Fork head prevents apoptosis and promotes cell shape change during formation of the Drosophila salivary glands

During embryonic development of many organisms, the movement and folding of epithelial layers give rise to a variety of tissues and organs. Epithelial invagination is a morphogenetic movement by which the epithelium folds inward through a process involving individual cell shape changes (reviewed by Ettensohn, 1985; Fristrom, 1988). Several general models have been proposed to explain how an epithelium invaginates. In the apical constriction model, nuclei migrate to a more basal position and the apical membranes constrict, thus turning a columnar cell into a wedge-shaped one. Changes in the underlying microfilament and microtubule systems are thought to drive this change in cell shape (reviewed by Ettensohn, 1985; Messier, 1978). In the second model for invagination, changes in cell-cell adhesion are proposed to drive internalization (Gustafson and Wolpert, 1962; Nardi, 1981; Mittenthal and Mazo, 1983). If an increase in adhesion between cells of an invaginating epithelium results in increased cell height while leaving the basal surface adherent to the substratum, the apical surface area would decrease, causing the epithelium to fold inward. During invagination of the mesodermal primordia in Drosophila embryos, changes in the location, size and morphology of the cadherin- and catenin-based adhesion junctions are associated with changes in the shape and internalization of the cells (Oda et al., 1998), supporting the cell-cell adhesion model for invagination. In the third model, changes in the extracellular matrix (ECM) are proposed to drive invagination (Lane et al., 1993). In gastrulating sea urchin embryos, invaginating cells are proposed to deposit a new hygroscopic layer of ECM between their apical surfaces and the older less hygroscopic layer. The new, more hygroscopic layer of ECM swells and increases in surface area due to its greater hydration, thus driving the bilayered ECM and underlying epithelial sheet to bend inward. Recent studies on epithelial invagination in genetically manipulatable organisms, such as C. elegans and D. melanogaster, have begun to identify mutations that affect invagination. A screen for mutations affecting vulval invagination in C. elegans identified the squashed vulva (sqv) genes that most likely encode components of a conserved glycosylation pathway (Herman et al., 1999; Herman and Horvitz, 1999). In sqv mutants, the vulval invagination partially collapses, resulting in a reduced invagination space (Herman et al., 1999). The sqv genes are proposed to establish and/or maintain the rigidity of the invaginating vulval epithelium. The collapsed phenotype could be either due to defects in adhesion between the invaginating cells or, perhaps more likely, due to defects in the rigidity of the ECM that lines the invaginating space.

In the gastrulating Drosophila embryo, epithelial invagination occurs during the internalization of the mesodermal primordium, and the anterior and posterior midgut primordia (reviewed by Leptin, 1999). Genetic studies in Drosophila identified two signaling molecules that regulate internalization of these tissues: Folded gastrulation (FOG), a putative secreted molecule, and Concertina (CTA), a putative Gα-like protein (Parks and Wieschaus, 1991; Costa et al., 1994). In fog mutant embryos, the mesodermal primordium is internalized, although in an uncoordinated manner, and the posterior midgut primordium completely fails to internalize (Parks and Wieschaus, 1991; Sweeton et al., 1991). The FOG signal is transduced via an unknown receptor to CTA, which then relays the signal to the cytoskeleton, causing cell shape change (Costa et al., 1994). The FOG/CTA signal most likely affects the actin cytoskeleton by activating RhoGEF2, which acts as a GTP exchange factor for a member of the Rho family of GTPases (Barrett et al., 1997; Hacker and Perrimon, 1998). Hyperactivation of the FOG/CTA signaling pathway causes ectopic apical constriction and cell shape change (Morize et al., 1998).

