Role of caudal in hindgut specification and gastrulation suggests homology between Drosophila amnioproctodeal invagination and vertebrate blastopore

Gastrulation begins at three hours of embryogenesis in Drosophila. During this process (described by Campos-Ortega and Hartenstein, 1997), cells of the ventral midline invaginate, forming the ventral furrow; the invaginated cells go on to form the mesodermal derivatives. At either end of the ventral furrow, cells invaginate to give rise to the anterior and posterior portions of the gut. At the posterior, a cap of approximately 450 cells begins moving dorsally, forming a plate and then a cup of cells referred to as the amnioproctodeal invagination. During invagination of the amnioproctodeum, the central domain of cells sinks in first, becoming the posterior midgut primordium. The ring of cells surrounding the posterior midgut primordium then comes together – forming the borders of a slit – and moves inside, completing gastrulation of the posterior gut primordia. The invaginating ring of approximately 300 cells is the hindgut primordium, which later gives rise to the hindgut proper and the Malpighian tubules.

Present understanding is that the patterning of the embryo posterior necessary for amnioproctodeal invagination and posterior gut development is initiated solely by activation of the maternal terminal system (reviewed by St Johnston and Nüsslein-Volhard, 1992; Perrimon et al., 1995). The critical component of the terminal system is the Torso receptor tyrosine kinase; activation of this receptor leads to transcription of only two known target genes, tailless (tll) and huckebein (hkb) (Weigel et al., 1990; Pignoni et al., 1992). Both of these genes encode transcription factors, and, acting together or separately, control the expression of additional transcription factor genes required for development of the posterior gut, namely fork head (fkh) (Weigel et al., 1989a,b), brachyenteron (byn, originally called Trg) (Kispert et al., 1994; Singer et al., 1996) and bowel (bowl) (Wang and Coulter, 1996).

tll and hkb, in addition to controlling genes required for specification and continued development of the posterior gut primordia, work together with fkh to activate folded gastrulation (fog) (Costa et al., 1994). fog encodes a secreted molecule that is required for gastrulation, specifically for coordinating the constriction of cells within the epithelial plate that gives rise to the posterior gut (Sweeton et al., 1991; Costa et al., 1994). Available data thus support the notion that the terminal system, acting solely through the genes tll and hkb, is responsible for initiating all of the changes in gene activity that are required to establish and maintain the posterior gut primordia, and to promote gastrulation.

A potential, but largely overlooked player in this network of gene activity controlling gastrulation and gut development is caudal. The caudal (cad) gene encodes a homeodomain transcription factor expressed, as a result of both maternal and zygotic transcription, at the posterior of the embryo (Mlodzik et al., 1985; Macdonald and Struhl, 1986). The domain of cad expression overlaps those of the above described genes, during the period when they are being transcriptionally activated.

While a requirement for cad in the development of epidermal, external structures that arise from the posterior of the embryo has been described (Macdonald and Struhl, 1986), the role of cad in gastrulation or in establishing posterior gut primordia has not been investigated.

To determine if Drosophila cad is required for gastrulation and/or posterior gut development, we generated embryos in which cad activity was removed maternally, both maternally and zygotically, or only zygotically. We find that embryos completely devoid of cad activity are unable to carry out gastrulation normally and, at older stages, lack the hindgut. By examining gene expression in these cad mutant embryos, we deduce that their gastrulation defect is due to a decrease in expression of fog, and their lack of hindgut due to loss (by apoptosis) of the misspecified hindgut primordium as a result of diminished expression of fkh and wg. These results establish cad as a key regulator in pathways required for gastrulation and gut development.

The strongest alleles available were used, namely: tll^l49 (Pignoni et al., 1990), hkb^A423R1(Gaul and Weigel, 1990), fkh^XT6 (Weigel et al., 1989a), wg^7L74 (Van den Heuvel et al., 1993), byn⁵(Singer et al., 1996), bowl² (Wang and Coulter, 1996), cad² and cad³ (Macdonald and Struhl, 1986), fog¹¹⁴(Schalet and Finnerty, 1968), and fog^4a6(Sweeton et al., 1991).

