Interacting functions of snail, twist and huckebein during the early development of germ layers in Drosophila

The germ layers in Drosophila are formed by two morphogenetic movements during gastrulation. The mesoderm and the anterior endoderm are formed by the invagination of the ventral furrow and the posterior endoderm is formed by the invagination of the amnioproctodeum. The cells remaining at the surface of the embryo constitute the ectoderm. They later will form mainly epidermis and neural tissue, while the endodermal cells will form most of the midgut and the mesodermal cells form muscles, heart, gonads and fatbody.

The primordia of the germ layers in the blastoderm have been mapped by various techniques (Poulson, 1950; Underwood et al., 1980; Hartenstein et al., 1985; Foe, 1989). The most ventral cells between 10 and 80% egg length contribute to the mesoderm while the posterior cap of the embryo constitutes the posterior endoderm primordium, also called the posterior midgut (PMG) primordium (Fig. 1A). Anterior to the PMG primordium resides the primordium of the proctodeum. It is thought to consist of a ring of cells around the entire circumference of the embryo which separate the PMG from the mesoderm primordium on the ventral side (Technau and Campos-Ortega, 1985). The proctodeal primordium invaginates together with the prospective PMG (Fig. 1B) and gives rise to anal pads, hindgut and Malpighian tubules, all considered to be ectodermal derivatives.

The anterior end of the ventral furrow is thought to be the primordium of the anterior endoderm, which forms part of the anterior midgut (AMG) (Poulson, 1950; Technau and Campos-Ortega, 1985). We will refer to it as endodermal AMG primordium to distinguish it from the stomodeal AMG primordium. The stomodeum, considered as ectodermal, begins to invaginate during stage 10 (Fig. 1B) and is the precursor of the foregut (which consists of pharynx, esophagus and proventriculus). In addition, the stomodeum provides the cells of the anterior part of the AMG (Technau and Campos-Ortega, 1985). Several zygotic genes, under control of different maternal morphogens, set up the primordia of the germ layers. The nuclear gradient of the maternal gene product dorsal (dl) directs the ventral expression of two zygotic transcription factors, twist (twi) and snail (sna) (Boulay et al., 1987; Thisse et al., 1988, 1991; Ip et al., 1992). Both twi and sna are required for the formation of the ventral furrow and the mesodermal germ layer (Simpson, 1983; Grau et al., 1984; NüssleinVolhard et al., 1984). They are thought to activate the genes responsible for ventral furrow formation and for mesodermal differentiation (Leptin and Grunewald, 1990; Leptin, 1991). In addition, sna represses genes in the mesodermal primordium that are involved in ectodermal development (Nambu et al., 1990; Leptin, 1991; Arora and Nüsslein-Volhard, 1992; Kasai et al., 1992).

The primordium of the posterior endoderm is set up by the maternal genes of the terminal system (Nüsslein-Volhard et al., 1987; Klingler et al., 1988). One of these genes, torso (tor), indirectly activates the expression of two zygotic transcription factors, huckebein (hkb) and tailless (tll), and thereby patterns the posterior region of the embryo (Weigel et al., 1990a; Pignoni et al., 1990; Casanova, 1990). hkb is expressed in the posterior cap of the blastoderm (Brönner and Jäckle, 1991) and is required for the invagination of the PMG primordium as well as for its specification (Weigel et al., 1990a). The invagination of the proctodeum proceeds normally in hkb embryos: this aspect of the amnioproctodeal invagination is dependent on tll (Strecker et al., 1988). tll is expressed in the same domain as hkb posteriorly (Pignoni et al., 1990; Brönner and Jäckle, 1991) and, in addition, in the primordia of the proctodeum and of the posterior segmented region of the embryo (Pignoni et al., 1990). tll is required for the formation of all structures posterior to abdominal segment 7 including anal pads, hindgut and Malpighian tubules (all derived from the proctodeum), but it is not required for the formation of the posterior midgut (Jürgens et al., 1984; Strecker et al., 1988).

