Control of Drosophila Gastrulation by Apical Localization of Adherens Junctions and RhoGEF2

See allHide authors and affiliations

Science  19 Jan 2007:
Vol. 315, Issue 5810, pp. 384-386
DOI: 10.1126/science.1134833


A hallmark of epithelial invagination is the constriction of cells on their apical sides. During Drosophila gastrulation, apical constrictions under the control of the transcription factor Twist lead to the invagination of the mesoderm. Twist-controlled G protein signaling is involved in mediating the invagination but is not sufficient to account for the full activity of Twist. We identified a Twist target, the transmembrane protein T48, which acts in conjunction with G protein signaling to orchestrate shape changes. Together with G protein signaling, T48 recruits adherens junctions and the cytoskeletal regulator RhoGEF2 to the sites of apical constriction, ensuring rapid and intense changes in cell shape.

Apical constriction of cells can contribute to the invagination of epithelia, such as during gastrulation or organogenesis, and the closure of wounds. In the Drosophila embryo, apical constrictions occur along the ventral side of the blastoderm epithelium, leading to the formation of the ventral furrow and the invagination of the mesoderm (1). Proteins necessary for the mechanics of these cell shape changes include the Rho guanosine 5′-triphosphate–exchange factor RhoGEF2 (2, 3) and a heterotrimeric G protein. Whereas RhoGEF2 is essential for furrow formation, disruption of the heterotrimeric G protein, such as by loss of its α subunit Concertina (Cta), leads to a delay but no lasting defects in mesoderm morphogenesis (4, 5). These maternally supplied proteins must be activated under the control of the zygotic genome in the embryo.

Twist is the zygotic transcriptional activator that is essential for the cell shape changes that produce the ventral furrow. One of its targets is the transcriptional repressor Snail, which is also essential for mesodermal morphogenesis (6).

However, the cell biological events responsible for the cell shape changes must ultimately be regulated by targets that are not transcription factors. Of the known Twist targets, only one, folded gastrulation (fog), is involved in mediating shape changes. Mutants in fog, which codes for a secreted peptide (7, 8), show the same defects as embryos lacking Cta. Fog is therefore thought to act in the same pathway as Cta, which we refer to as Fog/Cta signaling.

Fog/Cta signaling is thought to cause changes in the actin cytoskeleton in conjunction with RhoGEF2. Recruitment of myosin from basal to apical in constricting ventral cells is partly dependent on Fog/Cta and absolutely dependent on RhoGEF2 (8, 9). Furthermore, the mammalian homologs of RhoGEF2 and Cta interact (10). Finally, binding of Drosophila RhoGEF2 to microtubules by means of EB1 is disrupted by activated Cta (11). Given that myosin recruitment and apical constriction are reduced but not abolished in the absence of Fog/Cta (8), there must be other factors regulated by Twist that explain its effects on apical constriction.

In a screen for genes that mediate the zygotic control of gastrulation (12), we found the region uncovered by the chromosomal deficiency Df(3R)TlP to be necessary for the proper formation of the ventral furrow (fig. S1). Phenotypic analysis and molecular mapping of a set of overlapping deficiencies (table S1) identified the gene T48 as being responsible for the defects seen in Df(3R)TlP [supporting online material (SOM) text and Fig. 1, A and B]. T48 is expressed in the mesoderm (Fig. 1C) under control of Twist (13, 14). It codes for a predicted protein with a signal peptide and a potential transmembrane domain. When an internally hemagglutinin-tagged T48 protein (T48HA) was expressed in embryos, it localized at the peripheries of blastoderm cells, consistent with a close association with or insertion into the plasma membrane (Fig. 1D). Optical cross-sections showed that T48HA is targeted to the apical membrane (Fig. 1, E and F).

Fig. 1.

Characterization of T48. (A and B) Wild-type and age-matched Df(3R)CC1.2 (T48) embryos stained for Twist (brown) and Even-skipped (blue). For statistical evaluation, see figs. S2 and S3. (C) T48 RNA expression at blastoderm stage. (D to F) T48HA localization. (D) Surface view of stage 5 embryo; staining is at the cellular interfaces. [(E) and (F)] Optical sections through the blastoderm epithelium, apical is up. T48 is at the apical cell membranes, overlapping phospho-tyrosine (PTyr) staining, but not Bazooka (Baz). Scale bars, 10 μm. (G) Coimmunoprecipitates of 35S-labeled T48 and Crb with glutathione S-transferase fusions of PDZ domains (17). Coomassie-stained SDS–polyacrylamide gel electrophoresis bands of the PDZ-construct input shown below. (H to M) Co-localization of PDZ domains with T48HA. S2 cells transfected with GFP-tagged RhoGEF2-PDZ [(H) to (J)] or Scribbled-PDZ [(K) to (M)] alone or in combination with T48HA or T48HAΔITTEL. (See figs. S4 and S5 for statistical and graphic evaluation.)

