Repression of Wasp by JAK/STAT signalling inhibits medial actomyosin network assembly and apical cell constriction in intercalating epithelial cells

During gastrulation, embryonic tissues are shaped following stereotyped patterns. These complex morphogenetic processes require adhesive forces that maintain tissue cohesion, tension generation by myosin motors that drive the remodelling of cell contacts, and cell shape changes. Two main classes of cell morphogenetic processes have been reported. Apical cell constriction is required for local tissue bending and facilitates tissue invagination, such as in the Drosophila mesoderm or in vertebrate neural tube closure (Lecuit and Lenne, 2007; Leptin, 2005). In other tissues, cells exchange neighbours during intercalation, and drive convergence and extension movements that elongate the anteroposterior axis (reviewed by Keller, 2006).

Drosophila gastrulation vividly illustrates how tensile forces drive tissue remodelling. During mesoderm invagination, Myosin-II is apically recruited in ventral cells in response to the ligand Fog (Dawes-Hoang et al., 2005). Fog activates RhoGEF2 (Barrett et al., 1997; Hacker and Perrimon, 1998) and the Rho1/Rok1 pathway, and ultimately leads to Myo-II regulatory light chain (MRLC) phosphorylation and activation (reviewed by Tan et al., 1992); MRLC is called Spaghetti Squash (Sqh) in Drosophila (Karess et al., 1991). Although it is traditionally thought that apical constriction is driven by circumferential Myo-II at cell junctions, as for instance in vertebrate neurectodermal cells in response to the actin-binding protein Shroom (Hildebrand, 2005), recent studies have shown that constriction in the fly mesoderm is driven by apical medial actomyosin contractility (Martin et al., 2009). Pulses of contractility drive rapid, stepwise cell constriction events. In mesoderm cells, Myo-II is absent from the cortex and accumulates only in a central ‘web’ at the level of the adherens junctions (AJs).

After mesoderm invagination, ventrolateral (VL) ectodermal cells extend along the anteroposterior (AP) axis in a process called germband extension (GBE) (Bertet et al., 2004; Blankenship et al., 2006; Irvine and Wieschaus, 1994; Zallen and Wieschaus, 2004). GBE is driven by cell intercalation. During this process, cells exchange neighbours by the polarized remodelling of cell contacts. Local Myo-II enrichment leads to increased bond elastic tension and drives changes in cell contacts (Bertet et al., 2004; Rauzi et al., 2008).

This leads to a model in which different regions of the Drosophila embryonic epithelium undergo different morphogenetic processes owing to different subcellular regulation of actomyosin network contractility. However, the underlying mechanisms are still poorly understood. First, it is not known what regulates the planar polarized recruitment of Myo-II at the contacts between AP neighbours. More fundamentally, it is not even known what restricts Myo-II at the cortex instead of in the medial apical region in the VL ectoderm.

Myo-II subcelllar localization requires activation by the Rho GTPase and its target Rok, which phosphorylates the MRLC at conserved serine residues (Bertet et al., 2004; Jordan and Karess, 1997; Winter et al., 2001). Rok is the only kinase known to regulate MRLC in flies, although other kinases are likely candidates in other organisms. In conditions that alter MRLC phosphorylation, apical cell constriction is affected (Corrigall et al., 2007; Escudero et al., 2007; Lee and Treisman, 2004). In some cases, active RhoGTP accumulation occurs in regions matching the sites of Myo-II enrichment (Bement et al., 2005; Simoes et al., 2006). It has long been assumed that subcellular RhoGTP activation by RhoGEFs or inactivation by RhoGAPs provides the necessary and sufficient molecular signals that determine subcellular localization, especially when Myo-II cortical localization was shown to be independent of actin, such as during cytokinesis (Dean et al., 2005; Kamijo et al., 2006; Zang and Spudich, 1998).

Myo-II binds actin filaments and motor activity drives the movement and dynamics of actin networks. In vivo, actin networks are complex structures. Two types of proteins promote actin filament polymerization. Wasp and Scar, via the regulation of the Arp2/3 complex, promote the polymerization of branched actin filaments (Amann and Pollard, 2001; Pantaloni et al., 2000). The formin Dia promotes the polymerization of non-branched actin filaments (Romero et al., 2004). Non-branched networks facilitate contractility because filaments can move past one another. Indeed, Dia increases Myo-II accumulation at AJs (Homem et al., 2008).

JAK/STAT signalling controls many developmental processes, in particular morphogenetic events involving cell movements (Hou et al., 2002), such as border cell migration in Drosophila (Silver et al., 2005; Beccari et al., 2002), prestalk cell movements in Dictyostelium (Kawata et al., 1997) or convergence extension movements in the zebrafish (Yamashita et al., 2002). Here, we report that the JAK/STAT signalling pathway is required in early Drosophila embryos for germband elongation. We show that JAK/STAT signalling represses actomyosin accumulation in a medial apical ‘web’ in the VL ectoderm, thus blocking apical constriction in this region of the embryo. Although MRLC (Sqh) phosphorylation by the RhoGEF2/Rho/Rok pathway is necessary for the medial accumulation of Myo-II in JAK/STAT mutants, constitutive activation of Sqh and overexpression of the Myo-II heavy chain (MHC; also known as Zip) do not recapitulate the JAK/STAT mutant phenotype. This suggests that JAK/STAT uses additional mechanisms to control Myo-II recruitment in germband cells. We report that JAK/STAT represses Wasp, and that this is needed to repress the medial web accumulation of Myo-II and cell constriction in the intercalating ectoderm.