Embryos were derived from mosaic females carrying germline clones (GLCs) homozygous for either cad² or cad3, obtained by using the FLP-DFS technique of Chou et al. (1993), as applied to cad by Rivera-Pomar et al. (1995). To generate GLCs, males carrying the FRT ovoD chromosome and an X-linked hs-FLP chromosome were crossed to females carrying an FRT cad chromosome. Progeny of this cross were heat-shocked at 37°C for 2 hours on each of days 3, 4 and 5 after egg lay (AEL), and allowed to develop at 25°C. These GLC females were mated to males heterozygous for cad to obtain progeny embryos, half of which lacked both maternal and zygotic cad activity (denoted cadm^−z− embryos, and also referred to as ‘cad-deficient embryos’, or ‘embryos lacking cad activity’), and half of which lacked maternal cad activity but carried one zygotically active wild-type cad gene (denoted cadm⁻z+ embryos). These two genetic classes of embryos were easily distinguished at later stages on the basis of their previously described phenotypes (Macdonald and Struhl, 1986): cadm−^z− embryos have severe segmentation defects particularly in the posterior, while cad^m−z+ embryos look relatively normal but have smaller anal pads. Embryos lacking only zygotic cad were obtained by crosses between heterozygous parents, and are denoted cad^m+z−. Similar phenotypes were obtained with both cad alleles.

Antibody staining was performed using standard techniques (Ashburner, 1989). Anti-Crumbs antibody Cq4 (which labels the apical surface of all ectodermally derived epithelia; Tepass et al., 1990) was used at a dilution of 1:100 to outline the lumina of the hindgut and Malpighian tubules. Streptavidin conjugated to horseradish peroxidase secondary antibody was used at 1:2000. Horseradish peroxidase activity was detected by adding NiCl2 to 0.07% to the diaminobenzidine reaction, leading to a black precipitate. Embryos were dehydrated in ethanol and mounted whole in Epon/Araldite. Embryos were staged according to Campos-Ortega and Hartenstein (1997).

For in situ hybridization to whole embryos, embryos were collected, devitellinized, fixed and stored in 100% ethanol at −20°C until needed. DNA probes were labeled by random priming with digoxigenin-dATP (Genius kit, Boehringer Mannheim). Hybridization and detection was carried out by the procedure of Tautz and Pfeifle (1989).

Embryos were fixed, embedded and sectioned at 2 μm with a LKB 2088 Ultratome V, stained with toluidine blue (Ashburner, 1989) and mounted in Epon under coverslips. Whole-mount embryos and sections of embryos were photographed on a Zeiss Axiophot microscope equipped with differential interference contrast optics.

Maternally produced cad mRNA is deposited in the oocyte during oogenesis; very early during embryogenesis this uniformly distributed maternal mRNA is differentially degraded under control of the maternal effect gene bicoid to produce a posterior-to-anterior gradient (reviewed by Rivera-Pomar et al., 1996; Fig. 1A). By the beginning of the cellular blastoderm stage (stage 5), maternal cad mRNA has been largely degraded. As a result of zygotic gene activity, a posterior domain of cad mRNA now appears (Rivera-Pomar et al., 1995). Zygotic cad expression is highest at the posterior of this domain, in a stripe at approximately 11 to 17% egg length (EL, measured at 50% dorsal-ventral) (Fig. 1B), which corresponds with the position in the blastoderm fate map of the anlagen of hindgut and anal pads (Campos-Ortega and Hartenstein, 1997).

To assess the requirement for cad in establishing the structures that arise from the posterior ∼15% of the blastoderm embryo, namely the posterior midgut, Malpighian tubules and hindgut, we examined mature embryos lacking cad activity (maternal and/or zygotic contributions). Embryos lacking both maternal and zygotic cad activity (referred to as cadm−z−, or cad-deficient), as well as embryos lacking maternal cad activity but carrying one zygotically active wild-type cad gene (cad^m−z+) and embryos lacking only zygotic cad (cad^m+z−) were generated as described in Materials and Methods.

