Epithelium formation in the Drosophila midgut depends on the interaction of endoderm and mesoderm

The formation of epithelia is a fundamental step in metazoan development. After a phase of rapid division, cells of the early embryo form the blastoderm epithelium. Blastoderm cells, as well as all later formed epithelia, are characterized by their structural and functional apicobasal polarity and their two-dimensional arrangement. Some of the blastoderm cells are internalized during gastrulation and form the endoderm and mesoderm. In many metazoans, including Drosophila, cells of the endoderm and mesoderm lose their epithelial organization and become mesenchymal during or shortly after gastrulation, while the ectoderm remains epithelial. The various epithelia of the later embryo are formed by two different mechanisms. Primary epithelia originate directly (without any non-epithelial intermediates) from the blastoderm. In Drosophila, mainly the ectodermal epithelia (e.g. epidermis, fore- and hindgut, tracheal system) belong to this class. By contrast, secondary epithelia develop from mesenchymal intermediates by a process called mesenchymal-epithelial transition (Tepass and Hartenstein, 1993). In this case, cells have to undergo profound changes in their shape and spatial arrangement. A typical example of a secondary epithelium is the Drosophila midgut epithelium, the formation of which we have investigated in this study.

Recent studies carried out mainly in vertebrate model systems suggest that both cell-cell and cell-substratum adhesion triggers and stabilizes the reorganization from largely apolar mesenchymal cells into a polarized epithelium (for review see Rodriguez-Boulan and Nelson, 1989; Fleming, 1992). Evidence from Drosophila indicates that interactions at the apical surface of epithelial cells are also important for their development (Tepass et al., 1990; Tepass and Knust, 1990, 1993). These studies also show that the formation of primary and secondary epithelia may be governed by different mechanisms. The genes crumbs and stardust, for example, are required for the devel-opment of primary epithelia in the Drosophila embryo, but not for the formation of the (secondary) midgut epithelium (Tepass and Knust, 1990, 1993).

The development of the midgut epithelium in Drosophila represents a model system for a systematic genetic approach to study the mechanisms controlling epithelium formation. The analysis of midgut development in wild type and various mutants shows that the mesenchymal-epithelial transition that leads to the formation of the midgut epithelium requires inter-actions between the endoderm and the adjacent visceral mesoderm, as well as interactions among the endodermal cells themselves. In flies with mutations that result in a complete absence of visceral mesoderm (twist (twi); twi snail (sna) double mutants) the midgut epithelium does not form. In mutants with reduced visceral mesoderm or where the endoderm is spatially separated from the visceral mesoderm (tinman (tin); dorsal (dl) twi double heterozygotes; torso⁴⁰²¹ (tor⁴⁰²¹); folded gastrulation (fog)) only those endodermal cells that directly contact the visceral mesoderm form an epithelium. Finally, in shotgun (shg) mutant embryos the contact between endoderm and visceral mesoderm is properly established but endodermal cells do not form a columnar monolayer suggesting that shg might control adhesion between midgut epithelial cells.

The following mutations were used in this study. The strong twi alleles twi^1D96 and twi^HH07 (Nüsslein-Volhard et al., 1984), the strong sna allele sna^4.26 (Lindsley and Zimm, 1992), the strong dl allele dl¹ (Nüsslein-Volhard et al., 1980), the strong tin alleles Df(3R)GC14 and tin^EC40 (Mohler and Pardue, 1984), the strong fog allele fog^S4 (Wieschaus et al., 1984), the dominant tor allele tor⁴⁰²¹ (Klingler et al., 1988), the strong shg allele shg^g317 (U. T., E. Gruszynski de Feo, and V. H., unpublished data), and the intermediate shg allele shg^IH (Nüsslein-Volhard et al., 1984). As wild-type stock we used Oregon R.

Flies were grown under standard conditions and crosses were performed at room temperature or at 25 °C. Egg collections were done on yeasted apple juice agar plates. Embryonic stages are given according to Campos-Ortega and Hartenstein (1985). The cross and the egg collection to generate the dl twi double heterozygous embryos were done at 30 °C. Embryos were then stained with the anti-fasciclin III antibody (see below) and embryos that had small gaps in the visceral mesoderm were picked out for further examination.

