The Drosophila serine protease homologue Scarface regulates JNK signalling in a negative-feedback loop during epithelial morphogenesis

The trypsin- and chymotrypsin-like (S1 family) serine proteases (SPs) are mainly represented in animals and are implicated in food digestion, blood coagulation, immune response and development. They are almost exclusively localized outside the cytoplasm to perform their function and they are synthesized with a signal peptide for secretion. In addition, SPs are usually produced in an inactive form, which is cleaved upon a signal to form a functional protease. This mechanism favours a rapid and local response to a stimulus. For example, in mammals, cascades of localised SP activities control emergency blood clotting at a wound site (Davie et al., 1991). During Drosophila development, dorsoventral patterning of the embryo results from an extracellular cascade involving the SPs Nudel, Gastrulation defective, Snake and Easter and leading to the activation of the Toll pathway in the ventral region (Anderson, 1998).

SPs are endopeptidases that use a serine for their catalytic activity and are characterized by the presence of the residues histidine, aspartic acid and serine (HDS) in the active site (Rawlings and Barrett, 1993). Serine protease homologues (SPHs) are also found in various animal genomes and are defined by the lack of at least one of the essential residues of the catalytic triad HDS. Only a few SPHs have been characterized. One well-known example in mammals is the Hepatocyte growth factor or Scatter factor (HGF, SF) (Nakamura et al., 1989). In Drosophila, almost 30% of the 204 identified SP genes encode SPHs that are therefore most probably catalytically inactive (Ross et al., 2003; Shah et al., 2008). Only four of these SPH genes have been assigned a function. Mutations in masquerade (mas) lead to defects in somatic muscle attachment and in the formation of the nervous system during embryogenesis (Murugasu-Oei et al., 1996; Murugasu-Oei et al., 1995). The other three genes, spheroide, sphinx1 and sphinx2, were identified in an RNAi screen aimed at identifying genes involved in immunity (Kambris et al., 2006).

We isolated the SPH encoding gene scarface (scaf) in a microarray screen designed to identify new genes transcriptionally regulated by the JNK (c-Jun N-terminal kinase) pathway during dorsal closure (DC) of the Drosophila embryo. DC is a major morphogenetic movement that takes place after germ-band retraction to close the eye-shaped opening of the dorsal embryonic epidermis, over the transient tissue called the amnioserosa (AS). Cell elongation is responsible for the dorsalward spreading of the lateral epithelial sheets. The AS, conversely, progressively reduces through cell constriction and engulfment accompanied by cell death (Franke et al., 2005; Kiehart et al., 2000; Reed et al., 2004). Precise activation of the JNK pathway, specifically in the dorsalmost row of epithelial cells in contact with the AS (the leading edge, LE), is a prerequisite for DC progression (Bates et al., 2008; Byars et al., 1999; Glise et al., 1995; Glise and Noselli, 1997; Martin-Blanco et al., 1998; Reed et al., 2001). Several transcriptional targets of the JNK pathway have already been described during DC, such as the profilin-coding gene chickadee, the transcription factor cabut, the integrin-coding genes scab and myospheroid or the trafficking gene Rab30 (Homsy et al., 2006; Jasper et al., 2001; Munoz-Descalzo et al., 2005; Thomas et al., 2009). However, only two target genes are specifically expressed in the LE, decapentaplegic (dpp) and puckered (puc) (Glise et al., 1995; Glise and Noselli, 1997). Dpp, of the TGFβ family, participates in cell spreading of the lateral epithelium, cell constriction in the AS and connection of these two tissues (Fernandez et al., 2007; Wada et al., 2007). Dpp also regulates remodelling of the cytoskeleton in the segment border cells by activating Cdc42 and dPak (Ricos et al., 1999). The gene puc encodes a phosphatase that sets up a negative-feedback loop at the level of the c-Jun kinase Basket (Bsk) (Martin-Blanco et al., 1998). In addition, other JNK target genes are involved in different biological processes. For instance, JNK-induced expression of matrix metalloproteinases is required for disc eversion and tumour invasion (Srivastava et al., 2007); JNK protection against oxidative stress is realized through expression of autophagy-related genes (Wu et al., 2009); and during immune response, JNK participates in the expression of antimicrobial peptide genes (Delaney et al., 2006).