Staining with anti-Crb antibody, which labels the apical surface of the hindgut epithelium (Fig. 2A), reveals that cad activity, provided either maternally or zygotically, is essential for formation of the hindgut. When cad is provided only maternally (cad^m+z−), the posterior gut appears essentially normal (not shown). When cad is provided only zygotically (cad^m−z+), a hindgut tube is formed; this can range from almost normal to somewhat reduced in length and twisted (Fig. 2B). When cad activity is completely lacking (cad^m−z−), the hindgut (as well as the anal pads and the eighth abdominal segment) is almost entirely absent (Fig. 2C). We conclude that cad activity is essential for normal hindgut formation, and that while some of this activity must be provided maternally, most can be provided either maternally or zygotically.

The Malpighian tubules and posterior midgut also arise from domains expressing cad, but form relatively normally in cad-deficient embryos. The posterior midgut arises from the most posterior region of the embryo, where the maternally derived cad mRNA level is highest (Fig. 1A); additionally, starting during stage 10 and continuing to the end of embryogenesis, cad is expressed in a posterior-to-anterior gradient in the posterior midgut (Fig. 2D). Despite this suggestive expression pattern, however, in cad-deficient embryos the posterior midgut primordium forms normally, expresses cad in the characteristic posterior-to-anterior gradient and appears normal in size and development (Fig. 2E). The Malpighian tubules and the hindgut both arise from the same blastoderm primordium (Diaz et al., 1996); also, similar to what is seen for the posterior midgut, cad expression is reinitiated in a proximal-to-distal gradient in the tubules during stage 13 (Liu and Jack, 1992). Nevertheless, as revealed by their expression of fkh, the tubule primordia form relatively normally in embryos lacking cad activity (although their elongation is incomplete, as noted by Liu and Jack, 1992) (Fig. 2G).

Thus, even though it is expressed maternally in all of the primordia giving rise to posterior gut, in terms of the generation of a normal-appearing structure in the embryo, cad is only required in the hindgut primordium.

The absence of the hindgut primordium from cad-deficient embryos suggests that Caudal regulates genes required for establishing and/or maintaining the hindgut primordium. tll, fkh, byn, bowl and wg are likely targets for cad regulation, as all are required for some aspect of hindgut development: the hindgut is missing from both tll and fkh embryos, and severely reduced in wg, byn and bowl embryos (Fig. 3A-F; Pignoni et al., 1990; Skaer and Martinez Arias, 1992; Harbecke and Lengyel, 1995; Diaz et al., 1996; Singer et al., 1996; Wang and Coulter, 1996). Furthermore, since maternally provided Caudal, which persists only through the blastoderm stage (Macdonald and Struhl, 1986), is sufficient for essentially normal hindgut formation, the fact that all of these genes are expressed at the posterior of the embryo during the blastoderm stage (Fig. 4A-D,F; Baker, 1988; Pignoni et al., 1990; Weigel et al., 1990; Kispert et al., 1994; Wang and Coulter, 1996) means that they are potential targets for regulation by Caudal.

The effect of absence of maternal and/or zygotic cad activity on the expression of these genes was assessed by in situ hybridization with appropriate probes. For tll, byn and bowl, absence of cad activity did not result in a detectable effect on expression (Fig. 4A-C; data not shown). As described below, however, we found that cad activity is essential for expression of fkh and wg.

