The gene serpent has homeotic properties and specifies endoderm versus ectoderm within the Drosophila gut

The cell rearrangements of gastrulation generate the three germ layers, endoderm, ectoderm and mesoderm. In Drosophila, the mesoderm originates from the cells of the ventral furrow and forms musculature, heart, fat body and gonads. The endoderm is formed from two spatially separated invaginations. The anterior part of the endoderm is initially contiguous with the mesoderm and originates from the anterior tip of the ventral furrow, which is termed anterior midgut invagination. The posterior part of the endoderm (the posterior midgut) is internalized from the posterior pole of the embryo during the amnioproctodeal invagination. Both parts later become mesenchymal, migrate toward each other and form the epithelium of the midgut (for reviews on Drosophila development see Costa et al., 1993; Campos-Ortega and Hartenstein, 1985; Skaer, 1993).

The other major components of the digestive tract besides the midgut are the foregut (consisting of pharynx, esophagus and proventriculus), the Malpighian tubules and the hindgut. All these components are considered to be ectodermal in origin since they form cuticle (with the exception of the Malpighian tubules), have epithelial properties in common with the epidermis and never show any mesenchymal character. Hindgut and Malpighian tubules develop from the proctodeum whose primordium is anteriorly adjacent to the primordium of the posterior midgut. Both primordia are internalized together during the amnioproctodeal invagination. As a result, the prospective posterior midgut is located at the bottom of the invagination, while the proctodeum forms the connection to the surface of the embryo. At this time, there is no morphological difference between proctodeal cells and prospective posterior midgut cells. The two primordia become morphologically different when the pouches of the developing Malpighian tubules begin to evaginate from the proctodeum at the connection between the proctodeum and the posterior midgut. Slightly later, the posterior midgut cells lose their epithelial character and become mesenchymal.

The foregut develops from the stomodeum which invaginates in the region of the blastoderm anterior to the ventral furrow. This invagination begins almost 2 hours after the onset of gastrulation, long after endoderm and mesoderm have been established. For several hours it continues to internalize cells which contribute to the foregut. This distinguishes the stomodeal invagination from the amnioproctodeal invagination which is terminated after about 30 minutes and which forms the proctodeum (the structure in the posterior part of the embryo corresponding to the stomodeum).

Several zygotic genes act in the early determination and specification of the alimentary canal. The terminal gap gene huckebein (hkb) is required to establish the primordia of both anterior and posterior midgut (Weigel et al., 1990). In hkb embryos, the amnioproctodeal invagination is severely reduced and only the proctodeum invaginates from the posterior pole (Gaul and Weigel, 1991). In the anterior region of hkb embryos, neither the anterior midgut primordium is formed (Weigel et al., 1990) nor does the stomodeum invaginate (Reuter and Leptin, 1994). Being a transcription factor (Brönner and Jäckle, 1991), hkb presumably acts by activating and repressing downstream genes, which in turn drive the cell shape changes leading to the invagination of the posterior midgut or which initiate endodermal development.

The other terminal gap gene, tailless (tll), determines the primordium of the proctodeum and is required for the invagination of the hkb-independent aspect of the amnioproctodeal invagination (Jürgens et al., 1984; Strecker et al., 1988; Gaul and Weigel, 1991). Consequently, hindgut and Malpighian tubules are missing in tll embryos, but the midgut develops almost normally and is directly connected to the epidermis. tll also acts in the anterior region of the embryo where it is required for the development of part of the brain (Strecker et al., 1988). However, the development of the anterior part of the digestive tract is independent of tll function. Like hkb, tll is a transcription factor (Pignoni et al., 1990), which presumably activates the same set of downstream genes as hkb required for the cell shape changes that occur during the amnioproctodeal invagination. However, a different set of downstream genes is certainly responsible for the regional specification within the proctodeum.

One gene acting downstream of both hkb and tll is the putative transcription factor fork head (fkh), which at blastoderm stage is expressed at both poles of the embryo (Weigel et al., 1989; Casanova, 1990; Weigel et al., 1990). Later, fkh is found in stomodeum and proctodeum, and transiently also in the midgut (Weigel et al., 1989). It stays expressed in all proctodeal derivatives and the esophagus until the end of embryogenesis. In fkh embryos gastrulation begins normally, but then the proctodeum disintegrates, the stomodeum does not invaginate and finally anterior and posterior midgut decay (Weigel et al., 1989). Following this massive cell death within the digestive tract, rudimentary ectopic head structures develop in place of foregut and in place of proctodeal derivatives like the hindgut. This phenotype is thought to reveal the homeotic function of fkh to promote terminal versus segmental development in the ectodermal parts of the gut (Jürgens and Weigel, 1988).

