Apical accumulation of the Drosophila PDGF/VEGF receptor ligands provides a mechanism for triggering localized actin polymerization

Epithelial tissue functions depend largely on a polarized organization of the individual cells. The characteristic cell of a simple monolayered epithelium is polarized along an apicobasal axis, in that its plasma membrane is separated into functionally distinct apical and basolateral domains(Drubin and Nelson, 1996). The apical membrane domain communicates to neighboring cells via membrane proteins, such as cell-adhesion molecules and other cell-surface receptors. These cell-adhesion molecules are often organized in large specialized multi-protein complexes forming different cell junctions(Borrmann et al., 2000; Nagafuchi, 2001). In invertebrates, epithelial junctional complexes are composed of two types of junctions, the adherens junctions and the septate junctions(Muller, 2000; Tepass et al., 2001).

Among the molecular components of adherens junctions, cadherins and catenins comprise the core of an adhesive interaction that connects subapical actin belts of adjacent cells with each other(Tepass, 2002). This association with the cytoskeletal network is necessary for stable cell-cell adhesion and for the integration of these contacts with the morphology of epithelial cells. Several signaling pathways that are activated by cell adhesion have been identified, most notably those regulated by small Rho-like GTPases (Braga, 2002). However,the molecular mechanisms governing the interactions between adherens junction components and the actin cytoskeleton remain largely unresolved issues.

Septate junctions are maintained by localization of intracellular PDZ domain proteins such as Discs large (DLG) and Scribble. In combination with their ability to multimerize, PDZ proteins have the potential to assemble large multiprotein complexes at the cell membrane(Bilder, 2001).

The generation and maintenance of polarity in epithelial cells is intimately linked to the polarized distribution of transmembrane and secreted proteins. The junctions between cells form a barrier not only to diffusion of extracellular molecules between the cells, but also to movement of molecules within the plasma membrane. Thus, in order to localize a transmembrane protein to a particular cellular compartment (apical or basolateral), it must be targeted to secretory vesicles that will specifically fuse with that membrane compartment (Mostov and Cardone,1995; Simons and Ikonen,1997; Wandinger-Ness et al.,1990).

In this study, we identify broad expression of the DrosophilaPDGF/VEGF receptor (PVR) in epithelial tissues. We specifically examined possible roles of the PVR pathway in the wing disc epithelium. Two of the PVR ligands, PVF1 and PVF3, are deposited within the apical extracellular space,suggesting that polarized apical activation of the receptor may take place. Clones for null alleles of Pvr showed no phenotypes in the wing disc or pupal wing. However, when uniform activation of PVR was induced by a constitutively dimerizing receptor, loss of epithelial polarity, uniform appearance of adherens junctions and tumorous growth were observed in the wing disc. Elevation of full-length PVR levels also gave rise to prominent phenotypes, characterized by significantly higher levels of F-actin in the basolateral area of the epithelium. Taken together these results point to the importance of polarized apical PVR activation in maintenance of the wing disc monolayered epithelium, via regulation of localized actin microfilament polymerization.

The full-length Pvr construct was generated by combining the signal-peptide encoding region from EST SD02385 to the remaining Pvr-coding region encoded by EST SD03187, to generate a full length,intronless transcript. After verification of an intact ORF, the coding region was cloned into pUAST. To generate the UAS-dsRNA-Pvr construct, a cDNA fragment containing bases 825-2403 in an antisense orientation (numbers correspond to bases after the initiator ATG), followed by bases 1277-2403(sense orientation) were inserted into pUAST. The RNA transcribed from this construct should thus generate a stem of 1126 and a loop of 452 bases. Transgenic flies were generated from each construct. pUAST-Pvf1 was generated from EST LD28763, including its 5′ and 3′ non-coding regions. pUAST-Pvf3 was generated by PCR amplification of the full ORF of the Vegf 27Ca cDNA plasmid (GenBank Accession Number AY079183) provided by Exelixis. pUAS-sGFP was generated by fusing the Spitz signal peptide to EGFP.