In this paper, we report the characterization of scaf during Drosophila development. The catalytic triad of the SP domain of Scaf lacks two of the conserved HDS residues and Scaf was thus classified as a SPH (Ross et al., 2003). Although a previously isolated semi-lethal mutation led to adult escapers exhibiting a head scar under the proboscis (Bonin and Mann, 2004), no function has been attributed to scaf. Here we show that the JNK signalling pathway regulates scaf expression in the embryo and in imaginal discs. In S2 cells and in vivo, the protein Scaf is secreted, indicating that it behaves as an extracellular SPH. Loss-of-function and gain-of-function in vivo studies uncovered an essential role of scaf during epithelial morphogenesis through antagonistic regulation of JNK signalling.

The following fly lines were used: en-GAL4^e16E (gift from A. Brand, University of Cambridge, UK), AbdB-GAL4 (gift from E. Sanchez-Herrero, Universidad Autónoma de Madrid), ptc-GAL4 (gift from N. Perrimon, Harvard Medical School, USA), UAS-dpp (gift from K. Basler, University of Zurich, Switzerland), UAS-hep^4E (WT), hep¹ and hep^r75 (Glise et al., 1995), puc^E69 (puc-lacZ) (Ring and Martinez Arias, 1993), and UAS-puc2A (Martin-Blanco et al., 1998). The following lines from Bloomington were also used: w¹¹¹⁸ (#3605), 69B-GAL4 (#1774), act-GAL4 (#4414), pnr-GAL4 (#3039), UAS-hep^act (#9306), UAS-EGFP (#5431), UAS-CD8GFP (#5137), UAS-p35 (#5072), CyO,Kr-GAL4,UAS-GFP balancer (#5194) (Casso et al., 2000), and Δscaf deficiency (#8884); as well as UAS-scafRNAi (#13249) and UAS-Dicer2 (#60009) from VDRC.

scaf cDNA was obtained from the EST GH05918 (DGRC). The following recombinants were created for this study: (1) Rescue – scaf (scaf ^f01332, scaf ^PBss or scaf ¹¹),act-GAL4 and scaf (scaf ¹¹, scaf ^PBss or scaf¹¹),UAS-scaf; (2) RNAi – UAS-scafRNAi,UAS-Dicer2; and (3) Q-PCR – pnr-GAL4,puc^E69. The recombinants for the rescue were balanced with CyO,Kr-GAL4,UAS-GFP. We also made the following stocks: (1) overexpression in the disc (Fig. 4F) – ptc-GAL4;puc^E69//SM6a;TM6B; (2) overexpression in the disc (Fig. 7A) – UAS-EGFP;UAS-puc2A and UAS-scaf;UAS-puc2A; and (3) live imaging – UAS-CD8GFP/SM5;AbdB-GAL4/TM6B. The majority of the crosses were done at 25°C, the others were carried out at 29°C as indicated in the text.

scaf ^f01332 (Exelixis) and scaf ^PBss (Bonin and Mann, 2004) correspond to PiggyBac insertions. We used scaf ^KG05129 (Bloomington #13631) as a starter line for imprecise P element jump events by action of a P transposase. Two white minus lethal lines were selected: scaf ¹¹ and scaf ²⁷. Deletions were mapped by PCR on genomic DNA and sequencing.

The lethality associated with the scaf zygotic loss-of-function was determined as follows: eggs were collected at 25°C from CyO,Kr-GAL4,UAS-GFP balanced scaf lines or from the crosses scaf,act-GAL4/CyO,Kr-GAL4,UAS-GFP × scaf,UAS-scaf/CyO,Kr-GAL4,UAS-GFP for the rescue experiments. Ten hours after egg laying (AEL), GFP-minus dechorionated embryos were raised at 25°C, covered with Voltalef oil (Prolabo). We collected larvae that hatched and made cuticle preparations from the remaining dead embryos. We let the larvae grow to determine if they die during the larval or pupal stage, and if they give rise to some escapers. Experiments were done in parallel with a w¹¹¹⁸ line and we performed a statistical analysis (Student's t-test) on the different stages of lethality.

We used the Affymetrix DNA microarrays (DrosGenome1) to identify genes whose expression is regulated by the JNK pathway. Two genetic conditions were compared, wild-type and gain-of-function embryos, in which the JNK pathway is strongly activated by expressing Hep^act in the ectoderm using the UAS-GAL4 system (69B-GAL4). This allowed us to isolate 634 JNK target genes that are activated by a fold change of at least 1.8. Present among these genes were puc with a fold change of 2.0 and scaf with a fold change of 1.8. A secondary screen based on in situ hybridization led to the validation of scaf, as well as several other genes (R.R., F. Carballès and S.N., unpublished data). Microarray data have been deposited in the Gene Expression Omnibus (accession number GSE21805).