In this paper, I analyse the function of the gene serpent (srp) in the development of the midgut. srp had been identified in a screen for mutations with visible embryonic phenotypes due to its requirement for germ band retraction (Jürgens et al., 1984). Here, I show that srp is important for midgut development, and propose that srp acts as a homeotic gene promoting midgutspecific versus foregutand hindgut-specific differentiation. The genetic and phenotypic analyses suggest a role for srp downstream of hkb in midgut differentiation and predict the existence of a gene downstream of tll required for hindgutspecific morphogenesis and differentiation.

(1) srp^6G and srp^9L (Jürgens et al., 1984), apparently amorphic srp alleles. The phenotype of the homozygous embryos is indistinguishable from the phenotype of embryos hemizygous over the deficiency Df(3R)sbd¹⁰⁵. Cytologically, srp has been placed between 89A11 and 89B4 (Jürgens et al., 1984; Nelson and Szauter, 1992). (2) hkb², a hypomorphic hkb allele. (3) Df(3R)hkb^A, commonly used as hkb deficiency (Weigel et al., 1990). (4) Df(3R)tll^g, small deficiency of the tll locus; here used as tll null allele (Strecker et al., 1988). (5) tll^L10 (Jürgens et al., 1984), strong hypomorphic allele, termed tll¹ by Pignoni and coworkers (1988). (6) fkh^XT6: small deficiency of the fkh locus (Weigel et al., 1989). (7) Df(2L)TE116(R)GW11: small deficiency of the sna locus (Ashburner et al., 1990). The following enhancer trap insertions were used in the study: 1A121 (2^nd chromosome, Perrimon et al., 1991), A27 (2^nd chromosome, O’Kane and Gehring, 1987), A31 (1^st chromosome, O’Kane and Gehring, 1987; Ghysen and O’Kane, 1989) and A490.2M3 (3^rd chromosome, Bellen et al., 1989). A490.2M3 and 1A121 confer β-gal-expression mainly in prospective and definitive midgut, A27 in the hindgut and A31 in the Malpighian tubules. A31 is a P-element insertion at the fasciclin II locus (Grenningloh et al., 1991).

Double mutants were obtained by meiotic recombination. Single recombinant males were chosen following recessive markers and scored for their genotype by examining the embryonic phenotype of the progeny after successive crosses to two sets of appropriate tester females: hkb²srp^9L, hkb^Asrp^9L, srp^9Ltll¹, srp^6Gtll^g and srp^6Gfkh^XT6, in text and figures abbreviated with hkb²srp, hkb^Asrp, srp tll¹, srp tll^g and srp fkh.

The following antibodies were used in the study: anti-caudal (rabbit polyclonal; Mlodzik and Gehring, 1987); anti-crumbs (mouse monoclonal Cq4; Tepass and Knust, 1993); anti-cut (rat polyclonal F2; Blochlinger et al., 1990); anti-engrailed (mouse monoclonal 4D9; Patel et al., 1989); anti-fasciclin II (mouse monoclonal; Grenningloh et al., 1991); anti-fork head (mouse polyclonal; a gift from Yien Ming Kuo and Steve Beckendorf); anti-Krüppel (rabbit polyclonal; a gift from Chris Rushlow); anti-TN^1-123 (rabbit polyclonal; Kispert and Herrmann, 1993); anti-vasa (rabbit polyclonal; a gift from Ruth Lehmann); anti-wingless (rabbit polyclonal; Van Den Heuvel et al., 1989); anti-β-galactosidase (mouse monoclonal; Sigma, St. Louis); biotinylated anti-rabbit-, anti-ratand anti-mouse-IgG (goat; Jackson, Bar Harbor).

Embryos were fixed with 4% formaldehyde in PBS/heptane and immunostained following standard protocols. The bound antibodies were detected histochemically with the Vectastain ABC kit (Vector Labs, USA) using diamino-benzidine as chromogen. The embryos were mounted individually in Epon Durcupan (Fluka, Switzerland) and were photographed using either Kodak Ektachrome 160T or Agfapan APX 100 film on a Zeiss Axiophot equipped with Zeiss 20× (n.a. 0.50) or 40× (n.a. 0.75) Plan-Neofluar objectives and differential interference contrast optics.