The following Gal4 driver lines were used to express the various transgenic constructs: 69B-Gal4 (expressed uniformly in the embryonic ectoderm), actin-Gal4 (expressed ubiquitously), MS1096-Gal4(expressed in the pouch of the wing imaginal disc), patched-Gal4 and dpp-Gal4 (expressed along the anterior-posterior border of the wing disc), GMR-Gal4 (expressed in differentiating photoreceptor cells),and eyeless-Gal4 (expressed in all eye precursor cells). For generation of Pvf1 `flip out' clones the pUAST-Pvf1 line was crossed to a line containing hs-flp, UAS-GFP and actin>CD2>Gal4. To generate larval clones, a 30-minute heat shock at 37°C was applied at 72-96 hours AEL. Flies containingλ Pvr were obtained from P. Rorth.

The Pvr mutant Vegfr^C2195 [obtained from Exelixis (Cho et al., 2002)]was recombined with FRT 40A. Clones were generated following a cross to hs-flp; FRT 40A, ubi-GFP (Bloomington #5629) or hs-flp; FRT 40A, arm-lacZ (Bloomington #6579). Three heat shocks of 1.5 hours each at 37°C were applied, starting at 24-48 hours AEL, at 24 hour intervals.

EST SD02385 was used to insert the region encoding PVR C-terminal residues 1291-1509 into pRSETB. Recombinant protein was purified on a NINTA column, and injected into rats. Typically, the antibody is diluted 1/200 for immunohistochemical staining. EST LD28763 was used to insert the region encoding PVF1 residues 27-303 into pRSETA. Recombinant protein was injected into rats. The antibody is diluted 1/100 for staining. To generate antibodies recognizing PVF3, PCR amplification of the region encoding residues 22-359 was carried out. The template used was Vegf 27Ca cDNA. The fragment was cloned into pRSETA. The purified recombinant protein was injected into guinea pigs,and the antibody used at a dilution of 1/100 for staining.

Rabbit anti-DLG was obtained from P. Bryant, mouse anti-LGL antibodies were obtained from F. Matsuzaki, Profilin antibodies (chi1J) from the Developmental Studies Hybridoma bank, TRITC-labeled phalloidin was purchased from Sigma, and Alexa 488 phalloidin and fluorescein conjugate deoxyribonuclease I from Molecular Probes. Monoclonal anti-GFP antibodies were obtained from Roche, and diluted 1/10 for staining, and rabbit anti-beta galactosidase from Cappel(diluted 1/1000). Secondary antibodies were obtained from Jackson ImmunoResearch, and diluted at 1/300 for staining.

Antibody staining of wing discs was according to Neumann and Cohen(Neumann and Cohen, 1997). Staining of extracellular distribution of PVF1 and PVF3 was as described previously (Strigini and Cohen,2000). For extracellular staining, antibodies were used at a dilution of 1/30 for PVF1 or PVF3, and 1/3 for anti-GFP. Incubation with these antibodies was carried out on ice, to block endocytosis. Staining of pupal wings was as described previously (Fristrom et al., 1994) with the following modifications: Pupal discs were fixed at 4°C for 48 hours, and mounted in Aquamount after staining.

For sections, wing discs were fixed in 4% glutaraldehyde and 4%formaldehyde, embedded in JB4 (Polysciences), and stained in Hematoxylin and Eosin after sectioning.

For EM, discs were fixed by immersion in freshly prepared 3%paraformaldehyde, 2% glutaraldehyde in cacodylate buffer containing 5 mM CaCl₂ (pH 7.4) for 2 hours at room temperature and for 12 hours at 4°C. After washing, the tissue was post fixed in osmium tetroxide, 0.5%potassium dichromate and 0.5% potassium hexacyanoferrate in cacodylate buffer for 1 hour. The tissue was stained and blocked with 2% aqueous uranyl acetate followed by ethanol dehydration. The samples were embedded in EMBED 812. Thin sections were cut using a diamond knife, stained with 2% uranyl acetate and Reynolds's lead citrate, and examined with a transmission electron microscope(Philips, CM12) at an accelerating voltage of 120 kV.

Schneider S2 cells were transiently transfected with actin-Gal4 and pUAST-PVF1 or pUAST-sGFP plasmids, and grown in serum-free medium for 3 days. Medium was collected and incubated for 3 hours with Heparin-Sepharose cl-6B beads (Pharmacia). Beads were washed with PBS, and bound material eluted with 1.5 M NaCl. The presence of PVF1 or sGFP in the medium was monitored by western blotting with anti-PVF1 or anti-GFP, respectively.

