ABSTRACT
The endodermal midgut arises from two primordia, the anterior midgut (AMG) primordium and the posterior midgut (PMG) primordium, which are separated by almost the entire length of the Drosophila embryo. To form the midgut, these two parts have to extend towards each other and to fuse laterally on both sides of the yolk. Shortly before and during that movement, AMG and PMG are arranged as mesenchymal cell masses, but later the midgut cells form an epithelium. We show that these two aspects of midgut development, migration of AMG and PMG and transition to an epithelium, depend on the mesoderm. The extension of the midgut primordia is achieved by cell migration along the visceral mesoderm which forms a continuous layer of cells within the germ band. In mutant embryos lacking the entire mesoderm or failing to differentiate the visceral mesoderm, AMG and PMG are formed but do not migrate properly. In addition, they fail to form an epithelium and instead either remain as compact cell masses anterior and posterior to the yolk (in twist and snail mutant embryos) or only occasionally wrap around the yolk before embryogenesis is completed (in tinmandeficient embryos). We conclude that the visceral mesoderm serves as a substratum for the migrating endodermal cells and that the contact between visceral mesoderm and endoderm is required for the latter to become an epithelium.
INTRODUCTION
Gastrulation in Drosophila starts with the formation of the ventral furrow and the amnioproctodeal invagination of hindgut and posterior midgut primordia. While the cells of the posterior midgut invagination and at the anterior tip of the ventral furrow give rise to the endoderm, most of the cells of the ventral furrow form the mesodermal germ layer. The latter differentiates into two major mesodermal tissues: into the somatic mesoderm forming the muscles of the body wall and into the visceral mesoderm forming the musculature surrounding the gut.
Two zygotic genes encoding transcription factors, snail (sna) and twist (twi), are required for specification and differentiation of the mesoderm (Simpson, 1983; NüssleinVolhard et al., 1984; Boulay et al., 1987; Thisse et al., 1987; Leptin, 1991; Thisse et al., 1991). Both are expressed in the ventralmost region of the embryo, twi from the anterior to the posterior pole, sna from the anterior pole to about the border of the posterior terminal region of the embryo (Thisse et al., 1988; Alberga et al., 1991; Leptin and Grunewald, 1990). Mutant embryos lacking either of the two genes fail to form a ventral furrow and lack all mesodermal derivatives. Other aspects of gastrulation like amnioproctodeal invagination or germ band extension are not disturbed. A larger set of genes is thought to be involved in mesodermal differentiation. Among these is tinman (tin, formerly called msh-2 or NK-4), which codes for a transcription factor that appears to be required for positional specification within the mesodermal germ layer and differentiation of the visceral mesoderm (Bodmer et al., 1990; Bodmer, 1993; Azpiazu and Frasch, 1993).
The endoderm has two spatially separated primordia, the anterior midgut (AMG) and the posterior midgut (PMG). In order to form the endodermal part of the digestive tract, the midgut, these two parts have to join and enclose the yolk. At the onset of germ band retraction AMG and PMG extend towards each other and finally meet laterally on both sides of the yolk shortly before retraction of the germ band is completed (Fig. 1). After that, in a second phase of movement, the endoderm wraps ventrally and dorsally around the yolk and finally takes on the shape of a convoluted tube.
Two aspects of midgut development are of particular interest: (i) the mechanism of the movement of the two midgut parts towards each other and (ii) the apparently concomitant transition of the multi-layered, mesenchymal cell masses of AMG and PMG into a highly ordered epithelium. We present a morphological and genetic analysis of both processes and conclude that they require an interaction of the endodermal midgut cells with cells derived from another germ layer, the mesoderm. The arrangement of the midgut cells during the time of extension relative to the visceral mesoderm suggests that AMG and PMG migrate towards each other guided by the visceral mesoderm. This view is supported by the phenotypes of mutant embryos which fail to develop or differentiate a visceral mesoderm. In these cases, the migration of the midgut parts does not occur. The transition from the mesenchymal cell mass to an epithelium also does not occur and we propose that the visceral mesoderm not only serves as an adhesive substratum during the migration of the midgut parts but is also needed to induce the arrangement of the midgut as an epithelium.