Digoxigenin (DIG)-labelled RNA probe synthesis and in situ hybridization (ISH) experiments were performed using a standard protocol. Anti-DIG antibody conjugated to alkaline phosphatase (1/2000; Roche Diagnostics) was used and revelation was performed with NBT/BCIP reagent (Sigma). For fluorescent ISH coupled to immunostaining, the second fixation step was performed using freshly prepared paraformaldehyde and proteinase K treatment was not done. Primary antibodies were: rabbit anti-Scaf2 (1/50) and mouse anti-Ena (1/50). Sheep anti-DIG antibody coupled to horseradish peroxidase (1/2000; Roche Diagnostics) was joined to the secondary antibodies anti-rabbit Al488 (1/400; Molecular Probe) and anti-mouse Cy5 (1/100; Jackson ImmunoResearch). Revelation was done twice for 30 minutes with Tyramide Signal Amplification (PerkinElmer). Embryos were mounted in Mowiol 4-88 Reagent (Calbiochem).

We raised polyclonal antibodies against two different GST fusion proteins containing Scaf residues 138-270 (Scaf1 antibody) or 297-408 (Scaf2 antibody). Approximately 1 mg of the proteins was injected into a New Zealand White rabbit. The sera were affinity-purified on His-tagged peptide columns (Affi-gel 10; Bio-Rad) according to the manufacturer recommendations, except that we used the ImmunoPure Gentle Ag/Ab elution buffer pH 6.55 (Pierce). Finally, we tested the affinity-purified Scaf antibodies by western blotting and by immunostaining to ensure specificity.

Embryos were prepared as described for ISH and the incubation buffer was 0.1% PBS-Triton X-100 with 1% BSA. The following primary antibodies were used: rabbit anti-Scaf1 and 2 (1/200); mouse anti-En 4D9 (1/500; ascites fluid; DSHB); mouse anti-Ena 5G2 (1/500; DSHB); mouse anti-Lamin (1/50; DSHB); rat anti-Troponin H (1/300; Babraham Tech) and mouse anti-Golgi (1/500; Calbiochem). Nuclear staining was done with DAPI (10 μg/ml, Biochemika). For the actin staining, we used methanol-free formaldehyde (Polysciences, Inc.) in the fixation step and devitellinization was performed without methanol. We used instead chilled 80% ethanol. Actin was labelled with FluoProbes 547H or 647H phalloidin (Interchim). For the soft protocol, embryos were prepared as described above for the actin staining. In addition, the buffer did not contain Triton X-100, but Tween-20.

The genital discs were dissected from male L3 larvae in distilled water, transferred into cold PBS and fixed with 4% formaldehyde. Immunostaining was performed in 0.1% PBS-Triton X-100 and 1% BSA. We used the rabbit anti-Scaf2 (1/200) or the rabbit anti-En (1/200; Santa Cruz) and the chicken anti-βgalactosidase antibody (1/1000; Genetex).

S2 cells were transfected using the Effectene reagent (Qiagen). For overexpression experiments, we co-transfected a DNA mix made of pAct-GAL4 and pUAS-scaf or pUAS-scafGFP for about 24 hours. For RNAi experiments, we transfected 16 μg of double-stranded RNA directed against GFP or Syx5 in 1 ml of minimal Schneider's medium at 22°C for 1 hour. We then added 2 ml of complete medium and left the cells for 5 days at 22°C. Cell lysates were obtained by incubating S2 cells with RIPA solution containing a protease inhibitor cocktail.

For western blotting, we loaded 80 μg of proteins coming from the cell lysates and 25 μl of culture supernatant. We used the rabbit anti-Scaf2 or the rabbit anti-Scaf1 (1/700) and the mouse anti-βTub (1/200, DSHB). Revelation was performed with the Immobilon Western reagent (Millipore).

The Multigauge V2.3 program (Fujifilm) was used for signal quantification on the western blot. Quantification is based on pixel intensity. For each lane of the lysates, the background was subtracted to the signal intensity of the band of interest and of the βTub loading control, and the ratio of these two values was calculated.