Antibodies raised against the C-terminal tail of PVR were used to determine specific sites of expression and potential activity. The antibody detects PVR as a membrane-associated protein, throughout development. In embryos the protein is prominent in tissues that showed high levels of Pvr RNA expression, including hemocytes and midline glial cells. In addition,expression throughout the embryonic ectoderm, beginning at stage 14, was identified (Fig. 1A-C).

The broad epithelial distribution in the embryo prompted us to examine the expression in epithelial cells at later stages of development. Indeed, in the eye imaginal disc of the third instar larvae, the receptor is also broadly expressed around the cell circumference(Fig. 1D). In the wing imaginal disc, uniform expression in the wing pouch and notum was detected. Optical cross-sections show that the receptor is present on both basolateral and apical cell surfaces in this tissue. Uniform expression in the wing continues during pupal stages (Fig. 1E-G). We focused our subsequent analysis of PVR function on the wing disc epithelium.

Three putative PVR ligands have been identified following Drosophila genome sequencing. One of them, termed PVF1, was also identified in a functional ovary misexpression screen and shown to activate the receptor in cell culture assays. It represents the major PVR ligand in border cells, a migratory subset of ovarian follicle cells(Duchek et al., 2001). PVF2 was shown to be required for viability and proliferation of larval hemocytes(Munier et al., 2002). The third ligand, PVF3, has a redundant role with the other two ligands, in promoting embryonic hemocyte migration (Cho et al., 2002). We generated antibodies against PVF1 and demonstrated their capacity to recognize this protein in embryos, in a pattern similar to the one observed for the Pvf1 transcript by RNA in situ hybridization (not shown). These antibodies do not cross react with the other two ligands, as demonstrated by their failure to detect expression of PVF2 and PVF3 in the embryonic midline. In larval wing discs, PVF1 is specifically localized to the apical side of the disc epithelium(Fig. 2A). In this domain the disc forms a sac-like structure, with the layer of squamous peripodial cells on the opposite side. The pupal wing is a bilayered structure, with the basal side of each layer facing the other, and the apical sides facing outward. PVF1 localization in the pupal wing remained apical(Fig. 2I). Antibodies were also raised against PVF3, and shown to detect the expected embryonic midline expression (not shown). In the wing disc, they detect a distinct apical distribution, similar to PVF1.

As the endogenous PVF3 was readily detected, we were able to test its segregation between intracellular and extracellular compartments. Live discs were incubated with PVF3 antibodies, and fixed only subsequently. This protocol exclusively detects the extracellular ligand(Strigini and Cohen, 2000). A prominent apical staining is found, indicating that the majority of PVF3 detected in fixed discs is extracellular(Fig. 2B).

Overexpression of PVF1, using MS1096-Gal4, a strong wing-specific driver, retained the apical localization of PVF1(Fig. 2C). Despite the high levels of apical PVF1 (one to two orders of magnitude above endogenous levels), we did not detect any spread of the ligand to the basolateral side,indicating that the cell junctions form an efficient barrier to PVF1 diffusion towards the basolateral domain. Incubation of live discs overexpressing PVF1 with the antibody demonstrated that the majority of apical PVF1 is extracellular (not shown). The restricted apical localization of overexpressed PVF1 was maintained at the pupal stage (not shown).

Both endogenous PVF1 and PVF3 accumulate on the apical side of the disc epithelium. This is unusual, as other ligands used for patterning the wing disc (e.g. WG) were shown to be localized at the basolateral side(Strigini and Cohen, 2000). To ask if PVF1 and PVF3 use a common machinery for their secretion which results in apical accumulation, we overexpressed both ligands simultaneously. Overexpression of PVF1 alone did not perturb its apical accumulation. By contrast, overexpressed PVF3 was detected in a punctate pattern within the wing disc cells (Fig. 2D). The ligand failed to be secreted, probably owing to saturation of specific components of the secretory pathway, and no staining was detected when the extracellular PVF3 was probed by staining live discs (not shown). Co-expression of PVF1 with PVF3 led to a significant reduction in the level of apical PVF1, and elevated levels in punctate structures within the cells(compare Fig. 2C with 2E). In many of these puncta, colocalization of PVF1 and PVF3 was observed(Fig. 2F, inset).