Internalization of the Drosophila salivary gland also occurs by invagination through changes in cell shape (Myat and Andrew, 2000). The secretory tubes arise from two placodes of epithelial cells at the ventral surface (reviewed by Andrew et al., 2000). The columnar placode cells become wedge-shaped through the constriction of their apical surface membranes and the coordinated migration of their nuclei to the basal domain (Myat and Andrew, 2000). The cell shape changes and subsequent invagination in the secretory primordia are both regulated events. In embryos mutant for the transcription factor, Huckebein (HKB), secretory cells become wedge-shaped and invaginate, but are not internalized in the same order as in wild-type embryos. As a consequence, the secretory tubes are round and remain closely associated with the anterior embryo surface, instead of being cylindrical and associating with the lateral body wall as in wild-type embryos. hkb expression in the secretory primordia presages the order in which cells change shape and invaginate, suggesting that hkb may play an instructive role in determining the order of invagination.

fork head (fkh), like hkb, is among the earliest genes to be expressed in the secretory primordia. fkh encodes a transcription factor of the winged-helix family, and is homologous to mammalian HNF3β (Weigel et al., 1989b). In fkh mutant embryos, salivary glands do not form, even though the secretory primordia are established (Weigel et al., 1989a). In this report, we show that fkh has a dual role in salivary gland formation, one in secretory cell survival and a second in cell shape change, specifically in the constriction of the apical surface membrane.

The wild-type flies used in all experiments were Oregon R. The fkh⁶ allele and the Df(3L)H99 strain are described in Flybase (1999). The UAS-rpr, UAS-hid, UAS-grim flies (provided by H. Steller, Massachusetts Institute of Technology) and UAS-rpr; UAS-hid flies (provided by J. Nambu, University of Massachusetts) were crossed to fkh-GAL4 flies (Henderson and Andrew, 2000) to induce apoptosis in the salivary glands. The expression profile of GAL4 in fkh-GAL4 embryos was detected by antibody staining for GAL4 (Santa Cruz Biotechnologies, Santa Cruz). Df(3L)H99 was recombined onto a fkh⁶ mutant chromosome; the genotype of the recombinant fkh⁶Df(3L)H99 chromosome was confirmed by whole-mount in situ hybridizations with rpr and hid antisense RNA probes, which did not detect transcript in the homozygotes (see below).

The rat dCREB-A antiserum (Andrew et al., 1997) was used at a dilution of 1:5000. The rat CRQ antiserum (Franc et al., 1999) was used at a dilution of 1:1000. The rat PASILLA (PS) antiserum (Seshaiah, 2000) was used at a dilution of 1:5000. The mouse monoclonal β-galactosidase (β-gal) antibody (Promega; Madison, WI) was used at a dilution of 1:5000. Embryo fixation and staining were performed as described by Reuter et al. (1990). Embryos homozygous for fkh, fkh H99 or H99 were unambiguously identified by the absence of staining for β-gal, which detects the expression of the Ubx-lacZ insert on the TM6B balancer chromosome. Immunostained embryos were mounted in methylsalicylate (Sigma, Missouri, IL) and visualized and photographed on a Zeiss Axiophot microscope (Carl Zeiss, Thornwood, NY) using Nomarski optics and Kodak print film (Eastman Kodak, Rochester, NY). Embryonic mRNA was detected by whole-mount in situ hybridization as described by Lehmann and Tautz (1994). rpr and hid cDNAs (provided by E. Baehracke), and a grim cDNA (provided by J. Abrams) were used to make antisense riboprobes. Embryos were mounted in 70% glycerol and visualized and photographed as described above for antibody-stained embryos.

Embryos processed for antibody staining with α-dCREB-A and α-β-gal were embedded in Epon (Eponate Kit™, Ted Pella, Reading, CA) as previously described (Myat and Andrew, 2000). One μm thick sections, obtained on a Reichart-Jung Ultracut E (Leica Inc.) microtome, were counter-stained with 1% Toluidine Blue and 1% sodium borate (Sigma, Missouri, IL). Stained sections were mounted in Permount (Fisher Scientific; Pittsburgh, PA) and viewed and photographed as described above for antibody-stained embryos.

Embryos were processed for TEM by first dechorionating them in 50% bleach and then fixing them in 5% glutaraldehyde (Polysciences Inc, Warrington, PA) and heptane (Sigma). After manual devitellinization, embryos were fixed in 4% glutaraldehyde and 2% acrolein (Polysciences) in 0.1 M cacodylate buffer (Sigma). Devitellinized embryos were transferred to a chilled mixture of 1% osmium tetroxide (Polysciences) and 2% glutaraldehyde in 0.1 M cacodylate buffer, and then post-fixed in 1% osmium tetroxide in 0.1 M cacodylate buffer. Fixed embryos were dehydrated and embedded in Epon as previously described (Myat and Andrew, 2000). 65 nm thin sections, obtained on a Reichart-Jung Ultracut E, were stained with 2% uranyl acetate (Polysciences) and lead citrate (Polysciences) before viewing on a Phillips CM120 transmission electron microscope.