Embryos were collected for 24 hours and incubated at 25°C to let the wild-type larvae crawl away. The cuticles of unhatched (dead) embryos were prepared using a standard protocol.

Embryos from pnr-GAL4,pucE9/TM6B flies crossed to w¹¹¹⁸, UAS-hep^act or UAS-scafRNAi,UAS-Dicer2 homozygotes were collected at 25°C or 29°C. They were dechorionated, frozen in liquid nitrogen and stored at –80°C. They were then lysed with the TissueLyser LT (Qiagen) and RNA was extracted with the RNeasy Mini Kit (Qiagen). Reverse-transcription was performed with SuperScript III Reverse Transcriptase (Invitrogen) after DNase I digestion with a mix of oligo-dT and random primers. The following primers were used: CTTCATCCGCCACCAGTC (forward) and CGACGCACTCTGTTGTCG (reverse) for the endogenous control rp49, and GGTTACGATGCGCCCATCTA (forward) and ACGGCGGATTGACCGTAAT (reverse) for lacZ. Q-PCR was performed with the Mastermix Plus for SYBR Green containing Rox (Eurogentec) and standard curves showed a 100% efficacy for rp49 amplification and 99% for lacZ. For each condition we did three biological and four technical replicates. Results were analysed with the StepOne software v2.1 (Applied Biosystems).

To follow the genitalia rotation in vivo, male pupae were selected at the prepupal stage (0-12 hours after puparium formation) and incubated at 25°C or 29°C for 24 hours. The most-posterior part of the pupae was then cut in water. The pupae were stuck upside down in an agar plate and covered with water. Acquisitions of a z-stack (around 150-200 mm depth) were taken every 15 minutes over 15 hours with a Zeiss LSM 510 Meta confocal microscope.

We carried out a genomic analysis of JNK target genes during DC using the Affymetrix oligo-chips (R.R., F. Carballès and S.N., unpublished data). A secondary screen using in situ hybridization (ISH) on embryos led to the validation of several candidates. We particularly focused on scaf because of its strong and specific expression in the LE during DC, similar to the classical JNK target gene dpp (Fig. 1A,E). scaf mRNA is first visible during germband retraction (stage 12 of development) and is detected in the LE until the end of DC (stage 16; see Fig. S1 in the supplementary material). The expression is also noticeable in some cells of the AS, in particular at the posterior canthus (Fig. 1A), as well as in the ventral ectoderm (Fig. 1I). scaf mRNA is no longer present in the LE of embryos mutant for hep, although it is still visible in the ventral ectoderm (Fig. 1B). When we overexpressed the active form of JNKK (Hemipterous, Hep^act) in the posterior compartment of each segment (en>hep^act), we observed the appearance of ectopic stripes of scaf expression overlapping these compartments (Fig. 1C). JNK signalling is thus sufficient for scaf expression. In order to determine whether scaf is a primary target of JNK or a secondary one through the activation of the Dpp pathway, we expressed dpp in stripes (en>dpp) and observed that, in these embryos, scaf expression is wild-type (Fig. 1D). These data indicate that scaf is a novel primary JNK target gene in the LE of dorsally closing embryos, like puc and dpp (Fig. 1J).

In order to characterize the endogenous Scaf protein, we made affinity-purified antibodies directed against two distinct non-conserved regions (see Fig. S2A,B in the supplementary material). Immunostainings coupled to fluorescent ISH of wild-type embryos (using Triton) indicated that the protein was localised in the same scaf mRNA-expressing cells (see Fig. S2C in the supplementary material). In particular, Scaf is detected in the LE cells, with a stronger expression at both extremities (Fig. 2A). As Scaf contains a putative signal peptide, we tested the capacity of Scaf to be secreted. In S2 cells, the endogenous (Fig. 2B,C) and overexpressed (see Fig. S3C,D in the supplementary material) protein is detected in the culture medium, indicating that Scaf has indeed been secreted. Secretion is diminished when cells are treated by RNAi to silence Syntaxin 5 (Syx5), an essential protein of the secretory pathway (Fig. 2B,C) (Xu et al., 2002). In en>scaf embryos in vivo, the protein was detected within cells located outside the en-expressing cells, both in the anterior cells and in the AS cells (Fig. 2D). Co-stainings with markers of the nuclear envelope, of the endoplasmic reticulum and of the Golgi apparatus did not detect subcellular colocalization with Scaf in these cells (see Fig. S4 in the supplementary material). A similar pattern of Scaf localization outside of its expression domain, in the lateral ectoderm and the AS, was also revealed in wild-type embryos using a soft immunostaining protocol (Fig. 2E; see Fig. S5 in the supplementary material), although the signal was weaker compared with the overexpression condition (Fig. 2D). Apparent differences in stainings at the LE when comparing Fig. 2D and Fig. S2 with Fig. 2E are explained by the fact that two permeabilization protocols were used. In Fig. 2E, Triton was not used to allow detection of secreted Scaf, but this led to reduction of Scaf detection expressed at the LE. By contrast, in Fig. 2A and Fig. S2, Triton allowed detection of Scaf only in the LE. All together, these results indicate that Scaf is a secreted SPH both in cultured cells and in vivo, suggesting a function of Scaf in the extracellular space during DC.