The capacity of PVF3 overexpression to inhibit secretion of PVF1 is specific. When expressed alone, secreted GFP (sGFP) could be detected in both apical and basolateral domains (Fig. 2G). No alterations in this pattern were observed following coexpression with PVF3 (Fig. 2H). Optical sections show that sGFP was located between the cells in the basolateral region, while PVF3 was trapped within the cells(Fig. 2H, inset). These results indicate that PVF1 and PVF3 use a discrete secretory mechanism, leading to exclusive apical deposition, which is distinct from the bulk flow, represented by sGFP.

We next examined whether PVF1 is sequestered following secretion, or free to diffuse within the apical domain. As overexpressed PVF1 maintained its restricted apical localization, we expressed the ligand in clones of cells marked by GFP (`flipout' clones), and monitored its distribution over adjacent cells. Detection was performed under conditions of low laser illumination intensity, which identify only overexpressed PVF1 but not the endogenous ligand. Excess PVF1 could be uniformly detected within the entire apical space of the disc epithelium (Fig. 2J). This result indicates that following apical secretion of PVF1, the ligand is not trapped on the apical extracellular surface of the producing cells, but is capable of extensive lateral diffusion.

One of the hallmarks of vertebrate VEGF molecules is their capacity to bind heparin. This property is mapped to a distinct domain located C-terminal to the receptor-binding domain. Removal of this domain in one of the mammalian VEGF splice isoforms (VEGF¹²¹), generates a molecule that is capable of binding the receptor, while failing to bind heparin(Park et al., 1993a). Although PVF1 shows no distinct homology to the heparin-binding domain of VEGF, it may still have the capacity to bind heparin. To test for heparin-binding properties of PVF1, medium derived from Schneider S2 cells overexpressing PVF1 was collected. Incubation with heparin-Sepharose beads showed a specific and efficient binding of PVF1. Most of the protein was removed from the medium. The bound PVF1 was eluted from the beads at 1.5 M NaCl. No such heparin binding was detected for the control secreted GFP protein(Fig. 3).

A loss-of-function allele of Pvr was previously described(Cho et al., 2002). To asses the possible requirements for PVR during wing disc development, we used this allele to generate Pvr-mutant clones in the wing disc. The size of the clones was comparable with that of wild-type twin clones(Fig. 4A). Epithelial cells of the wing imaginal disc are elongated and form a continuous, highly polarized single-layered epithelium, characterized by apical localization of F-actin belts, at the adherens junctions. Proteins like Discs large (DLG) are localized at the septate junctions, which are positioned below the adherens junctions (Watson et al.,1994). No phenotype was identified in Pvr-mutant clones in terms of the epithelial organization of the tissue, or the polarity of the cells as monitored by apical localization of F-actin(Fig. 4B) and DLG (not shown). When clones were analyzed in the pupal wing, again no phenotypes could be detected using the same criteria (Fig. 4C-E).

Similarly, expression of UAS-Pvr RNAi constructs in wing discs using either MS1096-Gal4 or the spatially restricted driver ptc-Gal4, reduced the level of PVR (see Fig. 7A), but did not result in any phenotypes (not shown).

In view of the possibility of a redundant function for PVR in the wing, we sought to gain insights into the normal function through misexpression experiments. This is especially pertinent as the apical localization of the ligands may imply a spatial control of PVR activation. To examine the possible importance of polarized PVR activation, we expressed a constitutively dimerizing PVR construct, where the extracellular region was replaced by the lambda dimerization domain (λPVR). This construct was previously shown to be potent in the follicular epithelium(Duchek et al., 2001). Expression of UAS-λPvr induced by MS1096-Gal4 gave rise to a dramatic phenotype, which was completely penetrant even when the induction levels were compromised at 18°C. Most larvae continued to grow in size beyond the third instar and failed to pupariate. The wing discs of these larvae were several times larger than wild-type discs, and appeared completely disorganized. Sections of the proliferating discs showed multiple layers missing all aspects of polarity(Fig. 5A-E). Similar giant larvae and tumorous disc phenotypes have been described for loss-of function mutants in dlg, scribble and l(2)gl(Watson et al., 1994). In these mutants, imaginal discs lost apicobasal polarity, and continued proliferating as unorganized, multilayered structures. Consequently the larvae grew in size, and pupariation was not initiated. The level of F-actin within the λPVR-expressing cells was markedly elevated. In addition, instead of forming regular ring structures, microfilaments appeared disorganized. Finally, the apical restriction of F-actin was abolished, and a uniform distribution around the cell circumference was observed(Fig. 5G,H). EM examination of these cells revealed multiple adherens junctions instead of the typical single apical adherens junction per cell (Fig. 5I,J).