No other structural motifs are recognizable in the protein. However, the C-terminal amino acid sequence –Ile-Thr-Thr-Glu-Leu (-ITTEL) conforms to the class I consensus for peptides that interact with PDZ domains. T48 has no obvious human ortholog but shows some similarity to the intracellular part of Fras1 (15), which also has a PDZ-binding motif. To find candidates for PDZ domains that might interact with T48, we analyzed the putative PDZ-binding sequence with an algorithm designed to determine the PDZ domains that show the optimal fit for any given peptide (16, 17). Of the predicted interactors (table S2), RhoGEF2 was particularly interesting in view of its role in ventral furrow formation (2, 3). Furthermore, the mammalian ortholog of RhoGEF2 has been shown to bind to Plexin-B1 by means of a PDZ-binding motif (–Val-Thr-Asp-Leu) very similar to that of T48 (18).

We tested whether the C terminus of T48 is indeed able to interact with RhoGEF2. A 35S-labeled C-terminal peptide of T48 preferentially coprecipitated with the PDZ domain of RhoGEF2 rather than those of other PDZ domain–containing proteins, in contrast to Crumbs, which was used as a control and which preferentially coprecipitated with PDZ domains from its physiological interaction partner Stardust, as well as Bazooka (Fig. 1G). In Schneider S2 cells, a green fluorescent protein (GFP)–tagged RhoGEF2 PDZ domain or full-length RhoGEF2 was localized in the cytoplasm or formed intracellular aggregates (Fig. 1H) when expressed alone, but localized to the plasma membrane when coexpressed with T48 (Fig. 1I and figs. S4 and S5). In both assays, the interaction required the presence of the -ITTEL motif and was not seen with other PDZ domains (Fig. 1, J to M). Thus, T48 interacts with RhoGEF2 by means of its PDZ-binding motif and is able to enrich RhoGEF2 to the plasma membrane.

To understand the function of T48 during gastrulation, we studied the subcellular localization of RhoGEF2 and its dependence on T48 in the developing embryo. Before gastrulation, the apical surfaces of the blastoderm epithelium are dome shaped and the developing adherens junctions are located subapically. RhoGEF2 is associated with the basally located furrow canals, whereas Armadillo is found just below this site and at a subapical position of the lateral cell membranes (Fig. 2, A to C) (19, 20).

Fig. 2.

Redistribution of RhoGEF2 and Armadillo. Sections of wild-type [(A) to (L)] and T48 embryos [(M) to (Q)] stained for RhoGEF2 and Armadillo as indicated. (A to C and F) Stage 5 wild-type embryos; RhoGEF2 and Armadillo are lost from the basal end in ventral cells (arrowhead). (D, E, and G) Late stage 5: Disappearance of subapical Armadillo in ventral cells and first signs of apical localization of RhoGEF2 in cells with still-rounded surfaces [arrowheads in (E) and (G)]. [(F) and (G)] Details of embryos shown in (B) and (C) and in (D) and (E), respectively [marked by brackets in (B) and (D)]. (H to L) Stage 6: Strong apical localization of RhoGEF2 and Armadillo in constricting cells. (J) Detail of (H) and (I): Nonoverlapping localization of Armadillo (concentrated at cell junctions) and RhoGEF2 (apical surface). [(K) and (L)] Adherens junctions are apical throughout the mesoderm, including nonconstricting cells (arrowhead). (M to Q) Stage 6 T48 mutant embryos (17) show reduced apical localization of RhoGEF2 [(N) and (Q)]. Armadillo relocalization and apical flattening occur [(O), arrowhead], but apical constriction is delayed. Black-and-white fluorescence images were color-inverted in Photoshop.

After cellularization was completed, these distributions changed specifically in ventral cells (Fig. 2, B to E). Even before morphological changes occurred, RhoGEF2 and Armadillo disappeared from the basal ends (Fig. 2, A to C and F). Subsequently, Armadillo disappeared from its subapical site and accumulated apically (8) (Fig. 2, D and G). A weak association of RhoGEF2 with the apical plasma membrane was seen at this stage (Fig. 2, E and G).

As cells began to flatten apically, high levels of both RhoGEF2 and Armadillo accumulated apically (8, 20) (Fig. 2, H to L). Although they concentrated in the same region of the cell, Armadillo was restricted to the cell junctions, whereas RhoGEF2 was often more enriched between these sites (Fig. 2J). Notably, movement of the adherens junctions occurred not only in constricting cells but also in the more lateral mesodermal cells that flattened and became stretched on their apical sides (Fig. 2K).

To examine whether these processes depend on T48, we stained stage-selected T48 mutant embryos (17). Loss of RhoGEF2 and Armadillo from the basal side was unaffected in these embryos (Fig. 2, M and N), as was the apical concentration of Armadillo. The cells flattened apically (Fig. 2O) and lengthened, but the absence of constrictions resulted in a thick placode rather than an indentation (Fig. 2, P and Q). Localization of RhoGEF2 to the apical membrane was slightly delayed and possibly reduced (Fig. 2, N and Q). T48 therefore contributes to but is not essential for the recruitment of RhoGEF2 to the apical membrane. This is consistent with the observation that furrow formation is not completely abolished, but only delayed or weakened. We therefore examined other mechanisms that might participate in RhoGEF2 localization.