MATERIALS AND METHODS
Fly stocks
twiRY50: weak twi allele (Simpson, 1983); twiID and twiEY53: amorphic twi alleles (Nüsslein-Volhard et al., 1984; Thisse et al., 1987). twiRY50 and twiEY53 chromosomes have been cleaned by meiotic recombination from lethal mutations other than twi. Flies homozygous for these chromosomes can be fully rescued by 9 kb of genomic DNA containing the twi gene when these are transformed into flies mediated through a P-element vector (Thisse et al., 1991; our unpublished results). snaIIG and snaRY1: amorphic sna alleles (Nüsslein-Volhard et al., 1984; Grau et al., 1984; Boulay et al., 1987). ushIIA, ush19, tupIIIE and tupIIIB: homozygous mutant embryos fail to retract their germ band (Nüsslein-Volhard et al., 1984). A490.2M3: enhancer trap line (3rd chromosome) with β-gal expression in the endoderm (Bellen et al., 1989). A183.1F2 and fasIIIE25: null for fasciclin III protein (Wilson et al., 1989; Elkins, Ferres-Marco and Goodman, unpublished data). l(1)mysxG43: null mutation for β1 integrin (PS3, Leptin et al., 1989). Df(2L)DS6 and Df(2L)pr-A14: deficiencies with small overlap which uncovers the locus of the endoderm-specific βν integrin (S. E. Paine-Saunders and R. O. Hynes, pers. commun.; Yee and Hynes, 1993). Df(1)DlX43: small deletion of the Delta locus. Df(1)N8: small deletion of the Notch locus. Df(3R) eBS2, Df(3R) eF1 and Df(3R) GC14: deficiencies for the homoeobox gene tin (Bodmer et al., 1990). Df(3R) hkbA321: generally used as a hkb null allele (Weigelet al., 1990). torXR1: a 9.5 kb deletion of the tor locus (Sprenger et al., 1989). stg7B69 and stg7M53: amorphic alleles of string (Jürgens et al., 1984; Edgar and O’Farrell, 1989).
Immunohistochemistry and histology
Murine monoclonal anti-β-galactosidase (anti-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). Anti-Drosophila muscle myosin heavy chain antibody (rabbit) was a gift from Dan Kiehart (Kiehart and Feghali, 1986), murine monoclonal anti-fasciclin III and anti-Ubx antibodies from Mike Wilcox (Brower et al., 1980; White and Wilcox, 1985), anti-caudal antibody (rabbit) from Uwe Walldorf (Mlodzik and Gehring, 1987) and anti-twi antibody (rabbit) from Siegfried Roth (Roth et al., 1989).
Embryos were fixed and stained essentially 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). Whole-mount 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.
RESULTS
Midgut development
We followed the development of the midgut with the enhancer trap A490.2M3 which, beginning from stage 8, displays homogeneous gene activity throughout the cells that contribute to the midgut (Fig. 1, Bellen et al., 1989). Two hours after the onset of gastrulation, the midgut primordia are two separate groups of cells: the AMG primordium and the PMG primordium. At stage 10 the AMG primordium is contiguous with the invaginating stomodeum. The PMG is positioned at the end of the germ band and is contiguous with the prospective hindgut (Fig. 1A). While the AMG is already mesenchymal at this stage, the PMG will undergo a visible transition from an epithelium (Fig. 1A) to a mesenchyme at the end of stage 10 (Fig. 1B, see also Foe, 1989; Hartenstein et al., 1992). The cells are small and round and form a large mass which is several cells deep. At this time, the AMG begins to split into two arms which extend towards the posterior. Slightly later, during stage 11, this extension becomes more pronounced (Fig. 1C), and the PMG has also extended along the germ band in a similar fashion. The midgut primordia now span almost three times as much of the germ band (which is still almost fully extended) as they did a stage earlier. Moreover, both the AMG and the PMG are arranged in symmetrical bi-lobate structures with lobes right and left of the yolk (Fig. 1D). Shortly before the completion of germ band retraction the leading tips of AMG and PMG meet on both sides of the yolk, and by the time the germ band has retracted AMG and PMG will have fused (Fig. 1E,G). In a second phase of movement, the midgut cells then spread in transversal directions, both ventrally and dorsally, and completely enclose the yolk at the end of stage 14 (not shown). Two aspects of the midgut development are the topic of our paper: the transition of the mesenchymal cell mass of AMG and PMG into an ordered epithelium and the spatially accurate movement of the midgut parts towards each other.