In this study, we used four different scaf alleles: scaf ^PBss, scaf ^f01332, scaf ¹¹ and scaf ²⁷. scaf ^PBss and scaf ^f01332 correspond to two previously described PiggyBac insertions (Bonin and Mann, 2004; Thibault et al., 2004) and we generated scaf ¹¹ and scaf ²⁷ from P element imprecise excision that created small deletions within the locus (see Materials and methods; Fig. 3A). Whereas scaf ²⁷ is lethal, the other alleles are, like scaf ^PBss, semi-lethal and the adult escapers bear a scar on their inferior head with no apparent other defects (Fig. 4A) (Bonin and Mann, 2004). Based on the percentage of escapers (Table 1) and the lethality during the different stages of development (Fig. 3B), we established the strength of each allele, from the weakest to the strongest, as following: scaf ^f01332, scaf ^PBss, scaf ¹¹, scaf ²⁷. Interestingly, the dead embryos corresponding to the scaf ¹¹and scaf ²⁷ zygotic mutants displayed cuticle phenotypes corresponding to DC defects (Fig. 3C). Indeed, we observed opening of the anterior part of the embryo (`anterior open'), as well as puckering of the dorsal cuticles that resemble defects due to JNK hyperactivity (see below). Immunostainings did not detect the Scaf protein in these embryos, indicating the loss of Scaf expression and revealing the specificity of the antibodies (Fig. 3D). Importantly, rescue experiments saved the lethality and restored cuticle formation (Fig. 3B) (S.B.-L., unpublished data), therefore proving that the observed phenotypes are due to the mutations in the scaf locus.

To further confirm the scaf loss-of-function phenotypes, we expressed a hairpin RNA corresponding to scaf (along with Dicer to enhance RNAi) using the epidermal driver 69B-GAL4. scaf RNAi embryos resembled scaf ¹¹and scaf ²⁷ embryos (Fig. 3C). Efficiency of the RNAi was validated by immunostaining, which showed a loss of Scaf protein expression (Fig. 3D).

Diverse adult phenotypes were also noticed with other drivers (ptc-GAL4 and AbdB-GAL4). Scars under the proboscis confirmed the correct targeting of the RNAi against scaf (Fig. 4A). In addition, we observed loss of wing veins and thorax mechanosensory bristles, as well as male terminalia malformations (Fig. 4B) (R.R., unpublished data). The latter phenotype is of particular interest as JNK signalling has been previously implicated in male genital plate morphogenesis (Macias et al., 2004; McEwen and Peifer, 2005; Polaski et al., 2006). scaf RNAi led to genitalia rotation defects and partial dissociation of the genital plate from the abdomen, indicating that the scaf gene also participates in male terminalia morphogenesis (Fig. 4B). In conclusion, invalidation of scaf function affected morphogenetic movements controlled by JNK signalling such as DC and male genitalia formation.

