A targeted gain-of-function screen identifies genes affecting salivary gland. *Department of Physiology, Development and Neuroscience, University of

Transcription

1 Genetics: Published Articles Ahead of Print, published on December 8, 2008 as /genetics A targeted gain-of-function screen identifies genes affecting salivary gland morphogenesis/tubulogenesis in Drosophila Vanessa Maybeck* 1 and Katja Röper* *Department of Physiology, Development and Neuroscience, University of Cambridge, Cambridge CB2 3DY, UK 1 current address: Institute of Neuroscience and Biophysics, Molecular Biophysics (INB-2), Research Center Jülich, D Jülich, Germany 1

2 Running title: Salivary gland morphogenesis in Drosophila Keywords: morphogenesis, salivary glands, tubulogenesis, cell shape, cytoskeleton Author for correspondence: Katja Röper, Department of Physiology, Development and Neuroscience, University of Cambridge, Downing Street, Cambridge CB2 3DY, UK Phone: ++44 (0) Fax: ++44 (0)

3 ABSTRACT During development individual cells in tissues undergo complex cell shape changes to drive the morphogenetic movements required to form tissues. Cell shape is determined by the cytoskeleton and cell shape changes critically depend on a tight spatial and temporal control of cytoskeletal behaviour. We have used the formation of the salivary glands in the Drosophila embryo, a process of tubulogenesis, as an assay to identify factors that impinge on cell shape and the cytoskeleton. To this end we have performed a gain-of-function screen in the salivary glands, using a collection of fly lines carrying EP-element insertions that allow the overexpression of downstream-located genes using the UAS-Gal4 system. We used a salivary gland specific fork head-gal4 line to restrict expression to the salivary glands, in combination with reporters of cell shape and the cytoskeleton. We identified a number of genes known to affect salivary gland formation, confirming the effectiveness of the screen. In addition we found many genes not implicated previously in this process, some having known functions in other tissues. We report the initial characterization of a subset of genes, including chickadee, rhomboid1, egalitarian, bitesize, and capricious, through comparison of gain and loss-of-function phenotypes. 3

4 INTRODUCTION During development and organogenesis most tissues arise from layers of epithelial cells that reorganize through complex morphogenetic movements. Many adult organs consist of tubular arrangements of epithelial sheets, and these tubules form during development through a process called tubulogenesis. There are a number of ways to generate tubules (LUBARSKY and KRASNOW 2003). One important process is the direct conversion of epithelial sheets into tubules through wrapping (COLAS and SCHOENWOLF 2001) or budding (HOGAN and KOLODZIEJ 2002). Cells undergoing tubulogenesis change their shapes drastically, from a cuboidal or columnar epithelial shape to a wedge-shape or conical, and then back to a more columnar epithelial shape once positioned inside the tube. Cell shape is determined by the intracellular cytoskeleton, primarily actin and microtubules. The cytoskeleton is closely coupled to cell-cell adhesion as well as adhesion to the extracellular matrix. We are interested in understanding how the cytoskeleton and thus cell shape is regulated and coordinated during tubulogenesis. We chose to perform a gain-of-function screen rather than a mutagenesisbased loss-of-function screen as phenotypes observed in the latter might be subtle and thus missed or phenotypes in a given tissue might be obscured by disruption of other tissues and many genes might also have redundant functions. In contrast the gain-of-function/overexpression approach allows a particular tissue and gene to be targeted, and many such screens have been successfully conducted in the past (for examples see: BEJARANO et al. 2008; MOLNAR et al. 2006; RØRTH et al. 1998). The screen presented here uses the formation of the salivary glands in the Drosophila embryo as an assay system. The screen is based on a collection of transposable 4

5 elements (EP elements) generated by Rørth (RØRTH et al. 1998) that contain UAS sites that respond to the yeast transcription factor Gal4 followed by a promoter directing expression, when activated, of genes located downstream 3 of the EP insertion site. If combined through crosses with a tissue specific source of Gal4 (HENDERSON and ANDREW 2000; ZHOU et al. 2001) overexpression (and in some cases antisense expression) of a downstream gene will be activated only in the target tissue, in our case the embryonic salivary glands in the Drosophila embryo. Salivary gland formation in Drosophila is probably the simplest form of tubulogenesis (LUBARSKY and KRASNOW 2003). A patch of about a two hundred cells in the ventral epidermis of the embryo within parasegment two becomes specified to form a salivary gland primordium, the placode, with a hundred cells on either side of the embryo. This fate determination occurs through a combination of the activities of the homeotic genes sex combs reduced (scr), extradenticle (exd), homothorax (hth), and dorsal signalling by decapentaplegic (dpp) (HENDERSON and ANDREW 2000; HENDERSON et al. 1999; PANZER et al. 1992). Without scr, exd, hth function, no salivary glands form. Different subpopulations of cells are found in the invaginated gland, such as the secretory cells, and the common and individual duct cells. Their distinction depends on EGF signalling from the ventral midline (HABERMAN et al. 2003; KUO et al. 1996). Once the cells have been specified at stage 10 of embryogenesis no further cell division occurs within the primordium, and no cells are lost through apoptosis (BATE and MARTINEZ ARIAS 1993; CAMPOS-ORTEGA and HARTENSTEIN 1985; MYAT and ANDREW 2000a). Invagination initiates in the dorsal posterior corner of the primordium, with all future secretory cells invaginating in a precise order, followed by invagination of the duct cells and formation of the ducts 5

6 (MYAT and ANDREW 2000b). A key gene essential for the invagination is fork head (fkh). Fkh is a winged-helix transcription factor, and in its absence all of the cells fated to form the glands remain on the surface of the embryo as they fail to undergo apical constriction (MYAT and ANDREW 2000a). Once inside the embryo, the glands have to navigate their way through the surrounding tissues including the visceral mesoderm and CNS to reach their extended final position parallel to the midline and anterio-posterior axis. They are guided by cues from the surrounding tissues (HARRIS and BECKENDORF 2007; HARRIS et al. 2007; KOLESNIKOV and BECKENDORF 2005). Also, after initially invaginating in a posterior-dorsal direction, the glands turn and further extend into the embryo in a direction parallel to the anterio-posterior embryonic axis, in a process dependent on integrins and downstream signals (BRADLEY et al. 2003; VINING et al. 2005). A few factors have previously been identified that impinge on the cytoskeleton and cell shape during salivary gland morphogenesis. The actin cytoskeleton is modified through proteins such as Btk29/Tec29 in conjunction with Chickadee (CHANDRASEKARAN and BECKENDORF 2005). Small GTPases such as Rac and Rho affect the invagination of the glands (PIRRAGLIA et al. 2006; XU et al. 2008). Crumbs and Klarsicht affect the delivery of apical membrane and thus cell shape at late stages of morphogenesis (MYAT and ANDREW 2002). Nonetheless, how these factors work together throughout the whole process of invagination is still not clear and it is likely that many others remain to be identified. We have performed a gain-of-function/overexpression screen in the salivary glands with the aim to firstly identify more genes that are required for salivary gland tubulogenesis (and thus potentially also for tubulogenesis in general) and to 6

7 secondly use this system as an assay for factors affecting cytoskeleton and thus cell shape in general. The first aim assumes that genes that have a function in the morphogenesis of the glands and are endogenously expressed in the glands might perturb their invagination if overexpressed and if levels of expression are important. The second aim hypothesizes that overexpression of genes not endogenously expressed in the glands but important for cell shape coordination in other tissues will lead to identifiable phenotypes in this screen, as defects in cell shape changes resulting from the overexpression will affect the proper invagination of the glands. The orientation of some of the EP elements is also likely to lead to (over)expression of an antisense RNA, thus potentially inducing a tissue-specific loss-of-function effect. We identified seven genes that have previously been implicated in salivary gland morphogenesis or function confirming the effectiveness of the screen, and also 44 insertions that uncover genes with potentially novel roles in the salivary glands or functions in the regulation of cell shape and the cytoskeleton in other tissues. Of these genes 14 are previously uncharacterized genes. A selection of the genes that fall into the three categories discussed above (i.e. overexpression of a gene with a function in the glands, overexpression of a gene not expressed in the glands, revealing a function in cell shape coordination in other tissues, and loss-offunction of a gene through tissue-specific antisense RNA expression) recovered in the screen is examined in more detail below, including bitesize, egalitarian, chickadee, capricious and rhomboid1. 7

8 RESULTS Experimental design of the gain-of-function screen In order to address how the cytoskeleton and cell shape is regulated during such a process of tubulogenesis, we performed a gain-of-function screen in the salivary glands of the Drosophila embryo. We used a gland-specific Gal4-driver, fkhgal4 (HENDERSON and ANDREW 2000; ZHOU et al. 2001), to drive expression of either a marker of the microtubule cytoskeleton, GFP-EFGas2 (SUBRAMANIAN et al. 2003), or a marker of cell shape, SrcGFP (KALTSCHMIDT et al. 2000) in the glands only. Flies carrying these marker chromosomes (GFP-EFGas2 or SrcGFP maker plus fkhgal4; marker line ) were crossed to a collection of EP-element lines containing UASelements, leading to the expression of gene X located 3 downstream of the EPelement insertion site. We drove expression from 1001 EP elements specifically in the salivary glands and screened for any apparent problems in their morphogenesis (see Fig. 1 for wild-type morphogenesis and marker expression, and Fig.2A for a scheme explaining the screen set-up). It has previously been shown that the proper invagination and positioning of the salivary glands depends on the surrounding tissues such as the visceral mesoderm (VINING et al. 2005). The tissue-specific expression of genes in the screen allowed us to identify factors that acted within the glands themselves and did not affect functioning of the surrounding tissues, thus giving a phenotype due to a secondary defect. We crossed flies of marker to flies carrying an EP-insertion on the second or third chromosome (see scheme in Fig. 2A). The resulting offspring overexpressed a gene X specifically in the salivary glands. These embryos were collected early 8

9 (stage 10-13) and late (stage 13-15) during embryogenesis and analyzed live for any apparent defect in salivary gland invagination, positioning, and shape of salivary gland cells or gland lumen. When phenotypes were observed in more than 20% of embryos, embryos were collected again, fixed and counterstained for actin using phalloidin to analyze general morphology. Positive insertions were defined as having 20-90% of embryos showing a salivary gland phenotype after the second examination. The baseline rate of obtaining salivary glands with a phenotype in embryos expressing GFP-EFGas2 or SrcGFP in the glands under the control of fkhgal4 was ~4% (see Material and Methods). In some cases the position and orientation of EP-elements would be predicted to lead to the overexpression of an antisense RNA rather than coding sense mrna. Positive insertions resulting from presumed antisense RNA expression are indicated as such in Table 1 below. Phenotypes identified in the screen We screened 1001 EP element insertions on the second and third chromosome (a list of all lines screened can be found in the Supplementary Table 1). Overexpression in the salivary glands of genes located downstream of EP elements under the control of fkhgal4 led to a variety of phenotypes that could be classified into four major classes (Fig. 2): invagination defects ( failure to invaginate, wide invagination ; see Fig. 2 B, C), gland shape and lumen defects ( shepherd s crook, C-shape, S-shape, wrong length, enlarged lumen, aberrant lumen ; see Fig. 2 D-I), positioning defects ( wrong position, turning, budding, forking, hook, butterfly ; see Fig. 2 K-P) and gland sub-fate defects ( no proper duct ; see Fig. 2Q). 9

10 These overall phenotypes suggest that the screen detected interference at all stages of salivary gland formation. Although the phenotypes listed above were recurrently found in the screen, half of the positive insertions showed a variable phenotype, combining several of these individual phenotypes. The other half showed a consistent phenotype restricted to one class or even one specific phenotype (see Table 1, Phenotype in Salivary Glands ). This suggests that in the cases of genes showing variable phenotypes upon overexpression in the glands, a specific phenotypes was not necessarily reflective of only a certain process failing during the invagination, but rather indicates that many perturbances at the molecular level might lead to similar phenotypes. Genes identified in the screen Out of 1001 EP lines screened 51 showed a phenotype in the salivary glands when crossed to fkhgal4, equalling 5.09% of the total lines analyzed. The penetrance of phenotypes varied from weak (2.7% of EPs tested), to strong (1.7% of EPs tested) and very strong (0.7% of EPs tested; see Table). The genes affected could be classified according to their predicted function (see Table 1). Several of the genes identified by positive insertions have previously been implicated in salivary gland morphogenesis or function within the glands: chickadee (CHANDRASEKARAN and BECKENDORF 2005), tec29 (CHANDRASEKARAN and BECKENDORF 2005), doughnut on 2 (HARRIS and BECKENDORF 2007), rhomboid1 and spitz (KUO et al. 1996), tapδ (ABRAMS and ANDREW 2005), and slit (KOLESNIKOV and BECKENDORF 2005). Three of the insertions identifying these were potentially inducing antisense RNA expression 10