The following markers were used in this study. The enhancer-trap line B11-2-2 (Hartenstein and Jan, 1992) to label endodermal cells in wild type and in twi mutants. The enhancer-trap line A490 (Bellen et al., 1989) to label endodermal cells in wild-type and in fog and shg mutant embryos (data not shown). These enhancer-trap lines express β-galac-tosidase that was detected with a polyclonal anti-β-galactosidase antibody (Cappel; dilution 1:2000). The monoclonal anti-Fasciclin III antibodies mAb6D6 (kindly provided by Seymour Benzer) and mAb2D5 (Patel et al., 1987; kindly provided by Corey Goodman) were used to label the median portion of the visceral mesoderm in wild type and all examined mutants. Both antibodies were diluted 5-or 10-fold. A polyclonal anti-muscle myosin antibody (kindly provided by Dan Kiehart; dilution 1:200) was used to detect visceral muscle in wild type and in tor⁴⁰²¹ mutant embryos. Antibody stainings and sections of stained embryos were done as described previously (Tepass and Knust, 1993).

Embryos for examination in the transmission electron microscope were prepared as described previously (Tepass and Hartenstein, 1993). Embryos for semi-thin sectioning were prepared in the same way. 2 μm sections were cut on an LKB Ultrotom V and stained with a toluidine blue/methylene blue/borate solution.

The Drosophila larval midgut is composed of two tissue layers, an inner epithelial layer, which develops from the endoderm, and a mesodermally derived outer layer of visceral muscle. The primordia of these two tissues have established contact with each other at stage 11 (extended germ band stage). At this stage, the endoderm forms two mesenchymal cell masses called the anterior and posterior midgut rudiment, respectively (Fig. 1A).

Using cell-specific markers (U. T. and V. H. unpublished data) three different cell types can be distinguished in the midgut rudiments. The majority of cells of both midgut rudiments form an epithelium during germ band retraction. We suggest the name principle midgut epithelial cells (PMECs) for this cell type. A smaller fraction of cells do not become part of the midgut epithelium initially. This group of cells is composed of two cell types, the adult midgut precursors (AMPs) and the so called ‘large basophilic cells’ (LBCs). The LBCs, which derive only from the posterior midgut rudiment, and the AMPs occupy a position in the center of the midgut rudiments. When the PMECs organize into the midgut epithelium, AMPs and LBCs remain attached as mesenchymal cells to the apical surface of the epithelium (Fig. 2G,H,I). The LBCs integrate into the midgut epithelium at a later stage (stage 14/15; Reuter et al., 1990). The AMPs become transiently part of the midgut epithelium during late embryogenesis; in the larva, they assume a position at the basal surface of the epithelium (Hartenstein and Jan, 1992).

The midgut epithelium is surrounded by two layers of visceral muscles, an inner layer of circular (i.e. transversal) fibers, and an outer layer of longitudinal fibres (Tepass and Hartenstein, 1993). The circular fibers and their mesodermal precursors can be specifically labeled with an antibody that recognizes fasciclin III (Patel et al., 1987; Fig. 1G-J). The visceral mesoderm that forms the circular fibers first appears as metamerically repeated clusters in the dorsal mesoderm of trunk parasegments 2–13 (maxilla -seventh abdominal segment; Tremml and Bienz, 1989; Azpiazu and Frasch, 1993). During late stage 11, the clusters on each side of the embryo fuse into a continuous band at the interior-dorsal aspect of the mesoderm. At this stage, the midgut rudiments come into contact with the bands of visceral myoblasts and use them as tracks for migration and epithelium formation (see below). The fasciclin III-positive visceral mesoderm also covers the posterior portion of the foregut (Fig. 1I). The precursors of the longitudinal visceral fibres originate in the mesoderm at the tail end of the embryo (own unpublished observation). Finally, the anteriormost and posteriormost portions of the mesoderm give rise to visceral fibres associated with the foregut and hindgut, respectively. These cells can be distinguished from the visceral muscles of the midgut by the absence of anti-fasciclin III staining.

