Src kinases mediate the interaction of the apical determinant Bazooka/PAR3 with STAT92E and increase signalling efficiency in Drosophila ectodermal cells

Signalling pathways regulate development and homeostasis but their importance extends beyond normal physiological conditions, as mutations that ectopically activate or repress a signalling pathway are the cause of many human diseases. Protein tyrosine kinases are key elements of signal transduction in most pathways. In some cases, as in the EGF pathway, the transmembrane receptor itself has tyrosine kinase activity; in other cases, such as the JAK/STAT pathway, the intracellular domain of the receptor associates to a cytoplasmic tyrosine kinase.

The JAK/STAT pathway has been conserved during evolution, allowing Drosophila melanogaster to be used as a simplified model for signal transduction. In Drosophila, the pathway elements comprise only three ligands (Upd, Upd2 and Upd3), one homodimeric receptor (Dome), one JAK kinase (Hop) and one STAT transcription factor (STAT92E) (Arbouzova and Zeidler, 2006). This contrasts with the complexity found in mammals where multiple ligands and heterodimeric receptors, four JAK kinases and seven STAT proteins exist (Mohr et al., 2012). Despite millions of years of independent evolution, all studies show that the canonical signal transduction pathway is very similar in insects and mammals. In both systems, ligand binding to the receptor induces a change in conformation of the receptor complex that leads to the activation of its associated JAK kinases. These kinases transphosphorylate and phosphorylate the receptor on a tyrosine residue, creating docking sites for the inactive cytoplasmic STATs. STAT binding to the active receptor leads to its phosphorylation by JAK. Phospho-STAT proteins can homo- or heterodimerize through a conserved SH2 domain, allowing their stable translocation to the nucleus where they activate the transcription of target downstream genes (Mohr et al., 2012).

Despite this strong conservation, research in some model organisms has contributed information that still is not confirmed in others. For example, in mammals, where the pathway has been studied more extensively, it is known that unphosphorylated cytoplasmic STAT form dimers prior to their tyrosine phosphorylation. These STAT dimers have an antiparallel conformation different from the parallel conformation present in SH2 phospho-tyrosine dimers (Braunstein et al., 2003; Novak et al., 1998; Stancato et al., 1996). Moreover, in mammals, it has been shown that JAK is not the only kinase capable of STAT activation, but that Src tyrosine kinases can also activate STAT phosphorylation (Kazansky et al., 1999; Okutani et al., 2001; Silva, 2004). The Src family of non-receptor Tyr kinases comprises nine members in vertebrates, (reviewed by Thomas and Brugge, 1997), whereas in Drosophila only two members exist (Src42A and Src64B). The Src and STAT connection (Read et al., 2004; Silva and Shupnik, 2007), which has not been extensively studied in Drosophila, is especially interesting, as mutations in both proteins are linked to human cancer (Silva and Shupnik, 2007).

Studies in Drosophila have uncovered a strong relationship between ectoderm cell polarity and JAK/STAT signalling that have not yet been clearly established in vertebrate models (Sotillos et al., 2008). In Drosophila ectodermal tissues, the receptor, the JAK kinase and, surprisingly, the STAT transcription factor localize to the apical membrane. The apical polarization of the receptor and the JAK kinase affects the way the epithelium responds to the ligand, as cells can detect the ligands only when presented from the apical side but are unresponsive to ligands presented basally (Sotillos et al., 2008). In these ectodermal polarized tissues, the apical localization of STAT92E is functionally important, as it increases signalling efficiency, probably because the increased concentration on the apical side facilitates rapid translocation of the inactive cytoplasmic STAT92E to the activated receptor. It has been shown that the apical localization of STAT92E does not depend on DOME receptor binding or on pathway activation, but is due to STAT92E binding to the apical protein Bazooka (Baz, homolog of PAR3). Baz is a key component of the apical cell determining complexes in vertebrates and invertebrates, suggesting the possibility that JAK/STAT polarization may also occur in vertebrates (Hombría and Sotillos, 2008).

In this study, we have analysed the domains required for STAT92E and Baz interaction in Drosophila, and show that the Src tyrosine kinases are required for these two proteins to interact in vivo. This is the first time that a functional link between cell polarity complexes, the JAK/STAT signalling pathway and the Src kinases has been established in a whole organism. We also show data suggesting that, as in vertebrates, unphosphorylated cytoplasmic STAT92E could be forming dimers, extending the parallelism between the pathway in both lineages.