fkh is proposed to play a major role in salivary gland determination and morphogenesis based on the failure of the salivary glands to internalize in fkh mutant embryos (Weigel et al., 1989a) and on FKH’s repression of trachealess (trh), which encodes a duct-specific transcription factor (Isaac and Andrew, 1996), in the secretory cells (Kuo et al., 1996). To study FKH’s function in detail, we began by analyzing its expression in the salivary gland throughout embryogenesis. fkh mRNA was first detected during embryonic stage 9 in the ventral-posterior cells of the secretory placode (Fig. 1A). fkh mRNA was detected at approximately equivalent levels in all secretory cells from early embryonic stage 10 (Fig. 1B) and throughout embryogenesis (Fig. 1C,D and data not shown). FKH is expressed in the salivary glands of late third instar larvae (Kuzin et al., 1994) where it activates expression of the Salivary gland secretion protein 3 (Sgs3) and Salivary gland secretion protein 4 (Sgs4) genes (Lehmann and Korge, 1996; Mach et al., 1996). This late expression suggests that fkh is expressed in the salivary gland throughout the larval stages until the onset of metamorphosis.

To characterize the role of fkh in salivary gland morphogenesis, we analyzed both whole-mount embryos and histological sections of salivary glands from wild-type and fkh mutants stained with a secretory cell marker, dCREB-A (Andrew et al., 1997). In wild-type embryos, salivary glands are internalized through a series of coordinated cell shape changes that begin with cells in the dorsal-posterior region of the primordium (Myat and Andrew, 2000; Fig. 2A,B). A wave of cell shape changes extends to the dorsal-anterior, then the ventral-anterior, and, finally, the ventral-posterior cells of the placode, resulting in the internalization of the entire primordium (Fig. 2E,F). Once the ventral-posterior cells are internalized, the cells at the distal end of the secretory tube appear to turn and migrate posteriorly (Fig. 2F), ultimately positioning the salivary glands along the anterior-posterior body axis (Fig. 2I,J). In fkh mutants, relatively normal salivary primordia were observed. However, distinct invaginating pits, which are normally seen in wild-type salivary glands at this stage (Fig. 2A,B), were not visible in fkh mutants (Fig. 2C,D). At later stages, the size of the fkh secretory placode and the levels of dCREB-A staining were reduced (Fig. 2G,H). Even at late embryonic stages, dCREB-A-positive secretory cells were still located at or near the embryo surface of fkh mutants (Fig. 2K,L), confirming the failure to invaginate.

Comparisons of histological sections of wild-type and fkh embryos revealed three differences in the secretory primordia and surrounding tissues. (1) Wild-type salivary placode cells were uniformly columnar (Fig. 3A,B), whereas the fkh salivary placode cells were a mixture of round and columnar cells, with the ventral-most region consisting mostly of round cells (Fig. 3E-H). (2) wild-type secretory cells were always in a monolayer (Fig. 3A-D) while many of the fkh mutant cells were found in multiple layers with some cells apparently dissociating from the salivary epithelium (Fig. 3E-L). (3) Very few macrophage-like cells or pyknotic structures were observed near the wild-type secretory primordia (Fig. 3C,D), whereas numerous macrophage-like cells and pyknotic structures were observed surrounding the salivary primordia in the fkh mutants (Fig. 3G-L). Many of the macrophage-like cells in the fkh mutants (Fig. 3I,J, small arrows) showed variable levels of dCREB-A staining (Fig. 3I, arrowheads), suggesting that these cells may have engulfed dCREB-A positive secretory cells. Interestingly, fkh secretory cells in the dorsal-posterior region of the placode (Fig. 3F,H,J,L) had nuclei in a basal position, as in wild-type embryos (Fig. 3B,D). However, despite this nuclear translocation, most fkh cells remained columnar (Fig. 3H,J,L) rather than becoming wedge-shaped, as occurs in wild-type salivary glands.