srp embryos begin gastrulation normally and become morphologically distinguishable from wild-type embryos during stage 10 (Fig. 1). At this time, in wild-type embryos, the cells of the prospective posterior midgut, located at the bottom of the amnioproctodeal pocket, lose their epithelial character and form a mesenchyme which attaches to the visceral mesoderm (Fig. 1A). In srp embryos, the cells do not undergo the epitheliumto-mesenchyme transition (Fig. 1C). Instead they form a large cavity that is contiguous with the hindgut and is composed of a columnar epithelium resembling the epithelium of the hindgut (Figs 1D, 3F,H). This epithelial cavity appears to replace the posterior midgut. In the anterior part of srp embryos, the first morphological aberrations are detectable after the onset of the stomodeal invagination. No prospective anterior midgut cells (Fig. 2A), normally derived from the tip of the ventral furrow, attach to the posterior side of the stomodeum (Fig. 2C), and the anterior part of the midgut is not formed. The anterior part of the digestive tract later only consists of a blind-ended tube of ectodermal foregut (Figs 5C, 6D).

In addition to the lack of a morphologically distinguishable midgut in srp embryos, the cells of the prospective midgut do not express midgut-specific genes. For example, caudal (cad) protein is zygotically expressed in wild-type embryos in the developing posterior midgut, but not in the prospective hindgut (Macdonald and Struhl, 1986; Mlodzik and Gehring, 1987; compare Fig. 1A and B). In srp embryos, zygotic cad is not expressed in the developing posterior midgut. Like the hindgut, the entire epithelial cavity that appears to replace the posterior midgut is devoid of the protein. In contrast, cad expression in the posterior tip of the embryo, comprising the primordia of anal pads and some more anteriorly located epidermis, is not affected. The expression patterns of two other markers for midgut-specific cell differentiation, the enhancer trap insertions A490.2M3 (Bellen et al., 1989) and 1A121 (Perrimon et al., 1991), are similarly affected in srp embryos. They normally confer β-gal expression in both developing anterior and posterior midgut, but not in the developing hindgut (shown for A490.2M3 in Fig. 2A,B). In srp embryos, β-gal under the control of A490.2M3 (Fig. 2C,D) or 1A121 (data not shown) is not detected in prospective anterior or posterior midgut. Other aspects of the β-gal expression, like late expression in the epidermis for A490.2M3 or in the visceral mesoderm for 1A121, are not affected. Thus, there appears to be no endodermal differentiation in srp mutant embryos.

One aspect of ectodermal gut development is affected in srp embryos: morphologically the Malpighian tubules are difficult to distinguish in older mutant embryos. In addition, several genes normally expressed in the developing Malpighian tubules like cad or the enhancer traps A490 or 1A121 are not expressed in cells of the proctodeum of srp embryos (Figs 1D, 2D and data not shown). However, other observations lead to the conclusion that rudimentary Malpighian tubules are formed at the correct position between prospective hindgut and prospective posterior midgut. After stage 10 fasciclin II (fas II) is expressed in developing midgut and Malpighian tubules of wild-type embryos, but not in the hindgut (Grenningloh et al., 1991; Fig. 3A,B). In srp embryos, fas II is found in two groups of cells within the proctodeal epithelium at about the position where the primordium of the Malpighian tubules is located (Fig. 3C). There is no expression of fas II in prospective anterior or posterior midgut in srp embryos supporting the notion that no endodermal differentiation takes place. The fas II-expressing cells remain clustered posteriorly to the hindgut in later srp embryos (Fig. 3D) and are best seen when fas II expression is visualized by the enhancer trap insertion A31 (Ghysen and O’Kane, 1989; data not shown). In addition, these cells express the homeodomain protein cut (Fig. 3G,H) and the zinc finger protein Krüppel (data not shown) which are normally expressed in Malpighian tubules (Gaul et al., 1987; Blochlinger et al., 1990; see also Fig. 3C,D). Presumably, the lack of srp function permits the partial differentiation of Malpighian tubules which form bulges at the internal end of the proctodeum. This notion is supported by the finding that, in a fraction of old srp mutant embryos, uric acid (which normally is produced by terminally differentiated cells of the Malpighian tubules) is detectable within the lumen of the hindgut and the epithelial cavity (data not shown).