We used the following alleles: stat92E⁶³⁴⁶, stat92E³⁹⁷ (Brown et al., 2001; Hou et al., 1996; Silver and Montell, 2001); Src42A^26.1 and Src64B^P1 (Takahashi et al., 2005); Src64B^KO (O’Reilly et al., 2006) and STAT92E-GFP BAC [BAC CH321-73F24 P[acman] Genome Browser (Venken, 2009); a gift from Rebeca Spokony, modENCODE consortium]. This STAT92E-GFP BAC insertion is capable of rescuing the lethality and fertility of stat92E³⁹⁷/stat92E⁶³⁴⁶ or stat92E³⁹⁷/stat92E^83C9 flies.

Null embryos for STAT92E were generated by inducing germ line clones in hsFLP/+; FRT82B, stat92E⁶³⁴⁶/FRT82B, ovo^D1 females heat shocked for 1 hour at 37°C at second instar larva, mated to UAS-X; Klu-Gal4 stat92E⁶³⁴⁶/TM6B males (where UAS-X represents either UAS-stat92E-GFP; UAS-stat92E-Cterm-GFP; UAS-stat92E-Nterm-GFP; UAS-stat92E^Y711F-GFP; or a recombinant UAS-stat92E-Myc, UAS-stat92E-Nterm-GFP). Transgenic lines expressing UAS-baz (a gift from C. Doe, University of Oregon, USA), deletions of Baz (Krahn et al., 2010), UAS-Src42A^KM and UAS-Src42A^KY (Shindo et al., 2008), and UAS-STAT92E-GFP (Karsten et al., 2006) were used. We used 69B-Gal4, 24B-Gal4 and Klu-Gal4 as driver lines. The crb43.2-lacZ reporter gene has been described previously (Lovegrove et al., 2006) and ems0.35-lacZ is a fragment of ems1.2-lacZ (Jones and McGinnis, 1993).

Experiments in Fig. 5E were carried out by mating Src42A^26.1, UAS-STAT92E-GFP/CyO wg-lacZ females with Src64B^KO, 69B-Gal4/TM6B males. Those embryos expressing STAT92E-GFP were heterozygous mutant for Src42A^26.1 and Src64B^KO. For Fig. 5F, Src42A^26.1, UAS-STAT92E-GFP/CyO wg-lacZ females were mated to Src42A^26.1/CyO wg-lacZ; Src64B^KO, 69B-Gal4/TM6B males. As a result, embryos expressing STAT92E-GFP and not lacZ were homozygous mutant for Src42A^26.1 and heterozygous for Src64B^KO.

For Fig. 5H,I, Src42A^26.1, UAS-STAT92E-GFP/CyO wg-lacZ; Src64B^KO, UAS-baz/TM6B females were mated to Src42A^26.1/CyO wg-lacZ; 24B-Gal4/TM6B males. Embryos expressing lacZ and expressing STAT92E-GFP and Baz in the mesoderm are heterozygous mutant for Src42A^26.1 and Src64B^KO (Fig. 5H), whereas embryos expressing STAT92E-GFP and Baz in the mesoderm but not expressing lacZ are homozygous mutant embryos for Src42A^26.1 and heterozygous for Src64B^KO (Fig. 5I).

For Fig. 5J,K, Src42A^26.1, UAS-STAT92E-GFP/CyO wg-lacZ; UAS-baz, UAS-Src42A^KM/TM6B (J) or Src42A^26.1, UAS-STAT92E-GFP/CyO wg-lacZ; UAS-baz, UAS-Src42A^YF/TM6B (K) females were mated to Src42A^26.1/CyO wg-lacZ; Src64B^KO, 24B-Gal4/TM6B males. Embryos not expressing lacZ and expressing STAT92E-GFP and Baz in the mesoderm, are homozygous for Src42A^26.1 and heterozygous for Src64B^KO, and co-express the corresponding UAS-Src42A transgene.

For Fig. 6F,H,Src42A^26.1/CyO wg-lacZ; Src64B^KO/TM6B females were mated to Src42A^26.1/CyO wg-lacZ; crb-lacZ/TM6B males (Fig. 6F), or to Src42A^26.1/CyO wg-lacZ; ems-lacZ/TM6B (Fig. 6H). Embryos homozygous for Src42A^26.1 were identified by the absence of wg-lacZ staining pattern.

