Gprk2 adjusts Fog signaling to organize cell movements in Drosophila gastrulation

During morphogenetic movements in the course of development, cells dynamically change their shapes and positions (Leptin, 2005; Solnica-Krezel and Sepich, 2012). These cell movements are spatially and temporally organized for shaping tissues and embryos into their correct forms. Forces driving cell movements are generated by cytoskeletal filaments and motor proteins, such as actin filaments and myosin motors (Jacinto et al., 2002; Lecuit and Lenne, 2007; Martin, 2010). Cell-cell adhesions and intercellular signals regulate force generators in each cell and are thought to organize cell movements in a spatiotemporal manner. However, the cellular mechanics and signals governing cell movements at the tissue level are not fully understood.

Gastrulation of Drosophila melanogaster is an ideal system for studying morphogenetic movement, as cell movements occur in spatially distinct regions on a precise schedule (Costa et al., 1993). A cellular blastoderm embryo (3 hours after egg laying) consists of ∼6000 epithelial cells surrounding the outer cortex. About 1000 cells on the ventral side acquire mesodermal cell fate by means of the activity of Twist (Twi) and Snail transcription factors, and these cells then invaginate into the embryo. During mesoderm invagination, ventral mesodermal (VM) cells, which typically expand to a 12-cell width along the dorsoventral axis, constrict their apical cell surfaces. Consequently, the epithelial sheet bends and a ventral furrow is formed along the anteroposterior axis (see Fig. 1A, Fig. 2A). A recent report proposed that the constrictive force operates isotropically in each VM cell, but that a tissue-level tension along the longer anteroposterior axis restricts the apical constriction anisotropically to the dorsoventral axis (Martin et al., 2010). Apical constriction is not the only cell movement during mesoderm invagination. Lateral mesodermal (LM) cells, which are located on the sides of VM cells (typically a three-cell width on each side), do not undergo apical constriction, but involute toward the ventral furrow (see Fig. 3A,C). Finally, left and right LM cells meet at the midline and close the furrow (see Fig. 3E). These cell movements proceed sequentially and mesoderm invagination is completed within 15 minutes. The movement of LM cells has not been characterized in detail, and it has also remained unclear how the VM/LM area is defined in mesodermal cells (Leptin and Roth, 1994).

Previous studies have revealed molecular mechanisms regulating the apical constriction of VM cells (Lecuit and Lenne, 2007; Martin, 2010). According to a current model, the secreted signaling protein Folded gastrulation (Fog), which is expressed in a Twi-dependent manner, induces the apical constriction (Costa et al., 1994; Morize et al., 1998). The Fog receptor remains unknown, but is expected to be a G protein-coupled receptor (GPCR) because a heterotrimeric G protein Gα subunit [Concertina (Cta)] acts in the downstream signaling (Parks and Wieschaus, 1991). Fog signaling stimulates the apical localization of Myosin II protein, which generates a force that constricts the cell surface (Dawes-Hoang et al., 2005). In parallel with Fog signaling, other pathways coordinately regulate the actomyosin network to induce apical constriction. The transmembrane protein T48, which is another target of Twi, regulates the localization of myosin motors through direct binding with RhoGEF2 (Kölsch et al., 2007). Abelson kinase together with RhoGEF2 organizes actin filaments at the apical cell surface (Barrett et al., 1997; Fox and Peifer, 2007). These parallel pathways explain the incomplete progression of apical constriction in the absence of one pathway, as seen in fog mutant embryos (Sweeton et al., 1991; Costa et al., 1994; Kölsch et al., 2007).

Heterotrimeric G proteins are GTP-binding signaling complexes that consist of Gα, Gβ and Gγ subunits (Neer, 1995). They are involved in diverse biological processes, such as phototransduction and chemotaxis. Numerous studies have revealed that G protein signaling can be regulated positively or negatively by many factors. Such adjustments of cell signaling guarantee its sensitivity and robustness (Ross, 2008). The roles of G protein regulators in development are poorly understood. Ric-8 is a putative guanine nucleotide-exchange factor for G protein (Tall et al., 2003), and Drosophila Ric-8 plays important roles during mesoderm invagination (Kanesaki et al., 2013). GPCR kinase (also known as GRK) is another example of a G protein regulator (Pitcher et al., 1998; Penn et al., 2000). GPCR kinase phosphorylates ligand-stimulated GPCR and attenuates its signaling (Moore et al., 2007). Consequently, GPCR kinase provides a negative-feedback loop for G protein signaling. For instance, mouse Grk1 knockout retinal cells produce stronger and longer outputs upon receiving a flash of light (Chen et al., 1999), suggesting that GRK1 regulates the sensitivity and duration of signaling. Recently, it has been shown that a Drosophila GPCR kinase, Gprk2, plays diverse roles in various biological processes, such as egg morphogenesis, olfactory response and tissue patterning (Schneider and Spradling, 1997; Molnar et al., 2007; Tanoue et al., 2008; Chen et al., 2010; Cheng et al., 2012). However, the role of Gprk2 in embryonic development has remained unclear.