11 and could thus mimic a loss-of-function situation (chic, tec29, slit), three insertions would induce overexpression (dnt, rhomboid1, spitz). Overexpression of dnt could affect the positioning cues the migrating glands receive, whereas rhomboid1 and spitz overexpression would lead to excess Spitz ligand being provided, potentially overstimulating EGFR signalling (see below). These genes served as confirmation that factors impinging on salivary gland morphogenesis were picked up in this screen. The majority of positive insertions (44 out of 51 EPs), though, were inserted into genes that have not previously been implicated in salivary gland morphogenesis or tubulogenesis in general. Several of the encoded proteins have known functions in other tissues in flies such as Egalitarian (NAVARRO et al. 2004), Traf-4 (CHA et al. 2003), RanGAP (KUSANO et al. 2001), Smd3 (SCHENKEL et al. 2002), Nedd 8 (ZHU et al. 2005) and Tout-velou (THE et al. 1999). Fourteen out of the 51 hits were EPs inserted in previously uncharacterized genes, many with close orthologues in other species including vertebrates. Analysis of individual genes in detail In the following section we will discuss a subset of the genes identified in the screen. Three of these were genes endogenously expressed in the glands (chickadee, rhomboid1, and egl), and thus overexpression could have interfered with their proper function in the glands. One gene (btsz) was expressed in the glands and was identified through an insertion that would have lead to production of a tissue-specific antisense RNA, thus potentially mimicking a loss-of-function situation. The last gene (caps) was not endogenously expressed in the glands and 11

12 thus the overexpression identified a potential requirement elsewhere for proper cell and tissue shape. Genes endogenously expressed in glands identified through overexpression: Chickadee encodes the Drosophila Profilin protein. Profilins are actinmonomer sequestering proteins, that have been implicated in promoting both actin polymerization or depolymerization (YARMOLA and BUBB 2006). Drosophila Profilin fulfils essential functions at all stages of development, and also in the female germline (VERHEYEN and COOLEY 1994). Loss of profilin has been associated with the inability to constrict apical surfaces in the morphogenetic furrow of the larval eye disc (BENLALI et al. 2000). With respect to salivary gland morphogenesis it has been reported that tec29 chic double mutants show disorganized actin in the salivary gland placode and display a delay in invagination (measured by remaining placode area at stage 14). This study also reports that chic mutant embryos have normal glands (CHANDRASEKARAN and BECKENDORF 2005). Two EP insertions into chic showed phenotypes in the glands when driven by fkhgal4, EP(2)713 and EP(2)1011. EP(2)713 should overexpress the entire chickadee coding sequence (see supplementary Figure 1), whereas EP(2)1011 is inserted in the opposite direction and could drive expression of an antisense RNA to the 5 most 1kb of the chic pre-mrna (or it could drive expression of eif4a, situated 1.8kb away, see scheme in Fig. 3A). Expression of either EP led to aberrantly shaped glands (Fig.3 B-E), with EP(2)1011 giving frequent hook-shaped and shepherd s crook -shaped glands (Fig.3 D and E). Overexpression of a UAS-chickadee construct using fkhgal4 led to embryos showing a disorganized epidermis in the regions where 12

13 fkhgal4 was expressed, with a loss of apical Crumbs accumulation in the epithelial cells of the epidermis (Fig. 3 F-G ). The glands nonetheless invagianted and within the invaginated portion of the glands Crumbs was localized apically. This suggests that epithelial integrity and or polarity might be impaired if levels of Profilin are imbalanced. Effects on junctional Armadillo in the absence of Profillin have been described (TOWNSLEY and BIENZ 2000). In chic mutant embryos (either chic 221 or chic ) at a stage when the first cells had just invaginated from the salivary gland placode, cell shapes within the placode often appeared irregular compared to wildtype, though Crumbs was still localized apically at this stage (compare Fig. 3 H, H and I). At later stages the salivary glands invaginated but were irregular in shape and the placodal and surrounding epidermal cells on the surface of the embryo appeared disrupted with absent or mislocalized Crumbs labelling (Fig. 3, K-L and O- P at stage 12, M-N at stage 14, for comparison a matching wild-type epidermal scan at stage 14 is shown in Q-Q ). Other apical markers such as the spectraplakin Shot and DE-Cadherin also appeared disrupted at these stages (Fig. 3V-V for DE- Cadherin and data not shown). Nonetheless, within the invaginated portion of the glands Crumbs was localized apically, probably because early apical Crumbs localization in the placodal cells was unperturbed. Thus, in contrast to the previous report (CHANDRASEKARAN and BECKENDORF 2005) either elevation or disruption of Chickadee/Profilin levels appeared to affect salivary gland invagination to some extent. As Profilin has been shown to both promote actin polymerisation and depolymerisation, depending on the context and tissue (YARMOLA and BUBB 2006), an imbalance of Profilin levels (either lowered or increased) could affect critical 13

14 cortical function of actin during cell shape changes required to allow the invagination and/or the cell rearrangements on the surface of the embryo during invagination. Rhomboid1 and EGF-receptor signalling are known to influence cell fate decisions within the salivary gland primordium. The EGF-ligand Spitz is secreted from the ventral midline cells with Rhomboid being the intramembrane protease essential for its release (SHILO 2005). Spitz diffuses a few cell diameters laterally, to induce the most ventral cells within the salivary gland placode to become duct cells, whereas the other placodal cell become secretory cells (see scheme in Fig.4A). This switch in fate transmitted by EGF is in part achieved through repression of the fkh transcription factor. Fkh in turn represses three duct-specific genes, trachealess (trh), dead ringer (dri) and serrate (ser) (HABERMAN et al. 2003; KUO et al. 1996). Thus, by the end of embryogenesis rhomboid1 and other spitz group mutants have salivary glands that are entirely composed of secretory cells and are completely enclosed within the embryo without any ductal connection to the outside (Fig. 5 A-B ; (KUO et al. 1996)). Overexpression of rhomboid1 using EP(3)3704 in the glands led to the formation of glands that were positioned too anteriorly with no obvious distinction in shape between duct and secretory cells and no common duct-like structure formed at stage 15 (Fig.4 C-C ). An identical phenotype was observed when a rhomboid1 transgene under the control of the UAS promoter was expressed in the glands (Fig.4 D-D ). At earlier stages during the invagination process, when rhomboid1 was overexpressed in the gland primordium using fkhgal4 and either the EP(3)3704 or UAS-rhomboid1, aberrantly shaped lumena could be observed (Fig.4 E, E ) but most 14

15 prominently a large bulge of fkh-expressing ectopic cells seemed to arise between the already invaginated secretory gland portions at the position where usually the individual ducts would form (Fig.4 F, F ; similar ectopic cells could also be observed in other experimental situations, see below). Though the analysis of rhomboid1 mutant embryos has shown that EGFR signalling is necessary to induce duct fate in the most ventral cells of the placode (and thus loss of EGFR signalling leads to loss of ducts, see Fig.5 A-B ), this could indicate that activation of EGFR signalling throughout the placode was not sufficient to induce duct fate in all cells. To test this hypothesis, we labelled embryos overexpressing rhomboid1 with markers for duct (Eye gone, Eyg; JONES et al. 1998) and secretory (dcreb-a; ABRAMS and ANDREW 2005) gland fate and compared the expression to wild-type embryos. Both markers labelled the salivary placode (Fig. 4, G and K), the glands at stage 14 (Fig. 4, H and L) and stage 15 (Fig. 4, I and M) in a comparable pattern to wild-type placodes and glands. In addition, the bulge of potentially ectopic cells at the ventral surface that was observed at stages (bracket in Fig. 4L) strongly expressed the duct marker Eyg, indicating that cells fated to become duct have overproliferated. At stage 15-16, a large group of cells at the very anterior tip of the embryo that completely failed to invaginate expressed Eyg (Fig. 4 M and M ). These data suggest that, at least when the EGFR pathway is ectopically activated throughout the whole placode in a timeframe that mimicked the expression of fkh, this was not enough to convert secretory cells into duct cells. To test whether this failure in fate conversion could be due to the timing of the overexpression, we also expressed rhomboid1 under the control of armadillogal4 (armgal4) throughout the whole epidermis of the embryo and with expression starting at much earlier stages (Fig. 4 15

16 N-P ). The overactivation of EGFR signalling throughout the epidermis led to embryos with varying degrees of overall affected morphology (in many cases head involution failed, and general appearance of the epidermis was less organized compared to wild-type, though epithelial integrity/polarity appeared unperturbed judged by UAS-α-cateninGFP labelling that was also driven under the control of armgal4). In these embryos dcreb-a and Eyg were expressed in the placode at stage 11, though the dcreb-a expression domain appeared to extend beyond the placode area into the more anterior hemisegments (Fig. 4 N,N ). Salivary gland invagination was strongly affected in that only very short glands invaginated into the embryo (see Fig. 4 O and P for stage 13 and 14, respectively). Nontheless, these stubby glands expressed dcreb-a, the secretory fate marker, in the invaginated portion of the glands (Fig. 4 O and P, arrows), and expressed Eyg in a few cells that had invaginated but were still close to the surface of the embryo (Fig. 4 P ). In addition, a large bulge of Eyg expressing cells could be found at the surface of the embryo between the two invagination sides (Fig. 4 O, bracket), similar to what we observed when rhomboid1 was overexpressed under the control of fkhgal4. We also tested this hypothesis further by overexpressing additional components of the EGF pathway in the salivary glands: a constitutively active version of the EGFR, UAS-CA-EGFR, a secreted version of the ligand Spitz, UASsspi, and the negative regulator Argos, UAS-argos. Overexpression of argos using fkhgal4 led to a high proportion of glands that lost a ductal connection to the embryo surface, similar to the rhomboid1 mutant embryos (Fig.5 A-D ), though less penetrant (which is probably due to timing and/or expression levels of the transgene). When a secreted version of the EGFR ligand Spitz was expressed using 16

17 fkhgal4 the phenotypes observed appeared very similar to the ones seen in the rhomboid1 overexpressing embryos, namely ectopic cells and glands positioned too anteriorly without discernable duct (Fig.5 K-L ). Overexpression of a constitutively active form of the EGFR, UAS-CA-EGFR, in the salivary glands led to invagination of cells with slightly aberrant shapes, and an invagination hole that was too large. This led to fully invaginated glands that had a too large and aberrantly shaped lumen, though the individual and common ducts appeared normal (Fig.5 G-H ). The ectopic ventral cells observed upon expression of rhomboid1 or secreted spitz in the glands could be a result of overproliferation if EGFR signalling in the placodal cells is not only working as a fate switch but is also a mitogenic signal. We thus analyzed the amount of cell division in the placode at stage in embryos overexpressing rhomboid1 or secreted spitz under the control of fkhgal4 compared to wild-type placodes using an antibody against phosphorylated histone H3 (p- HisH3), a chromatin mark of mitotic cells (WEI et al. 1999). In control embryos, salivary placode cell nuclei at stage did not contain nuclei showing p-hish3 labelling (Fig. 5 M and P), whereas many placode cells overexpressing secreted spitz (Fig. 5 N and Q) or rhomboid1 (Fig. 5 O and R) showed the p-hish3 mark and were thus actively dividing. Also, labelling of microtubules with GFP-EFGas2 revealed mitotic figures (Fig. 5O and O, dotted lines) and remnant spindle midbodies (Fig. 5 O and O, arrows). The results presented above suggest the following: firstly, that EGFR signalling, though necessary for duct fate, is not sufficient to induce duct fate in all salivary placode cells, even though absence of EGFR signalling turns all cells into secretory cells. Secondly, an increase of EGFR signalling in the placode cells can 17