Fragments of STAT92E were amplified by PCR using High-Fidelity enzyme (Roche) using pBs-STAT92E as a template and the following primers: STAT92E Nterm 5′, ACGAATTCGAGCATGAGCTTGTGG; STAT92E Nterm 3′, ACGGATCCAGGTTCTTCTGCGCG; STAT92E Coiled-coil domain 5′, ACGAATTCATGCCGCTAAATCCC; STAT92E Coiled-coil domain 3′, ACGGATCCAGGTTCTTCTGCGCG; STAT92E DNA-binding domain 5′, CAGAATTCATGGCCTTTATCG; STAT92E DNA-binding domain 3′, ATGGGATCCATGTGATCTTTGG; STAT92E Cterm 5′, CAGAATTCATGGCCTTTATCG; STAT92E Cterm 3′, GTGAACAGCTCCTCGCCC; STAT92E CtermΔTAD 5′, CAGAATTCATGGCCTTTATCG; STAT92E CtermΔTAD 3′, CTGGATCCCTCTTCACGGGAC; STAT92E CtermΔ725-761 5′, CAGAATTCATGGCCTTTATCG; STAT92E CtermΔ725-761 3′, TCGGATCCTCTCCGTTTCTAC

After sequencing, the cDNAs were subcloned into the EcoRI-BamHI sites of pUAST-GFP. To generate the UAS-STAT92E-Myc the STAT92E cDNA was cloned in EcoRI in pUAST-Myc (Sotillos et al., 2008).

Embryos were fixed in 1:1 formaldehyde 4% in PBS:n-heptane for 20 minutes at room temperature and stained with the following primary antibodies: anti-Crb, anti-α-Spectrin, anti-Neurotactin (Developmental Studies Hybridoma Bank), anti-Myc (Cell Signaling), anti-Scrib (a gift from C. Doe, University of Oregon, USA), anti-Baz (a gift from A. Wodarz, GZMB, Göttingen, Germany), anti-Src42A (a gift from M. Takahashi, University of Tokyo, Japan) anti-aPKC C20 (Santa Cruz Biotechnology) and anti-βGAL (Promega). Secondary antibodies were coupled to Alexa488, Alexa555 or Alexa648 (Molecular Probes).

Cells were fixed in 4% formaldehyde, phosphate buffer (pH 7.4). Primary antibodies were mouse anti-Myc (9E10, Developmental Studies Hybridoma Bank, 1:100) and rat anti-HA (Roche #11867423001, 1:1000).

DNA was stained with DAPI (Molecular Probes). We used Cy3-conjugated (Jackson Laboratories) or Alexa647-conjugated (Invitrogen) secondary antibodies. Images were taken on a Zeiss LSM 510 Meta confocal microscope and processed using Adobe Photoshop.

Cuticle preparations were carried out as described previously (Hombría et al., 2005). The frequency of segmentation and spiracle defects observed in the late embryo cuticles presented in Fig. 6 are as follows: stat92E⁶³⁴⁶/+ females mated to Oregon R wild-type males (0%, n=86). Src42A^26.1/+ females mated to Src64B^P1/+ males (0%, n=145). Src42A^26.1/+; stat⁶³⁴⁶/+ females mated to Oregon R males (30%, n=92). Src42A^26.1/+; stat⁶³⁴⁶/+ females mated to Src64B^P1/+ males (30%, n=76).

For immunoprecipitation, 1 μg of mouse anti-GFP (Roche), mouse anti-βGAL (Promega), rabbit anti-βGAL (Cappel) or 5 μl of rabbit anti-Baz were added to UAS-STAT92EGFP/+; 69B-Gal4/UAS-baz embryonic extracts containing 3 mg of total protein. TNT buffer [1% Triton X-100, 150 mM NaCl, Tris-Cl (pH 7.5)] was used for immunoprecipitation with anti-Baz, and STAT buffer (0.1% Triton X-100, 300 mM NaCl, 50 mM Tris-Cl pH 7.5, 5 mM EDTA) for immunoprecipitation with anti-GFP. Buffers were supplemented with protease inhibitors (Roche), β-glycerophosphate and 20 mg/ml sodium orthovanadate. Antibody conjugation with protein A/G Plus-Agarose (Santa Cruz) was carried out overnight at 4°C. Embryos were lysed in the proper buffer and centrifuged for 30 minutes at maximum speed at 4°C to discard debris. Total protein (2 mg) was used in each assay and preabsorbed for 30 minutes with protein A-Agarose (Santa Cruz). Before incubation with the antibody, 5 μl of each assay was removed as loading control (input). Immunoprecipitations were performed at 4°C for 4 hours. Immune complexes were washed four times in washing buffer [300 mM NaCl, 50 mM Tris-Cl (pH 7.5), 5 mM EDTA and 0.1% Triton X-100] and once more in 50 mM Tris-Cl (pH 6.8) or PBS, boiled in 2×SDS sample buffer before SDS-PAGE and western blot.