To determine whether the oval-shaped cells observed in or near the secretory placodes of fkh mutants were macrophages, we stained for the macrophage marker protein, CROQUEMORT (CRQ; Franc et al., 1999). During embryonic stage 11, a few CRQ-positive cells were found in the region of the secretory primordia of wild-type embryos (Fig. 4A). In contrast, many more CRQ-positive cells were found in or near the secretory primordia of fkh mutants (Fig. 4C). At the end of invagination, we observed CRQ-positive cells near the internalized secretory cells of wild-type embryos (Fig. 4B), suggesting that low levels of apoptosis may occur normally in the region of the salivary glands. However, significantly more CRQ-positive cells were observed in the region of the secretory primordia in fkh mutants (Fig. 4D). The CRQ-positive cells observed in whole-mount embryos are likely to correspond to the oval-shaped cells observed in histological sections. This finding suggested an increased recruitment of circulating macrophages to the salivary glands of fkh mutants compared to wild type. Since macrophages are professional phagocytes that engulf dying cells (Tepass et al., 1994; Morrissette et al., 1999), their presence suggested a significant increase in cell death in the salivary glands of fkh mutants compared to wild type.

To verify the increased cell death of secretory cells in fkh mutants and to determine whether the cells were dying by apoptosis or necrosis, we examined the expression of three Drosophila death genes, reaper (rpr), grim and head involution defective (hid), in both wild-type and fkh mutant embryos. rpr, grim and hid encode proteins that function near the top of the Drosophila cell death pathway and induce a cascade of caspase activation (reviewed by Abrams, 1999; Song and Steller, 1999). Expression of rpr, grim and hid precedes apoptotic cell death and the absence of these genes eliminates the cell death that normally occurs during embryogenesis (White et al., 1994; Grether et al., 1995; Chen et al., 1996). In wild-type embryos, rpr RNA was initially detected in only a small group of secretory cells found posterior to the invaginating pit (Fig. 4E). At the end of invagination in wild-type embryos, rpr RNA was detected at relatively high levels in the few secretory cells remaining at the ventral surface, but not in the internalized secretory cells (Fig. 4F). In early fkh mutants, rpr RNA levels were significantly increased in all cells except those in a dorsal-anterior position in the primordia (Fig. 4G). Slightly later, high levels of rpr RNA were detected in all secretory cells of fkh mutants (Fig. 4H). hid RNA levels were also increased in fkh relative to wild type. Early fkh embryos showed increased hid expression levels in an apparently random array of secretory cells (Fig. 4K) compared to wild-type embryos at the same stage, in which hid expression levels were lower in the secretory cells than they were in the surrounding non-salivary gland tissues (Fig. 4I). At later stages, slightly increased levels of hid RNA were detected in wild-type embryos, particularly in the posterior cells that also had relatively high levels of rpr expression (Fig. 4J). High levels of hid RNA were detected at later stages in all fkh mutant secretory cells, with particularly high levels in the ventral primordia (Fig. 4L). Unlike rpr and hid, grim RNA was not detected in the secretory primordia of either wild-type or fkh embryos (data not shown).

The increased levels of both rpr and hid in the secretory primordia suggested that in the absence of fkh function, secretory cells die by apoptosis. In support of an apoptotic mechanism of cell death, TEM analyses revealed that fkh mutant salivary glands showed cytological changes characteristic of apoptosis (Abrams et al., 1993). For example, fkh secretory cells, particularly those in the ventral part of the placode, had separated from neighboring cells and demonstrated nuclear condensation, cytoplasmic shrinkage and fragmentation (Fig. 5C,D, white arrows and black arrowheads). In addition, numerous macrophages with ingested apoptotic bodies were observed near the secretory cells of fkh mutants (Fig. 5C,D, white arrowheads). This is in contrast to the secretory cells of wild-type embryos, which at an equivalent stage, were found as a monolayer of wedge-shaped cells surrounding a central lumen (Fig. 5A,B). We observed very few macrophages or apoptotic bodies adjacent to the wild-type salivary glands.