18 induce excessive proliferation in a part of these cells, probably resulting in the mishapen glands observed upon rhomboid1 or spitz overexpression. Thirdly, through modulation of fkh levels and downstream components within the placode, EGFR signalling might also impinge on the cell shape changes that invaginating cells undergo. Egalitarian (Egl) and BicaudalD (BicD) are two proteins that act together with cytoplasmic Dynein in the localization of mrnas in Drosophila embryos and the oocyte, with Egl interacting directly with Dynein light chain (BULLOCK and ISH- HOROWICZ 2001; NAVARRO et al. 2004). Overexpression of egl using EP(2)938 led to salivary glands that were C-shaped or shortened (Fig.6 B and C). Shortened glands could also be observed when egl was expressed in the glands using a UAS-egl construct (Fig.6 D and D ). GFP-positive cells could be observed that appeared to lose contact with the gland (arrow in Fig.6 D). Because both BicD and Egl have essential functions during oogenesis an analysis of egl or BicD null embryos is not possible. We therefore analyzed embryos from mothers carrying two hypomorphic alleles of egl that were mated to heterozygous fathers (see Materials and Methods for the exact genotypes). Embryos with reduced Egl function often showed a disrupted epidermis, with large patches that appeared to completely lack apical Crumbs labelling (Fig.6 F,F compared to Fig.6 G, G ). This phenotype was variable, though, and an example of an embryo with less disrupted epidermis is shown in Fig.6H. Also, during later stages of invagination at stage 13, the placode area was often disrupted and lacked apical Crumbs (Fig.6 K). Salivary gland morphogenesis was disrupted in egl mutant embryos in that the cells of the placode often did not 18

19 change their apices in a coordinated way (though crumbs still accumulated apically in the placode, see Fig. 6H ), the invagination hole appeared too large and extended (Fig.6 H) and the invaginated portion of the glands often had an irregular shape (Fig.6 I and K ). In the invaginated portion of a gland, Crumbs was not concentrated near the apical cell junctions as in the wild-type, where this accumulation appears as a honeycomb lattice, (compare Fig.6 L and Fig.6 M ). Instead, Crumbs was delocalized all over the apical surface and large accumulations could also be found intracellularly (arrow in Fig.6 L ). What could be the mechanism leading to a loss of apical surface identity or constituents? Egl and BicD together with Dynein act as minus-end directed microtubule motors, and as in most epithelial cells the minus ends of microtubules are located near the apical surface in the salivary glands (MYAT and ANDREW 2002). hairy mrna is one of the best understood cargoes of Egl and BicD mediated transport (BULLOCK et al. 2006; BULLOCK et al. 2003), and Hairy has been shown to be important for the regulation of apical membrane growth during salivary gland formation, in part through modulation of Crumbs (MYAT and ANDREW 2002). Thus, affecting hairy transcript localization through lowered levels of Egl and BicD could in turn affect the maintenance of apical membrane identity in the secretory cells. Alternatively, recent reports have shown that crumbs mrna itself, and also the RNA of the Crumbs-associated protein Stardust (Std), are apically localized and this apical localization is important for function (HORNE-BADOVINAC and BILDER 2008; LI et al. 2008). Thus, if crumbs mrna localization were dependent upon Egl and BicD, then reduction of Egl and BicD would result in loss of functional crumbs at the apical surface, leading to loss of epithelial characteristics. The apical localization at least of std mrna appears developmentally regulated in the embryo 19

20 (HORNE-BADOVINAC and BILDER 2008). Thus, it is possible to envisage that salivary gland cell apical maintenance is Egl and BicD dependent and especially sensitive to levels of Egl and thus Crumbs in comparison to other epithelial tissues at the same stage. We are currently investigating this issue in more detail. A gene endogenously expressed in glands identified through antisense expression: Bitesize (Btsz) is the sole Drosophila synaptotagmin-like protein. Its mrna is expressed in the salivary glands and also other epithelial tissues, with a strong apical enrichment (SERANO and RUBIN 2003). Btsz has recently been shown to control the organization of actin at adherens junctions in early embryos, though it might be dispensable in adult fly epithelia (PILOT et al. 2006). Recruitment of Btsz in early embryos to the apical junctional region is not dependent on E-Cadherin but on PIP 2 (phosphatidylinositol-(4,5)-bisphosphate) and Par-3/Bazooka, a protein of the Par-3/Par-6/aPKC apical complex (PILOT et al. 2006). Two mutant btsz alleles have been described, btsz K13-4 that deletes a portion of the N-terminus of some Btsz protein isoforms, and btsz J5-2 that introduces a stop codon due to a frameshift in the N-terminal portion of btsz (SERANO and RUBIN 2003). Expression driven by fkhgal4 from the EP-element identified in our screen, EP(3)3567, should lead to production of an antisense RNA to most of the btsz coding sequence and thus could downregulate endogenous btsz mrna levels (see scheme in Fig. 7A and Supplementary Figure 2). In embryos where EP(3)3567 is driven by fkhgal4 at stage 13, when most secretory cells have invaginated from the placode, the epidermis in the region where the antisense RNA was expressed was disrupted and 20

21 had lost apical Crumbs accumulation (Fig.7 B and B ). The glands themselves showed an irregular lumen (Fig.7 C and C ). btsz K13-4 mutant embryos that manage to cellularize and complete gastrulation showed a somewhat disrupted epidermis, with loss of apical Crumbs accumulation in patches, at later stages (see Fig.7 D for a stage 13 and Fig.7 F for a stage 14 embryo). Nonetheless, many salivary gland placode cells still showed enhanced Crumbs labelling, though shapes of apical circumferences of invaginating cells were irregular (Fig.7 D-D, compare to wild-type in Fig.7 E-E ). Also the apical accumulation of the fly spectraplakin protein Shot could not be observed in btsz K13-4 embryos in contrast to wild-type (Fig. 7 D versus E ). At stage 14 the secretory portion of the glands in btsz K13-4 embryos was often found to lose apical localization of Crumbs (and also show reduced apical actin enrichment; Fig.7 G-G ). Similar to what we observed upon expression of EP(3)3567 with fkhgal4, the epidermis in the region where the placodal cells were located previously was disrupted and lost apical Crumbs and also DE-Cadherin labelling completely (Fig.7 H versus wild-type in Fig.7 L; Fig.7 M). The disruption of the epidermis and failure in proper apical localization of Crumbs is to some extent reminiscent of the phenotypes observed in egl mutant embryos. As btsz mrna is another RNA that is localized apically in various epithelial cells (and the localization signal has been identified, SERANO and RUBIN 2003) one could speculate that its localization could also be dependent on Egl and BicD, thus explaining some overlap in the phenotypes. A gene not endogenously expressed in the glands identified through overexpression: 21

22 Capricious (Caps) belongs to the class of leucine-reach-repeat (LRR) transmembrane proteins, together with its close paralogue Tartan (Trn). Both proteins have been implicated in the formation of compartments of cells with different affinities in the wing disc (MILAN et al. 2001), modulation of epithelial integrity within the wing disc (MAO et al. 2008), correct targeting of a subset of photoreceptor axons to the correct layer within the optic lobe (SHINZA-KAMEDA et al. 2006), and also joining of tracheal branches over segment boundaries (KRAUSE et al. 2006). One study also showed that in tissue culture Capricious and Tartan are able to mediate homophilic cell adhesion, a molecular function that could explain their roles discussed above (SHINZA-KAMEDA et al. 2006). Overexpression of capricious using EP(3)552 led to glands with an enlarged lumen and very aberrant shapes at stage 15 (Fig. 8 B and C). The same phenotype was observed when a transgene of capricious was expressed under UAS control (Fig.8 D-E ). During early stages of invagination, the invagination hole appeared enlarged compared to wildtype and extended along the anterior-posterior axis (compare Fig.8 F-G, and Fig.8 H and H ), suggesting problems in the shape of invagination cells and the order of invagination. This disorganization could at later stages lead to the aberrant shapes of the secretory part of the glands observed. We then analyzed capricious and tartan single and capricious tartan double mutant embryos (MAO et al. 2008), as some previous studies have indicated redundancy between both molecules in some tissues (MAO et al. 2008). In all mutant situations invaginating glands often showed irregular lumens (Fig.8 I-M), indicating that both proteins might work together during salivary gland invagination. We next analyzed if and where capricious and tartan are expressed during salivary gland morphogenesis using P-element insertions into 22

23 each gene that carry a lacz gene leading to β-galactosidase expression under the endogenous expression control of each of the genes. Capricious was expressed in the embryo in the region of the salivary glands from stage 12-15, but appeared to be mostly excluded from the salivary glands themselves, though it was always expressed in cells in close contact with the glands (Fig.8 N-S). In contrast, tartan was expressed strongly in the salivary glands from placode stage on (Fig.8 T-U). Similar expression patterns could also be observed in in situ hybridization for capricious and tartan mrnas (see supplementary Figures 33 and 4). Thus, in analogy to the situation in the developing trachae where both protein are expressed in complementary tissues to allow for proper dorsal branch fusion (KRAUSE et al. 2006), the reciprocal expression of capricious and tartan in and around the salivary glands could play a part in the correct invagination and later positioning of the glands with respect to the surrounding tissues. 23

24 DISCUSSION Here we show that a gain-of-function screen looking for factors affecting a process of epithelial morphogenesis, the formation of the salivary glands in the Drosophila embryos, was efficient in identifying a range of known and new players. The screen was designed to identify genes that are endogenously expressed in the glands and where overexpression or antisense expression by an EP element interfered with endogenous function. The screen could also identify genes not endogenously expressed in the glands but with an apparent overexpression phenotype that uncovered a potential function for this gene in cell shape or cytoskeletal regulation elsewhere. Genes falling into all of these classes have been identified in the screen. We show that factors with a variety of proposed functions can affect salivary gland invagination, from cytoskeletal components, via signalling factors (some of which impinge on the cytoskeleton themselves), to micrornas and novel uncharacterized proteins. Most of the factors identified in this screen should affect salivary gland morphogenesis in a gland-autonomous fashion due to the restriction of overexpression to the salivary glands only, the only exception being the overexpression of secreted signalling factors. We could observe a variety of phenotypes from aberrantly-shaped glands to irregular lumena and wrongly positioned glands. In half of the cases, upon overexpression of a gene (or in a few cases potentially an antisense RNA) several different phenotypes could be observed, as opposed to a single dominant phenotype that was found in the other half. At the level of detail at which we analyzed the phenotypes (GFP-markers of cell shape or microtubules plus 24

25 phalloidin-labelling of actin), these seemed to fall into a limited number of classes, suggesting that several different perturbances of the system might lead to similar phenotypes. We also found no case that consistently led to a complete failure in salivary gland invagination. This is not completely surprising, as dominant effects or knock-downs that we would expect to see in our screen might not perturb the system enough to lead to a complete failure in all aspects of the cell shape changes required for the invagination. Also, when analyzing various mutants in the initial follow-up of a subset of the hits identified, even in situations where the embryonic epidermis seemed to be very disrupted and cell shapes of invaginating cells were highly irregular as for instance in btsz or egl mutants, the glands nonetheless managed to invaginate. These observations suggest that there is a strong drive for the invagination of the cells of the salivary gland primordium, with many different factors contributing at the effector level. Elimination or perturbance of only one of these factors will not prevent invagination completely, but will rather lead to a slightly disordered invagination process that in the end might result in the aberrant shapes and phenotypes we observed. This situation appears similar to the process of mesoderm invagination during gastrulation in the Drosophila embryo. Many factors contribute to this process, but loss of none apart from the most upstream transcription factor initiating the whole mesoderm invagination program, twist, will abolish invagination completely. In all other downstream mutants analyzed the mesoderm will nonetheless manage to invaginate, albeit in an uncoordinated and/or delayed fashion (LEPTIN 2005). Similarly, during salivary gland invagiantion, fkh appears to be the most upstream transcription factor initiating the invagination program for both the secretory and the ductal part of the glands. In the absence of 25

26 Fkh, invagination fails completely (MYAT and ANDREW 2000a). Several direct targets of Fkh have been identified, including the transcription factors senseless (CHANDRASEKARAN and BECKENDORF 2003) and sage (ABRAMS et al. 2006), PH4αSG2, a prolyl-4-hydroxylase (ABRAMS et al. 2006), and also creba, which in turns control the expression of secretory genes in the glands (ABRAMS and ANDREW 2005). Only one recent study has so far addressed direct targets of fkh in a genomewide manner (LIU and LEHMANN 2008). In this study, whole genome expression levels were compared between control pupae and pupae with forced expression of fkh. At the beginning of pupariation fkh controls both the expression of the salivary gland secretion proteins (ROTH et al. 1999) and also controls the cell death that occurs during pupariation in this tissue (MYAT and ANDREW 2000a). Downstream targets of fkh identified in this study included cell death genes, genes involved in autophagy, phospholipid metabolism, glucose and fatty acid metabolism and hormone-dependent signalling pathways, but also others. These other factors regulated by fkh included several proteins that we also identified in our screen: capricious, bitesize, gliotactin, ptpmeg, rhomboid1, and spitz. This overlap of fkhdependent factors and genes identified by us, that are also potentially fkhdependent, at different stages of development, i.e. embryo and pupa, could suggest that these overlapping factors are regulated by fkh independent of stage specific cofactors. The set of genes identified in our screen encode proteins with a wide range of potential functions: cytoskeleton or cytoskeleton-associated, signalling, nuclear or transcription factor, protein synthesis and degradation, membrane traffic, cell 26