In srp embryos, the entire midgut is missing and no signs of endodermal differentiation are detectable. Which fate do these cells acquire that normally would have formed the anterior midgut? wingless (wg) expression can be used as a marker for the part of the ectodermal foregut adjacent to the anterior midgut. At stage 10, wg is first expressed in a ring of cells that surrounds the opening of the stomodeum (Van Den Heuvel et al., 1989). Later (stage 11) these cells move inward with the stomodeum (Fig. 4A) and form part of the proventriculus (stage 13 and later) where the esophagus connects to the anterior midgut. In srp embryos, the entire stomodeum and a posteriorly adjacent mesenchymal mass of cells express wg (Fig. 4B). These mesenchymal cells are positioned where the anterior midgut cells derived from the tip of the ventral furrow are normally located at this time of development. They are close to the mesoderm and morphologically difficult to distinguish from the latter (see Fig. 2C). The identification of the conspicuous mesenchymal cell mass in srp embryos as prospective anterior midgut is corroborated by the phenotypic analysis of embryos that also lack sna function. sna has been shown to be required for specification and internalization of the endodermal anterior midgut cells, but not for stomodeal development (Reuter and Leptin, 1994, see also Fig. 4D). In sna srp double mutants only the stomodeum expresses wg, and a cell mass expressing wg and attached to the posterior stomodeal surface is not present (Fig. 4C). It is therefore reasonable to assume that in srp embryos the cell mass expressing wg and located posterior to the stomodeum at stage 10 is formed by cells that normally would give rise to anterior midgut.

Since wg expression can be considered as marker for ectodermal foregut, I conclude that in srp embryos the prospective anterior midgut has been transformed to ectodermal foregut. Two findings support this notion: the aberrant expression of fkh in the anterior gut of srp embryos and the epithelial properties that the prospective anterior midgut cells acquire. Normally, fkh is expressed in the primordium of the stomodeum at the onset of gastrulation and, while the stomodeum invaginates, also in the anterior midgut (Weigel et al., 1989; see also Fig. 5A). fkh expression persists in the esophagus, a stomodeal derivative, until the end of embryogenesis, but the expression in the anterior midgut is only transient and disappears during germ band retraction (Weigel et al., 1989; see also Fig. 5B). In srp embryos, fkh expression is maintained in the prospective anterior midgut, and a significantly larger, fkh-expressing esophagus is formed (Fig. 5D). In addition, the prospective anterior midgut of srp embryos apparently acquires epithelial properties as are normally seen in the foregut. In wild-type embryos, crumbs (crb) protein, an EGF-repeat-containing transmembrane protein, is localized in the apical membrane of epidermis and foregut, but not of the midgut (Tepass et al., 1990; compare Fig. 6A). In srp embryos, crb protein is also found on the inner surface of a set of epithelial spheres that were apparently formed from the originally mesenchymal cells of the anterior midgut primordium located posteriorly to the stomodeum (Fig. 6C). I suggest that these spherical structures as well as the inner part of the foregut of srp embryos are composed of cells that have acquired epithelial properties and have adopted the fate of foregut instead of anterior midgut. Therefore, srp seems to act as a homeotic gene and might promote anterior midgut versus foregut development.

The cells of the epithelial cavity, which replaces the posterior midgut in srp embryos, morphologically resemble the cells of the hindgut and show the same organization in a columnar epithelium. crb protein is localized to the entire inner surface of the cavity (Fig. 6D) but is normally found on the surface of hindgut or Malpighian tubules but not of posterior midgut cells (Fig. 6B). Consistent with this finding is the altered expression of fkh in the posterior gut of srp embryos. Before and during amnioproctodeal invagination, fkh is expressed in the primordium of both posterior midgut and proctodeum (Weigel et al., 1989). Later, during germ band retraction, fkh expression disappears from the midgut but continues to be expressed in proctodeal derivatives, in hindgut and Malpighian tubules (Weigel et al., 1989; see also Fig. 5B). In contrast, fkh expression in srp embryos is maintained at a high level in the prospective posterior midgut during germ band extension and later in the epithelial cavity (Fig. 5C,D). The shift in fate of the prospective posterior midgut becomes clearly defined when hindgut-specific gene expression is considered. The enhancer trap insertion A27 (O’Kane and Gehring, 1987) confers β-gal expression in the hindgut but not in the Malpighian tubules (Fig. 7A) while, in srp embryos, β-gal is found in almost the entire epithelial cavity in addition to the hindgut (Fig. 7B). This finding clearly shows a transformation of cell fate from posterior midgut to hindgut. The only areas that do not express β-gal are the cell clusters of the rudimentary Malpighiantubules (data not shown). Even more specifically the transformation of posterior midgut into hindgut is indicated by an antibody against the N terminus of the mouse T gene (Kispert and Herrmann, 1993), which recognizes an antigen termed DTRA (for Drosophila T-related antigen) exclusively in the hindgut of the Drosophila embryo (Fig. 8A). The hindgutspecific expression makes D-TRA an excellent marker for this organ. As in the case of the enhancer trap A27, in srp embryos, the entire epithelial cavity in addition to the hindgut expresses D-TRA (Fig. 8B). Again, only the central clusters of cells that constitute the rudimentary Malpighian tubules are excluded from the expression domain (data not shown). Therefore, according to all the tissue-specific markers used in this study, the prospective posterior midgut of srp embryos does not adopt any midgut properties, but instead shows hindgut-specific character.