Morphological analysis of the endodermal migration
How are AMG and PMG directed towards each other during their migration? The leading tips of the two midgut parts are precisely positioned below the narrow band of visceral mesoderm that forms one continuous layer of cells along the inner surface of the germ band on either side of the embryo (Fig. 1F,F′). The visceral mesoderm becomes distinguishable in such a form during the course of stage 10 (see below Fig. 4F). It therefore might fulfill the spatial and temporal requirement for a structure that could guide the midgut parts. If the visceral mesoderm had such a guiding function one would expect it to be in close contact with the midgut. In serial cross-sections (Fig. 2A-C) through the bilobate AMG of a stage 11 embryo, we observe a tight attachment of the endoderm along its entire length to the underlying mesoderm. The visceral mesoderm expresses fasciclin III, a cell surface molecule (Snow et al., 1989, see also Fig. 4F) before it becomes morphologically distinct from the somatic mesoderm. A cross-section through the AMG of a stage 11 embryo immunostained for the fasciclin III protein (Fig. 2G) reveals that indeed the entire visceral mesoderm makes contact with the endoderm, and that it is only the visceral mesoderm to which the endoderm attaches. This also holds true for the PMG (Fig. 2H). In the particular embryo shown, the PMG is attached to a more posterior (upper) stretch of visceral mesoderm but not yet to the more anterior (lower) region. Thus AMG and PMG are apparently moving towards each other along the cell layer of the visceral mesoderm.
Morphological analysis of the transition from mesenchyme to epithelium
During their extension, AMG and PMG each consist of a mass of cells with mesenchymal character and show little sign of specialization apart from their organization into two arms (Figs 1C.D, 2A-D). At stage 13, when the germ band is retracted, the midgut has acquired its typical arrangement of an epithelial layer of columnar cells between yolk and visceral mesoderm (Fig. 1G,H). While the columnar shape is only transient and disappears when the cells flatten to enclose the yolk in stage 14 (Fig. 1G), the epithelial character is maintained. When and how does the transition from mesenchyme to epithelium occur? The extending arms of the midgut are many cells thick during stages 10 to 12 and are not visibly organized in layers. The thickness of the arms slightly decreases towards the leading tips (Fig. 2A,B), but only at the very tip do they become merely one cell thick (Fig. 2C). Even the cells that are directly in contact with the visceral mesoderm and are arranged in a somewhat regular fashion remain small and round. Apparently throughout the time of migration the midgut cells stay mesenchymal. When AMG and PMG have joined in late stage 12, the midgut cells arrange into a columnar epithelium. Beginning from the anterior and the posterior end of the midgut, the cells in contact with the visceral mesoderm become elongated with their nuclei located towards the mesoderm (Fig. 2E). Concomitantly, the endoderm and the attached visceral mesoderm spread slightly dorsoventrally (compare the left and the right part of the midgut epithelium of the embryo in Fig. 2E which have developed slightly asynchronously). It should be pointed out that, in addition to the prominent layer of outer columnar cells, there remains a second layer of small round inner cells throughout the entire midgut (Figs 1H, 2E). These cells also express the enhancer trap A490.2M3, originate from AMG or PMG and presumably constitute the imaginal midgut precursors (Hartenstein and Jan, 1992). Furthermore, there is a cluster of large round PMG cells that had migrated just behind the leading tip and which also lie on the inside of the epithelium (Fig. 1G). It has been shown previously that both types of inner cells eventually become integrated into a single-layered midgut epithelium during later stages of development (Hartenstein and Jan, 1992; Reuter et al., 1990).