Here we show that Drosophila Gprk2 is essential for the process of gastrulation. In Gprk2 mutant embryos, the apical constriction was abnormally expanded to the whole mesoderm and the involution movement was abolished. We propose that two types of cell movement - apical constriction in VM cells and involution in LM cells - are directed by the adjustment of G protein signaling.

Drosophila melanogaster strains used in this study were gprk2⁶⁹³⁶ (Schneider and Spradling, 1997), cta^RC10 (Parks and Wieschaus, 1991), fog^4a6 (Costa et al., 1994), P{Moe::GFP} (Kiehart et al., 2000), P{Sph::GFP} (Royou et al., 2002) and P{Matα4-Gal4-VP16} (kindly provided by D. St Johnston, University of Cambridge, UK). w¹¹¹⁸ was used as the wild-type control strain. Flies were reared with a standard cornmeal medium at 25°C.

To make the UAS-Gprk2 transgenic strain, part of the coding region of Gprk2 was amplified by PCR from an EST clone (RE34982) and then subcloned into the pUASt vector. The UAS-Gprk2 K338R mutant was made by site-directed mutagenesis using a DpnI mutagenesis kit (Stratagene). Sequences of coding regions were confirmed for both constructs, and plasmids were injected together with helper plasmid (Δ2-3) into w¹¹¹⁸ embryos. Transgenic strains were established, and expression of transgenes was confirmed by in situ hybridization. We also made the UAS-Cta and the UAS-Cta Q303L transgenic strains in a similar way. An EST clone (LD04530) was used as starting material. Antibody staining confirmed the expression of the transgenes.

The antibodies used in this study were anti-Eve, anti-Dl, anti-Dlg, anti-Sxl, anti-DE-Cadherin (obtained from DSHB), anti-β-gal (mouse antibody from Promega; rat antibody kindly provided by T. Isshiki, National Institute of Genetics, Japan), anti-α-Tubulin (Sigma), anti-aPKC (Santa Cruz), anti-Zip (kindly provided by F. Matsuzaki, RIKEN, Japan), anti-Mir (Ikeshima-Kataoka et al., 1997), anti-Cta (Kanesaki et al., 2013) and anti-Twi (Roth et al., 1989).

To produce antibody against Fog, the full-length Fog ORF was amplified from an EST clone (SD02223) by PCR and subcloned into the pQE80 vector (Qiagen). The recombinant protein was expressed in E. coli and purified from lysate using Ni-NTA resin (Qiagen). The purified protein was injected into a rabbit and antiserum was prepared by MBL. The antiserum was affinity purified using the antigen. The antibody staining of embryos reproduced the pattern of fog RNA expression (see also Ratnaparkhi and Zinn, 2007).

Embryos maternally and zygotically mutant for Gprk2 were collected from mating between gprk2⁶⁹³⁶ homozygous females and males. For rescue experiments of the Gprk2 mutant, we set up a cross between P{Matα4-Gal4-VP16}/CyO; Gprk2/Gprk2 females and P{UAS-Gprk2 (wild-type or K338R)}; Gprk2/Gprk2 males, and collected embryos.

To make a double mutant for fog and Gprk2, embryos were collected from mating between fog/FM7 P{ftzlacZ}; Gprk2/Gprk2 females and +/Y; Gprk2/Gprk2 males, and were stained with β-gal antibody (to distinguish between the embryos with the fog and FM7 chromosome) and Sxl antibody (to distinguish between the embryos with + and the Y chromosome). The embryos negative for β-gal and Sxl were judged to be fog Gprk2 double mutants, and the embryos negative for β-gal and positive for Sxl were judged to be Gprk2 mutants carrying one copy of the fog gene. We also collected embryos from mating between fog/FM7 P{ftzlacZ} females and +/Y males for a control experiment.