There is one indication that the epithelial cavity actually constitutes a serially duplicated hindgut rather than part of an enlarged hindgut. The engrailed (en) protein is not only expressed in posterior compartments of all segments of the trunk region of the embryo but also in specific sites of head and tail that are derived from the fusion of several segmental primordia. In addition, there is a stripe of en-expressing cells within the proctodeum that during germ band extension runs along most of the ventrally located side of the proctodeum. This expression does not reach the base of the budding Malpighian tubules (Fig. 9A) and persists in the hindgut until the end of embryogenesis. No en expression is found in the posterior midgut. In contrast, in srp embryos, a separate, second en stripe is expressed within the epithelial cavity which has formed from the prospective posterior midgut (Fig. 9B). Apparently, the hindgut-specific aspect of en expression has been duplicated. It seems that the hindgut has not merely expanded into the primordium of the posterior midgut, but has been serially duplicated at the expense of the posterior midgut. This pattern duplication in srp embryos suggests a cryptic serial homology between posterior midgut and hindgut. In the posterior part of the embryo, the function of srp appears to be to promote posterior midgut identity versus hindgut identity. A loss of srp function leads to a homeotic transformation of posterior midgut to hindgut.

The gene hkb is known to be required for the specification of the prospective anterior and posterior midgut and acts in a gap-gene-like fashion. In hkb embryos, neither anterior nor posterior midgut are formed and from the amnioproctodeal primordium at the posterior pole only the proctodeal cells invaginate during gastrulation (Gaul and Weigel, 1991). If srp is required for promoting posterior midgut versus hindgut identity, one would expect srp to act downstream of hkb and to exert that function only when hkb has established the prospective posterior midgut. Consequently, the transformation of posterior midgut to hindgut in srp mutants should occur only in embryos with full hkb function. This is indeed the case, since hkb srp double mutants have the same phenotype in gut development as hkb mutants. No additional, homeotic hindgut or additional en expression pattern element is present (Figs 7C, 9E,G). The size and morphology of the proctodeum and later of the hindgut are the same in double null mutant embryos and in embryos mutant for hkb (Fig. 9G and data not shown). In the anterior region of the hkb srp embryos, the anterior midgut and the stomodeum are missing as observed in hkb embryos (Fig. 9E,G). Since in all aspects of gut development the hkb phenotype is epistatic to the srp phenotype, it is very likely that srp acts downstream of hkb.

The formation of the hindgut in wild-type embryos requires tll function and so does the homeotic hindgut in srp embryos. In srp tll double mutant embryos, neither a duplicated hindgut structure like in srp embryos nor a hindgut like in hkb srp embryos develops although the posterior midgut primordium invaginates. Depending on the residual dose of tll function, either no hindgut at all is formed (srp tll^g, Figs 7G, 8E) or only a very small, rudimentary hindgut is formed which does not express en (srp tll¹, Figs 7F, 8C, 9D). Thus, in these mutant embryos none or at best a few of the cells of the posterior midgut invagination appear to contribute to differentiated embryonic tissue. In srp tll^g embryos, the fate of these cells is not clear at all and they probably die. In the srp tll¹ embryos, however, a partial rescue of some terminal structures takes place. For example, a rudimentary hindgut is formed that would be absent from tll¹ embryos (Figs 7E,F, 8C,D). The residual function of tll in tll¹ embryos is not sufficient to initiate terminal development posterior to the abdominal segments when hkb and srp act and direct the formation of the posterior midgut. However, when srp function is absent, the residual tll activity directs rudimentary tail and proctodeal development in the primordium of the posterior midgut. While the formation of the homeotic hindgut in srp embryos requires tll function, there is no difference between the development of the anterior gut structures in srp and in srp tll embryos (Fig. 9B,D and data not shown). tll is neither expressed in their primordia after nuclear cycle 13 nor is tll required for their development.