Both maternal and zygotic cad contributions are necessary for posterior wg expression. During early stage 5, just prior to its expression in 14 stripes that are required to establish the segmental pattern, wg is expressed in two domains at the anterior, and in a broad posterior stripe (Baker, 1988; Fig. 4D). This terminal wg stripe is located at approximately 8-12% EL, overlapping with the posterior of the zygotic cad stripe and with the position of the hindgut and Malpighian tubule primordia in the blastoderm fate map (Campos-Ortega and Hartenstein, 1997). Expression of the wg terminal stripe has been shown to be independent of other segmentation genes, but has not been otherwise characterized (Ingham and Hidalgo, 1993). We find that all embryos from cad germline mothers (both cad^m−z− and cad^m−z+) fail to express the terminal stripe of wg (Fig. 4E). These results demonstrate that maternal cad activity is essential for the transcription of wg in the terminal stripe. Among embryos from wg heterozygous parents, approximately one-quarter (presumably those lacking only the zygotic component of cad expression) lack the terminal wg stripe (not shown). Thus both maternal and zygotic cad activities are required for expression of the terminal wg stripe. The expression of the early cap of fkh also requires cad activity; approximately half of the embryos from cad germline females mated to cad heterozygous males (i. e., cad^m−z− embryos) show a dramatic reduction in both the size and intensity of the posterior cap of fkh expression (Fig. 4G). If cad is supplied either maternally or zygotically, however, fkh expression is normal (cad^m−z+, cad^m+z−, not shown). Thus expression of the posterior cap of fkh requires cad activity, which can be provided either maternally or zygotically. Later, by stage 10 (Fig. 6D), fkh expression is expressed as strongly in cad-deficient as in wild-type embryos, indicating that this later expression is independent of cad activity. Since tll and hkb are required to activate early fkh expression (Weigel et al., 1990) but are not themselves regulated by cad (data not shown), cad must act combinatorially with these two genes to promote early fkh expression.

cad also regulates wg in combination with other genes. In addition to the demonstrated requirement for cad, expression of the posterior wg stripe requires positive input from fkh and tll, as the stripe is absent from the respective mutant embryos (data not shown). Since embryos lacking either maternal or zygotic cad fail to express the posterior wg stripe, but still express fkh and tll, cad must act combinatorially with fkh and tll to promote formation of the posterior wg stripe. Expression of the terminal wg stripe thus requires combinatorial action of cad, tll and fkh; the posterior limit of the stripe is defined by repression by hkb (Mohler, 1995).

The fact that tll, byn and bowl expression at the blastoderm stage are apparently normal in cad-deficient embryos suggests that a hindgut primordium is established in the absence of cad activity. The lack of proper blastoderm stage expression of fkh and wg, however, indicates that this hindgut primordium is not properly specified. We showed previously in byn mutant embryos that one of the earliest phenotypic manifestations of an abnormally specified hindgut primordium is ectopic expression of the cell death gene reaper (rpr) (Singer et al., 1996).

To ask whether the extremely reduced hindgut in cad-deficient embryos might result from a similar course of programmed cell death, we examined expression of rpr in embryos lacking cad. We observed a striking pattern of ectopic rpr expression in cad^m−z− embryos, beginning during stage 7 and continuing into stage 8 (gastrulation), in a ring at the circumference of the amnioproctodeal plate (Fig. 5B,D). The actual loss of cells that is presumably initiated by this ectopic rpr expression does not begin until after early stage 10, however, as the hindgut primordium is present at this stage in cad-deficient embryos, indicated by its expression of byn and fkh (see below). By stage 13, the cad-deficient embryo has a very short hindgut and no detectable anal pads (Fig. 2C,E); in sections of stage 13 embryos there are numerous apoptotic cells in the region of the hindgut remnant (Fig. 5F).

The amnioproctodeal plate of cad^m−z− embryos, distinguished by the fact that it ectopically expresses rpr, becomes increasingly retarded in its anteriorward movements during stages 7 and 8 (Fig. 5B,D, arrowheads). By early stage 10 (∼4.5 hours AEL), when the germ band in the wild-type embryo is almost fully extended, the germ band in the cad^m−z− embryo has extended only about 40% of the normal distance and the embryo appears twisted (Fig. 6B). This phenotype is particularly obvious when embryos are labeled for fkh expression (which is reestablished in cad-deficient embryos after the blastoderm stage): the hindgut primordium, which strongly expresses fkh, remains near the posterior of the embryo (Fig. 6D). Thus, in the absence of cad activity, there is a severe defect in germband extension.