Both genes, hkb and tll, are also expressed in the anterior region of the embryo (Brönner and Jäckle, 1991; Pignoni et al., 1990). There, tll is required for the formation of parts of the brain and head skeleton (Strecker et al., 1988), and hkb for the formation of the anterior midgut (Weigel et al., 1990a). In summary, the establishment of both anterior and posterior midgut primordia depend on hkb activity.

In this paper, we show that hkb sets the anterior and the posterior limits of the mesoderm primordium, acting in two different modes. hkb not only determines endodermal fate in both terminal regions of the embryo, but also prevents mesodermal development there. In addition, we show that hkb, sna and twi cooperate in the determination of anterior gut structures.

(1) bcd^E1: a small deletion of the bcd locus (Berleth et al., 1988). (2) bcd^E1tsl⁰³⁵: double mutant for bcd and the strong torsolike allele tsl⁰³⁵ (Nüsslein-Volhard et al., 1987; Frohnhöfer, 1987). (3) fkh^XT6: small deficiency of the fkh locus (Weigel et al., 1989b). (4) hkb^A and hkb²: a small deficiency of the hkb locus and a strong hypomorphic allele (Weigel et al., 1990a). The ventral furrow in embryos homozygous for hkb^A (for example Fig. 4B) extends to the anterior tip while, in embryos homozygous for the hypomorphic hkb² allele, the ventral furrow extends only about half the way to the anterior tip of the embryo, similar to embryos from tor mothers (Fig. 7F). In hkb²/hkb^A embryos, the extent of the phenotype varies between complete extension like in hkb^A or partial extension like in hkb². This indicates that no additional locus is removed by the deficiency hkb^A that would prevent ventral furrow formation in the anterior part of the embryo.(5) sna^IIG and sna^RY1: based on the cuticular phenotype amorphic sna alleles (Grau et al., 1984; Nüsslein-Volhard et al., 1984; Boulay et al., 1987); both alleles code for sna protein (data not shown), which seems to be defective in its C-terminal half (Boulay et al., 1987).(6) Df(2L)TE116(R)GW11: used as sna deficiency (Ashburner et al., 1990) and as a control for the specificity of our anti-sna antibody.(7)tll^g and tll^L10: a small deficiency of the tll locus and a strong hypomorphic allele (Jürgens et al., 1984; Strecker et al., 1988). Often we used the overlapping deficiencies tll^g and tll^e which give a small synthetic deficiency (Strecker et al., 1988). (8) hkb²tll^g: double mutant as described (Weigel et al., 1990a). (9) twi^EY53 and twi^ID: amorphic twi alleles (Nüsslein-Volhard et al., 1984; Thisse et al., 1987b). (10) sna^IIGtwi^IIH and sna^IIGDf(2R)S60: double mutants as described (Arora and Nüsslein-Volhard, 1992). (11) tor^TC/tor^QL: transheterozygous combination of strong alleles which do not express detectable tor protein (Sprenger, 1991). (12) tor^XR1: a deletion of the tor locus (Sprenger et al., 1989) (13) tor⁴⁰²¹: gain-of-function allele of tor (Klingler et al., 1988). (14) A490.2M3: enhancer trap insertion conferring β-gal expression in midgut and midgut primordia (Bellen et al., 1989). The insertion (third chromosome) was crossed into the background of twi^ID, sna^IIG and sna^IIGtwi^IIH. (15) 1A121: enhancer trap insertion conferring β-gal expression in the midgut and midgut primordia, and later also in the visceral mesoderm (Perrimon et al., 1991, our unpublished results). The enhancer trap insertion was transiently crossed into the background of hkb^A.

Oregon R flies were used as wild type.

The anti-sna antibody was raised in rabbits against the full-length sna protein expressed in E. coli in the PET3c vector. It was affinitypurified using the sna protein coupled to CNBr-activated Sepharose. Anti-twi antibody was a gift from S. Roth (Roth et al., 1989), antimuscle myosin heavy chain antibody (R722) from D. Kiehart (Kiehart and Feghali, 1986) and anti-fkh antibody from D. Weigel (Weigel et al., 1989a). Murine monoclonal anti-β-galactosidase (α-β-gal) antibodies were purchased from Sigma (St. Louis, USA) and biotinylated goat anti-mouse IgG antibodies or biotinylated goat anti-rabbit IgG antibodies from Jackson (Bar Harbor, USA).

