encodes a putative transcription factor containing a HNF3/forkhead DNA-binding domain (Weigel and Jäckle, 1990), it is possible that fkh directly regulates the transcription of hh and wg.

The most intriguing aspect of the spatial control of gene expression in the foregut is the splitting of the wg domain such that the gap between the two resulting wg domains coincide with the mesoderm-free zone. As wg is initially expressed as a contiguous band, there must be another factor that is responsible for repressing wg expression in the keyhole region.

The signaling mechanisms underlying proventricular cell movements in the foregut differ from that of another morphogenetic process in the gut, the formation of the second midgut constriction. For example, dpp acts to prevent proventricular cell movement, whereas in the midgut it acts to promote cells movements forming the second constriction. In addition, dpp activates wg in the midgut (Immerglück et al., 1990), whereas we do not observe activation of wg by dpp in the foregut. Recently several receptors for dpp have been identified that are differentially expressed in the embryo (Affolter et al., 1994; Nellen et al., 1994; Penton et al., 1994; Brummel et al., 1994). The different effects of dpp in the two developmental contexts could thus be mediated through different receptors.

A further difference is reflected in the differing roles that the various tissue layers play in the morphogenetic processes in the foregut and midgut. The second midgut constriction depends on cell signaling across germ layers from the visceral mesoderm to the underlaying epithelial endoderm: dpp and wg are expressed in a specific region of the midgut mesoderm and their gene products interact with cells across the germ layer to induce expression of various target genes in the adhering endoderm (Panganiban et al., 1990; Reuter et al., 1990; Reuter and Scott, 1990; Immerglück et al., 1990; Mathies et al., 1994). In striking contrast, the visceral mesoderm is completely absent in the keyhole region that will form the proventriculus. In addition, dpp, wg and hh are expressed only in the ectodermal layer of the foregut, even in regions outside the keyhole that are ensheathed in mesoderm. Furthemore, in embryos where the visceral mesoderm is missing due to specific mutations (twist and snail double mutants), the proventriculus develops normally and the inward cell migration takes place (Hartenstein et al., 1992; our unpublished observations). These observations indicate that the ectodermal cells of the proventriculus do not take cues from the mesoderm for their morphogenetic movements.

The observation that embryos lacking β integrin activity have very similar proventricular phenotypes to those of hh and wg^ts mutants suggests that integrins may be involved in mediating some of the activities of these secreted molecules. In cases where the mutant phenotypes of the Drosophila integrin genes have been analysed at the cellular level, a common theme has been the failure of different cell layers to attach or remain attached (Newman and Wright, 1981; Wilcox et al., 1989; Leptin et al., 1989; Zusman et al., 1990, 1993; Brown, 1994). By contrast, the proventricular phenotype that we observe in embryos that lack the PSβ subunit most likely arises from defects in cell migration: the cells are clustered on top of the proventriculus and do not move inwards (see Fig. 6C).

It is known that many types of cell movements require interactions with the extracellular matrix. The inside of the proventriculus is rich with extracellular matrix material (King, 1988), and it has been shown that there is an alteration in the accumulation of extracellular matrix material in mys mutants (Wright, 1960; Newman and Wright, 1981). It has also been demonstrated that the Drosophila integrins can interact with vertebrate extracellular components (Hirano et al., 1991: Bunch and Brower, 1992). Therefore, the proventricular cell migration defect in mys embryos is consistent with studies in vertebrate systems demonstrating the biochemical function of integrins in mediating cell-extracellular matrix interactions.

We have not observed proventricular defects in embryos lacking the α2 subunit. However, since different α subunits can form heterodimers with the same β subunit (Brower et al., 1984; Wilcox et al., 1984; Leptin et al., 1987), this could be due to the fact that one type of α subunit can substitute for another.

As mys is expressed uniformly in the entire foregut, hh and wg may provide the spatial cues for regulating the activity of the integrins. We do not know the mechanism by which this could occur, and we do not observe any differences in mys expression in hh mutants (data not shown). However, there exists a link between wg and another class of cell surface adhesion receptors, the cadherens. This link is provided by Armadillo, one of the components that transduce the wg signal: Armadillo is a cytosolic component of the adherens junction multiprotein complex which is associated intracellularly with the actin cytoskeleton (see Peifer et al., 1993 and references therein). The integrins are also associated with the cytoskeleton network through several cytosolic components (Luna and Hitt, 1992; Stossel, 1993, for reviews). Therefore, hh and/or wg may function through modulating the activities of various, as yet unidentified, cytosolic components that interact with the integrins.

The primitive gut of animals is essentially a closed sac, which has invaginated from one side of the body during gastrulation; the stomodeal and proctodeal openings were later evolutionary additions that facilitated ingestion and egestion. Wolpert (1994) has suggested that the openings to the primitive gut ‘could originally have resulted from the invaginating gut making contact with the ectoderm and thus providing a signal for making these cells different – possibly the first inductive event in development’. wg, hh and dpp, in addition to being expressed in the proventriculus, which forms at the junction between foregut and midgut, are also expressed at the junction between midgut and hindgut. Furthermore, the mesoderm-free zone that is found at the foregut/midgut junction is also found at the midgut/hindgut junction (Tepass and Hartenstein, 1994), suggesting that signaling mechanisms in the foregut and hindgut may be quite analogous. It is also noteworthy that the basic tubular structure of the gut extending from the mouth to the anus is found in almost all metazoans outside of Porifera, Plathelminthes and Cnidaria (see Remane et al., 1981). Recently, it has been shown that a mouse homologue of the fkh gene, HNF3β, is expressed in the mouse embryonic gut (Ang et al., 1993; Monaghan et al., 1993); a mutation in HNF3β affects the expression of Sonic hedgehog (Shh), a homologue of the Drosophila hh gene, and results in gut defects (Ang and Rossant, 1994; Weinstein et al., 1994). Therefore, it will be interesting to determine whether genetic interactions controlling gut development have been conserved between Drosophila and vertebrates.

We thank the following for providing us with fly stocks and anti-bodies: M. Affolter, D. Branton, D. Brower, T. Bunch, S. Cohen, C. Goodman, I. Guerrero, R. Holmgren, P. Ingham, D. Kiehart, E. Knust, K. Matthews, C. Nüsslein-Volhard, N. Patel, M. Peifer, G. Thomas, D. Weigel and the Indiana and Umea stock centers. We thank M. Tetzlaff for his help, many discussions and thank him and H. Jäckle for valuable suggestions on the manuscript, and S. Fellert for technical assistance. We would especially like to thank H. Jäckle for support and encouragement. M. J. P. would like to thank Dr. David King for materials and discussions, and gratefully acknowledge that G. Jürgens suggested the dyed yeast larval feeding assay. This work is supported by the Max Planck Society and the Deutsche Forschungsgemeinschaft (SFB 271 to M. H.).