The apoptotic cell death observed in the early secretory primordia of fkh mutants indicated that fkh was required for secretory cell survival (Figs 3-5). Thus, secretory cells may fail to invaginate in fkh mutants simply because the cells are dead or dying. Indeed, the ectopic expression of rpr and hid, but not grim, is effective in inducing early secretory cell death, which if extensive enough, prevents internalization (data not shown). Alternatively, fkh may have two separate roles in the salivary gland, one to promote cell survival and another to control invagination of the primordia. To distinguish between these possibilities, we rescued the apoptotic secretory cell death in fkh mutants by generating embryos that were mutant for fkh and also carried Df(3L)H99, a small deficiency that deletes rpr, hid and grim (fkh H99; White et al., 1994). Normal salivary glands were formed in embryos homozygous for Df(3L)H99 (H99; Fig. 6A,B). In the fkh H99 embryos, dCREB-A staining was detected in the entire secretory placode at early stages (Fig. 6C,D). This staining completely disappeared by embryonic stage 13 (Fig. 6E,F), suggesting that either fkh is required to maintain dCREB-A expression or that the fkh H99 secretory cells are still dying. To address this issue, we analyzed the expression of another secretory marker protein, PS, whose expression was thought to be fkh independent (Seshaiah et al., unpublished data). In wild-type embryos, PS was expressed at high levels in the salivary glands throughout embryogenesis (Fig. 6G; P. Seshaiah, B. Miller and D. J. A., unpublished data). In fkh mutant embryos, PS was initially expressed in the entire secretory placode (data not shown), and at reduced levels in the surviving ring of secretory cells (Fig. 6H). Importantly, PS was expressed to very high levels in all secretory cells throughout embryogenesis in the fkh H99 embryos (Fig. 6I,J). Nonetheless, the PS-expressing cells in the fkh H99 embryos were not internalized and remained at their site of origin on the ventral surface (Fig. 6J). Therefore, in addition to its early role in promoting secretory cell survival, FKH is also required for the invagination of the secretory cells.

Histological sections and TEM analyses of fkh H99 embryos confirmed the failure of the secretory cells to be internalized and revealed additional changes in the salivary glands (Figs 7 and 8). Cells of the fkh H99 placode were columnar like the wild-type cells (compare Fig. 7A,B with Fig. 3A,B), except for cells in the ventral-posterior portion of the placode, which were round and stacked on one another (Fig. 7B). Interestingly, although coordinate nuclear migration occurred in the secretory cells of all fkh H99 embryos in approximately the same temporal and spatial pattern as in wild type, the apical surface membranes failed to constrict even at very late stages (Fig. 7C-F). Consequently, the invaginating pits of fkh H99 salivary glands were wide and shallow compared to those of wild-type embryos. We confirmed the absence of apical membrane constriction by analyzing transmission electron micrographs of invaginating wild-type and fkh secretory cells. Prior to invagination, wild-type secretory cells have broad and flat apices (Fig. 8A, arrow). Wild-type secretory cells at the center of the invaginating pit have constricted apices, and showed numerous membrane protrusions at the apical surface (Fig. 8B, white arrowheads and black arrows). In contrast, secretory cells at the lateral edges of the invaginating pit, have flat and broad apices with few membrane protrusions (Fig. 8B, black arrowheads). In late fkh H99 embryos, the apices of the secretory cells were unconstricted and appeared flat and broad (Fig. 8C, arrows), similar to the apices of secretory cells in early wild-type embryos (Fig. 8A, arrow). Fewer apical membrane protrusions are found at the surface of fkh H99 secretory cells compared to wild type (Fig. 8D, arrows). Altogether, these findings suggested that fkh is required for secretory cells both to survive and to invaginate. Moreover, the absence of apical membrane constriction may be the underlying defect in the failure of fkh mutant salivary glands to internalize.

We have shown that salivary glands do not form in fkh mutants in part because the secretory cells die by apoptosis. Apoptosis of the fkh secretory cells was detected as early as stage 11 and was still apparent at stage 14. Cells in the ventral-most region of the fkh mutant placodes were the first to express rpr (Fig. 4G) and showed the earliest morphological abnormalities; cells were frequently round, instead of columnar, and were no longer in an epithelial monolayer (Fig. 3E-J). Cells in the more dorsal regions of the placode were elongated, were in a monolayer and underwent basal nuclear migration at approximately the appropriate developmental stage (Fig. 3E-J). However, at later stages, the dorsal cells of the fkh mutant salivary glands showed the same morphological abnormalities as observed in ventral cells at earlier stages (Fig. 3K,L), consistent with the relatively delayed expression of rpr in these cells (Fig. 4H). The time difference may reflect when FKH function is required for survival in the ventral versus dorsal primordia. Indeed, fkh expression is initiated in the ventral-most cells and only later expands to the rest of the secretory primordia (Fig. 1).