Wild-type stocks were Oregon-R. Deletion stocks Df(970) and Df(954) were from the Bloomington Stock Center. Mutant alleles used were: upd^YM55, hop^C111 and mrl⁰⁶³⁴⁶ (gift of N. Perrimon, Harvard Medical School, Boston), eve^1.27 (gift of J. P. Vincent, NIMR, London), runt³ (Bloomington), sqh^AX3 and sqhE20E21 (gift of R. Karess, Institut Jacques Monod, Paris), rho1^K2107B, dia⁵ (Bloomington) and wasp³ (gift of E. Schejter, Weitzmann Institute, Rehovot).

Time-lapse imaging was performed in E-cadGFP (gift of H. Oda, JT Biohistory, Osaka), sqhAX3; sqhGFP (gift of R. Karess), upd/FM7; E-cadGFP and upd/FM7;sqhGFP stocks, and in embryos laid by mothers expressing E-cadGFP and in which hop germline clones were induced.

Overexpression experiments were done at 18°C, by crossing females expressing mattubGal4VP16 (67 gal4) with the following UAS lines: UAS-upd, UAS-stat and UAS-hop^Tuml (gift of M. Zeidler, University of Sheffield, UK), UAS-sqhE20E21 (this work), UAS-zipGFP (gift of D. Kiehart, Duke University), UAS-myr-wasp (gift of K. Klaembt, Institute für Neurobiologie, Muenster) and UAS-diaCA NC (gift of P. Rørth, IMBC, Singapore).

The sqhE20E21Flag coding sequence was amplified by PCR from the p-Casper-genomic-sqhE20E21Flag vector (gift of R. Karess) as a KpnI-KpnI fragment and cloned into the pUAST vector. The construct was sequenced before being transformed into flies.

Expression of the construct was checked by RT-PCR on cDNA from embryos laid by 67gal4 females crossed with UASsqhE20E21Flag males, and by western blot using an anti-FLAG antibody (data not shown).

Embryos were fixed by heat-methanol treatment for stainings with rabbit anti-Zip (1:1000; gift from C. Field, Harvard Medical School, Boston), and mouse anti-Nrt (1:50; Developmental Studies Hybridoma Bank, DSHB). Standard 3.7% formaldehyde (FA) fixation was performed for stainings with rat anti-DE-Cad (1:10; DSHB), mouse anti-Arm/β-cat (1:200; DSHB), rabbit anti-Eve (1:1000; M. Frasch, Friedrich-Alexander University of Erlangen-Nuremberg, Erlangen), guinea pig anti-Runt (1:500; D. Kosman, SUNY, Buffalo), mouse anti-Dia (1:200; S. Wasserman, University of California, San Diego) and rabbit anti-Wasp (1:100; E. Schejter). For Alexa 546-phalloidin (Molecular Probes) labelling, embryos were hand devitellinized.

Y27632 and Latrunculin-A were injected at the end of cellularization at a concentration of 600 μM and 1 mM, respectively. The dilution factor in the embryo is around 1:50, so we estimated the local concentrations to be 12 μM for Y27632 and 20 μM for Latrunculin-A.

Embryos were prepared as detailed by Cavey and Lecuit (Cavey and Lecuit, 2008). Quantifications and statistical tests are detailed below.

We used a home-made set up as described previously (Cavey et al., 2008). Ablation pulses were performed using a 40× water immersion lens NA 1.2, at 370 mW at the back aperture of the objective, for 2.5 milliseconds.

For each genotype, approximately 50 μl of embryos (0-5 hours egg laying collection) were re-suspended in 100 μl of embryo lysis buffer, snap frozen on dry ice, ground, centrifuged for 10 minutes at 25,000 g, and the supernatant then boiled for 10 minutes with 5 μl loading buffer. The following antibodies were used: rabbit anti-Sqh (1:100; Cell Signaling Technology), rabbit anti-Dia (1:5000), rabbit anti-Rho1 (1:20; Cell Signaling Technology) and mouse anti-α-Tub (1:5000; Sigma), which was used as a loading control. Signal detection was performed using an ECL UV imager (Fisher), and quantification of the signal intensity was carried out using the Bio1D software (Vilber Lourmat).

Increasing concentrations of wild-type embryo extracts were used to determine the slope of the ECL signal versus protein concentration. This allowed us to determine how much protein (Sqh, Dia, Rho1 or α-Tub) was present per lane for the different genotypes. Protein levels were then normalized to α-Tub levels.

Embryos were filmed in phase contrast on a Zeiss microscope with a 20× objective. Time-lapse covered the end of cellularization to the end of germband elongation every 2 minutes. GBE measurements were done at the end of the intercalation process, i.e. 40 minutes after posterior midgut invagination. Percentage of GBE was calculated as the ratio of elongation of the germ band to the maximal elongation in wild-type embryos (cephalic furrow position).

Embryos were dechorionated with bleach, glued on a coverslip and covered with Halocarbon oil. upd;sqhGFP embryos were filmed with a LSM510 Meta confocal microscope, using an Apochromat 100×/1.4NA objective. Intercalating cadGFP, upd; cadGFP or hopGLC;cadGFP embryos were filmed using a Perkin Elmer spinning disk confocal microscope with a Plan Neo 40×/1.2NA objective. Movies used for measurements were made by projections of z-stacks of 1 μm step acquired at 30-second intervals. For each embryo, the surface of 150 cells was measured every 10 minutes, using the Metamorph software.