Embryos were fixed and stained following standard protocols, and the antibodies were detected histochemically using the Vectastain ABC kit (Vector Labs, USA). Sectioning of the embryos was performed as described (Leptin and Grunewald, 1990). Embryos were either mounted individually in Araldite or en masse in methyl salicylate and photographed with a Zeiss Axiophot equipped with Nomarski optics using either Kodak Ektachrome 160T or Agfapan APX 100 film.

The hkb cDNA used for the in situ detection of hkb mRNA was a gift from G. Brönner (Brönner and Jäckle, 1991). zfh-1 has been described by Fortini and coworkers (Fortini et al., 1991) and DFR1, one of the Drosophila homologues of the FGF receptor, by Shishido and coworkers (Shishido et al., 1993). The zfh-1 and the DFR1 probes used were a gift from J. Casal and originated from a molecular screen for ventrally expressed genes (J. Casal and M. Leptin, unpublished data). Probes were labelled with digoxigenin-dUTP (Boehringer) and used for in situ hybridization as described (Tautz and Pfeifle, 1989). For the simultaneous detection of sna and hkb gene products, embryos were first processed for in situ hybridization, subsequently hkb mRNA was detected using digoxigenin-dUTP-labelled cDNA probe and finally sna protein using the anti-sna antibody.

sna and twi are required for ventral furrow formation but their patterns of expression do not define the anterior and posterior borders of the ventral furrow. They both are expressed on the ventral side of the blastoderm embryo (Thisse et al., 1987a, 1988; Alberga et al., 1991), including the anterior 20% of the egg where no ventral furrow is formed (Fig. 2A-C). In the posterior part of the embryo, the border of sna expression sharpens after the onset of cellularization. It then coincides with the posterior border of the ventral furrow (see Fig. 7), while twi continues to be expressed beyond the border around the posterior pole. In a simple model (Fig. 2D) for how the anterior and posterior limits of the ventral furrow are set, a gene X active in the anterior region of the embryo counter-acts the effect of sna and twi on their target genes which otherwise would promote ventral furrow formation. In the posterior region, a postulated gene Y, presumably under the control of the terminal system, represses sna expression and thereby delimits the ventral furrow posteriorly.

The zygotic terminal gap gene hkb is expressed at both termini of the embryo at blastoderm stage (Brönner and Jäckle, 1991; see also Fig. 3A). At early gastrulation the spatial relationship of hkb expression to the ventral furrow is striking: anteriorly and posteriorly it lies directly adjacent to the ventral furrow (Fig. 3B,C). Therefore, hkb appeared as a good candidate for both the factor X and the factor Y (Fig. 2D). The phenotype of hkb embryos indicates that hkb does indeed have these functions. In hkb null mutant embryos, the ventral furrow extends all the way to the anterior tip of the embryo (Fig. 4B), now encompassing the entire snaand twi-expressing domain (see also below).

Mesodermal differentiation normally appears to be repressed by hkb, as the expression patterns of mesodermal genes are expanded in hkb embryos. An example for one such gene is zfh-1, which is normally expressed under the control of sna and twi only in those cells that form the ventral furrow (Lai et al., 1991 and Fig. 5A,E). In hkb embryos, the domain of zfh-1 expression is expanded anteriorly (Fig. 5B,F). Thus, hkb normally represses zfh-1 expression and therefore appears to antagonize the effect of the zfh-1 activators sna and twi. At later stages, cells expressing zfh-1 (Fig. 5G) fill the head of the hkb embryo (Fig. 5I) and seem to form disorganized muscle cells at the tip of the embryo (Fig. 5K,L). We take this as an indication for the expansion of the mesoderm at the expense of head structures in hkb embryos.