To examine the effects of overexpression of Cta and Gprk2, we set up a cross between P{Matα4-Gal4-VP16} females and P{UAS-Cta} males or P{UAS-Gprk2}; P{UAS-Cta} males, and collected embryos. Since P{UAS-Cta Q303L} was homozygous lethal, we used P{UAS-Cta Q303L}/CyO males for experiments. Embryos carrying P{UAS-Cta Q303L} were judged by severe malformation, which was never observed in embryos collected from P{Matα4-Gal4-VP16} females and the wild-type males.

For scanning electron microscopy (SEM), embryos were fixed as previously described (Kanesaki et al., 2013). Ion-coated embryos were observed using a JEOL JSM-7500F scanning electron microscope.

For immunostaining analysis, embryos were fixed and stained as described (Kanesaki et al., 2013). To acquire images, we mainly used confocal laser scanning microscopy (Carl Zeiss LSM5 Live or Olympus FV10). Alternatively, we used spinning disk microscopy (Olympus DSU system) for the images of Fig. 7 and supplementary material Figs S3 and S5. Stacked images along the z-axis were merged by maximum intensity projection using the LSM Image Browser (Carl Zeiss) or MetaMorph (Molecular Devices). Images were processed using ImageJ (NIH) and Photoshop (Adobe).

For in situ hybridization analysis, embryos were fixed twice and hybridized with Dig-labeled RNA probes (Roche) according to the standard protocol. Alkaline phosphatase was used to detect Dig probes, and embryos were stained with NBT/BCIP solution (Roche). Stained embryos were observed using a differential interference contrast (DIC) microscope (Olympus BX63).

Analysis of live embryos was performed as described (Kato et al., 2004). Briefly, dechorionated embryos were attached to a cover glass coated with rubber glue. A plastic slide with a 5-mm diameter hole was placed on the cover glass, and the hole was filled with silicon oil (Shin-Etsu Chemical). After incubation of embryos in a humidified chamber for an appropriate time, embryos were observed under a Zeiss LSM5 Live microscope at room temperature (controlled at 24±3°C). Images were processed using LSM Image Browser, ImageJ and Photoshop.

To quantify cell movements in the wild-type embryos, time zero was taken as the time when gastrulation started (some change in cell shape was visible). To compare Gprk2 mutant with wild-type embryos, we observed the cellularization process and determined time zero as the time when furrow canals of cells reached 12 μm in depth. This time zero corresponds to ∼5 minutes before gastrulation starts.

Quantitative analysis of cell movement was performed using the manual tracking plug-in of ImageJ. A clearly visible lateral vertex of a cell was chosen, and the trajectory of the cell vertex was tracked. The cell vertex frequently became unclear at the late stage of invagination. In such cases, we deduced the cell vertex by judging from the sequential timecourse images. The final destination of each cell before it disappeared from an image was defined as the end position, and distances of the trajectory to the end position were quantified. Data were processed using Excel (Microsoft) and R (http://www.r-project.org/).

A kymograph was constructed from time interval stacked images. A slit along the dorsoventral axis was chosen and the slits of images were aligned along the timecourse using the Reslice function of ImageJ.

To characterize the cell movements during mesoderm invagination, we first observed the cell morphology of embryos by SEM (Fig. 1A). As previously described (Costa et al., 1993), VM cells flatten and constrict their apical cell surface to make the ventral furrow (Fig. 1A, yellow), but LM cells do not undergo such apical constriction. We then observed that LM cells extended thin protrusions (several microns long) toward the mid-ventral region (Fig. 1A, red). The stretching of LM cells toward the mid-ventral region has been reported previously (Turner and Mahowald, 1977; Costa et al., 1993; Fox and Peifer, 2007), but the morphology and behavior of the protrusions have not been characterized in detail. The protrusions were not restricted to LM cells, as cells in more lateral regions, most likely ectodermal cells, also extended protrusions (Fig. 1A, blue), although these protrusions were much shorter than those of LM cells.