Gastrulation can be followed by using expression of byn as a marker for the hindgut primordium. In the wild-type embryo, byn is expressed in a ring at the circumference of the amnioproctodeal plate. The edges of this ring fold together as the posterior midgut primordium invaginates during stages 6 and 7; the ring of hindgut primordium then sinks inward during stage 8 and is completely internalized by the end of stage 9 (Fig. 7A-C,F,G). The zygotically expressed cad stripe and the posterior wg stripe are also expressed in the bordering ring (i.e., the hindgut primordium) of the invaginating amnioproctodeal plate (Fig. 7D,E). Strikingly, in cad-deficient (cadm−z−) embryos, the byn-expressing ring of hindgut primordium draws together, but fails to invaginate, remaining on the outside of the embryo (Fig. 7H). Thus, although internalization of the Malpighian tubule and posterior midgut primordia is normal in cad-deficient embryos (Fig. 2E,G), the gastrulation movements necessary for internalization of the hindgut primordium do not occur in embryos lacking cad activity.

The failure of the hindgut to become internalized in cad^m−z− embryos raises the question of whether cad might regulate a zygotically expressed gene required for the invagination of the amnioproctodeal plate. One gene known to be required for gastrulation is fog; fog mutant embryos lack not only the posterior midgut, but, as revealed by anti-Crb staining, the Malpighian tubules and hindgut as well (Fig. 8A). In the blastoderm stage embryo, fog expression is first activated in the region that will become the ventral furrow; shortly thereafter, expression is initiated in a posterior cap, in the region that will become the amnioproctodeal invagination (Costa et al., 1994; Fig. 8B). In cad^m−z− embryos, fog expression in the prospective ventral furrow is normal, but is significantly reduced in the posterior cap (Fig. 8C). Thus, cad is required for the normal level of expression of fog in the prospective amnioproctodeal plate; decreased fog expression in cad^m−z− embryos is likely responsible for the failure of the hindgut primordium to be internalized during gastrulation.

As fkh or wg mutant embryos do not display detectable defects in gastrulation (Weigel et al., 1989b; data not shown), fog is the only gene presently known to mediate effects of cad on gastrulation. In fog mutant embryos, none of the posterior gut primordia invaginate (Fig. 8D,E), while in cad-deficient embryos the posterior midgut and Malpighian tubule primordium do invaginate; thus, consistent with the in situ hybridization results (Fig. 8C), a low level of fog activity is present at the posterior of embryos lacking cad.

Using germline clones to remove both the maternal and zygotic contribution of the homeodomain transcription factor Caudal, we have shown that cad activity is required to establish the hindgut. This is explained, at least in part, by the required role of cad in activating blastoderm stage expression of the genes fkh and wg, which are required for hindgut commitment and development. cad is also required for invagination of the hindgut primordium, a function that is largely explained by its role inducing posterior expression of fog, a secreted protein required for cellular constriction during gastrulation. Thus via regulation of fkh, wg and fog, and possibly other downstream genes, cad plays an essential role both in specifying the hindgut and in its invagination.

The genetic data presented here demonstrate that cad acts together with a number of zygotically functioning genes to activate transcription at the posterior terminus. fkh expression requires input from cad (either maternal or zygotic) plus tll and hkb, while expression of the posterior wg stripe requires both maternally and zygotically provided cad, as well as tll and fkh. Previous work showed that transcription of fog requires tll, hkb and fkh (Costa et al., 1994); the significant decrease in fog expression in embryos lacking cad activity adds another required gene activity to this combination. Thus, as has been described for patterning of the central segmented trunk ectoderm (reviewed by Rivera-Pomar and Jäckle, 1996), different combinations of transcription factors – including Caudal – act combinatorially at the posterior of the embryo to pattern the hindgut and to control gastrulation.

The network of regulatory interactions controlling hindgut formation and gastrulation, deduced from previous work and results described here, is summarized in Fig. 9. The regulatory interactions involving Caudal, indicated by solid arrows, are most likely direct. First, all of the genes involved in these interactions encode transcription factors. Second, all of the interactions shown occur within a period of about 1 hour (stage 5, cellularization) allowing little time for the expression and action of unknown, potentially intervening transcription factor encoding genes.