27 surface and extracellular, enzymes, mitosis-meiosis-germline, and also uncharacterized genes and micrornas. Nonetheless, many of these have been implicated in some aspect of epithelial morphogenetic function, some even within the salivary glands themselves. Thus, we are confident that many of the factors identified in this gain-of-function screen will turn out to have a function in epithelial morphogenesis. The initial folllow-up of the subset of genes described in detail above also confirms that gain-of-function phenotypes can point to new factors involved in a process, but can also reveal new aspects of a function of a protein that were not previously appreciated, as in the case of rhomboid1. The overexpression phenotype of rhomboid1 in the salivary glands showed an intriguing phenotype; ectopic EGFR signalling throughout the part of the primordium that will constitute the secretory part of the gland does not simply induce these cells to switch to ductal fate. Instead it suggests that either other permissive factors are expressed in the duct primordium independent of EGFR signalling that are absent from the secretory primordium, or further inhibitory factors are at work in the secretory primordium in addition to EGF-induced fkh that prevent ductal fate. The timing of the overexpression of rhomboid1 in the screen, using part of the fkh promoter in the fkhgal4 line to drive the expression of the EP elements, did not seem to be crucial to the fate decision, as expression of rhomboid1 throughout the whole epidermis using and ealier Gal4-driver (armgal4) still lead to invagination of glands with identifiable secretory and ductal cells (Fig. 4, N-P). Thus it does not appear that the cells in the placode are responsive to the EGFR signal in terms of fate assignment only in a narrow time window, but supports the notion that other 27

28 factors are involved. It will be interesting to determine in the future what these factors are. The analysis of caps and tartan mutant phenotypes suggests a role for these genes in salivary gland morphogenesis. The molecular function of both Caps and Trn proteins is still unclear. They appear important to mediate interaction and disctinctiveness between groups of cells: neurons finding appropriate targets in the brain (SHINZA-KAMEDA et al. 2006; SHISHIDO et al. 1998), separation of ventral and dorsal compartment cells in the wing disc (MILAN et al. 2001), and tracheal morphogenesis across segment boundaries (KRAUSE et al. 2006). Caps and Trn have been suggested to act as homophilic or heterophilic adhesion receptor, or serve another unidentified function during adhesion. Our results indicate that salivary gland morphogenesis might be a useful system to address their molecular function in more detail. It is also interesting to note that not only Caps and Trn, but also Slit and the protein encoded by CG14351, which were also both identified as hits in the screen, belong to the family of LRR proteins, suggesting a general role for this class of surface receptors in salivary gland morphogenesis. Another gene identified in the screen with a highly penetrant phenotype (severely shortened glands) is TNF-receptor-associated factor-4 (traf-4, previously annotated as traf-1 in Flybase). Traf-4 has previously been shown to induce apoptosis via activation of JNK-kinase when overexpressed in other tissues (KURANAGA et al. 2002). Traf-4 has also been linked to the Ste20 kinase Mishapen (another gene identified in our screen) which in turn has been shown to be important for coordinated cell shape changes occurring for instance during dorsal closure in the fly embryo and epiboly in the zebrafish embryo (KOPPEN et al. 2006). It also 28

29 appears to have a role in mesoderm invagination (Maria Leptin; personal communication). Overexpression of traf-4 using EP(2)578 led to a severe reduction in the number of cells in the salivary glands at embryonic stage 15 (when counting cell numbers in fluorescence images taken through the middle of wild-type glands, these had /- 3.9 cells (n= 40) around the perimeter of the gland, whereas traf- 4 overexpressing glands had /- 5.0 cells (n=45) around the perimeter). We have not directly tested that the missing cells have died through induction of apoptosis, but would expect this to be the case in agreement with the earlier studies. To address whether traf-4 has a function linked to mishapen and cell shape changes in the glands we analyzed fly embryos lacking Traf-4 (this mutant was a kind gift of Maria Leptin) for any problems in the early cell shape changes occurring during salivary gland invagination, but could not find any strong defects (data not shown). Thus, despite many similarities between the epithelial morphogenetic processes of mesoderm invagination (which also starts with the invagination of an epithelial sheet) and salivary gland invagination, downstream effectors vary between the two systems. Another interesting group of hits identified in the screen are the EP-elements potentially driving overexpression of the mir-310 microrna cluster. This cluster contains the microrna genes mir-310, mir-311, mir-312, and mir-313. Overexpression of the cluster from three different EP-elements located just upstream of the cluster, EP(2)2536, EP(2)2586 and EP(2)2587, led in each case to glands with widened and irregular lumena, though with varying penetrance (data not shown). Antisense-mediated depletion of each microrna from this cluster has previously been shown to perturb dorsal closure and head involution in the embryo, 29

30 indicating that the inhibition of downstream targets of this cluster might be important for various epithelial morphogenetic events (LEAMAN et al. 2005). Thus, the micrornas in this cluster could be important to regulate targets that require downregulation to facilitate invagination during salivary gland morphogenesis. Two other micrornas have been shown to be expressed in the salivary gland in the embryo, mir-8 and mir-375, with mir-8 showing a dynamic expression pattern (ABOOBAKER et al. 2005), expression patterns for the mir-310 cluster have not been analyzed yet. This data together with our screen results suggests that micrornadependent control of gene expression might be an important factor in salivary gland morphogenesis. Thus in summary the gain-of-function screen for factors affecting cell shape during salivary gland morphogenesis in the Drosophila embryo presented here was successful in identifying a range of candidates. These candidates represent on the one hand genes that are endogenously expressed in the glands and thus are likely to be serve a role during salivary gland morphogenesis/tubulogenesis. On the other hand we identified genes that are not endogenously expressed in the glands but nonetheless interfered with their invagination, potentially through effects on cell shape or the cytoskeleton. These genes might therefore be important for the regulation of cell shape either in other tissues. It will be interesting to determine in the future which of the candidate genes of the first class serve a function only in the salivary glands, and which are required for tubulogenesis events in general, also in other species. For the second group of candidates an analysis of their role during cell shape changes in other morphogenetic events will be key to understanding how they affected salivary gland morphogenesis in our screen. As the screen identified 30

31 several uncharacterized genes whose expression gave strong phenotypes in the glands and that have close orthologues in mammals, the salivary glands appear to be a good model system to analyze the function of such genes. Finally, our followup investigation of a selected set of candidate genes through loss-of-function mutants demonstrates that the combination of functional screening and phenotypic loss-of-function analysis provides a useful approach to identify downstream effectors in a morphogenetic process. 31

32 ACKNOWLEDGEMENTS The authors would like to thank Nick Brown, Matthew Freeman, Simon Bullock, Deborah Andrew, Maria Leptin and the Bloomington and Szeged Stock Centres for fly stocks; Matthew Freeman, Simon Bullock, Sarah Bray, Deborah Andrew and the Developmental Studies Hybridoma Bank at the University of Iowa for antibodies; Maria Leptin for communication of results prior to publication; Nick Brown for use of his confocal microscope; Sean Munro and Nick Brown for helpful comments on the manuscript. This work was supported by the Biotechnology and Biological Sciences Research Council [grant number BB/B501798/1] and the Royal Society. 32

33 MATERIALS & METHODS Screen Design and Fly Husbandry Marker -lines: fkhgal4 (HENDERSON and ANDREW 2000; ZHOU et al. 2001) was recombined on the third chromosome with a UAS-construct containing GFP fused to the N-terminus of the EF-Gas2 region of Shot (SUBRAMANIAN et al. 2003) or the membrane targeting domain of src fused to GFP (KALTSCHMIDT et al. 2000). One or the other of these marker lines were crossed to 1001 EP lines from the Rørth collection obtained from the stock centres in Szeged (second and third chromosomes; and Bloomington (third chromosome; The fkhgal4 insertion was a gift from Deborah Andrew, the SrcGFP from Nick Brown. UAS-chickadee was from Lynn Cooley. rho[pδ5], argos[lδ7], flb[ik35], UAS-argos, UAS-CA-EGFR, UAS-sspi, UAScaps, caps[pb1], trn[28.4], caps[del1] trn[28.4] alleles were gifts from Matthew Freeman; egl[3e], egl[pr29], BicD[HA40], b BicD[18a], dp b Df(2L)TW119 and UASegl were gifts from Simon Bullock (the UAS-egl transgene leads to an approximate 3-fold increase in levels; Simon Bullock, personal communication). To analyze egl mutant embryos, egl[3e]/egl[wu50] females were mated to egl[pr29]/+ males, and to analyze BicD mutant embryos BicD[HA40]/+; b BicD[18a]/ dp b Df(2L)TW119 mothers were mated to BicD[18a]/CyO males. In the detailed analyses (apart from in the case of egl and BicD mutant embryos), mutant embryos were identified by the absence of green balancer (balancer chromosomes used were CyO Kr::GFP, TM3 Sb Ser twi-gal4 UAS-2x egfp, and TM6b Tb Sb df-gal4 UAS-YFP). All other stocks 33

34 used were from the Bloomington Stock Center. Crosses were maintained at 25ºC on cornmeal food, embryos were collected on apple or grape juice-agar plates. EP lines were determined to be heterozygous or homozygous for the EP insertion. In the absence of visible balancer chromosomes, lines were assumed to be homozygous. Homozygous lines were crossed to a homozygous driver line and embryos were collected over night on apple or grape juice plates with yeast paste. 20 embryos between stage 10 and 13 and 20 embryos between stage 13 and 15 (an early and a late sample) were scored in live mounts in halocarbon oil (Halocarbon Oil 27, Sigma) after dechorionation in 50% bleach. In heterozygous balanced lines, 40 embryos in each of the early and late group were scored, assuming equal fertilisation and survival from both genotypes through the end of embryogenesis. Lines with 20% salivary gland defects, or with potential defects that would require quantitative analysis (i.e. changes in length), were subjected to second pass screening. In the second pass, embryos were collected as above, fixed in 2:1 heptane:4% formaldehyde in PBS and stained with rhodamine phalloidin. A larger number of embryos was scored from these collections (average >80). First pass hits were also crossed to w f flies, embryos collected and stained with rhodamine phalloidin as above to check for dominant positional effects of the EP insertion. Immunohistochemistry, Widefield Fluorescence and Confocal Analysis Embryos were collected on grape-juice plates and processed for immunofluorescence using standard procedures. Briefly, embryos were dechorionated in 50% bleach, fixed in 4% formaldehyde and stained with phalloidin 34

35 or primary and secondary antibodies in PBT (PBS plus 0.5% bovine serum albumin and 0.3% Triton X-100). Crumbs and DE-Cadherin antibodies were obtained from the Developmental Studies Hybridoma Bank at the University of Iowa; the Shot antibody was raised in our lab and is identical in design to the one described in (STRUMPF and VOLK 1998); the anti-dcreba antibody was from Deborah Andrew (ANDREW et al. 1997); the anti-phospho-histone H3 and anti-gfp antibodies were from abcam (UK). Secondary antibodies used were Alexa Fluor 488-coupled (Molecular Probes) and Cy3- and Cy5-coupled (Jackson ImmunoResearch Laboratories Inc.), rhodamine-phalloidin was from Molecular Probes. Samples were embedded in Vectashield (Vector Laboratories). Widefield fluorescence images documenting the screen results were obtained on a Leica DMR (equipped with a MicroFire camera, Optronics), a Zeiss Axioplan 2 (equipped with a Princeton Instruments camera) and a Zeiss Axioskop Mot 2 (equipped with a Jenoptik C14 camera), using PictureFrame, Metamorph and Openlab software, respectively. Confocal images were obtained using an Olympus Fluoview Confocal laser, iris and amplification settings in experiments comparing intensities of labelling were set to identical values. Widefield fluorescence and confocal images were assembled in Adobe Photoshop, confocal z-stacks and z-stack projections were assembled in Image J. In Situ Hybridization In situ hybridization of whole-mount embryos was performed essentially as described by (TAUTZ and PFEIFLE 1989). To combine the in situ protocol with immunohistochemistry for GFP, the anti-gfp antibody was incubated together with 35