Since during cellularization the posterior border of sna expression coincides with the anterior border of hkb expression (Fig. 6), hkb also appeared as a good candidate for the zygotic gene Y that sets this border of sna expression. The phenotypic analysis of hkb mutants shows that indeed this function is fulfilled by hkb. In hkb embryos, the posterior boundary of the ventral furrow is shifted to the posterior (Fig. 5C,D) and sna protein is expressed up to the posterior pole (Fig. 7G,H). This shows that hkb also acts as the proposed factor Y that represses sna expression at the posterior terminal region and presumably thereby delimits the ventral furrow. Concomitantly with setting the border of the ventral furrow, hkb sets the border of the mesoderm primordium. The expression domain of zfh-1 (again serving as a mesodermal marker) extends in hkb embryos to the posterior pole during the blastoderm and gastrulation stage (Fig. 5B,F). Later, the posterior expression domain of zfh-1 within the mesoderm is enlarged indicating an expansion of the mesodermal germ layer at the expense of the endoderm (Fig. 5G,H).

tll, the other terminal gap gene, and fork head (fkh), which mediates some of the functions of hkb and tll (Jürgens and Weigel, 1988; Weigel et al., 1990a; Gaul and Weigel, 1991), are not involved in setting the borders of the mesoderm primordium. Neither the blastoderm expression of zfh-1 (Fig. 5C,D) and of sna (Fig. 8A,B) nor the position and form of the ventral furrow (data not shown) are changed in tll or fkh mutant embryos. The maternal terminal system of positional information appears to set the posterior border of the mesoderm primordium exclusively through the zygotic action of hkb. This is particularly obvious in embryos derived from tor gain-offunction mothers whose posterior border of the mesoderm primordium is shifted to about 30% egg length due to the hyperactivity of tor (Fig. 8C,E). hkb embryos derived from such tor mutant mothers express sna from pole to pole (Fig. 8F) and correspondingly form an extended ventral furrow (data not shown) similar to hkb embryos from tor⁺ mothers. The lack of tll function, however, has no consequence for the expression of sna, even in the background of the hypermorphic tor allele (Fig. 8D). Although tll mediates many effects of tor, tll is not responsible for delimiting the mesoderm primordium.

If hkb represses sna in the posterior region of the blastoderm why does it not have the same effect in the anterior? Apparently, bcd counteracts the inhibitory effect of hkb on sna expression. hkb itself is expressed anteriorly under the control of both bcd and tor. In embryos derived from tor lack-offunction mutant mothers (for simplicity called ‘tor embryos’) there is still some anterior expression of hkb (Brönner and Jäckle, 1991 and Fig. 9C). This expression is controlled by bcd as no blastoderm expression of hkb is detectable in embryos lacking both bcd and tor function (e.g. bcd tsl double mutants, Fig. 9D). In bcd embryos, hkb is expressed at both poles, but the anterior domain is smaller than in wild-type (Fig. 9B). Posteriorly, hkb is dependent only on tor (Brönner and Jäckle, 1991 and Fig. 9C). Thus, hkb is activated by both tor and bcd at the anterior pole and by tor alone at the posterior pole. However, tor and bcd direct hkb expression in specific subdomains in the anterior part of the embryo (compare Fig. 9B and C). Apparently, bcd and tor act maximally at slightly different points within the anterior pole, bcd at the apex and tor slightly dorsal to the apex. The tor-dependent (Fig. 9B) and the bcddependent (Fig. 9C) subdomain partially overlap, and their sum seems to give the distribution of hkb in the wild-type domain (Fig. 9A).

We now examined the sna expression in the various mutant embryos. In tor embryos (Fig. 7E) and in embryos lacking both bcd and tor function (data not shown), sna is expressed from pole to pole as in hkb embryos (Fig. 7G). In contrast, in bcd embryos (Fig. 7C,D) sna expression is not expanded at the posterior or the anterior pole and instead shows a sharp border of expression at a clear distance from the anterior pole. Not unexpectedly, this anterior border looks very similar to the posterior border, since in bcd mutant embryos the posterior terminal region is duplicated at the anterior pole (Frohnhöfer and Nüsslein-Volhard, 1986). In bcd embryos, the residual anterior hkb seems to repress sna, and we conclude that bcd normally inhibits hkb from repressing sna at the anterior pole. The ventral furrow in the mutant embryos follows the expression patterns of hkb and sna in agreement with the model depicted in Fig. 2D. It invaginates where sna and twi but not hkb are expressed. In tor embryos the ventral furrow extends towards the posterior pole of the embryo (Fig. 7F) as in hkb mutants (Fig. 7H), but it only partially extends towards the anterior pole of the embryo (Fig. 7F), unlike in hkb null mutants (Fig. 7H). hkb still represses ventral furrow formation in its anterior, tor-independent expression domain.