Z-stacks of 0.5 μm step were acquired in fixed embryos stained with anti-Zip and anti-Nrt antibodies. Embryos of different genotypes were fixed en masse, stained with the same antibody solution and processed in parallel. Imaging used the same confocal settings. Bulk amount of Myo-II was measured in the germband at different confocal planes starting from the apical surface to the adherens junctions (step size 0.5 μm). Integrated intensities were calculated in 58×58 μm images using the Metamorph software. Means±s.e.m. of different embryos were calculated. Student's t-tests were performed to compare the means between wild-type and mutant embryos.

Z-stacks were acquired in fixed embryos stained with anti-Nrt antibody. For each embryo, the apical surface of 40 cells was measured using the Metamorph software. Means±s.e.m. of different embryos were calculated. Student's t-tests were performed to compare the means between wild-type and mutant phenotypes.

Z-stacks were acquired in fixed embryos stained with anti-Zip and anti-Nrt antibodies. Quantifications were performed in the z-plane corresponding to the apical side of AJs (500 nm below the apical-most surface). Regions (adherens junctions and apical web) were drawn using Nrt staining and intensity measurements were performed on Zip staining. The mean intensity at adherens junctions (A; cortical) was measured by using the linescan function of the Metamorph software. For each embryo, about 200 junctions were measured. The results were expressed in arbitrary units per pixel and converted to arbitrary units per μm² taking into account pixel size for each image. The mean Myo-II intensity in the apical web (B; medial) refers to the average Myo-II integrated intensity per cell normalized to cell surface (expressed in arbitrary unit/μm²). For each embryo, around 40 apical webs were measured. Finally, the cortical/medial Myo-II intensity ratio was obtained by dividing A by B. For each ratio, standard errors were calculated and Student's t-test was performed.

All progeny embryos from heterozygous upd or Df mothers and +/Y fathers were genotyped based on the presence or absence of a GFP marker inserted on an FM7 balancer chromosome. The embryos were laid by mothers Df or upd/FM7,KrGal4,UASGFP crossed to males FM7,KrGal4,UASGFP/Y. upd/upd or Df/Df embryos were distinguished from heterozygous embryos based on the absence of GFP scored about 2.5 hours after the onset of gastrulation, at a time when GFP can be easily scored. The presence/absence of GFP and hence genotype was directly associated with a GBE phenotype recorded in time lapse because each embryo had recorded coordinates using a motorized platform and the Mark&Find module from AxioVision software (Zeiss).

For upd/FM7,KrGal4,UASGFP, 25/100 embryos had GBE defects and all were upd/upd. For Df/ FM7,KrGal4,UASGFP, 29 embryos had GBE defects and all were Df/Df.

Using the same GFP balancer chromosomes on living embryos expressing an EcadGFP transgene to label cell junctions, we found that all homozygous upd mutant embryos displayed characteristic apical cell constriction defects. No single heterozygous upd embryo and no yw control embryos expressing the same E-cadGFP transgene displayed these defects. For this reason, we used apical cell constriction as a proxy for genotyping upd homozygous mutants when embryos were fixed.

We overexpressed uniformly different components of the JAK/STAT pathway: upd, stat and a constitutively active form of JAK, hopTuml. Although all the embryos uniformly expressed these transgenes, we found borderline defects with upd (13%, n=100) and no defects with others (0%, n=100). In other contexts, based on the literature, this causes strong gain-of-function defects.

Dechorionated embryos were lysed in TRIzol (Invitrogen) and RNAs were prepared according to the manufacturer's protocol and quantified using nanodrop. For each sample, 1 μg of mRNA was reverse-transcribed using oligo-dT-VN primers and ImProm-II as the reverse transcriptase (Promega) in triplicate. Real-time quantitative amplification of RNA (q-PCR) was performed using the Sybr Green qPCR Super Mix (Invitrogen) and the iQ™5 Real-Time PCR Detection System from BioRad according to the manufacturer's protocol. The relative expression of indicated transcripts was quantified with the iQ5 Optical System Software using the 2^{[−ΔΔC(T)]} method. According to this method, the C(T) values for the expression of each transcript in each sample were normalized to the C(T) values of the control mRNA (RP49) in the same sample. The values of untreated cell samples were then set to 100% and the percentage transcript expression was calculated. The results are representative of three independent RT experiments. Primers used in these experiments were:

In order to identify new zygotic regulators of Myo-II cortical recruitment during GBE, we conducted a genetic screen looking at embryos homozygous mutant for genomic deficiencies. Time-lapse recording of 100 embryos undergoing GBE and quantification of tissue elongation 40 minutes after the onset of GBE (Fig. 1A) identified an important region that is required zygotically for GBE (Fig. 1B, see also Materials and methods). Mutants for unpaired (upd; os — FlyBase), a gene present in this interval and coding for a Drosophila ligand of the JAK/STAT signalling pathway (Harrison et al., 1998), display similar defects (see Movies 1 and 2 in the supplementary material): 25% of the progeny of heterozygous upd^YM55 parents, a null upd allele, had these defects and all of them were homozygous mutant, based on in vivo genotyping (see Materials and methods). These defects were rescued following expression of a wild-type upd cDNA (see Fig. S1 in the supplementary material; Materials and methods). These defects were noticeably stronger in severity than were those in eve mutants, where GBE defects have been previously characterized (Fig. 1C) (Irvine and Wieschaus, 1994).