36 the anti-dig antibody, followed by fluorescent secondary antibody incubation to reveal the GFP after the BCIP/NBT colour reaction. Images were obtained on a Leica DMR (equipped with a MicroFire camera, Optronics) and composites were assembled using Adobe Photoshop. The following primers were used to generate in situ probes: chic 5 TTTCCATCTACGAGGATCCC, chic 3 ATTTCGTTCAAAGCTGAGGAC; caps 5 CGGGCAATTACCATGTCGTTG, caps 3 GATGTGGCTGATGCGATTCTG; trn 5 GTGGGCATCTGGTGCATTTTG, trn 3 GATAAAGGATGCGCAACTGGG; for the btsz probe the cdna clone AY was used to transcribe antisense and sense probes. Statistics We determine the base rate of salivary gland defects observable in our experimental stocks by counting defects in the genotype: +/+; +/CyO; fkhgal4::uas- GFPmarker/+. The base rate in this genetic background was determined as 4.3% (n=748). A similar base rate was obtained in a genetic background where the CyO balancer chromosome was replaced by a GFP-marked chromosome: +/+;+/btlgal4::uas-gfp; fkhgal4::uas-gfpmarker/+; the base rate was 4.4% (n=878). Thus, we exclude any effect at least of a CyO balance chromosome present on salivary gland morphogenesis. To set a cut-off level for the rate of affected salivary glands counted in each experiment, above which we determined EP-elements driven by fkhgal4 affect the salivary gland morphogenesis, we chose an arbitrary 20% defects cut-off for the first pass analysis. This yielded 187 EP lines (=18.6% of the total lines screened) to be 36

37 re-screened in the second pass analysis, yielding 51 confirmed insertions overall. This equals ~5% of the total number of EP lines screened, which also equals two standard deviations form the mean of a normal distributed sample, indicating that we set our cut-off at a sensible level. 37

38 Table 1. Genes identified in the gain-of-function screen. This table lists all the genes identified in the screen, sorted according to their proposed function. EP number Cytology Gene affected Effect, Location in Gene, Direction? Penetrance of Phenotypes * Phenotype in Salivary Glands (fkhgal4) Function or Mutant Phenotype, Known Function in Flies or Glands Cytoskeleton & Cytoskeleton-associated EP(2)570 2L (26B4) CG13993 (actin/tubulin protein folding) overexpression, 5 end of gene EP(2)713 2L (26B1) chickadee (profilin) overexpression, in 5 region of gene upstream of most CDS EP(2)1011 2L (26B1) chickadee (profilin) antisense to 5 1kb of gene (or overexpression of eif4a, 1.8kb downstream) EP(2)938 2R (59F7) egalitarian (dynactinassociated) overexpression, directly 5 of gene EP(2)2047 2R (57E5) syndecan inserted in 5 end of sdc wrong strand, strong variable no mutant available strong variable lethal, Tec29 chic double mutants show salivary gland phenotype (CHANDRASEKARAN and BECKENDORF 2005) strong hooks, shepherd s crook lethal, Tec29 chic double mutants show salivary gland phenotype (CHANDRASEKARAN and BECKENDORF 2005) strong variable female sterile, lethal, microtubule-based transport strong variable lethal, works in conjunction with Slit 38

39 EP(3)3567 3R (88D5) bitesize (synaptotagmin-like protein) could drive antisense to 6kb out of the 90kb sdc locus or overexpress sara 7kb downstream middle of gene, could drive antisense to most isoforms (STEIGEMANN et al. 2004) weak variable lethal, actin organization at adherens junctions (PILOT et al. 2006) Signalling EP(2)578 2L (24E1) traf-4 (TNFreceptorassociated factor 4, previously called Traf-1 in flies) EP(2)2167 2L (29A1) btk29/tec29 (Btk family kinase) overexpression, middle of gene, upstream of most CDS 3 end of gene wrong strand, could drive antisense to >30kb of 40kb btk29a locus EP(2)1173 2L (37E1) rangap overexpression, middle of gene but upstream of CDS EP(2)2158 2L (37D2) doughnut on 2 (RYK family receptor tyrosine very strong less cells in glands, overexpression reportedly induces apoptosis (KURANAGA et al. 2002) larval lethal (KURANAGA et al. 2002) weak variable lethal, Tec29 is important for sal. gland invagination (CHANDRASEKARAN and BECKENDORF 2005) strong hook, crook viable, possible link between nuclear transport, actin and profilin (MINAKHINA et al. 2005) overexpression weak variable important for salivary gland positioning as is drl another RYK 39

40 kinase) EP(3)3542 3L (61C1) ptpmeg (tyrosine phosphatase) or mthl9 (G-protein coupled receptor) EP(3)3704 3L (62A2) rhomboid1 (EGF signalling, intramembrane protease) EP(2)2201 2L (37F2) spitz (secreted EGF ligand) EP(3)549 3L (62E7) misshapen (Ste20 kinase) overexpression of ptpmeg or antisense of mthl9 (intronic to ptpmeg) (HARRIS and BECKENDORF 2007) strong budding viable, Ptpmeg is FERM domain protein overexpression very strong aberrant duct morphogenesis, potentially due to overproliferation (see Figs 4 and 5) inserted into middle of spi, could drive antisense to 4kb of some spi mrnas, or overexpress msb1l 5kb downstream overexpression, directly 5 of gene mutations in rho or spitz lead to transformation of duct cells into secretory cells (KUO et al. 1996) weak variable mutations in rho or spitz lead to transformation of duct cells into secretory cells (KUO et al. 1996) weak too wide and lumpy lumen lethal, linked to nuclear movement via BicD (HOUALLA et al. 2005) and cell shape changes during morphogenesis (KOPPEN et al. 2006) Nucleus, Transcription (Factors) EP(2)2176 2R (48E4) Smd3 (snrnp, splicing) overexpression, 5 of gene EP(2)474 2L (21B5) kismet (helicase) overexpression, in 5 region of gene upstream of all CDS strong hooks lethal (SCHENKEL et al. 2002) weak too wide lumen lethal, segment specification (DAUBRESSE et al. 1999) 40

41 EP(2)993 2R (50E1) combgap (zincfinger protein) EP(3)486 3L (75E1) ftz-f1 (ftztranscription factor1) EP(3)711 3L (64E8) bre-1 (nuclear factor downstream of Notch) Protein Synthesis and Degradation EP(2)463 2R (47F7) Tapδ (transloconassociated protein δ) EP(2)2063 2L (37B7) nedd8 (regulation of proteolysis) EP(2)1187 2L (33C1) CG5317 (ribosomal subunit) Membrane Traffic EP(2)2028 2R (48F8) garz (arf-gef, GBF1) EP(2)2313 2L (35F1) syntaxin5 (SNARE protein) inserted into 5 end of CG30096 wrong strand, antisense to cg 2kb away inserted into locus, should overexpress longer isoform overexpression, directly 5 of gene overexpression, 5 end of gene overexpression, 5 end of gene overexpression, inserted in 5 end of JhI-21, wrong strand, should drive CG bp downstream overexpression, 5 end of gene overexpression, directly 5 of gene very strong variable lethal, hedgehog signalling in leg patterning (SVENDSEN et al. 2000) weak variable lethal (FLORENCE et al. 1997) weak variable lethal (BRAY et al. 2005) weak variable lethal, downstream of dcreb-a in the glands (ABRAMS and ANDREW 2005) weak butterfly lethal, ubiquitin-like, cooperates with cullin3 (ZHU et al. 2005) strong shepherd s n.d. crook weak severe hooks ER to Golgi trafficking in mammals (SZUL et al. 2007) weak degenerating glands lethal, membrane fusion, cytokinesis (XU 41

42 et al. 2002) Cell Surface & Extracellular EP(2)827 2R (58D4) CG3624 (Ig domain protein) EP(2)937 2R (52D1) slit (axon guidance receptor) EP(2)2120 2L (22A3) CG14351 (LRR and Ig domain transmembrane protein) EP(2)2463 2L (35D4) gliotactin (transmembrane protein of septate junctions) EP(3)552 3L (70A3) capricious (transmembrane LRR protein) Enzymes EP(2)2199 2R (51B1) tout velu (glucosaminyltransferase) overexpression, directly 5 of gene middle of gene, wrong strand, antisense to 20kb of 50kb gene? overexpression, directly 5 of gene overexpression, directly 5 of gene overexpression, directly 5 of gene inserted in intron of both ttv and lamc (which is intronic to ttv), could overexpress ~10kb of 60kb ttv (about 50% of CDS) or strong variable n.d. strong hooks lethal, slit has been shown to be involved in salivary gland positioning (KOLESNIKOV and BECKENDORF 2005) weak variable n.d., BLAST shows similarity to Slit weak variable lethal, important for tube size control in trachae (PAUL et al. 2003) very strong bizarrely branching and budding lumen lethal (SHISHIDO et al. 1998) weak variable lethal, mutans disrupt hh, wnt and dpp sigalling (BORNEMANN et al. 2004) 42

43 EP(2)1157 2R (59B6) CG9849 (potential protease of the subtilase family) EP(3)3639 3L (65A10) CG10163 (phospholipase A1) 1.1kb antisense to LamC inserted into 5 end of CG3800, wrong strand, overexpression of CG bp away inserted 5 of Best2, wrong strand, could drive antisense to CG bp away weak variable n.d. weak too large and irregular lumen n.d. Mitosis, Meiosis, Germline EP(2)812 EP(3)341 2L (35C1) 3R (82D2) vasa or vig (vasa intronic gene) tacc (centrosomal protein) in 3 region of vig which is intronic to vasa, would overexpress 3 1kb of vig or antisense to vasa middle of gene, could drive antisense to all large isoforms of tacc weak lumpy lumen vasa: lethal, germ cell determination/ vig: n.d. weak variable lethal (BARROS et al. 2005) Other EP(2)2356 2R (57A6) mir-310/-313 cluster overexpression, inserted 200bp 5 of cluster very strong to wide irregular lumen disruption of mir-310 cluster affects dorsal closure (LEAMAN et al. 2005) 43

44 EP(2)2586 2R (57A6) mir-310/-313 cluster EP(2)2587 2R (57A6) mir-310/-313 cluster overexpression, inserted 100bp 5 of cluster overexpression, inserted 100bp 5 of cluster EP(2)1221 2L (27F4) mir-275, mir-305 overexpression, 2kb upstream of genes EP(2)2083 2R (45F1) CG1888 >6kb away, overexpression EP(2)1163 2L (33E4) vir-1 (virus induced overexpression, RNA 1) directly 5 of gene EP(2)1239 2L (25F5) CG14005 or inserted in 5 end of CG7239 CG9171 wrong strand, antisense to CG bp downstream or overexpression of CG7239 2kb downstream EP(2)2219 2L (33E4) CG6405 overexpression, inserted 1.5kb upstream of CG6405 EP(2)2190 2R (55E1) CG30332 overexpression, 1kb 5 of CG30332 EP(2)2182 2R (54A2) CR30234 (cytosolic trna gene) EP(3)313 3R (98E5) CG1523 (related to WD40 repeatcontaining protein 32) strong irregular lumen disruption of mir-310 cluster affects dorsal closure (LEAMAN et al. 2005) weak irregular lumen disruption of mir-310 cluster affects dorsal closure (LEAMAN et al. 2005) strong shepherd s n.d. crook weak variable n.d. weak budding n.d. weak variable n.d., both CG14005 and CG7239 only conserved amongst Drosophilidae; weak variable n.d., two mammalian orthologues strong hooks n.d. overexpression very strong variable n.d. overexpression, inserted directly 5 of gene strong variable n.d. 44

45 EP(2)2269 2R (53D11) CG34460 EP(2)383 2L (23C4) nothing downstream for >10kb, next CG is CG kb away EP(2)2173 2L (35B2) nothing downstream for >10kb antisense to CG34460?, EP is ~2.5kb away inserted 3 of CG3558 inserted into 5 end of no ocelli, wrong strand EP(2)2146a???? genome position of EP unclear EP(2)2265???? genome position of EP unclear EP(2)985???? genome position of EP unclear weak variable n.d. weak short, expanded at turn n.d. weak variable n.d. strong searching cells n.d. very strong strong short, straight, expanded budding, branching n.d. n.d. * Penetrance of phenotypes: weak = 20-30% of embryos showing phenotype strong = 30-50% of embryos showing phenotype very strong = >50% of embryos showing phenotype (with 3 cases < 70% and 4 cases >90%) 45