We decided to use the genitalia as a second model of morphogenetic tissue to study the role of scaf and JNK signalling. The male genital plate derives from a unique, centrally located imaginal disc, called the genital disc. Three groups of cells originally coming from the abdominal segments A8, A9 and A10 of the embryo assemble to form the complete larval disc, each primordium containing an anterior compartment and a posterior one (drawing in Fig. 4C) (Chen and Baker, 1997; Freeland and Kuhn, 1996). During pupariation, the male genital disc undergoes partial eversion and metamorphoses. The disc also turns dextrally, accomplishing a full 360° rotation. We recently showed that the A8 segment is an organizer of this directional left-right morphogenesis through the activity of a novel type ID unconventional myosin (Speder et al., 2006; Speder and Noselli, 2007). We used the puc-lacZ reporter line to monitor the activity of JNK signalling in the male genital disc. Anti-β-Galactosidase immunostaining revealed a row of cells with a strong signal in the A8 segment (Fig. 4D), like previously reported (Macias et al., 2004). In addition, we show that this signal colocalises with the posterior determinant Engrailed (Fig. 4D). This indicates that JNK activity is high in the posterior compartment of the A8 segment. Moreover, we also detected some activity in the anterior compartment, as well as in the A9 segment. This weak staining was repressed by Puc overexpression (Fig. 4F) and thus corresponds to a basal activity of JNK signalling rather than a non-specific staining. The anti-Scaf antibody stained the same cells as the puc-lacZ-positive cells in the posterior A8 compartment, suggesting that scaf is also a JNK target gene in this tissue (Fig. 4E). Indeed, Scaf staining was lost when JNK signalling was inhibited in the A8 segment by Puc overexpression with ptc-GAL4 (Fig. 4F). Therefore, the JNK pathway controls scaf expression in the male genital disc, like in the embryo.

Previous studies based on puc ectopic expression revealed the involvement of JNK signalling in male terminalia formation (Macias et al., 2004; McEwen and Peifer, 2005). Consistently, mutations in hep (Holland et al., 1997) or the JNKKK-encoding slipper gene (Polaski et al., 2006) also affected male plate morphogenesis. To assess the effect of JNK inhibition in specific domains of the male genital disc, we used three different drivers to overexpress Puc: AbdB-GAL4 (expressed predominantly in the anterior compartment of the A8 segment), en-GAL4 (strictly in the posterior compartment) and ptc-GAL4 (in both compartments) (Speder et al., 2006). en-GAL4 and ptc-GAL4 drivers, but not AbdB-GAL4, also drive expression in the A9 and A10 segments (Speder et al., 2006). As previously described (Macias et al., 2004; McEwen and Peifer, 2005), two types of highly penetrant phenotypes were observed: an incomplete rotation of the plate (between 70° and 359° depending on the driver) and an incomplete fusion of the genital arch (Fig. 5A,B). In contrast to en-GAL4 and ptc-GAL4, which affected both processes, AbdB-GAL4 induced misrotation without arch defect (Fig. 5A,B). Whereas rotation, as already mentioned, depends on the A8 segment, genital arch formation derives from the A9 segment of the disc (Estrada et al., 2003). This indicates a requirement of JNK in the A8 segment for rotation and in the A9 segment for arch fusion. Similar results were obtained by overexpressing a dominant-negative form of Bsk. In addition, we never observed sinistral rotation owing to perturbation in JNK activity, going against a role the JNK pathway in left–right asymmetry establishment like myosin ID (R.R., J.-B. Coutelis and A. Petzoldt, unpublished data). Taken together, these data indicate that JNK activity is required in the A8 segment for full rotation of the male genital plate, as well as in the A9 for the correct fusion of the genital arch.

In order to better comprehend JNK-regulated genital plate formation, we developed live imaging of the male pupal disc undergoing rotation. AbdB-GAL4-driven expression of GFP enabled live visualization of the A8 segment of the male disc (Fig. 5C; see Movie 1 in the supplementary material). During metamorphosis of the genital disc, the A8 segment first flattens and then encloses the A9 and A10 segments (Keisman et al., 2001). We observed that, before the onset of rotation, the two lateral ends of the A8 segment have not yet encircled entirely the A9 and A10 tissues (Fig. 5C, left panels; see Movie 1 in the supplementary material). They progressively migrate towards each other and then meet. It is only when they have fused dorsally that rotation can occur. Interestingly, overexpression of puc in the A8 segment inhibits this fusion (Fig. 5C, right panels; see Movie 1 in the supplementary material). Despite this, rotation starts, but is incomplete, which gives rise to the adult phenotype described in Fig. 5B. This process of tissue sealing is analogous to other JNK-dependent epithelial sheet movements like DC of the embryo, thorax closure of the wing discs and wound healing. The merger also taking place dorsally in the male genital disc we named dorsal fusion of the male genital disc (not to be confused with genital arch fusion). Previous studies suggested that apoptosis promotes rotation and genital arch formation possibly through JNK signalling (Abbott and Lengyel, 1991; Macias et al., 2004; McEwen and Peifer, 2005). For example, suppression of apoptosis by overexpression of the caspase inhibitor p35 mimics the defects observed in plates overexpressing puc (Macias et al., 2004; McEwen and Peifer, 2005). However, our results strongly suggest that this is not the case for the dorsal fusion of the A8 segment, as p35 expression specifically in the A8 has no effect on its closing (Fig. 5D; see Movie 2 in the supplementary material). Like embryonic DC, JNK probably induces changes in cell morphology without triggering cell death. In conclusion, we show that JNK activity is required for complete rotation of the male genital plate by controlling closure of the A8 segment.