Upd binds the receptor Domeless (Dome) (Brown et al., 2001), and activates a single JAK kinase called Hopscotch (Hop) (Binari and Perrimon, 1994) and a single STAT transcription factor called Marelle (Mrl/Stat92E) (Hou et al., 1996; Yan et al., 1996) (Fig. 1C). All components of the pathway except for the ligand are maternally provided. Embryos coming from jak/hop^C111 and stat/mrl⁰⁶³⁴⁶ germline clones (both strong loss-of-function alleles) had GBE defects that were similar to those of upd mutants, and were rescued by the zygotic paternal wild-type allele (Fig. 1D).

All components of the JAK/STAT pathway are expressed uniformly prior to GBE, including the ligand Upd. At the onset of GBE, upd expression resolves into stripes (Harrison et al., 1998). Uniform activation of the JAK/STAT pathway did not affect GBE and rescued the upd mutant phenotypes (see Fig. S1A,B in the supplementary material), suggesting that patterned expression of upd might not be essential for GBE (see Materials and methods).

The JAK/STAT signalling pathway regulates the fifth stripe of the pair rule genes eve and runt (Binari and Perrimon, 1994; Harrison et al., 1998; Hou et al., 1996), both of which are also required for GBE (Irvine and Wieschaus, 1994). This suggested that JAK/STAT signalling could control GBE via the pair rule genes eve and runt. However, although the cuticle patterning defect was fully penetrant in jak/hop mutants (see Fig. S2A,A′ in the supplementary material), we failed to detect any obvious and penetrant defects in Eve and Runt expression during GBE (0 defects in 21 embryos laid by hop germline clone mothers; see Fig. S2B,C in the supplementary material), suggesting that the reported obvious loss of eve and runt stripe 5 might not be very penetrant (Binari and Perrimon, 1994; Harrison et al., 1998; Hou et al., 1996). This suggests that JAK/STAT signalling controls GBE independently of eve and runt. This is confirmed by the fact that cell shape defects observed in upd mutants (see below) are not observed in eve and runt mutants (data not shown).

We next addressed the cellular basis of GBE defects observed in JAK/STAT mutants. In JAK/STAT pathway mutants, although epithelial cells in the head and dorsal-most regions were normal, intercalating (ventrolateral) cells displayed a smaller apical surface than did controls at the same stage (Fig. 2A,B). Time-lapse imaging of upd mutant embryos expressing the fusion protein E-Cad::GFP showed that this was due to neither ectopic cell divisions (normally repressed at this stage), nor cell delamination (as is the case when junctions collapse), but was the result of an ectopic apical cell constriction (Fig. 2C,D; see also Movies 3 and 4 in the supplementary material). This defect was fully penetrant in upd hemizygotes (see Materials and methods) and in jak/hop mutants (data not shown). We quantified changes in the apical cell surface over time: in contrast to wild-type controls where the surface is nearly constant, in upd and jak/hop mutants the apical cell surface decreased gradually and became conspicuous when intercalation normally starts (Fig. 2E). At the end of GBE, the apical cross section area was reduced by 40% in upd mutants compared with in wild-type controls. In addition to apical cell constrictions, upd mutants exhibited cell intercalation defects, which were variable in strength (see Fig. S3 in the supplementary material). In the most extreme cases, cell intercalation was blocked (Fig. 2F and see Fig. S3 in the supplementary material). Thus, ectopic apical cell constriction and associated defects in cell intercalation provide a simple explanation for the failure to extend the germband.

These results point directly to potential defects in Myo-II localization in JAK/STAT pathway mutants. In control embryos, Myo-II is mainly localized at the apical cortex of intercalating cells (Fig. 3A-A″). In JAK/STAT pathway mutants, Myo-II is strongly enriched apically in a more medial position (Fig. 3B-B″,C-C″; see also Fig. S4A,B in the supplementary material) compared with in wild type (Fig. 3A-A″). Closer examination indicates a significant enrichment both at the AJs and within 1 μm apically of the AJs (Fig. 3E-E″, compare with Fig. 3D-D″; quantification in Fig. 3F, P<1×10⁻⁴). Besides the overall increase of Myo-II in the apical region of intercalating cells, a striking feature was a change in the subcellular localization of the actomyosin network. We measured the medial to cortical Myo-II ratio at the apical extent of AJs (500 nm below the apical-most region, see Materials and methods) in wild type and upd mutants. Whereas Myo-II is mostly cortical in wild type with only very low levels in the apical medial region (Fig. 3D-D″), in upd mutants (Fig. 3E-E″) Myo-II is strongly enriched in a medial meshwork, such that the ratio between cortical and medial Myo-II is significantly reduced compared with in wild type (P<6×10⁻⁵; Fig. 3G,H).