As in the case of T48, mutations in the Fog/Cta pathway delay but do not abolish apical constriction and furrow formation (4, 7). We therefore considered whether Fog/Cta signaling might cooperate with T48 to recruit RhoGEF2. In embryos lacking Cta, the recruitment of RhoGEF2 was weakened (Fig. 3B). Combining mutations in cta and T48 resulted in much more notable effects (Fig. 3D). These cta,T48 embryos failed to make a furrow; the lack of apical constrictions was mirrored by a failure to accumulate RhoGEF2 apically (Fig. 3 and fig. S6). Thus, T48 and Fog/Cta signaling act in parallel to concentrate RhoGEF2 apically.

Fig. 3.

Effect of Cta, Twist, and Snail on RhoGEF2 and Armadillo. Sections of stage 6 mutant embryos. (A and B) Embryos derived from homozygous concertinaR10 mothers. (C and D) Homozygous T48 embryos derived from homozygous concertina mutant mothers. See fig. S6 for a more extensive documentation of the cta;T48 mutant phenotype. (E and F) Homozygous twistEY53R1 mutant embryos. (G and H) Homozygous snail mutant embryos (Df(2L)TE116GW11). Fluorescent pictures were inverted.

We also observed severe defects in the behavior of the adherens junctions in the double-mutant embryos. Armadillo staining disappeared from its tight subapical localization but did not reaccumulate apically (Fig. 3C and fig. S6). Thus, movement of the junctions is not simply mediated by a tensile force from the constricting actin cytoskeleton—an independent step of at least partial disassembly must occur. We speculated that this might be controlled by Snail, which regulates the disassembly of cell junctions in vertebrates. We found that the disassembly of Armadillo from the subapical position was indeed blocked in snail (but not in twist) mutant embryos (Fig. 3, E and G). Thus, Snail acts in parallel to Twist to direct the disassembly of subapical junctions, a process to which currently unknown Twist targets may also contribute (SOM text).

Having observed that T48 and Fog/Cta activation are required for the apical localization of RhoGEF2 and Armadillo, we tested whether T48, like Fog/Cta signaling, was able to trigger their relocalization in other cells. Ubiquitous expression of T48 in the embryo led to a concentration of RhoGEF2 at the apical membranes of lateral cells (compare Fig. 4, A and B; fig. S7). Armadillo localization in ectodermal cells was no longer restricted to a distinct subapical domain but extended to the apical end of the lateral membranes in many cells. When T48 was coexpressed with activated Cta, this effect was slightly enhanced, and some embryos showed morphological defects (fig. S7).

Fig. 4.

Induced relocalization of RhoGEF2 and Armadillo, and a model for the control of furrow formation. (A and B) Details of sections of the wild-type (A) and T48-overexpressing (B) embryos shown in fig. S7. (C) Model for T48 and Fog/Cta function during gastrulation. (D) Genetic hierarchy of the genes acting downstream of Twist in regulating adherens junctions and cytoskeletal rearrangements.

With T48, we found a missing factor in the control cascade from transcriptional regulation by Twist to the cell biological mediators of furrow morphogenesis (Fig. 4, C and D). Two Twist targets, Fog and T48, appear to act in separate pathways that converge on RhoGEF2, which integrates the signal to activate myosin and modify the actin cytoskeleton (8, 9). Our model shows the maternally supplied RhoGEF2 as largely attached to microtubules by means of EB1 (11). The onset of Twist expression has two effects. Fog is synthesized, which triggers the activation of Cta. This in turn releases RhoGEF2 from the microtubules that, by analogy to its vertebrate homologs, may bind to Cta through its RGS domain (10), allowing some myosin activation and constriction. In parallel, T48 is synthesized and targeted to the apical membrane, where it acts to concentrate RhoGEF2 through its PDZ-binding motif. In the absence of Fog-mediated displacement of RhoGEF2 from EB1, T48 can probably still recruit sufficient freely diffusible RhoGEF2 to allow slow constriction. Only when both mechanisms fail are the downstream events of constriction and junction reassembly abolished completely.

The utilization of Gα12/13 proteins and a microtubule-bound RhoGEF have also been reported in vertebrate gastrulation (21, 22). The absence of an obvious homolog of T48 in vertebrates might suggest that this element of the control mechanism is unique to Drosophila gastrulation. However, the PDZ-binding motif in Plexin-B1 is similar to that of T48 and acts during neuronal growth cone remodeling by recruiting PDZ-RhoGEF (18). Therefore, this mechanism of controlling cell shape may operate in a variety of systems.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S7

Tables S1 to S3


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article