Are the activities of fkh, wg and fog sufficient to mediate the function of cad in hindgut specification and gastrulation? Reduction in the domain of expression of fog has been reported to cause a corresponding reduction in the size of the posterior gut primordium that is invaginated (Costa et al., 1994); the observed reduction in the domain of fog expression (Fig. 8C) seems sufficient to explain the partial gastrulation defects in cad-deficient embryos. Thus cad appears to be required together with tll, hkb and fkh to promote fog expression in the hindgut primordium, but is not necessary for the fog expression required to internalize the primordia of the posterior midgut and Malpighian tubules. The effect of cad on germband extension may be a result of its effect on gastrulation, since non-gastrulating (e.g., fog, tor and DRhoGEF2) embryos fail to elongate the germband (Fig. 8E; Schüpbach and Wieschaus, 1986; Häcker and Perrimon, 1998).

Whether the effect of cad on hindgut specification is mediated entirely through fkh and wg is less easily determined, since, late in embryogenesis, fkh and cad-deficient embryos lack most, and wg embryos lack much, of the hindgut. The reduced hindgut is likely a result of cell death in the misspecified primordium. While ectopic rpr expression is seen in the hindgut primordium of cad-deficient embryos beginning at stage 7, we did not detect this ectopic expression in either fkh or wg embryos (although there is ectopic expression of rpr later in the hindgut of stage 12 fkh embryos) (not shown). Thus cad may regulate at least one gene in addition to fkh and wg that is required for proper hindgut specification (and hence prevention of ectopic apoptosis).

cad homologs are highly conserved across several phyla. There is one gene known in C. elegans and Drosophila, while there are three cad homologs in vertebrates, known as X-cad in frog and Cdx in mouse and chick (reviewed by Marom et al., 1997).

Early expression of cad homologs is at the posterior of the embryo. In the 4-cell C. elegans embryo, the cad homolog (pal-1) is required in the most posterior cell for its correct specification (Hunter and Kenyon, 1996). In a number of other embryos, at the beginning of gastrulation, cad homologs are expressed at the posterior, in a ring or line of cells about to invaginate (see Fig. 10 for Drosophila and Xenopus). This region, which can be considered the ‘blastopore equivalent’, consists of the hindgut/posterior midgut primordium in Drosophila, the germ ring in fish, the marginal zone/blastopore lip in frog, and the node/primitive streak in chick and mouse. In chick and mouse, Cdx genes are expressed in the primitive streak, where cells are about to ingress into the interior of the embryo (reviewed by Marom et al., 1997). In a number of cases (Drosophila cad, chick Cdx-B and Cdx-C), the earliest cad expression is as a posterior-to-anterior gradient (Macdonald and Struhl, 1986; Marom et al., 1997).

An additional common feature in the expression of many cad homologs is a later expression in the intestine (gut). Drosophila cad is expressed in the posterior midgut and the Malpighian tubules of the older embryo (Mlodzik et al., 1985; Macdonald and Struhl, 1986; Fig. 2D); chick and mouse Cdx genes are expressed in the posterior gut, which is the only tissue of expression in the adult (reviewed by Marom et al., 1997). Echoing the early expression pattern, this gut expression is also usually found as a posterior-to-anterior gradient; this is seen for cad in the Drosophila midgut (Fig. 2D), and has been described for all three mouse Cdx genes in the adult intestine (James and Kazenwadel, 1991; James et al., 1994; Gamer and Wright, 1994).

In addition to cad, three other genes required at the posterior of the Drosophila embryo for formation of the hindgut, fkh, byn and wg, are related to genes found throughout the metazoa, known as HNF-3 (α, β? and γ?, Brachyury (also known as T) and Wnt, respectively. In many cases, these homologs are expressed in portions of the ‘blastopore equivalent’ at the posterior of the embryo that overlap with domains of expression of cad (Cdx).