Using a MRLC/sqh::GFP transgene, we found that the ectopic apical recruitment of Myo-II is detected just after cellularisation, preceding conspicuous apical constriction (see Fig. S4C,D in the supplementary material). Ectopic MRLC/Sqh::GFP is not cytoplasmic but is present on very dynamic structures, most likely actin fibers underlying the apical surface (see Movie 5 in the supplementary material; compare with Movie 6). In JAK/STAT pathway mutants, Myo-II is thus apically recruited prematurely, and in excess in a medial apical ‘web’ (Fig. 3H). This situation is reminiscent of the mesoderm cells, although an important difference is that Myo-II is still present at the cortex in upd mutants, whereas it is only present in a medial apical web in the mesoderm (Martin et al., 2009) (Fig. 3I,J). This resemblance with mesodermal cells strongly suggests that, in JAK/STAT mutants, the ectopic medial apical Myo-II pool might be responsible for ectopic apical cell constriction in germband cells. In order to test this directly, we performed focal ablation of the medial Myo-II ‘web’ with nano-scissors, a pulsed near IR-laser beam focused on a diffraction-limited spot (200-300 nm) for 2-3 milliseconds (see Rauzi et al., 2008). This led to the fragmentation of the medial actomyosin meshwork and to its redistribution away from the ablation points towards the cortex (Fig. 4A, yellow arrows). Importantly, this was also accompanied by a progressive relaxation of the apical cell surface within 30 seconds (Fig. 4A,B). When constriction was weak at the onset of ablation, the cell surface relaxed more modestly with a change in the curvature of the boundaries (Fig. 4A,F; see also Movie 7 in the supplementary material). The effects were stronger when constriction was more prominent (38% surface relaxation, Fig. 4B,F; see Movie 8 in the supplementary material) and were commensurate with apical surface constriction in upd mutants (−40%). Laser pulses outside of the apical Myo-II medial clusters did not cause Myo-II redistribution, nor area relaxation (Fig. 4C,F; see Movie 9 in the supplementary material), indicating that relaxation requires the ablation of tensile elements in the medial region. Together, these observations indicate that ectopic medial Myo-II is responsible for the abnormal cell constriction observed in upd mutants.

To test whether efforts to elongate the germband might affect non-autonomous Myo-II accumulation in the germband, we knocked down Krüppel by RNAi in upd embryos. GBE was strongly reduced in Krüppel RNAi embryos (data not shown), and Myo-II polarized accumulation at the cortex was severely perturbed in all embryos (see Movie 10 in the supplementary material). RNAi of Krüppel did not affect the penetrance of medial Myo-II accumulation in upd mutants (30%, n=23) and the defects were indistinguishable from those of upd mutants alone (see Movie 11 in the supplementary material). This suggests that Myo-II localization defects are likely to be cell autonomous.

Myo-II is present overall at higher levels apically in upd mutants (40% increase compared with wild type). Western blot analysis confirms this finding (Sqh levels are 20% higher in hopGLC in which 60% of embryos show GBE defects, compared with in wild type, not shown). This raises the question of whether elevated levels of Myo-II are necessary and sufficient to produce the changes in actomyosin localization and cell constriction in JAK/STAT pathway mutants. To test whether Myo-II levels are necessary for the aberrant apical constriction, we performed genetic epistasis experiments between upd and a protein null allele of sqh (sqh^AX3). Western blotting indicated that embryos laid by heterozygous upd, sqh/+ parents had on average a 38±9% reduction in Sqh compared with embryos laid by upd/+ parents (Fig. 5A, inset). This reduction of Sqh levels was sufficient to rescue the GBE defects normally observed in upd mutants (Fig. 5A, 24% defects in the progeny of upd heteterozygous parents compared with 4.5% on average within the progeny of upd, sqh^AX3 heterozygous parents). Indeed, apical cell constriction was significantly rescued in upd, sqhZ/Z,M/+ embryos [i.e. embryos removing the zygotic (Z) contribution of sqh and half of its maternal (M) contribution] compared with in upd mutants (Fig. 5B, n=200 cells in 5 embryos, P<6×10⁻¹⁶). Consistent with this, together with the expected globally reduced amount of Myo-II (Fig. 5C-E), the ratio between cortical and medial Myo-II was restored and similar to in wild type (Fig. 5F).

Heterozygosity for rho1, required for the activation of MRLC (Sqh) via Rok, also rescues the GBE phenotype of upd mutants (see Fig. S5A-C in the supplementary material; 24% and 6% defects in the progeny of, respectively, upd and upd, rho1/+ heterozygous parents). Rho1 is on average reduced by 60±9% in embryos coming from upd/+ and rho1/+ parents compared with those from upd/+ parents (Fig. S5A, inset). Injection of low concentrations of the Rok inhibitor Y-27632 into hop mutants also partially reduced the penetrance of the GBE defects observed compared with water-injected controls (37% compared with 75% GBE defects, respectively; Fig. S5D in the supplementary material). Thus, Myo-II activation is necessary for the ectopic apical constriction in upd mutants.

To test whether upregulation of Myo-II is sufficient to generate upd mutant phenotypes, we overexpressed MHC (Zip) or SqhE20E21, a phosphomimetic form of Sqh that renders it constitutively active (Winter et al., 2001). Ubiquitous expression of SqhE20E21 at similar levels to endogenous Sqh expression (not shown) using the sqh promotor (called endoSqhE20E21) rescued the phenotype of embryos injected with a Rok inhibitor, which removes endogenous Sqh from the cortex (Bertet et al., 2004). Additionally SqhE20E21 expression further enhanced the GBE defects in upd mutants (Fig. S5E in the supplementary material), clearly indicating that SqhE20E21 is active. In either case, GBE defects were not observed (Fig. 6A), despite a marked increase in the total levels of MHC (Zip) at the apical cell region when Zip was overexpressed, or upon overexpression of SqhE20E21 (Fig. 5B-D). In both cases, apical enrichment of MHC was significant (P<0.001 for ZipGFP and P<0.02 for SqhE20E21, Fig. 6E), although it was less pronounced in the case of SqhE20E21. Finally, we found that neither Zip nor SqhE20E21 overexpression induced apical cell constriction (Fig. 6F), and that the overexpression of these genes generated a relative increase in cortical Myo-II compared with the medial apical web pool (Fig. 6G; P<1.5×10⁻³ for ZipGFP and P<0.04 for SqhE20E21). This Myo-II cortical enrichment is the opposite of that observed in upd mutants.