To perform pull-down experiments, GST fusion proteins were obtained from subclones in pGEX4T1 (Amersham) of Baz^1-317 or Baz^1048-1464 (Krahn et al., 2010). Purification was carried out using standard methods. Drosophila protein extract (300 μg) in TNT buffer or 30 μl of ³⁵S -labelled STAT92E protein in STAT buffer were incubated for 5 hours at 4°C with 30 μg of Glutathione-S-transferase (GST) or with the different GST-fusion proteins bound to glutathione beads as described previously (Sotillos et al., 2004). Complexes were washed as immunoprecipitations. Western blotting was carried out according to standard procedures using rabbit anti-Baz and mouse anti-GFP (Roche) at 1:2000. When using ³⁵S-labelled STAT92E after electrophoresis, the gel was dried and detected by phosphorimager method.

Schneider 2R+ cells were transfected with the following constructs: Poly-Ubiquitin::Baz-GFP, Poly-Ubiquitin::Src42A-HA and Poly-Ubiquitin::STAT92E-myc. Cells were harvested 3 days after transfection and lysed in lysis buffer [1% Triton X-100, 150 mM NaCl, 50 mM Tris-Cl (pH 7.5)] supplemented with protease inhibitors. After centrifugation, 2 μl of rabbit anti-GFP (Molecular Probes #A11122) were added to the cell lysate. Immune complexes were harvested using protein A-conjugated agarose (Roche), washed five times in lysis buffer and boiled in 2×SDS sample buffer before SDS-PAGE and western blot. Western blot was performed using mouse anti-Myc (9E10, Developmental Studies Hybridoma Bank, 1:200), mouse anti-HA (Roche #11 583 816 001, 1:1000) or rabbit anti-GFP (Molecular Probes #A11122, 1:1000).

All vertebrate STAT proteins share a conserved domain structure whose function has been well studied in mammals. The domains include: a non-conserved N-terminal region that varies in different STATs, a coiled-coil, a DNA-binding, a linker, an SH2 and a C terminus trans activation domain (Fig. 1A) (Becker et al., 1998; Chen et al., 2002; Vinkemeier et al., 1998). The N-terminal domain is required for the formation of tetramers, for tyrosine dephosphorylation, for transcriptional activation and for protein-protein interactions (Chang et al., 2003; Chen et al., 2002; Murphy et al., 2000; Ota et al., 2004; Shuai et al., 1996; Vinkemeier et al., 1998; Xu et al., 1996). The helical coiled-coil domain mediates interactions with several proteins, including Jun (Zhang et al., 1999) and is also required for the formation of non-phosphorylated dimers (Mao et al., 2005; Neculai et al., 2005). The DNA-binding domain (DB) has limited contact with both the major and minor grooves of DNA (Chen et al., 1998). The linker domain modulates the rate of STAT:DNA interactions, ultimately controlling transcriptional activation of STAT target genes (Yang et al., 2002). The SH2 domain is required for the formation of an activated STAT dimer by mediating reciprocal interactions with a phosphorylated conserved tyrosine residue located around amino acid position 700 that exists in all STAT proteins (Chen et al., 1998; Levy and Darnell, 2002). Last, the C-terminal transactivation domain (TA) is required for transcriptional co-activation of mammalian STATs (Horvath, 2000).

The Drosophila STAT92E protein can be similarly subdivided (Fig. 1A) with the N-terminal domain spanning amino acid residues 1-139, the helical coiled-coil domain spanning residues 140-340, the DNA-binding domain spanning residues 340 to 588, the SH2 domain spanning residues 594-659 and the terminal TA domain spanning residues 659-761 (Hou et al., 1996; Yan et al., 1996). As the requirement of these domains has not been defined in relation to the apical localization of STAT92E, we analysed this specific aspect in the different fragments.

Using the Gal4/UAS (Brand and Perrimon, 1993) system, we expressed in the embryonic ectoderm different STAT92E fragments fused to GFP and analysed their membrane localization (Fig. 1; for comparison of expression levels, see also supplementary material Fig. S1). The full-length protein localizes to the subapical membrane of the cells (Fig. 1B,B′) as was previously described for the endogenous protein (Sotillos et al., 2008). Similar results were obtained with a STAT92E-GFP fusion under the control of the genomic region (STAT92E-GFP BAC, supplementary material Fig. S2) capable of rescuing STAT92E mutants to adulthood Deletion of the N-terminal non-conserved domain does not affect membrane localization (STAT92E-ΔN, Fig. 1C,C′). Constructs containing the coiled-coil domain plus the N-terminal domain (STAT92E-Nterm, Fig. 1D,D′), or the coiled-coil domain on its own (STAT92E-CC, not shown) localize to the membrane, although at much lower levels than the full-length STAT92E. By contrast, the DNA-binding domain shows no membrane localization (STAT92E-DB, Fig. 1E,E′). The C-terminal half of STAT92E, including the DNA-binding, the SH2 and the TA domains, also localizes to the membrane in ectodermal cells (STAT92E-Cterm, Fig. 1F,F′). These results indicate that both the N-and the C-terminal part of the protein have the potential to translocate STAT92E to the apical cell membrane.