Shortly before and during germ band retraction (late stage 11 and stage 12) the PMECs reorganize to form the midgut epithelium. Early morphogenetic movements during gastrulation and germ band extension have placed the midgut rudiments in a position where they are in contact with the anterior or posterior ends of the visceral mesoderm, respectively. The PMECs migrate along the visceral mesoderm (Fig. 1B,C) so that at about 50% germ band retraction, the arms of the anterior and posterior midgut rudiments meet and fuse with each other (Fig. 1D). By the end of germ band retraction (stage 13) two rectangular plates have formed (Fig. 1E).

At the same time when PMECs migrate along the visceral mesoderm, they become epithelial. The outermost PMECs, which directly contact the visceral mesoderm, are the first cells to undergo this mesenchymal-epithelial transition (Fig. 2A-C). They are organized in a regular monolayer and become columnar. Subsequently, the more interior PMECs, which originally had no contact with the visceral mesoderm, send processes between the already existing epithelial PMECs. After these processes contact the visceral mesoderm, the corresponding cells also become columnar and intercalate with the first formed epithelial cells (Figs 2B, 3B). At 50% germ band retraction, most PMECs have formed a monolayer of columnar cells covering the visceral mesoderm (Fig. 2D-F).

From mid stage 12 onwards the LBCs and the AMPs become morphologically distinct from the PMECs. The LBCs are larger, the AMPs are smaller than the PMECs. Both cell types are found at the apical side of the epithelium formed by the PMECs (Fig. 2D-I). At stage 13 all PMECs have assumed a highly columnar shape with the exception of those cells that lie beneath the LBCs, which form two clusters in the middle of the developing midgut, and attain a squamous cell shape during stage 13.

When PMECs move along the visceral mesoderm and start to form an epithelium, their surface is characterized by multiple slender processes (Fig. 3C). At this stage, no extra-cellular material or any kind of junctional specialization between endodermal and mesodermal cells could be detected, suggesting that the PMEC migration is mediated by direct cell-cell contacts. After germ band retraction, the number of filopodia between PMECs and visceral mesoderm diminishes and adherens junctions begin to form (Fig. 3G). These adherens junctions might be the precursors of the so-called connecting hemi adherens junctions (Tepass and Hartenstein, 1993). Between neighboring PMECs, scattered spot adherens junctions and gap junctions are the only cellular junctions present during midgut epithelium formation. A circumferential junction as seen in other epithelia is not differentiated by the PMECs until very late in embryogenesis (mid stage 17) and comprises then a smooth septate junction (Tepass and Harten-stein, 1993). Epithelium formation during stages 12 is accom-panied by the appearance of prominent apicobasal bundles of microtubules in the columnar PMECs (Fig. 3E,F).

In the hours following germ band retraction, the narrow band of visceral mesoderm expands dorsally and ventrally to form the circular visceral muscle fibres. The PMECs follow this movement and change in shape from columnar to cuboidal or squamous during stages 14 and 15 (Fig. 2J,K). Throughout this process, the basal surface of the epithelial cells contacts the visceral muscle. Fig. 4 provides a schematic overview of the formation of the midgut epithelium.

Our analysis of midgut epithelium formation in the wild type shows that all PMECs are in contact with the visceral mesoderm while converting into epithelial cells. This observa-tion suggests that the visceral mesoderm serves as a basal sub-stratum to which the PMECs must attach in order to form an epithelium. To test this hypothesis we analyzed midgut epithelium formation in the background of mutations that alter specific aspects of visceral mesoderm development.

In embryos mutant for twi¹⁰⁹⁶ or in double mutant embryos of the genotype twi^HH07sna^4.26 the mesoderm and mesodermal derivatives do not develop (Simpson, 1983; Grau et al., 1984). The endoderm becomes internalized during gastrulation in both mutants but does not form an epithelium (Fig. 5A,B).

To analyze midgut development in mutants that specifically lack visceral mesoderm we examined tin mutant embryos (Df(3R)GC14, tin^EC40; Mohler and Pardue, 1984; Bodmer, 1993; Azpiazu and Frasch, 1993). A complete absense of the fasciclin III-positive visceral mesoderm has been reported for both tin alleles and also associated defects in midgut morpho-genesis (Bodmer, 1993; Azpiazu and Frasch, 1993). We found, however, that small clusters of fasciclin III visceral mesoderm cells are present in tin mutants (Fig. 5E). Sectioned material demonstrates that only those PMECs in contact with these small islands of visceral mesoderm assume an epithelial phenotype (Fig. 5F,G).