The gene fkh, which is thought to promote terminal versus segmental development in the gut, is also required for the formation of the homeotic gut structures in srp embryos. The additional hindgut disintegrates like the other proctodeal structures in srp fkh double mutant embryos (Fig. 9F). The stomodeum does not invaginate and, as in fkh embryos, no foregut develops in the srp fkh double mutants (Weigel et al., 1989). Thus, with regards to gut development, the terminal phenotype of srp fkh embryos is morphologically indistinguishable from the phenotype of fkh embryos. In this respect, the lack of fkh is epistatic to the lack of srp: both the enlarged foregut and the homeotic hindgut of srp embryos require fkh function. This finding is consistent with the aberrant fkh expression in these structures in late srp embryos (Fig. 5).

srp embryos show striking and very specific defects in the development of their alimentary canal. The endodermal anterior midgut turns into a part of the foregut and the endodermal posterior midgut is replaced by a hindgut structure. The formation of this hindgut structure does not appear to be a mere enlargement of the original hindgut, but rather a serial duplication since a discrete stripe of en expression is found in the additional hindgut as is usually only found in the normal hindgut (Fig. 9A,B). There seems to be an intact boundary between hindgut and homeotic hindgut. This view is supported by the formation of Malpighian tubules from a normal position between the primordium of the hindgut and the primordium of the posterior midgut (Fig. 3). However, in srp embryos the primordium of the posterior midgut follows a developmental program which would be appropriate for the neighbouring hindgut. Such a phenotype can clearly be described in terms of a homeotic transformation which follows the rules for loss-offunction alleles of the Bithorax complex and transforms in an anterior direction within the fate map. Consequently, one possible interpretation is that in Drosophila the posterior midgut and hindgut exhibit a serial homology, which normally remains cryptic by the action of srp. Accepting the arguments for a segmental character of the proctodeum (Weigel et al., 1989) whose primordium is located within the fate map posterior to the other segmented body parts, srp might have the function of preventing the even further posteriorly located primordium of the posterior midgut from acquiring segmental properties.

Also, in the anterior region of the srp embryo, a fate shift occurs which can be described as homeotic in nature: the anterior midgut is transformed into foregut. However, there is no basis to postulate a serial homology between these two body parts since a pattern duplication, as observed for the transformation of the posterior midgut, was not observed in the transformed anterior midgut. Nevertheless, it is reasonable to propose that srp has the function to promote midgut development in the anterior as well as in the posterior primordium.

The gene fkh also fulfills a homeotic function during gut development of Drosophila which, in a sense, is complementary to the function of srp. fkh has been proposed to promote terminal development versus segmental development in the ectodermal parts of the gut (Jürgens and Weigel, 1988). In the anterior region of fkh embryos, a transformation occurs from foregut into head structures. In the posterior region, posterior tail structures, including hindgut and Malpighian tubules, are transformed into head structures and anterior tail structures (Jürgens and Weigel, 1988). Thus, lack of fkh mostly affects ectodermal gut structures, hindgut, Malpighian tubules and foregut, whose primordia are adjacent to the primordia affected in srp embryos. However, in contrast to srp embryos, the transformations seen in fkh embryos are accompanied by massive cell death within the transformed body parts. The homeotic structures found in the mutant embryos are therefore very small and might have been formed during a regeneration process from the neighbouring tissue. The homeotic phenotype of srp embryos suggests that genes exist that specifically direct either foregut or hindgut development and which are repressed by srp in the prospective anterior or posterior midgut (Fig. 10). These foregutor hindgut-specific genes should be different from fkh since fkh acts in both foregut and hindgut and more generally promotes ectodermal terminal development.

srp function is required for midgut-specific gene expression. In srp embryos, all of the genes normally expressed in the primordia of anterior and posterior midgut, but not in the hindgut, like cad, fas II or the genes detected by the enhancer trap insertions A490.2M3 or 1A121 are not expressed in these primordia. srp therefore might function as a key activator of midgut-specific genes. In contrast, srp does not appear to be involved in the establishment of the midgut primordia, in setting their borders or their early movements during gastrulation like the amnioproctodeal invagination. Interestingly, also the part of the anterior midgut primordium that is considered as ectodermal because it invaginates into the embryo with the stomodeum (Technau and Campos-Ortega, 1985) is affected in srp embryos like all the other midgut primordia, which are considered as endodermal. They all do not express midgut-specific genes and do not differentiate into midgut. This finding underlines the importance of srp for midgut development and questions the significance of the assignment of the different anterior midgut parts to different germ layers (see below).

srp embryos fail to retract their germ band (Jürgens et al., 1984). This failure is not an indirect consequence of the homeotic transformation within the alimentary canal due to a possible sterical hindrance by the additional hindgut. srp fkh embryos, for example, which do not form any gut structures also fail to retract their germ band (Fig. 9F). It is not known which mechanisms actually drive germ band retraction, but the amnioserosa might play a role in this process. srp seems to be required for the late differentiation of the amnioserosa (Fig. 2) which might be a prerequisite for the amnioserosa to serve as a flexible hinge between the dorsal edges of the epidermis along the entire body length.