In vertebrates, it has been described that non-phosphorylated STATs can form antiparallel dimers (Braunstein et al., 2003; Novak et al., 1998; Stancato et al., 1996). To test whether dimerization with the endogenous STAT92E could be the cause of the membrane localization of some of our constructs, we studied their distribution in embryos completely lacking the endogenous STAT92E protein. For that purpose, we studied the progeny of STAT92E mutants originating from STAT92E germ line clones to get rid of both the maternal and zygotic STAT92E protein. In these STAT92E-null embryos, tagged full-length STAT92E (Fig. 2A) and the STAT92E C-terminal half (Fig. 2C) can still localize to the membrane. By contrast, the N-terminal half loses its faint apical localization in STAT92E-null embryos (Fig. 2B), suggesting that its apical membrane localization in wild-type embryos is due to interaction with the endogenous STAT92E protein. Moreover, in STAT92E-null embryos STAT92E N-terminal localization to the membrane is restored when co-expressed with a full-length Myc tagged STAT92E (Fig. 2D). As the N-terminal half does not contain the SH2 domain, these results indicate that the Drosophila cytoplasmic STAT92E protein dimerizes prior to tyrosine 711 phosphorylation. However, although we cannot ignore the possibility that another protein is acting as a bridge between both STAT92E fragments, considering that vertebrate unphosphorylated STAT proteins dimerize, dimerization with the endogenous apically localized STAT92E is most probably responsible for the weak polarization observed in the N-terminal fragments (Fig. 1D).

These data also suggest that the C-terminal half of STAT92E binds to apically localized factors. As our previous work demonstrated that Baz is required for STAT92E membrane localization (Sotillos et al., 2008), we studied the ability of different STAT92E fragments to co-immunoprecipitate with Baz. Only STAT92E full-length and the STAT92E C-terminal half co-precipitate with Baz from embryonic extracts (Fig. 3A,B).

The C-terminal half of STAT92E contains the DB domain, the linker region, the SH2 domain and the TA domain. When the TA domain is removed from the STAT92E C-terminal half, the resulting fragment loses the membrane localization (Fig. 1G,G′ and supplementary material Fig. S3) and most of its capability of Baz binding (Fig. 3C). These results indicate that the TA domain is required for the interaction with Baz and the apical membrane localization of STAT92E.

Baz is a conserved scaffold protein involved in the establishment of apico-basal polarity in many organisms (Suzuki and Ohno, 2006). Baz contains three conserved regions (CR) (Fig. 4B) with established functions (Benton and St Johnston, 2003a; Izumi et al., 1998; Joberty et al., 2000; Lin et al., 2000; Morais-de-Sá et al., 2010; von Stein et al., 2005). The N-terminal CR1 is required for homodimerization; the CR2 includes three PDZ domains and is involved in the interaction with PAR6; the C-terminal CR3 interacts with aPKC. Apart from these, Baz contains two phosphorylation sites for PAR1, which serve as binding sites for 14-3-3 adaptor proteins, and the C-terminal region interacts with membrane lipids and mediates direct interaction with the plasma membrane (Benton and St Johnston, 2003b; Krahn et al., 2010). In the wild-type embryos, Baz is not expressed in the mesoderm and STAT92E does not localize to the membrane cortex in this tissue. However, we have shown (Sotillos et al., 2008) that STAT92E is translocated towards the membrane upon ectopic expression of Baz in the mesoderm (compare Fig. 4A,C and supplementary material Fig. S3A,B). To find out which region of Baz is required for STAT92E translocation to the cortex, we expressed different fragments of the Baz protein fused to GFP in mesodermal cells and observed which fragments could affect the localization of a full-length STAT92E-Myc-tagged protein. We find that the PDZ and the aPKC-binding domains are dispensable for STAT92E membrane translocation (Fig. 4E,F). However, when we remove either the most N-terminal (amino acids 1-317) or C-terminal regions (a complete deletion of the C-terminal region, amino acids 1001-1464; or a smaller deletion, amino acids 1221-1464) of the protein, Baz cannot translocate STAT92E towards the membrane (Fig. 4D,G,H). To confirm these results, we performed pull-down assays and found that both the N- (1-317) and the C-terminal (1048-1464) regions of Baz can co-precipitate with STAT92E from embryonic extracts (Fig. 3D). We obtained the same results with in vitro pull-down assays expressing STAT92E in rabbit reticulocyte extracts (Fig. 3E). Thus, our results point to a function of both, the N- and the C-terminal region of Baz in STAT92E membrane localization.