Is there any function of hkb anteriorly besides antagonizing the formation of the ventral furrow? Can any function be attributed to the expression of sna and twi anterior to the region which contributes to the ventral furrow? hkb embryos do not form an anterior midgut (Weigel et al., 1990a and Fig. 10C). Fig. 10 shows that they also have no stomodeal invagination, and therefore most of the ectodermal foregut of the embryo does not develop. Often, a rudimentary pharynx forms late during the aberrant head involution (Fig. 5L). Interestingly, embryos mutant for both sna and twi also do not form an AMG or a stomodeum (Fig. 11D), and usually they lack all anterior gut structures (Fig. 11F). This phenotype cannot be attributed to the lack of anterior hkb expression: in sna twi double mutant embryos, hkb expression is not changed (data not shown). Similarly, the expression of sna and twi does not require hkb function (Fig. 7F and data not shown), and thus sna, twi and hkb do not act in a simple epistatic pathway. Rather, they appear to act in a synergistic fashion.

fkh, one of the genes acting downstream of hkb, has been reported to be essential for a proper stomodeal invagination (Weigel et al., 1990b). fkh therefore might also be a target for sna and twi, and the phenotype of sna twi double mutants might be attributable to the dependence of fkh on sna and twi. This, however, is not the case. In wild-type embryos, fkh is expressed within the primordium of the stomodeum, in the invaginating stomodeum and later in the esophagus (Weigel et al., 1989a,b). In sna twi mutant embryos, early fkh-expression is normal (Fig. 11A-D). The group of fkh-expressing cells on the anterior surface of older mutant embryos (Fig. 11E) are probably the cells that would have invaginated and formed the esophagus in a normal embryo (Fig. 11F). Thus, at least some of the aspects of gene expression that depend on hkb, like the expression of fkh, are independent of sna and twi, and we conclude that the failure to form a stomodeum cannot be explained by an effect of sna and twi on fkh expression. Rather, sna and twi might act in parallel to hkb on other genes that are needed for the stomodeal invagination to occur. sna and twi appear to act in a redundant fashion since, in embryos lacking either sna or twi, we observe no defects in the formation of the stomodeum (Fig. 12B,C). This is in contrast to the formation of the ventral furrow where sna and twi exert different functions.

twi and sna have specific functions for the development of the endodermal AMG primordium. We followed the development of the AMG by the enhancer trap A490.2M3 which from the formation of the stomodeal plate onwards (stage 9) confers uniform β-gal expression in the primordia of AMG and PMG. In particular, β-gal is expressed both in the epithelium of the stomodeal AMG primordium and in the mesenchymal cells of the endodermal AMG primordium. These two primordia are morphologically well distinguishable at stage 10 (Fig. 12A,E) and are differentially affected by the lack of sna or twi function. In twi mutants, both parts express genes typical for the midgut as visualized by A490.2M3. However, early in gastrulation the endodermal primordium fails to invaginate (Fig. 12B,I), like the ventral furrow. This finding corroborates the notion that the endodermal midgut primordium is derived from the tip of the ventral furrow (Poulson, 1950; Technau and Campos-Ortega, 1985). Interestingly, during the course of germ band extension the cells ingress into the interior of the embryo and attach to the posterior surface of the stomodeum (Fig. 12F,J,K). Thus, the lack of twi function appears to be without consequence for the ultimate fate of the AMG, although the ventral furrow is not formed and thus the early morphogenesis of the AMG is abnormal.