Similar results were obtained with twi¹⁰⁹⁶ heterozygous mutant embryos derived from heterozygous dl¹ mothers. In these animals, variable amounts of mesoderm are lacking depending on the temperature at which the experiment is performed (Simpson, 1983). The reduction in the number of mesodermal cells leads to a reduction in cell number in all mesodermally derived organs. In embryos of the cross dl¹/+ × twi¹⁰⁹⁶/+ raised at 30°C many embryos show small gaps in the visceral mesoderm (Fig. 5C,D). Wherever these gaps occur, the overlying midgut epithelium is absent too.

In twi and twi sna mutant embryos, the primary epithelia of fore- and hindgut differentiate normally in the absence of mesoderm (data not shown). Interestingly, the ectodermally derived outer epithelial layer of the proventriculus, which is commonly considered as part of the midgut, also differentiates independently of visceral mesoderm (Fig. 5G).

In eggs collected from tor⁴⁰²¹ heterozygous mothers the segmented germ band is missing and tissues that derive from the anterior and posterior pole are expanded (Sprengler and Nüsslein-Volhard, 1993). The central part of tor⁴⁰²¹ embryos consists of hindgut epithelium whose apical (normally luminal) surface faces outward, since it does not invaginate during gastrulation. Similarly, the posterior endoderm remains at the surface at the tail end of the embryo (Fig. 6A). Visceral mesoderm of the trunk, which normally gives rise to the visceral muscles of the midgut, is absent (Fig. 6A). By contrast, the posteriormost portion of the mesoderm is spared and produces visceral muscle fibres which attach to the hindgut epithelium (Fig. 6B). This posterior mesoderm seems to be intrinsically different from the midgut associated mesoderm, as suggested by the fact that it does not express fasciclin III. Interestingly, the visceral muscles of the hindgut are able to support the formation of a rudimentary midgut epithelium. A fraction of the endoderm cells, which contacts the visceral mesoderm underlying the hindgut, takes on epithelial characteristics (Fig. 6D).

The amount of visceral mesoderm might play a critical role by emitting a diffusible factor in limited amounts that is required for epithelium formation. In an embryo with reduced visceral mesoderm the concentration of such a factor might be insufficient to support the mesenchymal-epithelial transition of all endodermal cells. To weaken this argument and to further support our conclusion that direct contact of visceral mesoderm and endoderm is required for midgut epithelium formation we studied midgut development in fog^S4. This mutation blocks gastrulation movements so that neither the ventral furrow, which internalizes the mesoderm and part of the anterior endoderm, nor the amnioproctodeal invagination, which internalizes the posterior endoderm, are formed (Zusman and Wieschaus, 1985; Sweeton et al., 1991). Despite these early defects, all tissues except for the posterior midgut epithelium develop normally. Thus, although the ventral furrow does not form, the cells of the mesoderm and the anterior endoderm are eventually internalized and assume a proper position in relation to the ectoderm. Two normally sized bands of visceral mesoderm are formed (Fig. 6E). Also the anterior endoderm usually ends up in its proper position and becomes attached to the visceral mesoderm. Subse-quently, it forms a normal epithelium (Fig. 6F). The posterior endoderm, on the other hand, fails to invaginate and remains at the surface at the posterior pole of the embryo. Not con-tacting the visceral mesoderm, posterior endodermal cells remain mesenchymal (Fig. 6G).

The gene shg (Nüsslein-Volhard et al., 1984) is required for several aspects of epithelial morphogenesis in the Drosophila embryo (U. T., E. Gruszinzky de Feo and V. H., unpublished data). In the context of this paper we focus on the defects in midgut development of shg mutant embryos. In embryos mutant for the strong allele shg^g317, the visceral mesoderm develops normally (Fig. 7A). The PMECs attach to and spread over the visceral mesoderm but they do not become columnar; instead, they maintain a rounded to cuboidal shape (Fig. 7B,C) and do not form a monolayer (Fig. 7B). Similar observations, although less well defined, have been made in embryos carrying the intermediate shg^IH allele. These findings suggest that there exist two independent adhesion systems required for midgut epithelium formation, one between PMECs, the other between PMECs and visceral mesoderm. shg mutations seem to delete specifically adhesion between PMECs, while the second system appears to be unaffected.