Genetic analysis
The morphological studies above have shown that extending AMG and PMG are in close contact with the visceral mesoderm. If this spatial arrangement had a functional role for the extension of the midgut parts and their transition from mesenchyme to epithelium, one would expect that, in mutant embryos lacking the mesoderm, both processes should be affected. twi and sna embryos fail to develop a mesoderm completely but they undergo normal posterior midgut invagination and form the normal number of PMG cells with endodermal fate as indicated by the expression of the enhancer trap A490.2M3 (Fig. 3B,C). Apparently, neither sna nor twi are required in the developing posterior midgut. Also the AMG develops, but in sna embryos there are slight aberrations from the normal developmental path: the AMG remains a simple epithelium until later stages of embryonic development (Fig. 3B, Reuter and Leptin, unpublished data). In spite of the normal early development, in these mutant embryos neither part of the midgut shows any sign of extension, nor do AMG and PMG form bi-lobate structures. Instead they remain as large separate mesenchymal cell masses immersed in the yolk at the ends of the embryo (Fig. 3B,C). AMG and PMG do not join; they do not spread transversally over the yolk and are obviously unable to engulf it (Fig. 2F). At no time during development do their cells arrange into an epithelium or even acquire the columnar shape that is typical for the endodermal layer in contact with the visceral mesoderm around stage 13 (Fig. 2E). Thus the formation of the simple epithelium of the midgut appears to depend on the presence of mesoderm. We assume that epithelial differentiation including shape changes to columnar cells is induced in the endodermal cells upon the attachment of the midgut endoderm to the visceral mesoderm. In addition, the extension of the midgut parts also requires mesodermal structures. This corroborates our notion of the migration of midgut cells needing the visceral mesoderm as a substratum.
The specific requirement for the visceral mesoderm rather than for the entire mesoderm is apparent in weak twi mutant embryos (twiRY50) which form a ventral furrow and initially also an apparently normal mesodermal germ layer (not shown). twi protein expression is normal during the early phases of development, in particular within the midgut primordia. Then, during germ band extension twi protein disappears from most of the mesoderm of the mutant embryos while in wild-type embryos it persists until after the onset of germ band retraction (Fig. 4A,B). Later, the mutant embryos partially differentiate somatic muscles (Fig. 4G) but no visceral muscles surrounding the midgut. The visceral mesoderm does not form a visibly discernible tissue. For example, fasciclin III is absent from the presumptive visceral region in the mutant embryos while all other aspects of its expression are undisturbed (Fig. 4E,F). In such embryos lacking the visceral mesoderm, the endoderm does not migrate (Fig. 4A) and AMG and PMG do not meet (Fig. 4C). They remain as compact, mesenchymal cell masses that insert into the yolk rather than enclose it. Both fail to form the typical bilobate structures, do not undergo the transition to an epithelium and ultimately no proper midgut is made.
Further support for our notion that the visceral mesoderm is required for endoderm migration and the transition of mesenchyme to epithelium comes from the phenotype of embryos deficient for the homoeobox gene tin. tin is expressed in the mesoderm in a twi-dependent fashion but is neither expressed in the primordia of the AMG nor of the PMG (Bodmer et al., 1990). It is required for the differentiation of the visceral mesoderm as embryos deficient for tin do not express fasciclin III within the mesoderm, and the mesodermal layer surrounding the yolk does not develop (Fig. 5E; Bodmer, 1993; Azpiazu and Frasch, 1993). Consistent with the postulated role of the visceral mesoderm, AMG and PMG of the tin-deficient embryos do not develop properly. They fail to form bilobate structures, do not migrate properly and are still at a considerable distance from each other when the germ band has been fully retracted (Fig. 5B). Later, AMG and PMG eventually join and appear to spread transversally as they often manage to wrap around the yolk ventrally (Fig. 5C). Only occasionally the midgut covers the yolk dorsally in a very irregular fashion at the end of embryogenesis. Since tin does not function directly in the midgut primordia (it is not expressed there), the failure of the midgut parts to migrate properly towards each other in the tin-deficient embryos strongly indicates that a differentiated visceral mesoderm is needed for this step in midgut organogenesis.
What are the molecules involved in migration and mesenchyme-to-epithelium transition?
It is reasonable to assume that cell adhesion molecules expressed on the visceral mesoderm and the midgut endoderm mediate the interaction between these tissues. So far, however, we have been unable to identify any molecules that are stringently and specifically required. For example, embryos lacking fasciclin III, a cell adhesion molecule with immunoglobulin-type extracellular domains (Snow et al., 1989), do not show any disturbances of midgut development (data not shown). Fasciclin III mediates homophilic cell adhesion (Snow et al., 1989), is expressed on the entire surface of the visceral mesoderm cells and therefore might be involved in keeping the visceral mesoderm together rather than in mediating attachment to the endoderm. The integrin subunit αPS2 is localized at the interface between visceral mesoderm and midgut endoderm (Bogaert et al., 1987), is expressed in the visceral mesoderm and is believed to heterodimerize with the ubiquitous β1 chain (PS3) to a functional integrin. However, in mutants for l(1)mys, the gene encoding β1, the endoderm migrates properly on the visceral mesoderm. This is even the case in embryos lacking maternal and zygotic contribution of β1 or in the additional absence of fasciclin III (data not shown). Also embryos transheterozygous for deficiencies that remove an endoderm-specific β integrin (βν; Yee and Hynes, 1993) develop a normal midgut (data not shown).