In sna mutants the endodermal part of the AMG primordium does not become specified. No cells corresponding to the group of ingressing cells in twi mutants are distinguishable (data not shown). In addition, there is no midgut-specific gene activity at the place where the anterior tip of the ventral furrow would have been located (Fig. 12C,G). Therefore, sna appears to function as an activator in the specification of the endodermal AMG primordium. This is in contrast to the proposed role of sna in the mesodermal primordium where it acts as a suppressor of lateral, neuroectodermal genes. As mentioned above, the stomodeum invaginates almost normally and expresses midgut-specific genes. However, no cells attach to its posterior surface, and the stomodeum stays epithelial for a long period of time (data not shown). Thus, sna and twi, previously considered as specifically required for mesodermal differentiation, are also involved in the specification of endodermal and ectodermal structures that are derived from the ventral side of the embryo anterior to the mesoderm primordium.

The expression of sna and twi is initiated in the ventral region of the blastoderm embryo by high doses of nuclear dl (Roth et al., 1989; Thisse et al., 1991; Ip et al., 1992). While there is no apparent restriction of twi expression along the anteroposterior axis (twi protein and RNA are found from the anterior to the posterior pole of the blastoderm embryo), sna RNA and protein become excluded from the posterior pole at the onset of cellularization (Kosman et al., 1991 and Fig. 7A). We have shown that hkb is responsible for this repression (Fig. 7G) and propose that hkb acts as a direct negative regulator of sna transcription (Fig. 13). Initially, before the end of nuclear cycle 13, sna is weakly expressed around the posterior pole (Alberga et al., 1991; Ray et al., 1991; data not shown). We hypothesize that at this time hkb has not yet been translated and therefore cannot be active as a transcriptional repressor.

Fig. 13. Model for the partitioning of the ventral side of the Drosophila embryo by sna, twi and hkb under the control of the maternal genes dl, tor and bcd. dl has the capacity to activate the expression of sna and twi over the entire length of the embryo. The terminal gap gene hkb is expressed under the control of tor and bcd anteriorly, and of tor posteriorly. hkb sets both the anterior and the posterior border of the ventral furrow: posteriorly by repressing sna and anteriorly by antagonizing sna and twi function. The effect of the maternal gene tor on sna expression can be fully accounted for by its action through hkb (see Figs 7 and 8) rather than through a reduction of maternal dl activity by tor at the termini of the embryo, as proposed previously (Casanova, 1991). The repression of sna expression in the terminal regions of the embryo appears to be strictly zygotic. In the anterior region of the embryo, although hkb expression is activated by bcd, bcd simultaneously prevents the repression of sna by hkb. hkb and, in a redundant fashion, sna or twi are needed for the specification of the stomodeum. With the exception of the regulation of twi by dl and of sna by dl and twi, none of the indicated interactions are known to be direct. The activation of hkb by tor cannot be direct since tor is a transmembrane tyrosine kinase, but the other interactions could be due to direct transcriptional regulation. We still do not know the downstream genes that are regulated by hkb, sna and twi and that actually trigger the morphogenetic cell shape changes leading to ventral furrow formation, for example. Unknown as well is the gene that determines the anteroposterior extent of the endodermal AMG, which otherwise is specified by hkb and sna. For further discussion see text.

Interestingly, sna and hkb gene products coexist during cellularization and early gastrulation in the anterior part of the embryo. Here, unlike at the posterior pole, hkb does not act as a transcriptional repressor of sna. We have shown that bcd is responsible for the inhibition of the repression (Fig. 7C), but we do not know the mechanism by which bcd modulates the action of hkb on sna. However, the inhibition by bcd of the repressive effect of hkb is apparently not complete. If it were complete, then wild-type embryos should express the same high level of sna at the anterior pole as is observed in hkb mutant embryos. Instead, in wild-type embryos, sna expression fades out in graded fashion around the anterior pole (Fig. 7A), while it is strongly expressed in this region in hkb mutants (Fig. 7G). There seems to be a delicate balance between the effects of bcd, tor and hkb on sna expression. sna is strongly expressed in the anterior pole of tor embryos (Fig. 7E), although there is still some weak hkb expression anteriorly (Fig. 9C). We suggest that, in tor embryos, bcd is able to inhibit completely the repressing effect of the low dose of hkb.