At later stages (stage 14–17) when the visceral mesoderm extends dorsoventrally, all PMECs of shg mutant embryos become attached to the visceral mesoderm and gradually adopt a more wild-type-like appearance. In our ultrastructural examination of the midgut in fully differentiated shg mutant embryos no difference from wild type could be found (data not shown; for a description of the ultrastructure of the larval midgut see Tepass and Hartenstein, 1993).

In the present study the mesenchymal-epithelial transition that leads to the formation of the midgut epithelium has been analysed in wild type and in a number of mutant embryos in which the visceral mesoderm is absent, reduced in size, or where endodermal and mesodermal cells are spatially separated (for a summary diagram see Fig. 8). Our findings demonstrate that endodermal cells have to establish local contact with the visceral mesoderm, which serves as a basal substratum for both migration and epithelium formation. Even small patches of visceral mesoderm are sufficient to induce endodermal cells contacting them to become epithelial as seen in tin mutant embryos. The fog mutant phenotype shows further that midgut epithelium formation does not depend on the fusion of the anterior and posterior midgut rudiments. This is consistent with our observation on normal midgut development where the transition of mesenchyme into epithelium has been almost completed before both midgut rudiments fuse.

In tor⁴⁰²¹ mutant embryos the segmented germ band, including the visceral mesoderm of the midgut, is absent (for review see Sprenger and Nüsslein-Volhard, 1993). In these embryos we found that ‘hindgut-specific’ mesoderm is able to induce epithelium formation in endoderm cells that come into contact with it. This observation bears on the question of the specificity of the interaction between endoderm and visceral mesoderm. As suggested by the analysis of embryos that lack visceral mesoderm, other tissues (e.g., epidermis, nervous system, and mesodermal derivatives such as somatic muscle and fat body) cannot serve as a basal substratum for the endo-dermal cells. This indicates a high degree of specificity in the interaction between the endoderm and the visceral mesoderm. On the other hand, within the visceral mesoderm, all portions may be capable of sustaining epithelium formation of the endoderm.

The interaction between the endoderm and the visceral mesoderm that leads to the establishment of the midgut epithelium is the first of several interactive processes between these two tissues. It has previously been reported that regional expression of a number of homeotic genes in the visceral muscle is critical for midgut morphogenesis, including the formation of the midgut constrictions and the gastric caeca (Bienz and Tremml, 1988; Tremml and Bienz 1989; Reuter and Scott, 1990). Furthermore, the visceral muscle emits diffusible signals that are involved in the late differentiation of midgut epithelial cells (Immerglück et al., 1990; Reuter et al., 1990; Panganiban et al., 1990).

The reported observations on midgut development suggest that the midgut epithelium undergoes three phases, which differ in the amount of intercellular adhesion (Fig. 9). The initial phase of midgut development, during which the highly columnar epithelium emerges from the mesenchymal midgut rudiments, may be characterized by strong adhesion. Experimental evidence from vertebrate systems (injection of the adhesion molecule N-cadherin into Xenopus embryos; Takeichi, 1991) indicates that high levels of adhesion are correlated with a columnar cell shape, while a low adhesion level correlates with a squamous or cuboidal epithelium. According to these findings, the highly columnar morphology of early midgut epithelial cells in Drosophila suggest a high level of intercellular adhesion. We propose that the strong intercellular adhesion between the PMECs depends on the function of the shg gene. The PMECs in shg mutant embryos do not become columnar but keep a round to cuboidal shape instead (Fig. 8F). shg might be involved in the adhesion process itself, or it might control the cellular response (i.e., polarization of the cytoskele-ton) to the adhesion event that leads to the reorganization of the cell structure into an asymmetric columnar shape. Surprisingly, the drastic early defects in epithelial cell morphology in shg mutants have no apparent consequences for the terminal differentiation of the midgut epithelium. The functional significance of the highly columnar cell shape of the PMECs might be related to the limited substratum surface provided by the visceral mesoderm that forms a narrow band at stage 13. Thus, a small basal cell surface is a prerequisite for the PMECs to fit onto the surface area provided by the visceral mesoderm. This view is corroborated by the observation that in shg mutant embryos where the PMECs remain rounded or cuboidal, many PMECs do not initially establish contact with the visceral mesoderm.