We next performed live imaging of the cell movements in embryos expressing Moe::GFP, a fluorescent marker that binds to actin filaments (Kiehart et al., 2000) (supplementary material Movie 1). From the sequence of images, we defined VM cells (apical constricting ventral cells) and LM cells (involuting lateral cells), and analyzed their morphologies and movements. As the apical constriction proceeded in VM cells, we observed thin protrusions extending from the LM cells, as also seen by SEM (Fig. 1B). The protrusions shook and moved dynamically back and forth (supplementary material Movies 1, 2; compare left and right images focused on the apical surface and at 9 μm depth of the same embryo). Since the protrusions were observed only at the plane of the most apical surface, we termed this structure the ‘apical protrusion’. We also found that the apical protrusions could be visualized by immunostaining for α-Tubulin (supplementary material Fig. S1), suggesting that cytoskeletal filaments (microfilaments and microtubules) actively support the protrusions on the apical cell membrane.

We also quantified the cell movements using Moe::GFP embryos (Fig. 1C-E; supplementary material Movie 3). While the apical constriction proceeded, LM cells moved toward the mid-ventral position with a mean speed of 2.9 μm/minute (Fig. 1F,G, phase 1), which roughly corresponds to the speed of VM cells (2.5 μm/minute). After the apical constriction was completed, the movement of LM cells accelerated to 12.7 μm/minute (Fig. 1F,G, phase 2). These observations demonstrate that there are two phases of LM movements (Fig. 1G, inset). The apical protrusions became apparent just before LM cells started the accelerated movement, suggesting the possibility that the protrusions contribute to the involution movement of LM cells (see Discussion). Thus, during mesoderm invagination, two cell groups (VM and LM cells) display different morphologies of the apical cell surface (flat surface and apical protrusion, respectively) and undergo distinct cell movements (apical constriction and involution, respectively). It was hitherto unknown how these distinct cell movements are spatially organized in presumptive mesodermal cells.

To elucidate the molecular mechanism underlying these cell movements, we searched for new factors involved in the G protein signaling pathway during gastrulation, and found that a Drosophila GPCR kinase, Gprk2, acts in the process. gprk2⁶⁹³⁶ is a female-sterile mutant carrying a P-element insertion that abolishes the maternal expression of Gprk2 in developing eggs (Schneider and Spradling, 1997) (supplementary material Fig. S2E). We collected embryos maternally and zygotically mutant for gprk2⁶⁹³⁶ (referred to simply as Gprk2 mutant hereafter) and observed them by SEM (Fig. 2). Although the Gprk2 mutant embryos displayed a rounded overall shape, as reported previously (Schneider and Spradling, 1997), cellularization proceeded normally in the mutant. We then observed abnormal gastrulation in Gprk2 mutant embryos (Fig. 2E-G): the apical constricting area represented by a flattened surface was abnormally expanded (Fig. 2H, yellow bars). Although the cells lateral to the furrow extended apical protrusions (Fig. 2H, arrow), these cells were thought to be ectodermal cells, as described below. At a later stage (germ band extension), about two-thirds of the mutant embryos (11 of 15) exposed the mesoderm to the outside (Fig. 2G), although other embryos did finally close the furrow. The Gprk2 mutant also displayed abnormal positioning of germ cells (Fig. 2F, arrow). This phenotype might be due to the expansion of the posterior midgut invagination together with the ventral furrow.

Staining for Dorsal (ventral region), Even-skipped (segmental stripes) and Twi (mesoderm) proteins showed that the Gprk2 mutant establishes axial patterning and allocates mesodermal cells normally (supplementary material Fig. S2). Staining for DE-Cadherin (Shotgun - FlyBase) and aPKC proteins showed that the Gprk2 mutant has normal adherens junctions and apicobasal cell polarity (supplementary material Fig. S2). These results indicate that the gastrulation phenotypes in the Gprk2 mutant are due to impaired cell movements.

We established transgenic fly lines carrying UAS-Gprk2 and observed that the exogenous Gprk2 expression rescued the expansion of the ventral furrow in the Gprk2 mutant embryos, confirming that Gprk2 is required for mesoderm invagination (supplementary material Fig. S3). In this condition, the apical constriction was occasionally blocked in some regions (supplementary material Fig. S3C, arrow), suggesting that excess Gprk2 activity might inhibit the apical constriction.

To examine the role of the kinase activity of Gprk2 in vivo, we made a Gprk2 construct carrying a point mutation (K338R), which is analogous to the kinase dead mutations of other GPCR kinases (Lee et al., 2004). In contrast to the wild-type Gprk2 protein, the K338R mutant protein did not rescue the Gprk2 mutant phenotypes (supplementary material Fig. S3D). Thus, the kinase activity of Gprk2 is essential for its function during gastrulation.