To test whether Baz is sufficient for STAT92E membrane localization, we co-expressed both proteins in S2R+ cell lines. Surprisingly, in these ex vivo conditions we were unable to see STAT92E membrane translocation (Fig. 5A; 82% of cells show cytoplasmic STAT92E and only 18% of cells show cortical STAT92E, n=50). As we had observed that Baz and STAT92E in vivo co-precipitation is detected only in the presence of phosphatase inhibitors (this work) and we had previously shown that JAK phosphorylation is dispensable for STAT92E-Baz interaction [STAT92E still localizes to the membrane in null mutants for JAK or when the Tyr phosphorylated by JAK is mutated (Sotillos et al., 2008)], we suspected that other kinases might be involved in this process. Because members of the Src kinase family have been previously related with JAK/STAT signalling in vertebrates and invertebrates (Read et al., 2004; Silva and Shupnik, 2007), we investigated their possible involvement in this process. We observe that expression of Src42A or Src64B did not affect STAT cytoplasmic localization (Fig. 5B, 72% of cells show cytoplasmic STAT92E). However, when we express either Src42A or Src64B in the presence of Baz, this leads to STAT92E cortical translocation in S2R+ cells (Fig. 5C, 86% of cells show STAT92E cortical colocalization and only in 14% STAT92E remains in the cytoplasm). Moreover, we were able to co-immunoprecipitate Src42A with Baz from transfected S2R+ cells (Fig. 5D).

In wild-type ectodermal cells, STAT92E and Src42A colocalize at the apical junctional region. Moreover, in mesodermal cells, Src42A also colocalizes with STAT92E after Baz overexpression (supplementary material Fig. S5). To confirm that Src is required for STAT92E cortex localization, we studied the epidermal membrane distribution of STAT92E in Src mutants, and the ability of Baz to translocate STAT92E to the membrane in mesodermal cells with depleted Src levels. In Src42A/+; Src64B/+ double heterozygous ectodermal cells, STAT92E-GFP membrane localization is not noticeably affected (Fig. 5E). However, in Src42A/Src42A; Src64B/+ embryos, there is a reduction of STAT92E at the membrane without any effect on Baz localization (compare Fig. 5E,F, and supplementary material Fig. S6). The requirement of Src for STAT92E cortical membrane localization is observed more clearly in the pharynx, where there is a gradual reduction of STAT92E membrane localization that correlates with the number of functional Src alleles in the embryo: high levels when Baz is expressed in the wild type; intermediate levels when expressed in double heterozygous Src42A/+; Src64B/+ embryos; and residual STAT92E localization when expressed in Src42A/Src42A; Src64B/+ embryos (compare Fig. 5G with 5H,I, and supplementary material Fig. S6). These experiments show that Src is required for Baz-mediated recruitment of STAT92E.

To test whether the abnormal STAT92E membrane localization in Src mutants correlates with a decrease of JAK/STAT signalling in vivo, we examined whether Src mutations increase the phenotype of a stat92E partial loss of function. During embryogenesis, STAT92E is required for head involution, segmentation and posterior spiracle formation. Heterozygous stat92E/+ mutant embryos do not present any defects (Fig. 6A) but provide a sensitized background to screen for mutations affecting JAK/STAT signalling efficiency (Sotillos et al., 2008). Similarly, heterozygous embryos for Src42A/+ or Src64B/+, or double heterozygous Src42A/+; Src64B/+ have normal cuticles (not shown and Fig. 6B). However, Src42A/+; stat92E/+ heterozygous embryos present defects in head formation, segmentation and posterior spiracle development (Fig. 6C), and these phenotypes are enhanced in Src42A/+; Src64B/+; stat92E/+ embryos (Fig. 6D). We also studied how the levels of Src affected the expression of the spiracle crumbs enhancer crb43.2, a direct target of the JAK/STAT pathway (Fig. 6E) (Lovegrove et al., 2006). The levels of activation of crb43.2 are severely reduced in Src42A/Src42A; Src64B/+ embryos (Fig. 6F), whereas the expression of ems0.35, another spiracle enhancer activated independently of the JAK/STAT pathway is unaffected (compare Fig. 6I with 6H). In parallel to the observed reduction of JAK/STAT signalling, these Src mutant embryos show altered spiracle morphology (Fig. 6G).