As previously mentioned, we noticed that scaf loss of function in the embryo (Fig. 3C) did not lead to holes in the dorsal part of the cuticle like the ones due to mutations in positive regulators of the JNK pathway. Rather, scaf mutations induced holes in the head resembling to the ones associated with anterior open (aop) mutations, as well as puckering of the dorsal cuticle also observed in puc mutants, two genes that negatively regulate JNK activity (Martin-Blanco et al., 1998; Riesgo-Escovar and Hafen, 1997). Ectopic activation of JNK in the embryonic epidermis by expression of either the active (hep^act) or the wild-type form of hep gave rise to similar anterior open and puckered phenotypes (Fig. 6A). Thus, the scaf phenotypes might be due to hyperactivity of JNK signalling, suggesting that, like puc and aop, scaf might antagonize the activity of the JNK pathway.

The puc-lacZ staining and the effect of Puc expression in the imaginal disc (Figs 4, 5) suggest that, like in the embryo, JNK signalling needs to be precisely controlled for proper male genital plate morphogenesis. Hence, JNK hyperactivity should also trigger plate malformation. To address this question, we overexpressed hep during male genital disc development. The active form of hep induced lethality with all the drivers that were tested; therefore, we used the wild-type form, which allowed the eclosion of adult flies. Like for scaf loss-of-function, the male escapers displayed misrotation defects and plate malformations, showing that excess JNK activity is also detrimental to genital plate formation (Fig. 6B).

To further confirm the negative action of scaf, we co-expressed Scaf along with Puc in the genital disc with en-GAL4. We observed that scaf enhanced the effect of puc, as rotation was more incomplete with Puc and Scaf expression compared with Puc and GFP (Fig. 7A). This result was confirmed by live-imaging of the A8 dorsal fusion. Indeed, the A8 extremities did not migrate dorsally when Scaf was co-expressed at 25°C, increasing the defect observed with Puc expression alone at the same temperature (Fig. 7B; see Movie 3 in the supplementary material). Consistently, this severe phenotype resembles the one seen when Puc overexpression is enhanced at 29°C (Fig. 7B; see Movie 3 in the supplementary material). This corroborates the effect of Scaf overexpression on adult male genital plates described above and validates scaf as a negative regulator of JNK-controlled epithelial morphogenetic movements.

Scaf could act directly by restraining JNK signalling activity, behaving as an inducible antagonist of the pathway, or indirectly by affecting another event (i.e. another pathway). To test the former possibility, we used quantitative PCR on scaf RNAi embryos (pnr-GAL4,puc-lacZ>scafRNAi) and measured the level of puc-lacZ expression. Although no difference with the control was observed at 25°C, scaf loss-of-function at 29°C induced an increase of puc expression in the LE (Fig. 7C), indicating that JNK activity was upregulated. In conclusion, our work shows that scaf regulates the JNK pathway, acting in a negative-feedback loop during epithelial morphogenesis.

We isolated scaf in a genome-wide microarray screen aiming at identifying new JNK target genes in the Drosophila embryo (R.R., F. Carballès and S.N., unpublished data). Here, we showed that scaf is specifically induced by the JNK pathway in the LE cells that drive embryonic DC. To our knowledge, this is the third gene, after dpp and puc, showing this very specific site of expression under the control of JNK. In addition, we demonstrated that scaf expression is regulated by JNK in the A8 segment of the male genital disc. Scaf is also detected in the JNK-responsive cells (expressing puc-lacZ) of wing and leg discs (R.R., unpublished data), indicating that scaf can be considered as a general JNK target gene during epithelial morphogenesis. We determined that, like other proteins of the SP family, Scaf is a secreted SPH. Importantly, our results point to a negative role of scaf against JNK activity during DC in the embryo and during male terminalia formation in the pupa. Therefore, in addition to the inducible antagonist puc, the JNK pathway restrains its own activity during epithelial morphogenesis through the expression of the SPH gene scaf.