The positional information provided by the maternal terminal system of genes, is mediated in the zygote, at least within its posterior region, by the two gap genes hkb and tll (Casanova, 1990; Weigel et al., 1990). The highest doses of tll activity specify the proctodeum, while hkb determines the posterior midgut primordium in a region posteriorly adjacent to the proctodeal primordium. Both genes, hkb and tll, are responsible for the amnioproctodeal invagination and presumably regulate within different, but adjacent domains the same set of genes responsible for the cell shape changes that lead to the invagination (Fig. 10). The genes responsible for the specific differentiation and morphogenesis of the posterior midgut or the proctodeum are certainly different.

The phenotypes of srp, hkb and hkb srp embryos are best explained by a model in which srp acts downstream of hkb as the mediator of the midgut-specific differentiation and morphogenesis (Fig. 10). Apart from its role in germ band retraction, srp appears to act predominantly in the midgut primordia of the embryo, which are established by hkb. When these primordia are not established by hkb an additional effect of the lack of srp on gut development is not observed. Furthermore, a gene complementary to srp must exist, here termed Z, that specifically promotes hindgut development (Fig. 10). Z should be repressed by srp in the prospective posterior midgut, transcriptionally or functionally and, thus, in srp mutant embryos, the ectopic action of Z in the posterior midgut primordium would provide the basis for the homeotic phenotype. Most likely tll itself cannot fulfill the function of Z, since tll expression disappears immediately after the onset of gastrulation, long before Z might act and any sign of regional differences between posterior midgut and hindgut can be detected. This view can also explain the phenomenon that the lack of srp partially restores terminal development in embryos carrying a hypomorphic tll allele (Figs 7F, 8C). tll¹ embryos form the abdominal segment 7, but all structures derived from primordia posterior to this segment (except the posterior midgut) are missing. However, a tll-dependent gene like Z might still be transiently expressed overlapping with the hkb domain before it is repressed there by srp. When srp function is lacking, Z expression might be maintained in the prospective posterior midgut and drive the differentiation of the primordium to a rudimentary hindgut.

The model for the positional specification of the posterior part of the alimentary canal, with the separation between the function of the terminal gap genes hkb and tll in establishing the primordia and the function of terminal homeotic genes srp and Z in specifying the identity of posterior midgut and hindgut, is formally equivalent to the model for the positional specification of the trunk of the embryo. There, gap genes partition the embryo, establish the respective regions and initiate the region-specific expression of the homeotic genes of the Antennapedia or Bithorax complex. These homeotic genes in turn promote the segmentor region-specific differentiation (for a review see McGinnis and Krumlauf, 1992). However, it remains to be seen whether srp is indeed sufficient to promote region-specific differentiation, for example by testing whether ectopic expression of srp within the proctodeum will repress proctodeal development and direct differentiation towards midgut, as predicted from the model.

fkh is regulated by hkb and tll and has been postulated to mediate some of their functions in the developing alimentary canal (Casanova, 1990; Weigel et al., 1990; Gaul and Weigel, 1991). However, fkh certainly does not act upstream of srp since midgut-specific genes like the one detected by the enhancer trap insertion A490.2M3 are transiently expressed in the developing midgut of fkh embryos before the tissue disintegrates (Weigel et al., 1989). In contrast, β-gal-expression in srp embryos conferred by A490.2M3 is never detected in the prospective or definitive midgut. Therefore, within the midgut fkh appears to act in parallel to or downstream of srp and might be necessary to maintain the state of determination and differentiation achieved by srp. Later in development (after germ band retraction), srp is responsible for the repression of fkh in the midgut, a finding that places fkh clearly downstream of srp at this time of development. Interestingly, the expression of fkh in the peripheral yolk nuclei (Weigel et al., 1989; compare Fig. 5B) is dependent on srp (Fig. 5D). This observation suggests that in the yolk srp acts as an activator of fkh, and indicates a function of srp for the normal development of the yolk.