In Gprk2 mutant embryos, the ventral furrow was abnormally expanded (Fig. 2H). There could be two possible explanations for this: more cells might undergo apical constriction and/or the apical constriction might be incomplete. To examine the first possibility, embryos were stained for Twi and Zipper (Myosin heavy chain) and transverse sections of the embryos examined (Fig. 3). In the wild-type embryos, ∼12 cells in the middle of the Twi-positive (VM) cells constrict their surfaces to form the ventral furrow, and two to three cells on the sides of the Twi-positive region (LM cells) do not descend into the furrow (Fig. 3A,C). In VM cells, Myosin was translocated from the basal (inside) to the apical (outside) surface (Fig. 3G, arrowhead). In Gprk2 mutant embryos, however, the ventral furrow was expanded to include all of the Twi-positive cells (Fig. 3D) and the apical Myosin was also expanded to all cells in the furrow (generally 16 to 18 cells; Fig. 3H, arrowhead), indicating that the apical constriction is expanded to the whole mesoderm in the mutant (Fig. 3B). This notion was also supported by two additional features of apically constricting cells: expansion of the basal cell surface and the basal displacement of nuclei (Costa et al., 1993; Gelbart et al., 2012). At later stages, the ventral furrow frequently remained open as a U-shaped furrow in the mutant (Fig. 3F). Thus, all of the mesodermal cells undergo apical constriction in the Gprk2 mutant, suggesting that Gprk2 normally inhibits the apical constriction in LM cells, such that apical constriction is restricted to VM cells.

To examine whether the apical constriction is incomplete in the Gprk2 mutant, we performed live imaging of the embryos using Moe::GFP (Fig. 4; compare supplementary material Movie 5 with Movie 4 of the control embryo). We confirmed that all mesodermal cells (an ∼18-cell width) start to constrict their apical cell surfaces mostly simultaneously in the mutant (Fig. 4C′). Ectodermal cells on the sides of the mesoderm extended apical protrusions but never underwent involution movement. We made kymographs from time interval images to characterize the trajectory of cell movements, and this analysis also showed two phases of movement of the wild-type LM cells (Fig. 4B, red arrows; compare with Fig. 1F). In the Gprk2 mutant, the width of the mesodermal cell region was approximately halved within 3 minutes (roughly estimated speed of cell movement: 6 μm/minute), which was almost twice as fast as that of the wild type (see Fig. 4B,D, yellow arrows; slopes of cell trajectory represent the speed of cell movements). Thereafter, the apical constriction ceased prematurely and never resumed in the mutant (Fig. 4C′,D, yellow arrowhead), suggesting that Gprk2 is required for completion of apical constriction.

We examined the dynamics of Myosin, which is a primary force generator for apical constriction, using Sqh::GFP [Myosin light chain fused to GFP (Royou et al., 2002)]. In this experiment, we started the observations from the cellularization process and took time zero as that when cellularization was completed (∼5 minutes before gastrulation; see Materials and methods) in order to examine the timing of changes in Myosin localization. It has been reported that Myosin is initially localized as particles and later forms a supracellular meshwork to generate tension acting throughout the tissue (Martin et al., 2009; Martin et al., 2010) (Fig. 5A; supplementary material Movie 6). In the Gprk2 mutant, the Myosin accumulation started earlier than in the control (at 2 minutes in the Gprk2 mutant versus 5 minutes in the control; Fig. 5B; supplementary material Movie 7), and Myosin continued to accumulate on the apical surface of the mutant cells (Fig. 5B′, inset). This aberrant accumulation of Myosin might cause the premature termination of apical constriction in the mutant. Thus, Gprk2 is required for the localization and kinetics of Myosin accumulation in apically constricting cells.

The expansion of the apical constricting area and the abnormal accumulation of Myosin in the Gprk2 mutant are similar to the phenotypes of Fog overexpression (Morize et al., 1998; Dawes-Hoang et al., 2005). However, there is a significant difference between the two: apical Myosin accumulates only in mesodermal cells in the Gprk2 mutant (Fig. 3H) but in all cells in the embryo with Fog overexpression (Dawes-Hoang et al., 2005). These observations suggest that Fog signaling might be activated in mesodermal cells of the Gprk2 mutant.