46 FIGURE LEGENDS Figure 1. Salivary gland development visualized using fkhgal4-driven GFPmarker expression. Salivary gland morphogenesis from embryonic stage10-15 is shown. A Schematic of salivary gland invagination, ventral view. B-F shows low magnification confocal sections of embryos stained with phalloidin to reveal actin (red) and expressing GFP-EFGas2 under the control of fkhgal4 in the salivary glands (green). C-E show lateral views, B is a ventral and F a dorsal view. G-K Close up confocal sections of salivary glands labelled with phalloidin to reveal actin (red) and expressing SrcGFP under the control of fkhgal4 (green, and as a single channel in G -K )). All panels show lateral views. Note that GFP-EFGas2 labels microtubules, whereas SrcGFP is targeted to the membrane and thus reveals cell shape. Figure 2. Phenotypes observed upon overexpression of genes in the salivary glands using fkhgal4. A Schematic of the set-up of the screen. The phenotypes observed in the screen could be classified according to the categories depicted in this figure. Broad categories are invagination defects (B, C), gland shape & lumen defects (D-I), positioning defects (K-P) and gland fate defects (Q). All panels show the GFPmarker expression in green and phalloidin staining to reveal actin in red. Lateral or dorsal views are indicated in each panel. The line in C indicates the area of the too wide opening of the invaginating gland shown; the arrow in D points to where proximal and distal cells of the gland touch due to excessive bending; the double 46

47 arrows in H indicate the too wide width of the gland shown; the arrows in M point to two buds emerging from the side of the gland shown; the arrows in N point two the two ends of a fork; the arrows in O point to cell of the glands that appear to touch across the midline. B-G and K-Q are widefield fluorescence images, H, I and Q are confocal sections. Embryonic stages of embryos shown are indicated in the panels. Figure 3. Chickadee (Profilin) is important for salivary gland invagination. chickadee encodes the Drosophila Profilin protein. A Scheme of the chic locus indicating the position and orientation of the two EP lines that showed phenotypes when driven in the salivary glands. B, C and D, E show phenotypes observed in the screen for EP713 and EP1011 respectively, widefield fluorescent images of live embryos are shown. F-G overexpression of chickadee using a UAS-chickadee construct led to invagination problems and aberrantly shaped glands. F show an internal confocal stack of a gland (14µm thick), labelled with Crumbs to reveal the apical surface/lumen of the glands (red) and showing the SrcGFP marker in green. G, G show a surface stack of the same embryo (with Crumbs in red and SrcGFP in green in G and crumbs as a single channel in G ; 3µm thick). The arrow points to disrupted epidermis in the region of the placode from where the glands have started to invaginate. Note the absence of Crumbs from the apical surface of cells in this region. H, H shows a low and high magnification view of the slightly disorganized epidermis of a chic mutant embryo labelled for crumbs, with H showing the salivary gland placode and gland in a projection (18µm thick stack). I shows a wild-type placode and gland (17µm thick stack) at the same stage as in H. Note that the highly organized arrangement of apical constriction of the placodal cells is less 47

48 apparent in the chic mutant (bracket in H and I). K-L and O-P show the aberrant glands and disrupted epidermis in two different chic alleles (chic and chic 221 ) at stage 12, labelling for Crumbs is in green (and also shown as a single channel in K and O ) and for phalloidin is red in K, L, O, P. Note the disruption and absence of apical Crumbs labelling in the region of the salivary gland placode (arrows in K, L, O and P point to these areas). K is a projection of a 35µm thick stack, L is a 5µm thick surface projection, O is a projection of a 26µm thick stack, and P is a 3µm thick surface projection. M-N show the disrupted epidermis in the region from where salivary gland cells invaginated in a stage 14 embryo (Crumbs is in green in M and as a single channel in M, and phalloidin is in red in M and as a single channel in N; M is a projection of a 34µm thick stack, N is a 5µm thick surface stack). For comparison a stage 14 wild-type embryo is shown in Q-Q (Crumbs is in green in Q and as a single channel in Q, and phalloidin is in red in Q and as a single channel in Q ; Q is a 5µm thick surface stack). The arrows in M and N point to the disrupted region, the white lines in M and Q indicate the ventral midline (the view in M-N is slightly oblique). R-V show chic 221 mutant embryos at stage R, R are lateral views of a placode, whereas S shows an internal stack of the gland. T, T are ventral views of the two placodes, with U showing an internal stack of the glands. V-V show a surface stack of a mutant embryo (5µm thick). Crumbs is in green in R, T and V and as a single channel in R,T and V, phalloidin is in red; both S and U show Crumbs labelling to outline the lumen of the gland. DE-Cadherin (DE-Cad) labelling is in red in V and as a single channel in V. 48

49 Figure 4. rhomboid1 overexpression disrupts salivary gland morphogenesis, but is not sufficient to induce salivary duct fate. A Scheme of the rho locus indicating the position of the EP identified in the screen. B Scheme depicting the known involvement of EGF signalling in salivary gland morphogenesis. EGF is released from the midline (red line) and induces, in the cells close to the midline (light green), the repression of fkh which in turn leads to suppression of secretory fate in these cells, inducing them to adopt duct fate. Fkh expression remains high in the remaining salivary gland primordium (dark green), thus inducing these cells to form the secretory part of the gland (KUO et al. 1996). C- C Overexpression of rhomboid1 in the salivary glands using EP(3)3704 led to glands that, at stage 15 of embryogenesis, were located too far anterior with secretory cells that appeared cuboidal instead of columnar, no proper duct connecting the secretory portions to the outside and a aberrantly shaped lumen. The SrcGFP marker is in green in C and as a single channel in C, phalloidin is in red in C and as a single channel in C. D-D The same phenotype as in C is observed when a UAS-rhomboid1 construct is expressed in the glands using fkhgal4. The GFP-EFGas2 marker is in green in D and as a single channel in D, Crumbs is in red in D and as a single channel in D. E-F Already at stage 13 the invaginated portion of the gland shows aberrant morphology ( ectopic lumen indicated by the arrow in E ), and the amount of cells remaining at the surface appears too large (bracket in E and F ). The SrcGFP marker is in green in E and F (and as a single channel in E and F ), crumbs is in red. G-P analysis of dcreb-a and Eyegone expression: markers of secretory and duct fate, respectively. G-I Control glands expressing only the GFP-EFGas2 marker labelled with antibodies 49

50 against dcreb-a and Eyg at stage 11 (G-G ), stage 14 (H-H ) and stage 15 (I-I ). K-M Glands expressing UAS-rhomboid1 in the salivary glands using fkhgal4 labelled for dcreb-a and Eyg at stage 11 (K-K ), stage 14 (L-L ) and stage 15 (M- M ). Note that despite the irregular shape and ectopic cells (bracket in L) dcreb-a is strongly expressed in the early invaginated part of the glands (L and M ). Eyg is expressed in the most anterior cells of the invaginated glands (L and M ), as in the control, and also in the ectopic cell bulge on the surface of the embryo (L, bracket in L denotes the bulge, the dotted line indicates the ventral midline). N-P Glands expressing UAS-rhomboid1 using armgal4. N-N dcreb-a and Eyg expression in the placode at stage 11. O, O Ventral view of the remaining placode (O) and invaginated glands (O ) at stage 13. More ectopic cells expressing Eyg are found on the ventral surface (bracket in O). Small stubby glands have invaginated and express dcreb-a (arrows in O ). P-P Lateral view of glands at stage 14. More cells have invaginated and express dcreb-a (arrow in P), and the most ventral cells on the surface still express Eyg (P ). GFP makers are in green, dcreb-a is in red and Eyg in blue in G-P, dcreb-a is shown as a single channel in G -N and P, and Eyg as a single channel in G -N and P. All panels are projections of confocal stacks that cover the whole thickness of either the invaginated glands or of the placode at earlier stages. Figure 5. EGFR signalling is necessary but not sufficient to induce salivary duct fate, and overactivation leads to ectopic cell divisions. Analysis of components of the EGFR signalling pathway in the glands. It has been reported previously that salivary glands in rhomboid/spitz-group mutant embryos do 50

51 not specify any ductal portion of the glands (KUO et al. 1996). A-B This phenotype was confirmed in a rho PΔ5 embryos, a null allele of rho (FREEMAN et al. 1992). A, A At stage 13 most of the secretory cells of the gland have invaginated, leaving two large holes visible at the surface of the embryo (arrows in A; A shows the surface of the embryo, A shows an internal confocal stack to reveal the shape and location of the invaginated glands). B, B At stage 14 the glands have fully invaginated and detached from the surface of the embryo, leaving no ductal connection to the outside and a large hole on the surface (B shows the surface of the embryo, B shows an internal confocal stack to show the blunt ended gland; arrow in B points to the blunt end, arrow in B points to the hole). A-B show labelling for Crumbs. C shows an amended gland fate specification scheme as introduced in Fig. 4 to illustrate the altered signalling in rho mutant embryos, where absence of EGFR signalling induces the entire salivary gland primordium to adopt secretory fate. D-F Overexpression of argos, an extracellular inhibitor of EGFR signalling (SCHWEITZER et al. 1995), using fkhgal4 caused a similar phenotype as that seen in rho mutants: D-D At stage 13 the invaginated secretory portion of the glands detaches from the surface of the embryo (arrow in D ), and no duct is formed. E, E The salivary gland primordium at stage 11 appears normal. F Scheme showing that downregulation of EGFR signalling leads to conversion of presumptive duct cells into secretory cells. G-I Overexpression of an activated form of the EGF receptor (CA-EGFR) using fkhgal4 leads to glands with highly disorganized and aberrant lumen from stage 12 on. G-G shows a confocal stack of stage 13 embryo, the arrow in G points to the lumen marked by Crumbs. H, H Cell shapes marked by crumbs in the salivary gland primordium at stage 11 appear normal. I Scheme showing that elevated 51

52 EGFR signalling throughout the primordium does not induce duct fate in all cells. K-I Overexpression of a secreted and active form of the ligand Spitz using fkhgal4 leads to glands that very much resemble those seen upon overexpression of rho (compare Fig. 4). K-K shows a confocal stack of stage 15 embryo. No ductal structures are formed, and the shape of the secretory cells and the lumen is aberrant. L, L. The salivary gland primordium at stage 11 appears disrupted with irregular and too large apices of the invaginating cells. The arrow in L points to a group of cells that show midbodies left by mitotic divisions marked by the GFP- EFGas2 microtubule marker. Ectopic mitoses can also be observed when CA-EGFR is expressed using fkhgal4. M Scheme showing that elevated EGFR signalling through overexpression of secreted Spitz throughout the primordium does not induce duct fate in all cells. The SrcGFP or GFP-EFGas2 markers are green in D, E, G, H, K, L, and shown as a single channel in D, G, K, Crumbs labelling is red in D, E, G, H, K, L and shown as a single channel in D, E, G, H, K, L. M-O Analysis of ectopic cell divisions induced by activation of EGF signalling using phospho-histone H3 (p-hish3) as a marker of mitosis. In the marker-expressing control p-his3 labelling is restricted to the area outside the placode (M) and invaginated gland (P). When UAS-secreted spitz (N, Q) or UAS-rhomboid1 (O, R) are expressed in the glands, many mitotic cells can be found in the placode and invaginated gland. GFP-markers are in green, p-hish3 in red in M-R and O, Crumbs in blue in M-Q and R, and DAPI in blue in O and O. O -O show a higher magnification of the dividing cells in O. GFP-EFGas2 is shown as a single channel in O, p-hish3 as a single channel in O and DAPI as a single channel in O. The 52

53 dotted line in O -O highlights a cell in anaphase (note the spindle in O ), the arrow in O -O points to a midbody in telophase, similar to the ones indicated in L. Figure 6. egalitarian overexpression reveals a potential role for egalitarian and BicD in salivary gland morphogenesis. A Scheme of the egalitarian (egl) locus indicating the gene structure and the position of the EP identified in the screen. B, C Two phenotypes observed in the screen upon overexpression of egl using EP(2)938: bent (B) and shortened (C) glands. The GFP-EFGas2 maker is in green, phalloidin labelling in red. B and C are widefield fluorescent images. D, D Overexpression of a UAS-egl construct using fkhgal4 frequently led to short glands at stage 14, with some GFP-positive cells losing contact with the glands (arrow in D). E, E shows a comparable wild-type embryo. Crumbs is in red in D and E and as a single channel in D and E, SrcGFP is in green in D, and Shot is in green in E. F-G Dorso-lateral views of stage 14 embryos. F, F egl mutant embryos often show a disrupted epidermis with mislocalized Crumbs labelling (arrows in F point to areas where Crumbs is completely absent), whereas in wild-type embryos Crumbs is localized apicalcircumferential in all epithelial cell (G, G ). H, H egl mutant embryo showing a disorganized salivary gland placode, with a too large and extended invagination hole. I, I A stage 13 egl mutant embryo with a gland that appears to wide and short, showing mislocalized Crumbs labelling. K, K Ventral view of a stage 13 egl mutant embryo. K is a surface confocal stack, showing two disrupted areas in the epidermis where the glands invaginated (arrows). K shows an internal confocal stack of the same embryo, with a too wide and aberrant gland (the red dotted line traces the 53