Data from vertebrates have shown that Src kinases are able to activate STAT independently of JAK function by direct phosphorylation of the same conserved Tyr involved in STAT dimerization (Okutani et al., 2001; Silva, 2004). We excluded the possibility that Src phosphorylation of this tyrosine is responsible for STAT92E membrane translocation, as we can detect the mutated STAT92E-Y711F at the membrane cortex of ectodermal cells in STAT92E-null embryos (Fig. 5L). This agrees with our previous observation that STAT92E membrane localization also occurs in cells where the pathway is not active (Sotillos et al., 2008). To test the requirement of the kinase activity of Src to mediate STAT92E-Baz interaction, we studied the rescue capability of a Src kinase dead (Src42A^KM) versus a constitutively activated Src42A (Src42A^KY) in mesodermal cells. Both forms were able to rescue to the same extent (Fig. 5J,K and supplementary material Fig. S6), indicating that the kinase activity of Src42A is not required for this interaction.

Thus, our data uncover a novel function for the Src kinases in JAK/STAT signalling where Src regulates a Baz-STAT92E membrane interaction required for efficient signal transduction.

Maintenance of cell polarity is crucial for the correct cellular function. Not only most of the cellular processes are compartmentalized but the cell membrane polarity also affects signal transduction in several pathways. Here we uncover the relationship between Baz, a key molecule involved in the establishment and maintenance of cell polarity; STAT92E, a transcription factor of a major signalling pathway; and the Src kinases, which play a role in signal transduction and cell polarity. All three proteins have been associated with different kinds of tumours (Huang and Muthuswamy, 2010; Silva and Shupnik, 2007) including epithelial tissue carcinomas and breast cancer. However, this is the first time that these three proteins have been found to function together in a cellular process in the whole organism and a common mechanism of action is proposed.

In vertebrates, there is strong evidence showing that, besides JAK, the Src kinases can activate STAT signalling. First, STAT3 and STAT5 are crucial downstream factors in Src-induced transformation (Bromberg et al., 1998; Kloth et al., 2003). Second, Src kinase activation has been shown to result in STAT tyrosine phosphorylation (Okutani et al., 2001; Silva, 2004). By contrast, in Drosophila the available data suggest that the involvement of Src in STAT92E activation is probably marginal. This is shown by the fact that in Drosophila the phenotypes of null stat92E alleles are very similar to those of mutants in which the function of the JAK kinase, the receptor or of all ligands, is abolished, indicating that STAT92E activation via alternative kinases occurs in a minority of tissues. One of the few cases reported of JAK-independent STAT92E activation occurs during the proliferation and migration of the embryo pole cells where STAT92E is activated by the Ras/Raf pathway downstream of the Torso tyrosine kinase receptor (Li et al., 2003). The study of mutations in the C-terminal Src kinase (Csk), a negative regulator of Src signalling, offers indirect evidence to suggest that Src may activate STAT92E in the Drosophila eye. Eyes that lack Csk are larger than normal, a phenotype also observed when the canonical JAK/STAT pathway is ectopically activated during eye development. Moreover, Csk clones exhibit higher levels of STAT92E expression, which has been interpreted as being due to the induction of a STAT92E positive-feedback loop (Read et al., 2004). However, these results do not clarify whether Src-induced STAT92E activation is due to direct STAT92E phosphorylation or caused by indirect regulation.

In the ectoderm, where the JAK/STAT pathway is highly polarized, we have shown that efficient signalling requires STAT92E apical localization achieved through interaction with the membrane-associated apical polarity protein Baz (Sotillos et al., 2008). Here, we have obtained several pieces of evidence showing that Src is also required for the correct STAT92E membrane localization and signalling (Figs 5, 6). First, although heterozygous stat92E or Src mutant embryos are normal, double heterozygous stat92E and Src mutant embryos display phenotypes that resemble a partial JAK/STAT signalling failure. Second, the expression of a direct target of STAT92E is downregulated in a Src mutant background. Third, decrease of Src gene activity affects STAT92E localization to the membrane of epithelial cells or to the membrane of mesodermal cells expressing Baz. Fourth, in S2R+ cells STAT92E co-localizes with Baz only when co-expressed with Src42A or Src64B. Our data also show that tyrosine 711 is not required for membrane localization, demonstrating that this function of Src is independent of STAT92E activation.