During stage 14 the cell shape of the PMECs changes from columnar to cuboidal or squamous, marking the beginning of a phase of reduced intercellular adhesion that extends until mid stage 17. During this phase, cells of the midgut epithelium may detach from each other (Reuter and Scott, 1990; own unpub-lished observations). The monolayered arrangement of the cells is ensured by the adhesion of the endodermal cells to the surrounding visceral muscle. The phase of reduced intercellu-lar adhesion between midgut epithelial cells might be important for (i) the intergration of the LBCs and the AMPs, which initially remain mesenchymal, into the midgut epithe-lium (Reuter et al., 1990; Hartenstein and Jan, 1992) and (ii) the rearrangement of the midgut epithelium from a short sac-like structure into the slender, elongated tube, the larval midgut. A strong adhesion to the visceral muscle and a reduced adhesion among the midgut epithelial cells themselves might play a permissive role during these morphogenetic changes. This view is corroborated by the observation of Newman and Wright (1981) that midgut morphogenesis is arrested when the midgut constrictions form in embryos mutant for the gene lethal(1)myospheroid [that encodes the β chain of the Drosophila PS-integrins (MacKrell et al., 1988)], where the attachment between visceral muscle and midgut epithelium is disrupted.

The third phase of midgut development is again characterized by strong cell-cell adhesion. It begins during mid stage 17 with the differentiation of a smooth septate junction that extends over the apicolateral 40–60% of the cell surface (Tepass and Hartenstein, 1993). This junctional specialization is typical for the arthropod midgut epithelium and it is generally believed to provide strong intercellular adhesion. The strong adhesion among epithelial cells is evidently necessary to resist the stretching forces which arise during the peristaltic movements of the midgut.

In vertebrates it has been shown that cell-cell and cell-sub-stratum adhesion systems are involved in both formation and maintenance of the epithelial structure and act independently in organizing the epithelial phenotype (Rodriguez-Boulan and Nelson, 1989; Wang et al., 1990; Fleming, 1992). Like epithelia in vertebrates, the formation of the midgut epithelium in Drosophila depends apparently on cell-substratum and cell-cell adhesion events. Preliminary observations (U. T., unpublished) suggest that the Drosophila PS-integrins, which are expressed at the interface of the midgut epithelium and the visceral mesoderm (Leptin et al., 1989), provide cell substratum adhesion during midgut epithelium formation. A further candidate molecule is a newly identified integrin β chain that is specifically expressed in the midgut epithelium (β_ν; Yee and Hynes, 1993). That cell-cell adhesion, which might depend on shg function, is involved in the formation of the midgut epithelium has been discussed above. Cell-cell and cell-substratum adhesion events presumably initiate the reorganization of the cell structure in a directed assembly process that leads, for example, to the polarization of the microtubules that has been observed in vertebrate epithelia (e.g. Bacallao et al., 1989) as well as in the Drosophila midgut epithelium (this work). A polarized array of microtubules, in turn, appears to be important for the selective transport of proteins and organelles to the apical or basolateral membrane domains that is critical for the maintenance of the epithelial phenotype. The formation of the midgut epithelium in Drosophila provides us with a model system to study the geneyic mechanisms that control early events in the reorganization of mesenchymal cells into polarized epithelial cells.

We are grateful to Silvia Yu for technical assistance, to Dan Kiehart, Seymor Benzer and Corey Goodman for providing antibody probes and to Judy Lengyel, Steve Crews, Lisa Fessler, Christiane Nüsslein-Volhard, Mary-Lou Pardue, Rolf Bodmer, and the Bloom-ington Stock Center for providing fly stocks. We thank Dorothea Godt and members of the Hartenstein laboratory for critical reading of the manuscript. This work was supported by NIH Grant NS29367 to V. H. and by a HFSPO fellowship to U. T.