Altogether these results lead us to conclude that although Myo-II is upregulated in upd mutants and is required to mediate apical constriction, neither its concentration, nor its activation by Rok are sufficient to explain the formation of a medial apical meshwork and the associated changes in cell shape. This suggests that JAK/STAT signalling controls Myo-II subcellular localization via additional mechanisms.

Despite several reports that, at least during cytokinesis, Myo-II cortical enrichment is independent of F-actin (Dean et al., 2005; Kamijo et al., 2006; Zang and Spudich, 1998), Myo-II co-localized with F-actin in the apical region of intercalating cells (see Fig. S6A in the supplementary material) and disappeared from regions of the cortex where F-actin was removed upon depolymerization by Latrunculin-A (Fig. S6A,B in the supplementary material). Thus, during GBE, junctional Myo-II requires its binding to F-actin.

This led us to test the role of regulators of actin polymerization in the control of Myo-II recruitment. We first tested the role of actin polymerization controlled by the formin Dia (Fig. 7A). We overexpressed a constitutively active form of Dia, Dia^CA (Somogyi and Rorth, 2004), that localizes to the apical cell cortex (Fig. 7B). In this situation, Myo-II was increased apically, consistent with previous reports (Homem and Peifer, 2008) (Fig. 7D,F, P<0.004), but the ratio of cortical to medial Myo-II was not significantly altered (Fig. 7H). Consistent with this, in embryos laid by dia/+ mothers, in which dia is on average reduced by 65±7% (mean±s.e.m., Fig. S6B in the supplementary material) compared with wild-type controls, Myo-II levels were decreased and the cortex/medial web ratio was unaffected (Fig. 7H, P<0.1, not significant; see Fig. S6C,D in the supplementary material). Dia^CA led to a collapse of the epithelium, which explains the apical cell surface reduction observed in these embryos (Fig. 7G, P<7×10⁻²²). Thus, Dia-dependent actin polymerization is required to regulate the amount of Myo-II recruited in apical actin networks.

Testing the role of actin polylmerization by Wasp led to a very different result. We first expressed a myristoylated form of Wasp (myr-Wasp) (Bogdan et al., 2004; Bogdan et al., 2005), which strongly targets Wasp to the plasma membrane (Fig. 7C). In mammalian cells Wasp is normally recruited and activated at the plasma membrane by Cdc42, although in flies, binding to Cdc42 can be dispensable (Tal et al., 2002). The myristoylated group bypasses this regulatory mechanism and leads to constitutive activation at the plasma membrane. Interestingly, not only did this fail to increase Myo-II at the cortex, but Myo-II strongly accumulated in a medial network (Fig. 7E, P<6×10⁻⁴; see also Fig. 7H, and Movie 12 in the supplementary material), although it was still present, albeit at low levels, at the cortex. This led to ectopic apical constriction (Fig. 7G, 28% surface area reduction, P<4×10⁻¹²). Focal ablation in a diffraction limited spot for 2-3 milliseconds of the medial Myo-II web led to Myo-II redistribution at the cortex together with surface area relaxation (Fig. 4D-F; see also Movie 12 in the supplementary material), suggesting that, as for upd mutants, medial Myo-II accumulation in myrWasp-expressing embryos causes apical constriction. Wasp was shown not to be essential for viability at early stages of development (Ben-Yaacov et al., 2001). Consistent with this, we could not reveal a role of Wasp during GBE (not shown). This can be simply explained by the fact that, as shown below, Wasp activation is normally downregulated specifically in the germband.

Together these data reveal a central role of actin polymerization in the control of Myo-II cortical or medial localization. They also indicate different roles for Dia and Wasp in the regulation of Myo-II localization. Dia controls the total amount of Myo-II present apically in both the cortex and the medial web, but up- or downregulation of Dia does not affect the ratio between the cortex and the medial web. However, activation of Wasp has a strong impact on this ratio and leads to cell constriction. Moreover, these data show that Wasp activation bears strong similarities with upd inactivation, in that it causes both apical constriction and medial accumulation of Myo-II.

This led us to investigate whether JAK/STAT controls Myo-II apical medial recruitment by inhibiting Wasp. We first investigated the localization of Wasp. In wild-type embryos, Wasp was detected at the apical cell cortex at the end of cellularisation and in early gastrulating embryos during mesoderm invagination. However, at the onset of GBE, the levels of Wasp at the cell cortex dropped, especially in the VL intercalating regions where Myo-II was largely cortical. Wasp was still detected more dorsally at the cortex (Fig. 8A). Thus, Wasp was no longer detected and presumably was not activated at the cell cortex in intercalating cells. This downregulation of Wasp required JAK/STAT signalling. Indeed, in hop mutants, at the onset of GBE, Wasp was still strongly enriched at the apical cell cortex in the VL regions of the embryo, where cells undergo abnormal apical constriction (Fig. 8B). Thus, JAK/STAT signalling downregulates the cortical recruitment of Wasp in intercalating cells.