Staining of the DLG protein showed a uniform distribution around the cell circumference (Fig. 5K),suggesting that in discs expressing λPVR septate junctions are intact,although not correctly polarized. Localization of a second marker, LGL,allowed us to verify this interpretation. LGL is a protein that exists in a soluble, cytoplasmic form and an insoluble form thought to be associated with membranes and the actin cytoskeleton(Strand et al., 1994). Association of LGL with the membrane requires intact septate junctions,because in dlg or scribble mutants, LGL is found only in the cytoplasm (Bilder et al.,2000). In λPVR-expressing cells, LGL retained its membrane association, possibly to even higher levels than normal(Fig. 5L-N). These observations demonstrate that septate junction components are retained in the membrane ofλPVR cells, and furthermore, they maintain at least some of their roles.

The dramatic effect of λ-PVR expression prompted us to study the phenotypic consequences resulting from overexpression of PVR. Ubiquitous expression of a full-length construct of PVR in the wing disc results in pupal lethality. When we analyzed the wing disc upon overexpression of PVR in a specific domain along the anteroposterior border using the dpp-Gal4driver, we noticed a significant elevation in the level of F-actin along the basolateral area of the epithelium (Fig. 6B,C). Conversely, the levels of G-actin were slightly lower than in the surrounding cells (Fig. 6D) indicating that the elevation in PVR levels enhanced actin polymerization, rather than induced actin transcription. The elevation in actin polymerization was accompanied by elevation in the Chickadee (Profilin)protein that binds actin monomers (Fig. 6F). Consequently, an irregular fold was formed within the domain expressing PVR (Fig. 6H).

As PVR overexpression was lethal, we looked for a way to express the native protein in lower amounts. While analyzing PVR expression in the different lines expressing Pvr-RNAi, we noted one instance in which the levels of PVR in the domain expressing ds-RNA were slightly higher than normal, by approximately twofold (Fig. 7B). We termed this line RNAi-gain of function(RNAi-GOF). The insertion of the UAS construct in RNAi-GOFis in cytogenetic band 27A1-27A2, 1.5 Mb from the endogenous Pvr gene(28F3-28F5). Although the insertion is not in very close proximity, it still raises the possibility of cis-interaction between the loci, upon binding of the Gal4 transcriptional activator, thus accounting for the unexpected rise in PVR levels. In contrast to the other RNAi lines in which PVR levels were reduced (Fig. 7A) that gave no wing phenotype, induction of RNAi-GOF by MS1096-Gal4 resulted in wing blisters(Fig. 7I,J).

We analyzed wing discs that expressed RNAi-GOF under the control of ptc-Gal4. In accordance with the phenotypes caused by UAS-Pvr expression, F-actin organization was altered specifically at the basal side of the epithelium (Fig. 7C). Elevation in Chickadee levels was also observed (not shown). Examination of components of the adherens junction (phospho-tyrosine) as well as septate junctions (DLG protein) revealed no change in the localization or the levels of these proteins (not shown).

We also examined the effects of expressing RNAi-GOF in the pupal wing, to determine the basis for the adult blistering phenotype. In the pupal wing, F-actin in each layer is normally enriched not only at the apical side as in larval discs, but also in the basal region(Fig. 7D,F). Although the organization of apical F-actin was unaffected, an ectopic and highly irregular accumulation of F-actin at the basolateral side of MS1096Gal4/RNAi-GOF pupal wings was observed(Fig. 7E,G).

The specificity of the gain-of-function effect was strengthened by genetic interaction experiments. Females heterozygous for the driver MS1096-Gal4 and RNAi-GOF show a mild wing phenotype(Fig. 7I). When the ligand PVF1 was overexpressed, accumulation of excess ligand at the apical side was observed (Fig. 2C),but no apparent phenotype ensued (Fig. 7K). If expression of the RNAi-GOF construct leads to a gain-of-function phenotype, excess apical PVF1 should result in a more severe phenotype. Indeed, co-expression of RNAi-GOF with Pvf1 gave rise to an enhanced wing phenotype that was fully penetrant. Females containing a single copy each of MS1096-Gal4, UAS-RNAi-GOF and UAS-Pvf1 showed a dramatic enhancement. The wings were not only blistered, but also reduced in size and highly folded (compare Fig. 7L with 7I,K). Thus, the excess apical PVF1 enhanced the effect of RNAi-GOF.