54 outline of the gland). L, L Higher magnification of the gland shown in K in a smaller confocal stack. Note the mislocalized Crumbs protein at the lateral sides of cells and internally (arrow in L ) that cannot be seen in wild-type glands (compare to M ). H-L Crumbs labelling is in red in H, I, K, L, and as a single channel in H, I, K, L, phalloidin labelling is in green in H, I, K, L, and as a single channel in L. M, M Magnification of a section through a stage 13 wild-type gland, SrcGFP is in green in M, Crumbs is in red in M and as a single channel in M. Figure 7. bitesize overexpression reveals a potential role for bitesize in salivary gland morphogenesis. A Scheme of the bitesize (btsz) locus indicating the gene structure and the position and orientation of the EP identified in the screen. B-C Potential knock-down of Bitesize through overexpression of antisense RNA using EP(3)3567 lead to epithelial defects in the overexpressing cells (arrow in B ) and glands that invaginate with aberrant morphology. B, B shows a surface confocal stack of a stage 13 embryo, C, C shows the corresponding internal stack to reveal the glands. Note the absence of Crumbs labelling in the area that shows GFP-EFGas2 marker expression (arrow in B ). GFP-EFGas2 is in green in B and C, Crumbs is in red in B and C and as a single channel in B and C. D-E show ventral views of btsz K13-4 mutant versus wild-type embryos at stage13. The arrow in D point to the disrupted epidermis in the mutant embryo. Note the disorganization of the placode area compared to wild-type (indicated by the brackets in D ad E), and the failure to accumulate Shot apically (arrow in E indicates the accumulation in the wild-type). Crumbs is in red in D and E, and as a single channel in D and E, Shot is in green 54

55 in D and E and as a single channel in D and E. F-I show examples of lateral views of btsz K13-4 mutant versus wild-type embryos at stage 14. F-F Highly disrupted and disorganized epidermis in the btsz K13-4 mutant (arrows point to areas lacking apical circumferential Crumbs labelling, compare to the wild-type epidermis in I-I ). G-G show an internal stack of the same embryo as in F (the corresponding internal stack for the wild-type embryo in I is shown in K-K ). Note that the salivary gland of btsz K13-4 mutant embryo is losing apical Crumbs accumulation (G ) compared to the wild-type (K ), the phalloidin labelling in G still shows cell outlines, but these also lack apical actin accumulation as seen in the wild-type (K ). Crumbs labelling is in green in F, G, I, K, and as a single channel in F, G, I, K, phalloidin is in red in F, G, I, K and as a single channel in F, G, I, K. H and L show confocal stacks of the embryos in F and I at the level where the salivary duct reaches the epidermis labelled for Crumbs, H shows the of btsz K13-4 mutant and L the wild-type. Note that the duct shown in H lost apical Crumbs accumulation (the arrow points to the remnants of Crumbs labelling in the duct) and that the epidermis at the point from where the glands invaginated is disrupted and lacks apical Crumbs (indicated by the bar in H). M-N The btsz J5-2 mutant at stage 14 also shows disrupted epidermis and loss of Crumbs (M ) and also DE-Cadherin (M ) in the area where the placode was previously located. Crumbs labelling in the invaginated gland is aberrant (N ) whereas DE-Cadherin appears still apical (N ). M-M is a projection of a 5µm thick confocal surface stack, whereas N-N shows the projection of a 20µm thick internal stack covering the whole gland. DE-Cadherin is in green in M and N and as a single channel in M and N, Crumbs is in red in M and N and as a single channel in M and N. 55

56 Figure 8. capricious overexpression reveals a potential role for capricious and tartan in salivary gland morphogenesis. A Scheme of the capricious (caps) locus indicating the gene structure and the position of the EP identified in the screen. B, C show live images of the GFP- EFGas2 marker of the caps overexpression phenotype using EP(3)552 observed in stage 15 embryos in the screen. D-E Confocal stacks of two examples of aberrantly shaped lumen of salivary glands at stage 15 upon overexpression of a UAS-caps construct using fkhgal4. The lumen is highlighted by crumbs labelling in D and E and very much resembles the defects observed in the screen. F-H Shown are examples of invaginating glands at stage 12. F, F is a surface stack of the primordium upon UAS-caps overexpression. Note that the hole at the invagination point is too extended and not positioned completely within the primordium (as highlighted by the GFP marker) compared to the wild-type primordium shown in H, H, G, G show a complete stack of the glands at stage 12 upon UAS-caps overexpression. Note that the size of the invagination hole (marked by the red dotted lines in G ) is again too large and irregular compared to wild-type, and the invaginated portion of the glands shows a too wide and irregular lumen. The GFP- EFGas2 marker is green and Crumbs labelling in red in D, E, F, G, H and Crumbs is shown as a single channel in D, E, F, G. I-M caps PB1 single, trn 28.4 single and caps Del1 trn 28.4 double mutants (all are null mutations, (MAO et al. 2008)) often show defects in salivary gland morphology, i.e. irregular lumen at different stages of invagination. Staining for the fly spectraplakin Shot is in green in and Crumbs in red in I, K, L and M, and Crumbs is shown as a single channel in I and K. N-S A lacz 56

57 containing P-element insertion into the caps locus reveals that caps is not expressed in most cells of the salivary glands. N and M show β-galactosidase (βgal) labelling at stage 12 and 15, respectively. P-S show cross sections of a gland at stage 14. Note that the glands are surrounded by cells expressing caps. N-S The outline of the glands is marked by a white dotted line, β-gal labelling is in green and Crumbs in red. T-U A lacz containing P-element insertion in the trn locus reveals that trn is expressed in salivary gland cells at all stages. T, T Most cells of the salivary gland placode at stage 11 express trn at varying levels (border of the placode is marked by dotted lines). U At stage 14 trn is still expressed strongly in all salivary gland cells including the duct. The outline of the gland is indicated by a dotted line. β-gal labelling is in green and Crumbs in red in T and U, β-gal is shown as a single channel in T. Supplementary Figure 1. In situ hybridization for chic mrna in control and chic-overexpressing embryos. A-B chic mrna expression pattern in wild-type marker-line embryos at stage 14 (A-A ) and stage 15 (B-B ). A and B show overlays of the in situ signal (that is shown as a separate channel in A and B ) false-coloured in red with the SrcGFP marker signal in green (also shown as a single channel in A and B ). C-E show chic mrna levels upon chic overexpression using EP(2)713 x srcgfp fkhgl4 at stage 12 (C), stage 14 (D) and stage 15 (E). Supplementary Figure 2. In situ hybridization for btsz mrna in control and E(3)3567-overexpressing embryos. 57

58 btsz mrna expression pattern in control (A and D; srcgfp fkhgl4) and EP(3)3567 x fkhgal4 (B and C; E) embryos at stage 13 (A-C) and stage 15 (D and E). Note that btsz mrna is strongly enriched at the apical surface of the salivary gland cells in the control (arrows in A and D) and appears reduced in the glands in the EP(3)3567-expressing embryos (arrows in B, C and E). Supplementary Figure 3. In situ hybridization for caps mrna. caps mrna expression pattern in wild-type marker-line embryos at stage 14 (A- A ) and stage 15 in a ventral (B-B ) and lateral (C-C ) view. A, B and C show overlays of the in situ signal (that is shown as a separate channel in A, B and C ) false-coloured in red with the SrcGFP marker signal in green (also shown as a single channel in A, B and C ). Supplementary Figure 4. In situ hybridization for trn mrna. trn mrna expression pattern in wild-type marker-line embryos at stage 13 (A-A ) and stage 15 (B-B ). A and B show overlays of the in situ signal (that is shown as a separate channel in A and B ) false-coloured in red with the SrcGFP marker signal in green (also shown as a single channel in A and B ). Supplementary Table 1. Table of all EP lines analysed in the screen. This table list all EP lines that were analyzed in the screen. Lines without a phenotype are marked in blue as generally unremarkable, genes picked up in the first pass of live screening that were not confirmed in the second examination are marked in light yellow as HIT first pass, not confirmed, and the confirmed genes 58

59 are marked in bright yellow as HIT. Green labelling of a gene indicates that several P-insertion into this locus were analyzed in the screen. The existing data column lists which gene the EP is annotated to be inserted in according to Flybase ( but as this does often not represent the gene that would be overexpressed, we have added the expression information where possible. 59

60 LITERATURE CITED ABOOBAKER, A. A., P. TOMANCAK, N. PATEL, G. M. RUBIN and E. C. LAI, 2005 Drosophila micrornas exhibit diverse spatial expression patterns during embryonic development. Proc. Natl. Acad. Sci. USA 102: ABRAMS, E. W., and D. J. ANDREW, 2005 CrebA regulates secretory activity in the Drosophila salivary gland and epidermis. Development 132: ABRAMS, E. W., W. K. MIHOULIDES and D. J. ANDREW, 2006 Fork head and Sage maintain a uniform and patent salivary gland lumen through regulation of two downstream target genes, PH4alphaSG1 and PH4alphaSG2. Development 133: ANDREW, D. J., A. BAIG, P. BHANOT, S. M. SMOLIK and K. D. HENDERSON, 1997 The Drosophila dcreb-a gene is required for dorsal/ventral patterning of the larval cuticle. Development 124: BARROS, T. P., K. KINOSHITA, A. A. HYMAN and J. W. RAFF, 2005 Aurora A activates D-TACC-Msps complexes exclusively at centrosomes to stabilize centrosomal microtubules. J. Cell Biol. 170: BATE, M., and A. MARTINEZ ARIAS, 1993 The Development of Drosophila melanogaster, Volume II. Cold Spring Harbour Laboratory Press. BEJARANO, F., C. M. LUQUE, H. HERRANZ, G. SORROSAL, N. RAFEL et al., 2008 A gainof-function suppressor screen for genes involved in dorsal-ventral boundary formation in the Drosophila wing. Genetics 178:

61 BENLALI, A., I. DRASKOVIC, D. J. HAZELETT and J. E. TREISMAN, 2000 act up controls actin polymerization to alter cell shape and restrict Hedgehog signaling in the Drosophila eye disc. Cell 101: BORNEMANN, D. J., J. E. DUNCAN, W. STAATZ, S. SELLECK and R. WARRIOR, 2004 Abrogation of heparan sulfate synthesis in Drosophila disrupts the Wingless, Hedgehog and Decapentaplegic signaling pathways. Development 131: BRADLEY, P. L., M. M. MYAT, C. A. COMEAUX and D. J. ANDREW, 2003 Posterior migration of the salivary gland requires an intact visceral mesoderm and integrin function. Dev. Biol. 257: BRAY, S., H. MUSISI and M. BIENZ, 2005 Bre1 is required for Notch signaling and histone modification. Dev. Cell 8: BULLOCK, S. L., and D. ISH-HOROWICZ, 2001 Conserved signals and machinery for RNA transport in Drosophila oogenesis and embryogenesis. Nature 414: BULLOCK, S. L., A. NICOL, S. P. GROSS and D. ZICHA, 2006 Guidance of bidirectional motor complexes by mrna cargoes through control of dynein number and activity. Curr. Biol. 16: BULLOCK, S. L., D. ZICHA and D. ISH-HOROWICZ, 2003 The Drosophila hairy RNA localization signal modulates the kinetics of cytoplasmic mrna transport. Embo J. 22: CAMPOS-ORTEGA, J. A., and V. HARTENSTEIN, 1985 The embryonic development of Drosophila melanogaster. Springer-Verlag, Berlin. 61