In C. elegans, a Wnt homolog is expressed, and required for proper posterior development, in the same posterior blastomere where the cad homolog pal-1 functions (reviewed by Han, 1997). In sea urchin, HNF-3 and Brachyury homologs are expressed in the vegetal plate just prior to gastrulation (Harada et al., 1996). In fish and frog, Caudal, Brachyury and Wnt (Wnt8 and Wnt11) are initially expressed around most or all of the blastopore lip while HNF-3 expression is dorsally localized (Joly et al., 1992; Kelly et al., 1995; Schulte-Merker et al., 1992; Strähle et al., 1993; Northrop and Kimelman, 1994; O’Reilly et al., 1995; Horb and Thomsen, 1997; Smith and Harland, 1991, Christian and Moon, 1993; Ku and Melton, 1993) (see Fig. 10 for expression in Xenopus). As gastrulation proceeds, the expression of these genes becomes more restricted and non-overlapping, with HNF-3 and Brachyury expression becoming localized to the notochord and Wnt8 expression retreating from the dorsal position and becoming exclusively ventral. Patterns of expression of HNF-3 and Brachyury consistent with this general description have been found in ascidians, amphioxus, chick and mouse (Olsen and Jeffery, 1997; Yasuo and Satoh, 1993; Corbo et al., 1997; Holland et al., 1995; Zhang et al., 1997; Shimauchi et al., 1997; Shimeld, 1997; Tam and Behringer, 1997; Ruiz i Altaba et al., 1993).

Required roles for some of these genes have been demonstrated by analysis of mutants: mouse HNF-3ß knockouts reveal requirements in formation of the node, notochord and head process (reviewed by McMahon, 1997); fish no tail and mouse T mutants reveal a requirement for Brachyury in migration of mesoderm through the primitive streak and in formation of the notochord (reviewed by Hermann and Kispert, 1994; Tam and Behringer, 1997).

There is thus a constellation of conserved genes – cad (Cdx), fkh (HNF-3), wg (Wnt8 and Wnt11) and byn (Brachyury) – whose overlapping expression patterns in the blastopore equivalent suggests function in a related process. The phenotypes of the available mutations in these genes suggest that the common function is to specify cell fate at the blastopore; in most cases, an essential part of this fate is internalization and forward migration during gastrulation.

The striking conservation in expression, and likely in function, of cad suggests that the regulation of posterior terminal development in Drosophila by Caudal may represent a more ancient regulatory mechanism than the tor receptor and the two genes that it activates, tll and hkb. Of these three genes, a vertebrate homolog is known only for tll; the function of this vertebrate gene, Tlx, is related to that of Drosophila tll not in the posterior, but rather in the anterior, in the establishment of the brain (Pignoni et al., 1990; Monaghan et al., 1997). Thus the Torso receptor pathway and its activation of tll and hkb has probably been superimposed relatively recently (in evolutionary terms) upon a more ancient, Caudal-regulated network of gene activity controlling gastrulation and gut formation.

The fact that the same four genes are expressed at the blastopore equivalent of chordates and the amnioproctodeal invagination of Drosophila suggests that these two highly dynamic domains are homologous. Given the regulatory hierarchy that we have demonstrated in Drosophila, we propose that, in embryos of the proximate ancestor to arthropods and chordates (reviewed by DeRobertis and Sasai, 1996), the posterior was defined by a posterior-to-anterior gradient of Cad activity. Cad then activated expression of downstream network of genes that controlled invagination (gastrulation) and gut specification. Cad expression in the archenteron probably continued and played an essential role, as this structure differentiated into the gut. Going beyond the bilaterian ancestor to chordates and arthropods, it is worth considering that this nexus of gene expression may have evolved even more basally in the metazoa.

The foregoing, by homologizing the insect amnioproctodeal invagination with the echinoderm and vertebrate blastopore, does not fit with the classical definition of protostomes and deuterostomes. This view categorizes arthropods as protostomes, in which the mouth is derived from the primary invagination of gastrulation, and chordates as deuterostomes, where the mouth arises from a secondary invagination (reviewed by Willmer, 1990). More recently, comparisons of gastrulation patterns in many different species, as well as construction of molecularly based cladograms, has called into question the utility of these classically defined groups (Willmer, 1990; Kirschner and Gerhart, 1997). While there continues to be uncertainty in our understanding of ‘protostome’ and ‘deuterostome’ phyla, the significant conclusion of the data presented here is that there may be a homology between the blastopore of vertebrates and the amnioproctodeal (posterior) invagination of insects.

We thank Eddy DeRobertis, Karen Lyons and Dave Jacobs for helpful comments on the manuscript, Jonathan Singer for providing fkh and byn probes, and Silvia Yu and Ron Togelang for excellent technical assistance. This work was supported by NIH grant GM 51270 to J. A. L.