hkb determines the PMG primordium (Weigel et al., 1990a) and is expressed in the region of the embryo (Brönner and Jäckle, 1991) to which this primordium has been mapped. We propose that hkb delimits the posterior border of the ventral furrow and of the mesoderm primordium by setting the posterior border of sna expression. tll, the other terminal gap gene, is not involved in the regulation of sna or of the boundaries of the ventral furrow (Figs 5, 8). This is surprising when we consider the blastoderm fate map for the posterior primordia and compare it with the expression domains of the genes that have been suggested to set up these primordia. The map places the primordium of the proctodeum, which is set up by tll, as a ring between the primordium of the PMG and the primordium of the mesoderm (Hartenstein et al., 1985). In contrast, hkb and sna are expressed adjacent to each other (Fig. 6) and are not required for proctodeal development. If the region expressing sna constitutes the mesoderm primordium one has to conclude that mesoderm and PMG primordium are juxtaposed without an intervening proctodeal primordium. This view is consistent with the expansion of the mesoderm at the expense of the posterior midgut in hkb embryos (Fig. 5) and with the finding that tll function is irrelevant for the posterior border of the mesoderm primordium. Thus, all these data suggest that mesodermal and PMG primordia are not separated by the proctodeal primordium. This view does not necessarily contradict the data that were used to compile the fate map (Technau and Campos-Ortega, 1985), which suggests that there is an intervening proctodeum. These authors transplanted HRPlabelled cells after the formation of the ventral furrow, so that the resulting fate map was actually a fate map of the early gastrula. Therefore, the deduction of the blastoderm fate map required the assumption that the ventral furrow forms anterior to the cells contributing to the proctodeum. (An exact judgment in live embryos is very difficult, since with the extension of the germ band the ventral furrow joins the amnioproctodeal invagination). However, the results are equally consistent with mesoderm forming ventrally at the same position alongthe anteroposterior axis where more laterally and dorsally located cells form the proctodeum.

sna, twi and hkb function in the development of structures that are derived from the anterior pole of the embryo. We relate these functions to the anterior expression domains of the genes and the anterior aspect of the fate map.

The ventral furrow ends anteriorly in a Y-shape, which is visible when viewed from the ventral side. This structure is presumably formed for topological reasons when a long rectangular array of cells invaginates. Without hkb function the ventral furrow extends to the tip of the anterior pole (Figs 4, 5). We observe this phenotype in hkb null mutants as well as in embryos from bcd tsl mothers (data not shown). We propose that hkb sets the anterior border of the ventral furrow by antagonizing the action of sna and twi on those target genes that actually drive the cell shape changes leading to the ventral furrow (Fig. 13). The expression of hkb is compatible with this proposed function (Fig. 3B,C). hkb is expressed within the cells that form the anterior rim of the Y-shape, invaginate with the ventral furrow, but do not constrict apically (Fig. 3C). Presumably these cells are pulled in passively by the action of the immediately neighbouring ventral furrow cells. Two mesoderm-specific genes, which are expressed under the control of sna and twi within the entire mesoderm primordium, are indeed repressed by hkb: zfh-1 (Fig. 5) and DFR1 (data not shown). They are repressed by hkb within the anterior ventral region of the embryo including the anterior rim of the Y-shape (Fig. 5 and data not shown). This part of the rim presumably does not belong to the mesoderm primordium, but constitutes the primordium of the endodermal AMG primordium (see below). In addition, in the head of hkb mutants, we find supernumerary muscle cells at the surface of the embryo (Fig. 5K). These findings indicate that hkb not only sets the anterior limit of the ventral furrow but also of the mesoderm primordium.