Of the cell surface molecules tested only Notch (N) and Delta (Dl) are required for migration of the endoderm and its transition to an epithelium. In N and Dl embryos, AMG and PMG express midgut-specific genes but remain as large mesenchymal cell masses immersed in the yolk as they do in twi embryos (shown in Fig. 6B,D for Dl; see also Harten stein et al., 1992). They do not assume a bilobate structure and do not attach to the visceral mesoderm (Fig. 6B). In contrast, N and Dl do not seem to be required for the development of the visceral mesoderm. In N and Dl embryos, the mesoderm expresses fasciclin III and forms an apparently normal, continuous layer of cells along the germ band (Fig. 6F). It remains to be seen how N and Dl are involved in the process of migration.
How is the direction of endodermal migration determined?
It is conceivable that specific cues might be present on the anterior visceral mesoderm to direct the AMG posteriorly and on the posterior visceral mesoderm to direct the PMG anteriorly. We believe that this is not the case. Embryos from torso (tor) mothers develop an apparently normal AMG primordium and differentiate the visceral mesoderm, but no PMG is formed (Fig. 7A and data not shown). The AMG migrates far into the posterior region of the embryo, makes its transition to an epithelium, and later spreads transversally to enclose most of the yolk even in the absence of the PMG (Fig. 7B). In a wild-type embryo at stage 13, the junction between cells derived from AMG and cells derived from PMG is well visible (Fig. 1E). The visceral mesoderm just posterior to this junction, i.e. visceral mesoderm that touches the most anterior part of the PMG, expresses Ultra bithorax (Ubx) over the length of one parasegment (Bienz and Tremml, 1988; Tremml and Bienz, 1989; Reuter and Scott, 1990; data not shown). When we use Ubx expression as a landmark in the visceral mesoderm of embryos from tor mothers, we unambiguously see that the AMG migrates posteriorly beyond this point (Fig. 7C,D). Thus, the AMG migrates over visceral mesoderm that normally would contact PMG, and it migrates over this mesoderm in the direction opposite to the one that the PMG would normally take. This shows that the mesoderm does not contain any directional cues pointing towards the position where AMG and PMG normally meet; there is also no specific signal at this point that stops the migration. We propose that AMG and PMG merely spread over the available surface of the visceral mesoderm starting from their positions at opposite ends of the germ band.
DISCUSSION
The requirement of the visceral mesoderm for midgut morphogenesis
Our data on midgut development in wild-type and mutant embryos indicate that the visceral mesoderm is needed (i) for the migration of AMG and PMG towards each other and (ii) for the establishment of the midgut epithelium. We assume that the visceral mesoderm guides the migrating endoderm and that the contact to the visceral mesoderm is required for the endoderm to become an epithelium. We can only draw these conclusions if neither sna, twi nor tin have a direct function in the development of the midgut primordia and the observed midgut defects therefore have to be of indirect nature and due to the missing mesoderm. Certainly for tin this is the case. The midgut primordia never express tin (Bodmer et al., 1990) and their development proceeds normally until the time when migration should begin. Since tin codes for a transcription factor, like sna and twi, and therefore acts cell-autonomously, the defects seen in midgut development of tin embryos must be of indirect nature. We consider it unlikely that the transheterozygous deficient embryos that we used for our studies are lacking other genes directly involved in endodermal development since the same phenotype in midgut development has been observed in embryos mutant exclusively for tin (Bodmer, 1993; Azpiazu and Frasch, 1993).
twi protein is expressed in the primordia of AMG and PMG for a short period of time. We do not believe that this expression has a function for midgut development, although we have not proven this directly. However, a number of findings support our view. First, in twi embryos, early midgut development proceeds normally, up to the stage when migration should start. Secondly, weak twi embryos initially express twi in the normal pattern and perform early twi-dependent events, like ventral furrow formation, only slightly slower than normal. Thus, the twi protein is apparently functional in these mutants. However, its later expression is impaired and the defects in the differentiation of somatic and visceral mesoderm can be attributed to the disappearance of the twi protein from the mesodermal cell layer after the onset of germ band extension (Fig. 4E,G). In these weak twi mutant embryos, twi protein is expressed normally in the midgut primordia. If twi had a function in these primordia it should fulfill it, as it fulfills its function in ventral furrow formation in these mutants. Any later defects in midgut development seen in these embryos should be indirect in nature. Since, with respect to gut development, the phenotypes of twi embryos homozygous for weak (Fig. 4C) or for amorphic alleles (Fig. 3C) are indistinguishable, we assume that also twi does not have a direct function in the midgut primordia.