The early expression of rpr and hid in the secretory primordia and the pattern of cell death in fkh mutants indicate that cell death is not induced because of a failure of fkh secretory cells to change shape and invaginate. Expression of both rpr and hid, and morphological signs of apoptosis were detected in the fkh secretory placode well before these cells would normally change shape. Moreover, apoptosis in fkh embryos occurred first in the ventral secretory cells which are normally the last cells to change shape and invaginate in wild-type embryos. In the absence of fkh, expression of rpr and hid are elevated, suggesting that FKH prevents apoptosis by suppressing rpr and hid expression, either directly or indirectly, and that expression of both rpr and hid is responsible for the extensive secretory cell death observed in fkh mutants. The normal expression of rpr and hid in the posterior-most secretory cells in wild-type embryos suggests that some regulated cell death may normally occur, even though fkh levels remain high in all secretory cells. The posterior-most secretory cells that first express rpr (Fig. 4E) and that later express both rpr and hid (Fig. 4F,J) may remain at the embryo surface during the internalization of the remainder of the secretory primordia to provide an anchor to the surrounding ectoderm. At the end of invagination when all cells of the placode have been internalized, we postulate that these posterior-most cells may undergo apoptosis, which would separate the internalized secretory cells from the epidermis. This separation would then allow the posterior secretory cells to form contacts with the duct cells, which internalize after the secretory cells and form the tubes that connect the secretory cells to the larval mouth (for review, see Andrew et al., 2000).

fkh may promote cell survival not only in the salivary gland, but also in the anterior and posterior midgut. In fkh mutants most of the midgut epithelium degenerates and is infiltrated by a large population of macrophage cells (Weigel et al., 1989; Tepass et al., 1994). Interestingly, the C. elegans HNF-3/FKH homolog, daf-16, acts downstream of the insulin-receptor signaling pathway to double life span (Lin et al., 1997; Hsin and Kenyon, 1999). Perhaps certain members of the FKH family of transcription factors function to promote survival both at the cellular and whole organismal levels.

The formation of an ectodermally derived organ, such as the Drosophila salivary gland, requires that the primordia initially found at the surface of the embryo invaginate to form an internalized structure. Our previous work demonstrated that salivary secretory cell invagination occurs in a regulated and sequential manner, most likely by an apical constriction mechanism (Myat and Andrew, 2000). In contrast to HKB, a transcription factor that regulates the order of cell shape changes in the secretory placode (Myat and Andrew, 2000), FKH appears to mediate the actual cell shape changes. In fkh H99 embryos, where secretory cells are kept alive by the removal of death genes, secretory cells fail to constrict their apical surface membranes and, therefore, do not change shape and internalize (Figs 7, 8). In contrast, cells in the dorsal-posterior region of the fkh placode did occasionally constrict their apical surface membranes, thus forming a distinct invaginating pit like those of wild-type salivary glands (Fig. 3H,L). The failure of secretory cells in fkh H99 embryos to constrict their apical surface membranes, while their counterparts in fkh embryos are occasionally able to do so, could be explained by the extensive cell death in the secretory placode of fkh mutants. By the time fkh secretory cells start to constrict, a significant number of secretory cells have already died and been cleared away by macrophages. The presence of fewer secretory cells in the placode could reduce the surface tension in the epithelium, thus facilitating apical constriction of the remaining secretory cells. In contrast, when the normal number of secretory cells is restored in fkh H99 embryos by blocking apoptosis, the same physical constraints exist in the secretory epithelium as in wild-type embryos, and due to the absence of fkh function, cells fail to constrict their apical membrane. Alternatively, the cells that appear wedge-shaped in the fkh mutants may actually be dying cells that are not able to maintain their shape. These dying cells may collapse and remain broad only in basal regions, where the nuclei are found.