In order to understand how JAK/STAT regulates the cortical localization of Wasp, we performed quantitative RT-PCR experiments. These experiments revealed that the levels of wasp transcripts were the same in wild type and in hop mutants (data not shown). We conclude that the JAK/STAT signalling pathway does not repress the cortical recruitment of Wasp by regulating wasp transcription, but presumably by as yet unknown post-transcriptional mechanisms.

To test whether Wasp overactivation is required to mediate Myo-II medial accumulation in JAK/STAT mutants and apical constriction, we performed an epistasis experiment. We investigated whether a reduction in wasp gene dosage rescues apical constriction and the medial accumulation of Myo-II observed in upd mutants (Fig. 8C-E). As shown in Fig. 8E, the characteristic medial apical accumulation of Myo-II in upd mutants was lost in upd mutant embryos laid by mothers heterozygous for a null allele of wasp, wasp³. The global enrichment of Myo-II apically was similar in upd mutants and upd; wasp³/+ mutants (Fig. 8F, P>0.05), but the cortical/medial Myo-II ratio was rescued in upd; wasp³/+ mutants, with significantly more Myo-II at the cortex (Fig. 8G, P<0.006). This was accompanied by a rescue of the apical constriction phenotype (Fig. 8H, P>0.05). We conclude that, at the onset of GBE, the JAK/STAT signalling pathway represses Wasp activation. This repression is essential to prevent medial Myo-II accumulation and promotes the enrichment of Myo-II at the cortex, which is characteristic of intercalating cells. Note that Myo-II cortical localization is not totally normal in upd; wasp/+ mutant embryos; its planar polarized distribution is affected (Fig. 8E). Consistent with this, GBE extension is defective in these double mutants.

Myo-II subcellular localization controls different cell shape changes such as cell constriction or intercalation. The data we report shed new light on the mechanisms of the subcellular localization of actomyosin networks in the early Drosophila ectoderm. VL ectodermal cells intercalate via the cortical recruitment of Myo-II at AJs, which drives polarized junction remodelling (Bertet et al., 2004; Blankenship et al., 2006; Rauzi et al., 2008). This contrasts with the behaviour of immediately adjacent cells in the mesoderm, which undergo apical constriction and recruit Myo-II into a medial apical web (Martin et al., 2009). Our data indicate that the cortical enrichment of Myo-II in ectodermal intercalating cells is not a ‘default pathway’, and requires at least activity of the JAK/STAT pathway. Indeed, in JAK/STAT pathway mutants, Myo-II is aberrantly recruited in a medial apical meshwork and cells consequently undergo apical constriction. This is surprising, as apical constriction is normally only observed in mesodermal ventral cells and is considered to be a unique attribute owing to their selective expression of Twist and Snail. Twist and Snail induce expression of the ligand Fog in the ventral cells only, which activates RhoGEF2, Rok and Myo-II. It also regulates expression of the transmembrane protein T48, which participates in the apical recruitment of RhoGEF2 and contributes to apical constriction (Kolsch et al., 2007). However it is not clear whether activation of the RhoGEF2 pathway is sufficient to drive the apical medial recruitment of Myo-II (Martin et al., 2009). Here, we show that apical constriction is not simply induced in mesodermal cells by Fog, but is also prevented in ectodermal cells by activity of the JAK/STAT pathway and that this is essential for GBE. In JAK/STAT pathway mutants, ectodermal cells undergo apical constriction despite the absence of ectopic Twist expression (see Fig. S7 in the supplementary material). Note, however, that apical constriction is not as rapid in these mutants as in mesodermal cells, so Twist and Snail accelerate or render more efficient the capacity to apically constrict. Moreover, the fact that Wasp mediates JAK/STAT function in the ectoderm but is not required in the mesoderm indicates that the mechanisms promoting medial Myo-II in mesoderm cells are likely to be different.

Our findings provide a novel opportunity to investigate the regulation of cortical or medial Myo-II localization in the ectoderm. Our data document two novel features of this regulation.

MRLC (Sqh) phosphorylation by the RhoGEF2 and the Rok pathway are both necessary for apical constriction, as lowering the dose of RhoGEF2, Rho or Rok suppress the apical constriction observed in upd mutants. However, neither constitutive activation of this pathway by expression of a phosphomimetic form of Sqh, ShqE20E21, which rescues Rok inhibition (Bertet et al., 2004), nor overexpression of MHC (Zip) is sufficient to promote medial accumulation of Myo-II. The medial accumulation of Myo-II requires additional regulation apart from the activation of Myo-II. As RhoGEF2 and Rok are key regulators of Myo-II, this suggests that activation of the RhoGEF2/Rok pathway is necessary but not sufficient to explain medial Myo-II accumulation and apical constriction.

Our analysis of the JAK/STAT mutant phenotypes indicates a key role of Wasp in this process. Wasp is shown to be necessary for medial Myo-II accumulation, at least in ectodermal cells, and very strong activation of Wasp at the cortex (myrWasp) also causes medial Myo-II accumulation (Fig. 6). Moreover, although Wasp is normally downregulated in VL ectodermal cells, in JAK/STAT pathway mutants Wasp is strongly recruited and hence activated at the plasma membrane, which suggests that JAK/STAT signalling represses the membrane activation of Wasp. Importantly, lowering the dose of Wasp maternally suppresses medial accumulation of Myo-II in upd mutants, and restores prominent accumulation at the cortex, as in wild-type embryos. Consistent with this, ectopic apical constriction is completely rescued in these double mutant embryos.