The amount of trypsin- and chymotrypsin-like SP enzymes in metazoan genomes is highly variable, from a dozen in Caenorhabditis elegans to more than two hundred in dipteran species like Drosophila melanogaster and about a hundred in mammals (Pils and Schultz, 2004). Gene expansion that occurred in flies is remarkable and concerns both the active enzymes and their homologues, indicating specific adaptations to environment and life strategies. Catalytic domains of SP enzymes share high structural similarities and substrate specificity is insured by a binding cleft located near the active site, as well as surface loops present in the SP domain (Perona and Craik, 1995; Perona et al., 1997). SPH proteins have lost their catalytic activities and it is assumed that they have adopted new functions based on novel binding capacities (Pils and Schultz, 2004). For instance, SPH could compete with an active SP for specific substrates.

Classical genetics has revealed the implication of various genes in DC (Harden, 2002; Noselli and Agnes, 1999) and other studies have identified several JNK target genes (Glise et al., 1995; Glise and Noselli, 1997; Homsy et al., 2006; Jasper et al., 2001; Munoz-Descalzo et al., 2005; Thomas et al., 2009). However, scaf is the first member of the SP family with a role in this epithelial sealing movement. Like for the Drosophila SPH gene mas (Murugasu-Oei et al., 1996; Murugasu-Oei et al., 1995), mutations in scaf did not show complete expressivity and penetrance in the embryo. One straightforward explanation is the presence of a redundant protein among the various other SPH of the Drosophila genome. Compensatory mechanisms, which are known to be important during DC (Hutson et al., 2003), could also hide the impact of scaf removal. Nevertheless, our results showed that scaf is important for morphogenesis by directly modulating the activity of the JNK pathway. How scaf negatively acts is still an open question, but it is probable that Scaf plays a role in the extracellular space, in contrast to the strong intracellular inhibitor Puc (see model in Fig. 7D). The signal that activates JNK signalling (and therefore DC) has not yet been discovered, preventing a straightforward analysis of Scaf molecular action. SP proteins are well-described for their role in activating signal transduction pathways or in degrading the extracellular matrix (ECM) (Lopez-Otin and Matrisian, 2007), and defects in muscle attachment observed in mas mutations might be due to a decrease in cell-matrix adhesion (Murugasu-Oei et al., 1995). Accordingly, possible roles of Scaf could be to act locally on an unknown extracellular signal or at the level of the putative receptor of the JNK pathway, through interaction with the ECM. Functioning in a negative-feedback loop, scaf would allow a fine tuning of signalling in JNK-active tissues.

The JNK signalling pathway is a major player of epithelial morphogenesis in many organs and developmental stages and its activity must be precisely controlled to coordinate tissue sealing. Two JNK target genes, puc and scaf, code for antagonists and therefore participate in this control. Whereas Puc acts intracellularly, Scaf is likely to operate in the extracellular space and represents the first secreted regulator of JNK-dependent epithelial sealing. Further studies of the role of scaf will highlight how a signal transduction pathway synchronizes collective cell behaviour in a three-dimensional environment, and how these cell rearrangements are controlled extracellularly to lead to a precise coordination of the movement.

We thank R. Mann, A. Brand, E. Sanchez-Herrero, N. Perrimon, K. Basler, the Bloomington Drosophila Stock Center (http://flystocks.bio.indiana.edu/bloomhome.htm), the Vienna Drosophila RNAi Center (VDRC; http://www.vdrc.at/) and the Exelixis Collection (https://drosophila.med.harvard.edu/) for fly stocks; the Drosophila Genomics Resource Center (DGRC; https://dgrc.cgb.indiana.edu/) for the scaf EST, as well as Developmental Studies Hybridoma Bank (DSHB; http://dshb.biology.uiowa.edu/) for antibodies. We are grateful to A. Petzoldt and J.-B. Coutelis for help in dissecting male genital discs and for dissections of adult male genital plates; to D. Cerezo for help in recombinant screening; to R. Delanoue for help in Q-PCR; to F. De Graeve and S. Schaub for help in statistical analyses; and to laboratory members for helpful discussions. Work in the S.N. laboratory is supported by CNRS, ANR, ARC, CEFIPRA, EMBO Yip and FRM.