To examine the relationship between Fog and Gprk2, we made a double mutant for fog and Gprk2 (Fig. 6; see Materials and methods). Embryos were stained with β-gal and Sxl antibodies (chromosome markers) to judge the genotype of the embryos. The fog Gprk2 double-mutant embryos showed only a few apically constricting cells at random positions (Fig. 6F), which is essentially the same phenotype as that of the fog single mutant (Fig. 6C). This indicates that the Gprk2 phenotypes are dependent on the fog gene, and is consistent with our idea that Gprk2 acts on Fog signaling. In addition, we examined Gprk2 mutant embryos carrying one copy of the fog gene. In these embryos (+/fog; Gprk2/Gprk2 genotype), apical constriction was substantially disorganized and asynchronous among mesodermal cells, compared with that in the Gprk2 single mutant (Fig. 6E compared with 6D). By contrast, the fog heterozygous embryos with the Gprk2-plus background (+/fog genotype) showed mostly normal invagination, as in the wild-type embryos (Fig. 6B compared with 6A). These results indicate that the Gprk2 mutant is sensitive to the dose of the fog gene and suggest that Gprk2 adjusts Fog signaling to an appropriate level from either one or two copies of the fog gene.

We next examined the expression of Fog in the Gprk2 mutant. In wild-type embryos, fog RNA is expressed at the ventral region, mostly in VM cells (Costa et al., 1994). However, we observed that not all VM cells express fog RNA; rather, fog-positive and fog-negative cells were intermingled in the VM area, and the same was true in the LM area (supplementary material Fig. S4A; the fog-expressing area was ∼18 cells wide). This expression pattern of fog RNA was not altered in the Gprk2 mutant (supplementary material Fig. S4C). We then raised an antibody to examine the localization of Fog protein. At the subcellular level, Fog was detected mostly in puncta in the cytoplasm (supplementary material Fig. S4B), as previously reported (Dawes-Hoang et al., 2005). At the tissue level, Fog protein was detected in a mosaic pattern in mesodermal cells, consistent with its RNA expression. In the Gprk2 mutant, Fog was expressed in a similar pattern to that in the wild-type embryo (supplementary material Fig. S4D), although, at the late stage of gastrulation, somewhat more Fog staining was seen than in the wild type. From these data, we suggest that Gprk2 primarily regulates Fog signaling rather than the expression of Fog protein.

To obtain further evidence for the function of Gprk2 in Fog signaling, we examined the functional interactions between Gprk2 and Cta, a downstream component of Fog signaling. We made transgenic fly lines carrying UAS-Cta and observed the effects of Cta expression driven by Matα4-Gal4-VP16 (hereafter termed as Maternal-Gal4). Overexpression of Cta compromised the gastrulation process, causing, for example, partial disruption of the ventral furrow (Fig. 7C), although the phenotypes were varied among the embryos (e.g. delay of furrow closure and ectopic folding of the lateral ectodermal cells). In such embryos, Myosin had accumulated abnormally on the apical surface of the lateral ectodermal cells (Fig. 7D), similar to in Gprk2 mutant mesodermal cells (Fig. 5B). This suggests that Cta overexpression induces ectopic signaling in ectodermal cells, as previously reported (Morize et al., 1998; Rogers et al., 2004). We then examined simultaneous overexpression of Cta and Gprk2. The effects of Cta overexpression were largely suppressed by simultaneous expression of Gprk2 (Fig. 7G,H). Overexpression of Gprk2 alone did not severely affect furrow closure or overall Myosin localization (Fig. 7E,F), although it partially inhibited apical constriction, as seen in the rescue experiment (supplementary material Fig. S3). Furthermore, the lethality of Maternal-Gal4/UAS-Cta individuals was suppressed by simultaneous expression of Gprk2 [from 73% (n=131) to 27% (n=135), as judged by the ratios of adults with or without the transgene]. These results suggest that Gprk2 attenuates Cta signaling.

We next performed a similar experiment using a point mutant of Cta (Q303L), which is analogous to the constitutively active mutant in which Cta is fixed to the GTP-bound form of Gα protein (Grishina and Berlot, 1998). Overexpression of Cta Q303L resulted in severe malformation of the embryo (supplementary material Fig. S5A). Myosin accumulated strongly in the lateral cell cortex, and the epithelial architecture seemed to have collapsed (supplementary material Fig. S5B). These phenotypes could be explained by the perturbation of signaling due to constitutive Cta activity. Overexpression of Gprk2 did not suppress the effects of Cta Q303L (supplementary material Fig. S5C,D), suggesting that Gprk2 regulates Fog-Cta signaling before the activation of Cta effectors. Together, these data suggest that repression of Cta activity might be one mode by which Gprk2 regulates Fog signaling.