In view of the retention of PVF1 and PVF3 on the apical extracellular surface of the wing imaginal disc, we examined whether a similar localization also takes place in other epithelial tissues. PVF1 was overexpressed in the embryonic ectoderm and the eye imaginal disc. In both cases, no polarized localization of the ligand could be observed either within or outside the cells (Fig. 8A,B). Furthermore,no deleterious phenotypes resulted from this manipulation(Fig. 8A). As the uniform distribution of ectopic PVF1 in the embryonic ectoderm and eye disc is expected to trigger the receptor uniformly in the responding cells we concluded that apical localization of PVF1 in these tissues is not necessary.

The constitutively dimerizing form, λPVR, provides an alternative method to activate the receptor uniformly. We expressed λPVR in embryonic tissues and other larval imaginal discs. No defects were observed in polarized embryonic epithelia (e.g. ectoderm or trachea). Although expression of λPVR results in embryonic lethality, cuticular defects appear to be restricted to holes at the termini, without any overall abnormalities(Fig. 8C). Similarly, in the eye disc, no apparent abnormalities in tissue organization or F-actin distribution were observed following λPVR expression(Fig. 8D,E). These results imply the presence of specific components in the wing disc, which may transmit or facilitate the intracellular responses to polarized activation of PVR, only in tissues where the ligand is apically restricted.

This work examined the involvement of PVR and its ligands in the maintenance of epithelial cell polarity in the wing imaginal disc. Although the receptor is broadly distributed, the ligands are apically restricted, and could lead to polarized receptor activation. Phenotypes resulting from PVR gain-of-function circumstances imply that apically polarized PVR activation is normally essential for regulated polymerization and bundling of actin microfilaments.

Examination of PVR protein revealed a broad expression in epithelial tissues in the embryo from stage 14, and in the imaginal discs. PVR expression is not confined along the apicobasal axis of the cells. In contrast to the uniform distribution of the receptor, there is restricted apical localization of the ligands PVF1 and PVF3 within the wing disc epithelium.

The mechanism responsible for the apical accumulation of PVF1 and PVF3 in the wing disc is intriguing. As cell junctions are likely to form barriers that can not be bypassed by exogenous ligand, apical accumulation may imply preferential secretion of PVR ligands at the apical compartment. It is possible that PVF1 and PVF3 are targeted to vesicles that are specifically marked for secretion at the apical surface. The observation that PVF3 overexpression compromises the secretion of PVF1 but not that of sGFP,supports such a possibility. The presence of distinct secretory vesicles which are targeted to apical versus basolateral compartments has been previously observed (Jacob and Naim,2001).

What further interactions do PVF1 and PVF3 undergo, once secreted to the apical extracellular compartment? One possible interaction involves binding to heparan-sulfate proteoglycans on the cell surface. The vertebrate VEGF proteins have a defined heparin-binding domain at the C terminus that is distinct from the receptor-binding moiety(Park et al., 1993b). The equivalent C-terminal domain of PVF1 does not show a distinct homology to the heparin-binding domain of VEGF. We have shown, however, that PVF1 secreted by S2 cells can bind heparin beads. To examine if PVF1 is trapped on the cell surface following secretion, we created marked clones of cells overexpressing PVF1. We find that the ligand is uniformly redistributed along the entire apical surface, including the surface of cells not secreting the ligand. Although the ligand is capable of spreading readily within the apical plane,it is incapable of crossing the cell junctions, and is thus excluded from the basolateral extracellular compartment.

Accumulation of PVF1 and PVF3 at the extracellular apical compartment implies that the PVR receptor is activated in a polarized fashion. Does such an apically polarized pattern of activation play a role in shaping the wing disc epithelium? The most direct way to examine PVR function in the wing disc is to generate clones for null Pvr alleles, and follow their phenotype. The Pvr-mutant clones were similar in size to their wild-type twins, and within the clones no aberrant morphology or misorganization of actin was detected. We thus conclude that PVR has a redundant role in the wing. Nevertheless, a series of dramatic wing phenotypes is induced following expression of various PVR constructs. Our analysis leads us to propose that these phenotypes represent gain-of-function circumstances following inappropriate activation of PVR on the basolateral side of the wing disc epithelium.