sna protein is only very weakly expressed in the PMG primordium during cycle 13 (Alberga et al., 1991; Ray et al., 1991) before sna transcription is repressed by the activity of the terminal system in this region (Casanova, 1991; Leptin, 1991; Ray et al., 1991). Therefore it appears very unlikely that sna itself functions directly in posterior midgut development. Only the proper development of the AMG requires sna function (Reuter and Leptin, unpublished data). Therefore, the prerequisite for our conclusions, the indirect nature of the defects in midgut development, is fulfilled for tin embryos and most likely for the PMG of sna embryos and for AMG and PMG of twi embryos.
twi and sna embryos often fail to retract their germ band and then form a U-shaped cuticle (compare Fig. 3B and C). It has been speculated that germ band retraction plays a key role in bringing together the midgut parts (Campos-Ortega and Hartenstein, 1985). Thus, the failure of germ band retraction in sna or twi embryos might be responsible for the defects in midgut development. However, this is not the case. First, in wild-type development the extension of the midgut is already very far advanced when the germ band starts to retract (Fig. 1C). And secondly, in embryos with defective germ band retraction but normal mesoderm differentiation like u-shaped (ush) and tail-up (tup) embryos (Nüsslein-Volhard et al., 1984), the midgut primordia nevertheless extend normally along the visceral mesoderm and the yolk is completely engulfed by the endoderm (data not shown). Therefore, it is not germ band retraction alone which brings AMG and PMG closer to each other. In this context, it is worth mentioning that in hkb embryos, which do not form endodermal derivatives (Weigel et al., 1990), germ band retraction is properly executed (data not shown). Thus, the two processes, germ band retraction and endoderm migration, occur completely independent from each other. Other mechanisms that have been proposed to cause the movement of the AMG are its growth by cell division and pressure from the ingrowing foregut (Campos-Ortega and Hartenstein, 1985). While these may contribute, they are neither sufficient (mitoses and pressure from the foregut) nor necessary (mitoses). For example, in twis, the AMG proliferates and the foregut forms normally (Figs 3C, 4C), but the AMG does not extend; in string mutants, the AMG extends even in the absence of mitoses after cycle 14 (Edgar and O’Farrell, 1989; our data not shown).
The phenotype of tin-deficient embryos in midgut development is less severe than the one observed in twi or sna embryos. While bilobate structures are not formed and migration of the endoderm is not supported by visceral mesoderm, the two parts of the midgut nevertheless eventually join after stage 14, and the midgut partially encloses the yolk at the end of embryogenesis (Fig. 5B,C). We do not have an explanation for this phenomenon but we speculate that the same movement of the endoderm that normally leads to the transversal spreading also closes the gap in the mutant. The endoderm of the deficient embryos is in fact capable of the transversal spreading (Fig. 5C), and we assume that this movement is not directional by itself but merely involves spreading over the available surface of the yolk. During their spreading the cells never assume the organization of a columnar epithelium, and the resulting layer of endodermal cells is irregular unlike the wild-type epithelium. It is not clear whether the formation of this layer is aided by some residual function of rudimentary visceral mesoderm in tin-deficient embryos or whether it is completely independent of mesoderm.
The midgut does not seem to need its mesodermal component for the transversal spreading. However, the endodermal component is clearly required. Normally both the midgut endoderm and the adhering visceral mesoderm spread dorsoventrally so that the yolk is enclosed by the two tissues simultaneously. However, in hkb embryos that lack the entire endoderm the yolk does not become enclosed by the visceral mesoderm. Instead, a rudimentary net of muscles spreads over part of the yolk sac and partitions it into several irregular lobes (data not shown). We conclude that of the two tissues only the midgut endoderm can migrate on the yolk membrane in Drosophila.
How does the directional migration occur?