We report two different roles for Dia and Wasp in the regulation of Myo-II localization. Consistent with previous data, Dia controls the amount of apical Myo-II, but the specific localization of Myo-II at the cortex or in the medial network is not affected by loss of Dia. Dia promotes polymerization of non-branched filaments, and might control the formation of a good substrate for the stable association of Myo-II minifilaments. The fact that in dia heterozygotes the amount of apical Myo-II is reduced indicates that the amount of actin filaments might be limiting and controlled. Indeed, more F-actin is detected at the cortex of intercalating cells, preceding by a few minutes the enrichment of Myo-II (Blankenship et al., 2006). The role of Wasp is more surprising and unique, as it is shown to mediate specifically repression of medial Myo-II accumulation and, hence, cell constriction in the germband. Because activation of Wasp leads to activation of medial web formation and reduction of Wasp dosage rescues cortical Myo-II in JAK/STAT mutants, we conclude that Wasp controls an essential feature of Myo-II subcellular localization that is essential for the regulation of apical constriction. How does Wasp control Myo-II localization? We can envisage two non-exclusive models. In the first model, Wasp controls actin branching through activation of the Arp2/3 complex (Ben-Yaacov et al., 2001). Because Wasp has been implicated in endocytosis via Arp2/3 in Drosophila (Georgiou et al., 2008; Leibfried et al., 2008), Wasp could promote Myo-II web formation indirectly by regulating endocytosis of a surface protein required to anchor the medial actomyosin network at the membrane, such as E-cadherin. Consistent with this, downregulation of E-cadherin by RNAi disrupts the faint medial Myo-II pool (M.R., unpublished). In mesodermal cells, E-cadherin appears to anchor the strong medial Myo-II pool (Dawes-Hoang et al., 2005; Martin et al., 2009). In the second model, Wasp might act more directly via the regulation of actin network architecture and its impact on the dynamic interactions between the medial web and the cortex, and thereby might affect the steady-state distribution of Myo-II exchanging between these two pools. Although Wasp uniquely mediates Myo-II regulation via JAK/STAT in the ectoderm and not in the mesoderm, regulation of Arp2/3 might be more generally implicated in the control of Myo-II regulation.

Although wasp is an important mediator of JAK/STAT function in the ectoderm, it is unlikely to be the only one. Indeed wasp mutants rescue the cortical accumulation of Myo-II and apical constriction in upd mutants, but GBE is still strongly affected (data not shown); we noticed that cortical Myo-II distribution was not properly polarized in the plane of the epithelium. This suggests that other subcellular processes are also perturbed in the mutant. The fact that a reduction of Myo-II levels suppresses the upd defects indicates that the overall dosage of Myo-II is important as well. Identifying the transcriptional targets of JAK/STAT might shed light on its complex regulatory role during embryonic morphogenesis.

Finally, although this work identifies an important regulator of Myo-II network subcellular distribution in epithelial cells, it is still not clear what regulates the polarized distribution of Myo-II at the cortex.

JAK/STAT signalling controls a number of developmental processes. Importantly, this pathway has been implicated in diverse morphogenetic processes, such as convergent extension movements in the zebrafish embryo, hindgut elongation in Drosophila embryos, which probably involves intercalation movements as well (Johansen et al., 2003), and posterior spiracle morphogenesis in Drosophila embryos (Sotillos et al., 2008). JAK/STAT signalling also controls border cell migration (Beccari et al., 2002; Brown et al., 2006; Ghiglione et al., 2002; Silver et al., 2005). Our data indicate that JAK/STAT signalling plays an important and hitherto unappreciated morphogenetic function in gastrulating embryos. We document that, JAK/STAT controls, via Wasp, a morphogenetic switch based on the regulation of medial or cortical Myo-II distribution. Interestingly, dorsal cells do not undergo apical constriction in JAK/STAT mutants. Indeed, dorsal cells exhibit neither cortical nor medial web Myo-II and are thus unable to participate in profound tissue remodelling. It appears that DV patterning provides a first general subdivision within the embryonic epithelium whereby Myo-II is globally repressed dorsally, and activated laterally and ventrally. Cortical or medial web distribution then results from the combinatorial input of Fog and JAK/STAT.

We thank Jean-Marc Philippe for performing the QRT-PCR experiments and Ahmed Fatmi and Virginie Thomé for assistance. We are grateful to all those who generously provided us with reagents, especially C. Field for the gift of Zipper antibody at a crucial time, and E. Schejter, M. Zeidler, S. Wasseman, N. Perrimon, R. Karess, D. Kiehart, K. Klaembt, P. Rørth, H. Oda, J. P. Vincent, A. Gautreau, M. Frash, D. Kosma and the Bloomington Stock Center. We thank the members of the Lecuit lab for fruitful discussions, as well as S. Kerridge, M. Cavey, S. Merabet, B. Charroux and P. F. Lenne for useful comments on the manuscript. This work was supported by the CNRS, the ACI-BCMS (Ministère de la recherche), the Association pour la Recherche sur la Cancer (ARC), the Fondation Schlumberger pour l'Education et la Recherche (FSER), and the EMBO Young Investigator Programme. C.B. was the recipient of a fellowship from the ARC.