We characterized the cell movements in Drosophila mesoderm invagination and examined the roles of a GPCR kinase, Gprk2, in this process. We found that two cell groups (VM and LM cells) among mesodermal cells undergo distinct cell movements, and Gprk2-adjusted Fog signaling defines the areas of these two cell groups. Thus, these two different cell movements in mesoderm invagination are not predetermined, but rather are organized by the adjustment of Fog signaling.

In the Gprk2 mutant embryos, cell movements triggered by Fog signaling were compromised. We found that fog is genetically epistatic to Gprk2, indicating that Gprk2 functions by acting on Fog signaling. LM cells undergo apical constriction in the Gprk2 mutant, suggesting that Gprk2 normally inhibits Fog signaling in LM cells. We also observed premature termination of apical constriction and abnormal accumulation of Myosin in the Gprk2 mutant, suggesting that Gprk2 adjusts Fog signaling to an appropriate level in VM cells. Thus, Gprk2 regulates Fog signaling in a cell group-dependent manner. But what are the underlying molecular mechanisms?

It is known that GPCR kinase phosphorylates the C-terminal region of GPCR, and regulates GPCR signaling by multiple mechanisms (Penn et al., 2000; Moore et al., 2007). The phosphorylated GPCR dissociates from the G protein and is internalized from the plasma membrane. This produces a negative-feedback loop for GPCR signaling. Theoretically, the negative-feedback loop stabilizes the signaling and generates biphasic output from fluctuating inputs: OFF for low inputs and ON for high inputs (Brandman and Meyer, 2008). We speculate that Gprk2 might phosphorylate a GPCR and might generate biphasic output for Fog signaling in a spatial manner: OFF in LM cells and ON in VM cells. The Fog receptor is expected to be a GPCR, since a G protein (Cta) functions downstream of Fog (Parks and Wieschaus, 1991). Identification of the Fog receptor would help to clarify the molecular functions of Gprk2.

The kinase activity of Gprk2 is essential for gastrulation. Although we do not yet know what substrates are phosphorylated by Gprk2 in this process, one might be Gprk2 itself because we observed that Gprk2 protein was phosphorylated in S2 cultured cells and that the phosphorylation was abolished in the K338R mutant of Gprk2 (N.F., unpublished data). Autophosphorylation of other GPCR kinases has been demonstrated previously and is thought to stimulate their binding to GPCR (Pitcher et al., 1998). Autophosphorylation of Gprk2 might play a similar role.

In addition to its kinase activity, GPCR kinase has an RGS domain, which exhibits GAP (GTPase activating protein) activity and functions in recycling of the Gα protein (Pitcher et al., 1998; Penn et al., 2000). Therefore, whether Gprk2 exhibits GAP activity for Cta is an intriguing issue. Indeed, this possibility was supported by our genetic data showing that Gprk2 suppresses the effect of Cta overexpression, but not that of Cta Q303L, the GTP-bound form of Cta protein. Cta Q303L might not be subject to the inhibitory effect (GAP activity) of Gprk2, although we have not ruled out the alternative explanation that the inhibition of Cta Q303L might require more Gprk2 protein than does the inhibition of wild-type Cta. Considering that GPCR kinase regulates GPCR signaling by multiple mechanisms, we suggest that the repression of Cta activity might be one of several means by which Gprk2 regulates Fog signaling.

Fog signaling stimulates the apical localization of Myosin, which generates a force to constrict the apical cell surface. Martin et al. (Martin et al., 2009) observed that, in the wild-type embryo, Myosin protein appears and disappears at the apical surface in a dynamic pattern that they described as ‘pulsed coalescence’. In the Gprk2 mutant, Myosin continued to accumulate on the entire apical surface of mesodermal cells. We also observed similar phenomena in Cta-overexpressing ectodermal cells, and this phenotype was suppressed by simultaneous expression of Gprk2. We suggest that Gprk2 normally attenuates Fog-Cta signaling to an appropriate level, and such refinement might contribute to controlling the dynamics of Myosin protein.

Previous studies showed that Gprk2 acts in Hedgehog (Hh) signaling for imaginal disc patterning (Molnar et al., 2007; Chen et al., 2010; Cheng et al., 2012). In this process, Gprk2 phosphorylates a GPCR, Smoothened, and potentiates Hh signaling. Thus, Gprk2 plays roles in multiple signaling pathways in various contexts during development.