62 CHA, G. H., K. S. CHO, J. H. LEE, M. KIM, E. KIM et al., 2003 Discrete functions of TRAF1 and TRAF2 in Drosophila melanogaster mediated by c-jun N-terminal kinase and NF-kappaB-dependent signaling pathways. Mol. Cell Biol. 23: CHANDRASEKARAN, V., and S. K. BECKENDORF, 2003 senseless is necessary for the survival of embryonic salivary glands in Drosophila. Development 130: CHANDRASEKARAN, V., and S. K. BECKENDORF, 2005 Tec29 controls actin remodeling and endoreplication during invagination of the Drosophila embryonic salivary glands. Development 132: COLAS, J. F., and G. C. SCHOENWOLF, 2001 Towards a cellular and molecular understanding of neurulation. Dev. Dyn. 221: DAUBRESSE, G., R. DEURING, L. MOORE, O. PAPOULAS, I. ZAKRAJSEK et al., 1999 The Drosophila kismet gene is related to chromatin-remodeling factors and is required for both segmentation and segment identity. Development 126: FLORENCE, B., A. GUICHET, A. EPHRUSSI and A. LAUGHON, 1997 Ftz-F1 is a cofactor in Ftz activation of the Drosophila engrailed gene. Development 124: FREEMAN, M., B. E. KIMMEL and G. M. RUBIN, 1992 Identifying targets of the rough homeobox gene of Drosophila: evidence that rhomboid functions in eye development. Development 116: HABERMAN, A. S., D. D. ISAAC and D. J. ANDREW, 2003 Specification of cell fates within the salivary gland primordium. Dev. Biol. 258:

63 HARRIS, K. E., and S. K. BECKENDORF, 2007 Different Wnt signals act through the Frizzled and RYK receptors during Drosophila salivary gland migration. Development 134: HARRIS, K. E., N. SCHNITTKE and S. K. BECKENDORF, 2007 Two ligands signal through the Drosophila PDGF/VEGF receptor to ensure proper salivary gland positioning. Mech. Dev. 124: HENDERSON, K. D., and D. J. ANDREW, 2000 Regulation and function of Scr, exd, and hth in the Drosophila salivary gland. Dev. Biol. 217: HENDERSON, K. D., D. D. ISAAC and D. J. ANDREW, 1999 Cell fate specification in the Drosophila salivary gland: the integration of homeotic gene function with the DPP signaling cascade. Dev. Biol. 205: HOGAN, B. L., and P. A. KOLODZIEJ, 2002 Organogenesis: molecular mechanisms of tubulogenesis. Nat. Rev. Genet. 3: HORNE-BADOVINAC, S., and D. BILDER, 2008 Dynein regulates epithelial polarity and the apical localization of stardust A mrna. PLoS Genet. 4: e8. HOUALLA, T., D. HIEN VUONG, W. RUAN, B. SUTER and Y. RAO, 2005 The Ste20-like kinase misshapen functions together with Bicaudal-D and dynein in driving nuclear migration in the developing drosophila eye. Mech. Dev. 122: JONES, N. A., Y. M. KUO, Y. H. SUN and S. K. BECKENDORF, 1998 The Drosophila Pax gene eye gone is required for embryonic salivary duct development. Development 125: KALTSCHMIDT, J. A., C. M. DAVIDSON, N. H. BROWN and A. H. BRAND, 2000 Rotation and asymmetry of the mitotic spindle direct asymmetric cell division in the developing central nervous system. Nat. Cell Biol. 2:

64 KOLESNIKOV, T., and S. K. BECKENDORF, 2005 NETRIN and SLIT guide salivary gland migration. Dev. Biol. 284: KOPPEN, M., B. G. FERNANDEZ, L. CARVALHO, A. JACINTO and C. P. HEISENBERG, 2006 Coordinated cell-shape changes control epithelial movement in zebrafish and Drosophila. Development 133: KRAUSE, C., C. WOLF, J. HEMPHALA, C. SAMAKOVLIS and R. SCHUH, 2006 Distinct functions of the leucine-rich repeat transmembrane proteins capricious and tartan in the Drosophila tracheal morphogenesis. Dev. Biol. 296: KUO, Y. M., N. JONES, B. ZHOU, S. PANZER, V. LARSON et al., 1996 Salivary duct determination in Drosophila: roles of the EGF receptor signalling pathway and the transcription factors fork head and trachealess. Development 122: KURANAGA, E., H. KANUKA, T. IGAKI, K. SAWAMOTO, H. ICHIJO et al., 2002 Reapermediated inhibition of DIAP1-induced DTRAF1 degradation results in activation of JNK in Drosophila. Nat. Cell Biol. 4: KUSANO, A., C. STABER and B. GANETZKY, 2001 Nuclear mislocalization of enzymatically active RanGAP causes segregation distortion in Drosophila. Dev. Cell 1: LEAMAN, D., P. Y. CHEN, J. FAK, A. YALCIN, M. PEARCE et al., 2005 Antisensemediated depletion reveals essential and specific functions of micrornas in Drosophila development. Cell 121: LEPTIN, M., 2005 Gastrulation movements: the logic and the nuts and bolts. Dev. Cell 8:

65 LI, Z., L. WANG, T. S. HAYS and Y. CAI, 2008 Dynein-mediated apical localization of crumbs transcripts is required for Crumbs activity in epithelial polarity. J. Cell Biol. 180: LIU, Y., and M. LEHMANN, 2008 Genes and biological processes controlled by the Drosophila FOXA orthologue Fork head. Insect Mol. Biol. 17: LUBARSKY, B., and M. A. KRASNOW, 2003 Tube morphogenesis: making and shaping biological tubes. Cell 112: MAO, Y., M. KERR and M. FREEMAN, 2008 Modulation of Drosophila retinal epithelial integrity by the adhesion proteins capricious and tartan. PLoS ONE 3: e1827. MILAN, M., U. WEIHE, L. PEREZ and S. M. COHEN, 2001 The LRR proteins capricious and Tartan mediate cell interactions during DV boundary formation in the Drosophila wing. Cell 106: MINAKHINA, S., R. MYERS, M. DRUZHININA and R. STEWARD, 2005 Crosstalk between the actin cytoskeleton and Ran-mediated nuclear transport. BMC Cell Biol. 6: 32. MOLNAR, C., A. LOPEZ-VAREA, R. HERNANDEZ and J. F. DE CELIS, 2006 A gain-offunction screen identifying genes required for vein formation in the Drosophila melanogaster wing. Genetics 174: MYAT, M. M., and D. J. ANDREW, 2000a Fork head prevents apoptosis and promotes cell shape change during formation of the Drosophila salivary glands. Development 127: MYAT, M. M., and D. J. ANDREW, 2000b Organ shape in the Drosophila salivary gland is controlled by regulated, sequential internalization of the primordia. Development 127:

66 MYAT, M. M., and D. J. ANDREW, 2002 Epithelial tube morphology is determined by the polarized growth and delivery of apical membrane. Cell 111: NAVARRO, C., H. PUTHALAKATH, J. M. ADAMS, A. STRASSER and R. LEHMANN, 2004 Egalitarian binds dynein light chain to establish oocyte polarity and maintain oocyte fate. Nat. Cell Biol. 6: PANZER, S., D. WEIGEL and S. K. BECKENDORF, 1992 Organogenesis in Drosophila melanogaster: embryonic salivary gland determination is controlled by homeotic and dorsoventral patterning genes. Development 114: PAUL, S. M., M. TERNET, P. M. SALVATERRA and G. J. BEITEL, 2003 The Na+/K+ ATPase is required for septate junction function and epithelial tube-size control in the Drosophila tracheal system. Development 130: PILOT, F., J. M. PHILIPPE, C. LEMMERS and T. LECUIT, 2006 Spatial control of actin organization at adherens junctions by a synaptotagmin-like protein Btsz. Nature 442: PIRRAGLIA, C., R. JATTANI and M. M. MYAT, 2006 Rac function in epithelial tube morphogenesis. Dev. Biol. 290: ROTH, G. E., S. WATTLER, H. BORNSCHEIN, M. LEHMANN and G. KORGE, 1999 Structure and regulation of the salivary gland secretion protein gene Sgs-1 of Drosophila melanogaster. Genetics 153: RØRTH, P., K. SZABO, A. BAILEY, T. LAVERTY, J. REHM et al., 1998 Systematic gain-offunction genetics in Drosophila. Development 125: SCHENKEL, H., S. HANKE, C. DE LORENZO, R. SCHMITT and B. M. MECHLER, 2002 P elements inserted in the vicinity of or within the Drosophila snrnp SmD3 66

67 gene nested in the first intron of the Ornithine Decarboxylase Antizyme gene affect only the expression of SmD3. Genetics 161: SCHWEITZER, R., R. HOWES, R. SMITH, B. Z. SHILO and M. FREEMAN, 1995 Inhibition of Drosophila EGF receptor activation by the secreted protein Argos. Nature 376: SERANO, J., and G. M. RUBIN, 2003 The Drosophila synaptotagmin-like protein bitesize is required for growth and has mrna localization sequences within its open reading frame. Proc. Natl. Acad. Sci. USA 100: SHILO, B. Z., 2005 Regulating the dynamics of EGF receptor signaling in space and time. Development 132: SHINZA-KAMEDA, M., E. TAKASU, K. SAKURAI, S. HAYASHI and A. NOSE, 2006 Regulation of layer-specific targeting by reciprocal expression of a cell adhesion molecule, capricious. Neuron 49: SHISHIDO, E., M. TAKEICHI and A. NOSE, 1998 Drosophila synapse formation: regulation by transmembrane protein with Leu-rich repeats, CAPRICIOUS. Science 280: STEIGEMANN, P., A. MOLITOR, S. FELLERT, H. JACKLE and G. VORBRUGGEN, 2004 Heparan sulfate proteoglycan syndecan promotes axonal and myotube guidance by slit/robo signaling. Curr. Biol. 14: STRUMPF, D., and T. VOLK, 1998 Kakapo, a novel cytoskeletal-associated protein is essential for the restricted localization of the neuregulin-like factor, vein, at the muscle-tendon junction site. J. Cell Biol. 143:

68 SUBRAMANIAN, A., A. PROKOP, M. YAMAMOTO, K. SUGIMURA, T. UEMURA et al., 2003 Shortstop recruits EB1/APC1 and promotes microtubule assembly at the muscle-tendon junction. Curr. Biol. 13: SVENDSEN, P. C., S. D. MARSHALL, M. KYBA and W. J. BROOK, 2000 The combgap locus encodes a zinc-finger protein that regulates cubitus interruptus during limb development in Drosophila melanogaster. Development 127: SZUL, T., R. GRABSKI, S. LYONS, Y. MOROHASHI, S. SHESTOPAL et al., 2007 Dissecting the role of the ARF guanine nucleotide exchange factor GBF1 in Golgi biogenesis and protein trafficking. J. Cell Sci. 120: TAUTZ, D., and C. PFEIFLE, 1989 A non-radioactive in situ hybridization method for the localization of specific RNAs in Drosophila embryos reveals translational control of the segmentation gene hunchback. Chromosoma 98: THE, I., Y. BELLAICHE and N. PERRIMON, 1999 Hedgehog movement is regulated through tout velu-dependent synthesis of a heparan sulfate proteoglycan. Mol. Cell 4: TOWNSLEY, F. M., and M. BIENZ, 2000 Actin-dependent membrane association of a Drosophila epithelial APC protein and its effect on junctional Armadillo. Curr. Biol. 10: VERHEYEN, E. M., and L. COOLEY, 1994 Profilin mutations disrupt multiple actindependent processes during Drosophila development. Development 120:

69 VINING, M. S., P. L. BRADLEY, C. A. COMEAUX and D. J. ANDREW, 2005 Organ positioning in Drosophila requires complex tissue-tissue interactions. Dev. Biol. 287: WEI, Y., L. YU, J. BOWEN, M. A. GOROVSKY and C. D. ALLIS, 1999 Phosphorylation of histone H3 is required for proper chromosome condensation and segregation. Cell 97: XU, H., J. A. BRILL, J. HSIEN, R. MCBRIDE, G. L. BOULIANNE et al., 2002 Syntaxin 5 is required for cytokinesis and spermatid differentiation in Drosophila. Dev. Biol. 251: XU, N., B. KEUNG and M. M. MYAT, 2008 Rho GTPase controls invagination and cohesive migration of the Drosophila salivary gland through Crumbs and Rho-kinase. Dev. Biol. YARMOLA, E. G., and M. R. BUBB, 2006 Profilin: emerging concepts and lingering misconceptions. Trends Biochem. Sci. 31: ZHOU, B., A. BAGRI and S. K. BECKENDORF, 2001 Salivary gland determination in Drosophila: a salivary-specific, fork head enhancer integrates spatial pattern and allows fork head autoregulation. Dev. Biol. 237: ZHU, S., R. PEREZ, M. PAN and T. LEE, 2005 Requirement of Cul3 for axonal arborization and dendritic elaboration in Drosophila mushroom body neurons. J. Neurosci. 25:

70

71

72

73

74

75

76

77

78

79

80

81