In our experiments, although we are able to co-precipitate Baz with STAT92E and Baz with Src42A, we could not co-precipitate Src42A with STAT92E. This suggests that the interaction between Src42A and STAT92E is either too labile to be detected or that Src modifies Baz and that this allows the recruitment of STAT92E to Baz. Our data also suggest that the kinase activity of Src42A is dispensable for STAT92E-Baz interactions, as a kinase-dead isoform of Src42A is able to rescue membrane localization in Src mutant backgrounds to the same level that a constitutively activated form can. Although we cannot discard a direct activation of STAT92E in response to Src and growth factors in the minority of cell types, our genetic interactions do not reveal any phenotype apart from those of the canonical JAK/STAT pathway.

In addition to the parallel dimers formed by SH2 interaction with phosphor-Tyrosine in activated STATs (Mohr et al., 2012), vertebrate STAT proteins have been shown to form ‘inactive’ antiparallel dimers mediated by the region that includes the N-terminal and the coiled-coil domains (Mao et al., 2005; Neculai et al., 2005; Ota et al., 2004). In Drosophila, active phosphorylated STAT92E also forms parallel homodimers through an SH2-phosphotyrosine interaction. Our study suggests that STAT92E may also form inactive homodimers mediated through the N-terminal coiled-coil domain (Fig. 2). Thus, formation of STAT92E dimers prior to pathway activation could be an ancestral STAT characteristic, reinforcing Drosophila as a model for studying vertebrate STAT signalling.

We have previously described the requirement of a STAT92E-Baz interaction for efficient JAK/STAT signalling (Sotillos et al., 2008). Now, we have uncovered the domains in both molecules involved in this physical interaction. In the case of STAT92E we have narrowed down the region to the transactivation (TA) domain of the protein: only the most C-terminal part of the molecule is able to localize to the apical membrane cortex on its own (Fig. 2C) and to co-precipitate with Baz (Fig. 3A-C). Moreover, when the TA domain is removed, the C-terminal domain is unable to localize to the cell membrane where Baz is located (Fig. 1G; supplementary material Fig. S3) and reduces its ability to co-precipitate with Baz (Fig. 3C), probably owing to the loss of binding to the N-terminal part of Baz (supplementary material Fig. S4). However, this construct is still able to bind to the C-terminal part of Baz (supplementary material Fig. S4), indicating that another domain in this fragment is also involved in this interaction. In vertebrates, the TA domain has been shown to be crucial for the regulation of the activity of STAT through the interaction with several proteins (Decker and Kovarik, 1999; Horvath, 2000). Here, we have added a new function to this domain as a mediator of the interaction of STAT92E with Baz.

Baz is a scaffolding protein that is able to interact with various proteins and lipids through different regions. We have shown that interaction with STAT92E requires both the Baz N-terminal region (1-317) that includes the oligomerization domain and the C-terminal region (1048-1464) that includes the phosphatidylinositol-binding site (Fig. 3D,E; Fig. 4). Given that both Baz N- and C-terminal domains are conserved, and that STAT92E TA domain is conserved in vertebrate STATs, it is possible that the PAR3-STAT interaction is a conserved feature of JAK/STAT signalling. Our S2 cell and embryo experiments show a requirement of Src in the STAT92E-Baz interaction. Paradoxically, we are able to precipitate STAT92E in the absence of Src using the N- and the C-terminal fragments of Baz in vitro. As in vitro we are using Baz fragments, which most probably have a different conformation from the full-length protein, this paradox can be explained if, in vivo, Src function was required to change the conformation of Baz or to displace another protein that interferes with the binding, events that would not take place in the in vitro binding.

In summary, our results show that Src42A and Src64B are required redundantly in the ectoderm to allow STAT92E to bind to Baz. This interaction leads to the priming of JAK/STAT signalling by concentrating inactive STAT92E dimers apically near the polarized receptor kinase complex, contributing in this way to the efficient canonical JAK/STAT signalling. Although we cannot completely discard that future studies in Drosophila may find some specific cell type in which direct STAT92E activation by Src kinases exists, our data indicate that, in general, there is no direct STAT92E activation by Src in Drosophila. We would like to speculate that the existence of complexes where STAT, PAR3 and Src interact might have allowed the evolution of STAT-activation shortcuts that in vertebrates would have led to Src directly phosphorylating STAT in tyrosine 700 without the intervention of JAK or the canonical receptors. Considering the relevance of these proteins in development and disease, future studies should address whether apical STAT localization through PAR3-Src activity is also functioning in the vertebrate lineage.

We thank Erika Bach, Nicolas Tapon, Alana O’Reilly, Rebecca Spokony, Shigeo Hayashi and Tetsuya Kojima for reagents; and Andreas Wodarz, Joaquim Culí, Martin Zeidler and Sonsoles Campuzano for suggestions to the manuscript.