While apical membranes failed to constrict in the secretory cells of fkh H99 embryos, nuclear migration was not affected (Figs 7 and 8). It was previously not possible to discern whether the basal migration of nuclei and the constriction of the apical membrane were two separate events that contributed independently to cell shape change, or whether one event depended on the other. Live imaging of cell shape changes during Drosophila gastrulation using three-dimensional fluorescence microscopy showed that in some cases, nuclei moved basally in the absence of apical constriction, suggesting that these two processes can be independent (Kam et al., 1991). Our histological and TEM data of fkh H99 embryos confirm that nuclear migration and apical membrane constriction are indeed two separate events and that apical constriction does not occur simply because of basal nuclear movement (Figs 7 and 8). Ultrastructural and immunocytochemical data from a variety of epithelia that undergo apical constriction indicate that the apical contractile machinery consists of an actin-myosin network (Bement et al., 1993; Young et al., 1991, 1993). This model is supported by experiments where cytochalasin, an actin-depolymerizing agent, arrests or reverses invagination (Wrenn, 1971; Karfunkel, 1972; Spooner and Wessells, 1972; Morriss-Kay, 1981). Similar effects are observed with calcium treatment of invaginating epithelia (Ash et al., 1973; Moran, 1976; Moran and Rice, 1976; Brady and Hilfer, 1982). Since the basal movement of nuclei is linked to the microtubule network (Messier, 1978), and the constriction of the apical membrane is linked to the actin filament system, it is possible that FKH functions in a signal transduction pathway that specifically targets actin. TEM data of wild-type embryos at different stages of salivary secretory cell invagination demonstrated that constriction of the apical surface membrane was accompanied by dynamic membrane protrusions at the apical surface (Fig. 8B and data not shown). These data support the apical constriction model for invagination which predicts that apical membrane protrusions form as a result of the actomyosin-driven contraction of the apical domain. However, it is still possible that active secretion and cell-cell adhesion also contribute significantly to the inward folding of the epithelium. Thus, the fkh H99 embryos, in which apical membrane constriction is specifically abrogated, provide a valuable tool with which we can further address the roles of apical constriction, apical secretion and cell-cell adhesion in epithelial invagination.

Our studies on Drosophila FKH demonstrate two roles during development of the embryonic salivary glands, one in cell survival and another in apical membrane constriction. Among the known targets of FKH in the salivary gland are two genes that also encode transcription factors, trh (Isaac and Andrew, 1996; Kuo et al., 1996) and dCREB-A (this work), and two salivary gland glue protein genes, Sgs3 and Sgs4 (Lehmann and Korge, 1996; Mach et al., 1996). trh is expressed and required in the duct cells and its initial expression in the secretory cells is later repressed by FKH (Isaac and Andrew, 1996; Kuo et al., 1996). Early expression of dCREB-A is FKH-independent; however, later expression of dCREB-A requires FKH, especially in cells rescued from programmed cell death (Fig. 6). None of these downstream target genes is likely to mediate FKH’s roles in cell survival and/or apical membrane constriction. Sgs3 and Sgs4 are not expressed until several days after salivary gland internalization (Lehmann, 1996). Secretory cell survival and internalization are completely normal in embryos lacking function of trh or dCREB-A (Isaac and Andrew, 1996; Andrew et al., 1997) and in embryos where trh is continuously expressed in the secretory cells (D. D. Isaac and D. J. A., unpublished observation). Thus, it is important to identify the target genes that mediate FKH’s role in early secretory cell survival and morphogenesis. To this end, we are characterizing two recently identified genes whose expression in the salivary gland is FKH-dependent (E. Abrams and D. J. A., unpublished observation).

We wish to extend our thanks to P. Bradley, C. Machamer, and K. Wilson for their constructive criticisms of the manuscript. We thank J. Abrams, E. Baehracke, J. Nambu and H. Steller for fly stocks and cDNA constructs. We thank K. White for the α-CRQ antisera. We thank C. Cooke for training M. M. M. in TEM. We thank the C. Goodman Laboratory for TEM fixation protocols. We thank the other members of the Andrew lab for their material and intellectual contributions to the work. This work was supported by an NIH grant to D. J. A. (RO1 GM51311) and a Jane Coffin Childs postdoctoral fellowship to M. M. M.