We propose that the visceral mesoderm serves as an adhesive guiding tract for the migration of AMG and PMG towards each other (Fig. 8). At the moment we can only speculate on the cellular mechanism of this migration. We assume that midgut cells have an affinity for cells of the visceral mesoderm and, as motile cells in their mesenchymal state, distribute over the surface of the visceral mesoderm. At the same time adhesion among the endodermal cells and among the mesodermal cells ensures the integrity of the respective tissues. The direction of migration would then result from the initial positions of AMG and PMG at opposite ends of the centrally located midgut visceral mesoderm and the availability of free mesoderm surface to migrate on. Thus, the migration would be only virtually directional and would not require specific cues for the AMG to migrate posteriorly or the PMG anteriorly. The finding that embryos from tor mothers extend their AMG far into the posterior region of the embryo supports this view (Fig. 7).
Molecules involved in endodermal migration or the transition from mesenchyme to epithelium
Of the cell adhesion molecules tested, only N and Dl appear to have a function in endodermal migration and/or the transition from mesenchyme to epithelium. The others (fasciclin III, PS3, βν integrin) might participate in these processes only in a redundant fashion. In N and Dl embryos AMG and PMG do not contact the apparently intact visceral mesoderm, remain mesenchymal and do not enclose the yolk (Fig. 6). The latter two aspects strikingly resemble the phenotype observed in mutants lacking the visceral mesoderm. However, N and Dl are nearly ubiquitously expressed (Haenlin et al., 1990; Kooh et al., 1993) and involved in so many processes during development (for a discussion see Hartenstein et al., 1992) that we cannot exclude an indirect effect on endodermal migration. In particular, it is difficult to decide whether N and Dl might not merely be required for a developmental decision within the endoderm, which would lead to the mesenchyme-to-epithelium transition (Hartenstein et al., 1992). For example, the expression of genes mediating the contact to the visceral mesoderm could depend on N and Dl. Without the attachment, the migration and the transition to an epithelium would not occur (like in the mutant embryos lacking the visceral mesoderm). Alternatively, it is possible that Dl and N directly mediate cell adhesion between visceral mesoderm and endoderm. It is tempting to speculate that N and Dl could even be involved in transmitting the signal to the endoderm, which induces the transition to a columnar epithelium upon adhesion to the visceral mesoderm.
The midgut as a model system for the interaction between germ layers during organogenesis
During the development of the midgut, two tissues derived from different germ layers, the visceral mesoderm and the midgut endoderm, interact at various times and in several ways so that different steps in morphogenesis are taken or regional specification is obtained. (i) We have shown that the layer of visceral mesoderm cells serves as a substratum for the migration of the endodermal midgut cells, the most prominent example of cell migration in Drosophila that leads to the fusion of the two previously separate primordia. We assume that such a mechanism is used in a wide range of insect species since the outgrowth of bilobate midgut primordia from cells close to stomodeum and proctodeum is commonly observed (for a review see Siewing, 1969). (ii) The formation of the endodermal columnar epithelium immediately after the migration is also dependent on the contact to the visceral mesoderm. (iii) Later the midgut develops several well-defined constrictions, which transiently partition the organ. Their formation is not only regulated by homoeotic genes expressed in the visceral mesoderm (Bienz and Tremml, 1988; Tremml and Bienz, 1989; Reuter and Scott, 1990), but they also are imposed by this tissue on the underlying endoderm (Reuter and Scott, 1990). (iv) The mesoderm also directs gene expression in the endoderm through appositional induction (Immerglück et al., 1990; Panganiban et al., 1990; Affolter et al., 1993). The genetic and molecular tools available for Drosophila therefore make midgut development a suitable model to study developmental mechanisms that are widely used in morphogenesis and differentiation of various organs in Drosophila and in other species.
ACKNOWLEDGEMENTS
Thanks to Karin Ekström, Walter Gehring, Corey Goodman, Kathy Matthews and Pat Simpson for sending fly stocks. We are grateful to Dan Kiehart, Siegfried Roth, Uwe Walldorf and Mike Wilcox for the gift of antibodies and to Stephenie Paine-Saunders and Richard Hynes for communicating data prior to publication. We thank José ‘Lince’ Casal and Helen Skaer for improving the manuscript and Katrin Brenner, Gertrud Scheer and Sandra Schäfer for assistance with the photographic reproductions. This work was supported by the Human Frontier Science Program and the MaxPlanck-Society.