To design a model for the specification of the anterior gut analogous to the model shown in Fig. 10 is presently not possible. The ontogenesis of the anterior gut is more complicated and much less is known about the genes involved. It is not clear how the regionalization within the anterior alimentary canal is achieved: no zygotic genes are known that have a role in the determination of the foregut equivalent to that of tll and possibly Z for the determination of the hindgut. Presumably also in the anterior region of the embryo, srp acts downstream of hkb in midgut differentiation and represses foregut-specific genes, while other functions of hkb such as the setting of the anterior border of the mesoderm primordium are not mediated by srp.

It has been convenient to describe development as a progressive sequence of determinative events that subdivide the body of the embryo until the definitive regional specification of the adult is achieved. Within this sequence, gastrulation is the phase of morphogenetic movements that introduces the three germ layers, ectoderm, mesoderm and endoderm and which, in the end, generates the so-called ‘Körpergrundgestalt’. These germ layers are not only characterized by their initial position relative to each other, but also by the common prospective fate of cells within a germ layer. In insects, for example, only those cells that will form the epithelium of the midgut are generally considered as endodermal (for reviews see Siewing, 1969; Anderson, 1972a,b). In some species, these cells arise from the most anterior and the most posterior regions of the mesendoderm. They are internalized by the ventral furrow and subsequently attach to stomodeum and proctodeum. In other insect species, the endoderm appears to originate from specialized regions contiguous with the stomodeum or proctodeum, which following the invagination separate from the stomodeal or proctodeal epithelium. In this case, the distinction between endodermal and ectodermal regions of the gut would not be based at all on the initial positions of the germ layers relative to each other but merely on the prospective fate of the cells. (Structures originating from stomodeum and proctodeum themselves, like foregut and hindgut, are considered as ectodermal because they have a number of properties in common with the epidermis and form cuticle, for example.) During the development of Drosophila both alternatives for the formation of the midgut epithelium are used. The posterior part of the midgut invaginates with the proctodeum early in gastrulation, and the definition of posterior midgut and proctodeum as endoderm and ectoderm, respectively, is based on their prospective fates. The situation is more complex for the anterior midgut. Part of the anterior midgut originates from a mesendodermal primordium which is formed by the anterior tip of the ventral furrow. The other part of the anterior midgut invaginates with the stomodeum, several hours after the invagination of the proctodeum. In contrast to the posterior midgut, this part of the anterior midgut primordium has been viewed as ectodermal based on its external position after the onset of gastrulation. In this case, its prospective fate as midgut has not been considered. Thus, in Drosophila, the distinction of the various midgut primordia as endodermal or ectodermal is indeed arbitrary and not consistent. However, several observations support a common classification of these primordia as endoderm. Despite their different origins, all the midgut parts have one morphological feature in common that clearly distinguishes them from ectodermal gut parts: their transient mesenchymal state during the phase of cell migration that precedes the formation of the midgut epithelium (Campos-Ortega and Hartenstein, 1985; Hartenstein et al., 1992). Another indication is the apparent equivalence of cells from the different midgut primordia when they are exchanged by transplantation (Technau and Campos-Ortega, 1986). A third unifying criterion is genetic. On the one hand, all primordia of the midgut are established by the terminal gap gene hkb (Weigel et al., 1990). However, hkb embryos also fail to form the stomodeum (Reuter and Leptin, 1994) and thus, the function of hkb is not confined to the midgut primordia. On the other hand, the function of the gene srp is specifically required for the formation of the entire midgut and no part of its primordium, no matter where it is located, develops into midgut in srp mutant embryos. This supports the view that all the parts of the midgut primordium are essentially equivalent and therefore, it might be reasonable to propose that they all belong to one and the same germ layer, the endoderm. The molecular analysis of srp will provide the means for testing whether srp indeed has the capacity to determine the typical properties of the cells of the endodermal germ layer.

I am indebted to Maria Leptin for supporting this work in her laboratory through grants from the Behrens-Weise-Stiftung and the Human Science Frontier Program. I would like to thank Andreas Kispert, Elisabeth Knust, Yien Ming Kuo, Paul Lasko, Ruth Lehmann, Roel Nusse, Chris Rushlow, Helen Skaer and Uwe Walldorf for providing antibodies, and Günter Brönner, Jordi Casanova, Walter Gehring, Cahir O’Kane, Norbert Perrimon and Helen Skaer for sending fly stocks. The manuscript was significantly improved by insightful comments from José Casal, Camila Esguerra, Andreas Kispert and Maria Leptin. Katrin Brenner, Gertrud Scheer and Sandra Schäfer helped with the photographic reproduction of the figures.