We characterized the movements of LM cells and found that they extended apical protrusions. Some examples have been documented of the extension of protrusions by epithelial cells, such as dorsal ectodermal cells of embryos and wing disc cells of larvae in Drosophila (Jacinto et al., 2000; Bosch et al., 2005). However, the mechanisms that induce the protrusion and the roles of protrusion in directional cell movement are not understood. Since we observed that apical protrusions in LM cells always pointed toward the ventral furrow and that cells close to the furrow extended longer protrusions than cells distant from it, we speculate that the apical protrusion might be induced by the apically constricting neighbors. Indeed, in cta mutant embryos the apical protrusions did not always point toward mid-ventral, but rather frequently pointed toward the slight depressions that were formed at random positions by uncoordinated apical constriction [data not shown; see also figure 4E in Kanesaki et al. (Kanesaki et al., 2013)]. One possibility is that mechanical or chemical signals that emanate from apically constricting cells might induce apical protrusions in surrounding cells.

Apical protrusions became apparent when LM cells started to accelerate toward the ventral furrow. From this observation, we suppose that the directional protrusion might contribute to the movement of LM cells. Given that the apical protrusion of LM cells is analogous to the pseudopod of cultured cells (Sánchez-Madrid and del Pozo, 1999), the apical protrusion might act as a scaffold for pulling the cell body into the furrow. The fact that the apical protrusion was also observed in some ectodermal cells, which never undergo involution movement, suggests that the apical protrusion is not sufficient to induce involution movement and that other mechanisms might regulate the cell movement in parallel.

Drosophila mesoderm invagination is driven by sequential movements of different cells (Costa et al., 1993). The apical constriction of VM cells is one of the essential movements in this process. We expect that the involution movement of LM cells might be another of the cell movements driving mesoderm invagination. The movements of different cells would probably influence each other in a complex manner. Our observations of LM movements might be explained by such a coordination of cell movements. For example, the apical constriction of VM cells might stretch LM cells and thereby prevent LM cell apical constriction, as previously suggested (Leptin and Roth, 1994). VM cells might then continue to move inward and pull LM cells toward the ventral furrow. In addition, ectodermal cells might generate a force to push mesodermal cells inward. These possibilities are not mutually exclusive. Further analyses are required to clarify the role of each cell movement and the effect of coordinated movements in mesoderm invagination.

In the Gprk2 mutant, LM cells underwent apical constriction instead of involution movement. Given the inhibition of Fog signaling in the wild-type LM cells, involution movement might be a default state of mesodermal cells without Fog signaling. As noted above, in the fog and cta mutant, apical constriction occurs in some VM cells, and involution-like movement operates in an uncoordinated manner [data not shown; see also figure 4E in Kanesaki et al. (Kanesaki et al., 2013)]. These uncoordinated cell movements finally result in disorganized, but nearly complete, mesoderm invagination (Sweeton et al., 1991). Thus, apical constriction and involution movements seem to be alternative choices for mesodermal cells, as previously suggested (Leptin and Roth, 1994), and robust mesoderm invagination might progress via either type of cell movement. In normal Drosophila embryos, cell movements are spatially and temporally organized, and such organization might ensure the correct shape of gastrulae.

Cell movements in gastrulation show diversity among insects (Roth, 2004). For example, mosquito embryos undergo only apical constriction and no apparent involution process (Goltsev et al., 2007). Locust embryos undergo neither apical constriction nor involution, but instead utilize the delamination of individual mesodermal cells. Compared with gastrulation in these insects, Drosophila gastrulation is a more complex process and is completed within a shorter time (15 minutes compared with hours). The highly organized cell movements in Drosophila might enable this rapid completion of gastrulation. The molecular mechanisms underlying the evolution of insect gastrulation are an intriguing issue for future studies.

We thank Dr T. Kanesaki for helping with some experiments; Drs A. Spradling, E. Wieschaus, S. Roth, R. Karess, F. Matsuzaki, T. Isshiki, D. St Johnston and S. Hayashi for providing valuable materials; colleagues at NIG for suggestions and discussions; Drs S. Hayashi and E. Nakajima for critical reading of the manuscript; and the Kyoto Drosophila Stock Center, Bloomington Drosophila Stock Center, The Developmental Studies Hybridoma Bank, and The Drosophila Genomics Resource Center for providing fly strains, antibodies and DNA constructs.