We suggest that the expression domains of the three genes hkb, sna and twi are sufficient to position ventral furrow and mesoderm primordium along the anteroposterior axis (Fig. 13). The mesoderm primordium is limited to the region where sna and twi but not hkb are expressed. Of the two genes, sna and twi, the expression of sna appears to be more indicative for mesodermal fate. The expression domain of sna is contained within the twi domain, and its lateral boundaries define the precise lateral limits of the ventral furrow (Kosman et al., 1991; Leptin, 1991) as is also the case for the posterior limit. In contrast, twi expression is initially significantly wider than sna both laterally and at the posterior border of the ventral furrow (Kosman et al., 1991; Leptin, 1991) and becomes restricted to the mesodermal germ layer when the germ band has extended about half-way.

hkb is required for the development of the entire AMG (Weigel et al., 1990a, see also Fig. 9) and sna for the specification of the endodermal AMG primordium (Fig. 12C,G). Are the expression patterns of hkb and sna compatible with the location of the endodermal AMG primordium at the tip of the ventral furrow? sna is expressed in the entire anterior ventral region of the embryo and hkb in all the cells of the anterior cap including those which form the anterior rim of the Y-shaped end of the ventral furrow (Fig. 3C). These latter cells, which invaginate with the ventral furrow, might therefore constitute the endodermal primordium of the AMG. However, the two genes hkb and sna cannot be sufficient to give the spatial information for the specification of the AMG primordium since their expression domains cover a much larger region. Thus, to determine the extent of the endodermal AMG primordium along the anteroposterior axis, a third factor has to be proposed: either a particular concentration of the morphogen bcd itself or a zygotic gene under the control of bcd. We then would suggest that the positional specification for the endodermal primordium is given by sna, hkb and the unknown third gene in a combinatorial fashion.

The other structure that is formed anteriorly from the ventral side of the embryo is the stomodeum. We have shown that hkb is required for the stomodeal invagination (Fig. 10) and thus for the proper development of stomodeal derivatives like the esophagus. In addition, sna or twi are required for the stomodeal invagination as well, since in sna twi double mutants this morphogenetic movements does not occur (Fig. 11). However, several aspects of the functions of hkb, sna and twi in stomodeal development remain unclear. First, we do not have an explanation for the redundancy shown by sna and twi. Possibly, sna and twi have merely permissive roles in foregut development, for example by repressing more lateral and dorsal fates. One such case, where sna and twi act redundantly, has indeed been observed. The anterolateral mitotic domain δ9, which flanks the stomodeal primordium, is expanded around the ventral side of sna twi double mutant embryos, but not of sna or twi embryos (Arora and Nüsslein-Volhard, 1992). Moreover, it is not known whether hkb has an instructive role for the stomodeal invagination or whether hkb is merely required to repress mesodermal development in the stomodeal primordium. This question might be answered by experimental manipulation of the hkb expression. Finally, we do not yet know any genes involved in stomodeal development that are regulated by twi and sna in a redundant fashion and might also be regulated by hkb.

In summary, we propose that the ‘ground’ fate of the ventral cells of the blastoderm embryo is to form a ventral furrow and the mesodermal germ layer (Fig. 13). For this determination, the combination of twi and sna is required. Since sna is expressed only in a subset of the blastoderm cells expressing twi, it is sna, but not twi, that is spatially limiting for mesoderm determination. By the action of the terminal gap gene hkb, the two ends of the embryo follow a different fate. Anteriorly mostly ectodermal and partly endodermal structures of the digestive tract are formed. Here, hkb, sna and twi are involved in the formation of the stomodeum and sna and hkb in the determination of the endodermal AMG primordium. In the posterior region of the embryo, hkb determines endodermal gut structures, represses sna expression and excludes the mesoderm primordium.

We are indebted to José Casal, Andreas Kispert and Mary Mullins for critically reading the manuscript. Thanks to Nick Brown and José Casal for reagents, to Gos Micklem for constructing the sna expression vector, to Günter Brönner for sending the hkb cDNA and to Dan Kiehart, Siegfried Roth and Detlef Weigel for antibodies. We thank Günter Brönner, Jordi Casanova, Karin Ekström, Walter Gehring, Kathy Matthews, Norbert Perrimon and Pat Simpson for sending fly stocks and Katrin Brenner, Gertrud Scheer and Sandra Schäfer for assistance with the photographic reproductions. KlausPeter Rehorn helped by performing some of the hkb mRNA localizations and Barbara Grunewald by preparing the sections. This work was supported by the Max-Planck-Society.