We were first drawn to this interpretation by the dramatic effects of non-restricted and constitutive receptor activation, achieved by expression ofλPVR in the wing disc epithelium. The epithelium lost its polarity,multiple cell layers were generated, and giant tumorous discs were formed. Ectopic accumulation of F-actin around the circumference of the cells was observed, and corroborated by the identification of multiple adherens junctions in EM images.

The phenotype created by expressing λPVR in the wing disc is reminiscent of the phenotype described for loss of the septate junction proteins DLG and Scribbled, as well the LGL protein. We believe that the alterations in cell polarity following λPVR expression are less severe than the dlg, scribble or lgl mutant phenotypes. Although excess adherens junctions are established and the septate junctions are mislocalized, the LGL protein, which requires intact septate junctions for its insertion into the membrane, is found associated with the membrane in wing discs expressing λPVR. The tumorous growth of the cells is believed to be a secondary consequence of the loss of polarity, which may lead to impairment of cell-cell communication.

We examined also the consequences of misexpressing full-length PVR. We noticed significantly higher levels of F-actin in the basolateral area of the cells expressing PVR, while the level of actin monomers was lower. This indicates that PVR has a localized effect on actin polymerization, rather than a general role in actin monomer synthesis. We also noticed elevation in Profilin (chickadee) protein levels. Profilin binds actin monomers in a way that inhibits nucleation and elongation of pointed ends but promotes rapid elongation of uncapped barbed ends, leading to depletion of the actin monomer pool (Amann and Pollard,2000).

Overexpression of the ligands alone did not lead to any phenotype, while even mild overexpression of the receptor resulted in pronounced phenotypes. Moreover, these phenotypes were strongly enhanced by elevating the levels of PVF1 in the apical domain. There are two possible explanations for the overexpression phenotype. The overall levels of receptor activation may be important. The receptor could be present in limited amounts, so that increasing its levels allows more ligand at the apical side to bind and activate receptors. Alternatively, polarized activation of the receptor that normally takes place is disturbed, because of redistribution of the ligand. This may happen by recycling of the ligand-bound receptor inside the cells. The fact that elevation in PVR levels resulted in basolateral polymerization of actin, while the ligand is normally found on the apical side supports this possibility. Mislocalized activation of the receptor may also take place because of spontaneous dimerization caused by the higher levels of the receptor.

It is interesting to note that while λPVR gave rise to a dramatic phenotype when expressed in the wing disc or the follicular epithelium, no apparent phenotypes were observed following expression in the embryonic ectoderm or the eye disc. Some of the intracellular elements that may be essential for relaying the signals resulting from PVR activation could thus be expressed or active only in a restricted set of tissues.

In the embryonic ectoderm and eye disc where λPVR was inactive, we also failed to see apical accumulation of PVF1. The correlation between the capacity of the wing epithelium to localize the ligands apically, on the one hand, and to respond to uniform PVR activation, on the other, strengthens the notion that apical activation of PVR is instructive in this tissue.

What can the ectopic phenotypes teach us with regards to the normal downstream responses to PVR activation in the wing epithelium? The primary defect upon overexpression of PVR is misorganization of the actin cytoskeleton at the basolateral side. In addition, expression of the constitutively active receptor results in multiple adherens junctions. We thus suggest that apically restricted PVR activation provides signals that facilitate the formation of F-actin at the adherens junctions. This role is reminiscent of the activity of PVR in the border cells of the ovary, where polarized activation by PVF1,expressed in the oocyte, participates in guiding the migration of the border cells (Duchek et al., 2001). It is tempting to suggest that polarized PVR activation regulates migration or cell polarity, using a common set of intracellular responses leading to localized actin polymerization.

We thank Ayelet Schlesinger for providing the sGFP flies, Limor Landsman for the initial PVR disc stainings, Ilana Sabanay for EM preparations and Ruth Tal for technical help. We also thank P. Bryant, F. Matzuzaki, P. Rorth, the Berkeley Drosophila Genome Project, Exelixis, the Developmental Studies Hybridoma Bank, and the Bloomington stock center for antibodies, cDNAs and fly strains. We are grateful to D. Bilder, R. Fehon, P. Rorth and E. Zelzer for fruitful discussions. This work was funded by grants from the Israel Science Foundation, the Minerva Foundation and the Forchheimer Center to B.S., who is an incumbent of the Hilda and Cecil Lewis professorial chair in Molecular Genetics.