The Drosophila gap junction channel gene innexin 2controls foregut development in response to Wingless signalling

Cell-to-cell communication is essential for the generation of the diverse tissues and organs of the body in multicellular organisms(Gurdon, 1992). One class of molecules important for such interactions are secreted signalling molecules that are involved in intercellular signalling processes. Several conserved families of secreted growth and differentiation factors have been identified and it has been shown that assorted combinations of these factors control cell differentiation and morphogenesis in a variety of organisms (for reviews, seeMassague, 2000;Taipale and Beachy, 2001;Wodarz and Nusse, 1998;Grapin-Botton and Melton,2000). The Drosophila gene hedgehog, which encodes a secreted protein, and the genes wingless and decapentaplegic, which encode members of the WNT and TGFβfamilies of growth factors, respectively, have been studied extensively during early embryogenesis; limb and eye patterning, and gut development(Affolter and Mann, 2001;Bienz, 1997;Dahmann and Basler, 1999;Lawrence and Struhl, 1996;Stringini and Cohen, 1999). In the ectodermal fore- and hindgut of Drosophila, Wingless,Decapentaplegic and Hedgehog have been shown to coordinate morphogenesis in signalling centres that become established at the junctions of the ectodermal and endodermal tissue layers (Hoch and Pankratz, 1996; Pankratz and Hoch, 1995). At the junction of the foregut and midgut, they direct the formation of the proventriculus, a gut-associated organ that mediates food passage in the larva(Campos-Ortega and Hartenstein,1997; King, 1988;Snodgrass, 1935; Strasburger,1992; Skaer, 1993). It has remained largely elusive, however, how the signals are transformed into local activation of target genes that induce the morphogenetic movements.

Another class of molecules that is involved in the direct communication between neighbouring cells are transmembrane channel proteins that assemble into gap junctions (for reviews, see Kumar and Gilula, 1996; Goodenough et al., 1996). Gap junctions are composed of two hexameric hemichannel subunits that are transported to the plasma membrane, following the secretory pathway for transmembrane proteins (for reviews, seeFalk and Gilula, 1994;Zhang et al., 1996;Yeager et al., 1998). If hemichannels of adjacent cells interact, they form a functional dodecameric gap junction channel directly linking cytoplasm of neighbouring cells(Unger et al., 1997). This allows cells to exchange ions and small molecules that can participate in signalling events. Gap junctions are clusters of intercellular channels and may consist of tens or thousands of channels forming so-called plaques in the membranes of cells. It is believed that the permeability of the intercellular channels for ions and small molecules depends on their molecular composition and on the size and charge of the permeant molecules (for reviews, seeKumar and Gilula, 1996;Goodenough et al., 1996). How plaque formation and localisation is controlled during developmental processes is largely unknown. In vertebrates, gap junction channel proteins are coded by the connexin multigene family (Willecke, 1991). Many tissues express several different connexin isoforms with complex and overlapping temporal and spatial profiles. It has been shown that heteromeric channels can occur in vivo and that heterotypic ones can be formed in vitro, suggesting that the normal physiological functioning of cells may require expression of multiple connexins (for a review, see Yeager et al., 1998). In invertebrates, a separate gene family encoding gapjunction channel proteins has been identified, the innexinmultigene family (Phelan et al.,1998a) (for review, see Phelan and Starich, 2001). Innexins are structurally and functionally analogous to connexins; however, they share no sequence homology (for a review, see Phelan and Starich,2001). Electrophysiological evidence obtained in paired Xenopus oocytes suggests that innexins, like connexins, form heteromeric channels (Landesman et al.,1999; Phelan et al.,1998b; Stebbings et al.,2000). The innexin mutants that have been identified in C. elegans and Drosophila display defects in electrical signal transmission. Mutations in the C. elegans unc-7 and unc-9 genes cause uncoordinated phenotypes(Starich et al., 1996;Barnes and Hekimi, 1997), and eat-5 mutants display an impairment of electrical coupling between pharynx muscles (Starich et al.,1996). In Drosophila, shaking-B mutants are characterised by a loss of electrical signalling at synapses of the giant fibre system(Phelan et al., 1996), and ogre mutants display a reduced optic lobe primordium (Watanabe et al., 1990). Most of the six other known Drosophila innexin genes(Adams et al., 2000) are expressed in a complex and overlapping temporal and spatial profile, with several members showing high levels of expression in the developing embryonic gut (Curtin et al., 1999;Stebbings et al., 2000).

In a screen of P-element insertion lines to identify novel regulators of gut development, we isolated four P alleles of a locus that we named kropf. The mutant phenotype of the different P alleles is highly similar. Most of the mutant animals are embryonic lethal; however, a few mutants reach the first instar larval stage and display a feeding defect. Molecular analysis has revealed that the kropf locus corresponds to the innexin 2 transcription unit and that at least two P alleles are innexin 2 transcript null alleles. Expression studies indicate that innexin 2 is expressed maternally and zygotically. We now report that the feeding defect of kropf mutant larvae is due to a malformation of the proventriculus, an organ that controls food passage at the foregut/midgut boundary of the larval gastrointestinal tract. It originates during embryogenesis by the migration and subsequent folding of ectodermal and endodermal tissue layers. In kropf mutants, the formation of the keyhole, a primordial structure that was shown to be generated by the Wingless- and Hedgehog-dependent outfolding, and subsequent migration of proventriculus primordial cells, does not occur properly. Genetic lack- and gain-of-function studies and experiments in Drosophila tissue culture cells demonstrate that innexin 2 transcription is induced by Wingless signalling in the foregut.

Embryo hosts for injection were w¹¹⁸. We used Oregon-R as wild-type flies. wg^CX4, dpp^S4, hh^IJ35 and twi-Gal4 were obtained from the Bloomington stock centre. The 14-3fkh-Gal4 driver, which mediates Gal4 expression in the proventriculus and other regions of the gut, has been described previously(Fuß and Hoch, 1998). UAS-wingless and UAS-decapentaplegic were gifts from S. M. Cohen (EMBL, Heidelberg). We refer to the P lines l(1)G0016, l(1)G0034,l(1)G0036 and l(1)G0188 that were isolated by screening the Göttingen X-chromosome collection of P lines(Peter et al., 2002), as kropf^P16, kropf^P34, kropf^P36 and kropf^P188 alleles, respectively. We refer to the hemizygous kropf mutant male embryos in this manuscript as`kropf mutants'. Two-thirds of these mutants are embryonic lethal and one-third shows a lethality in the first instar larval stage (feeding defect). All four alleles showed similar mutant phenotypes in the embryo and all displayed a larval feeding defect. Using in situ hybridisation analysis, we determined that kropf^P16 and kropf^P36were innexin 2 transcript null alleles(Fig. 1). Expression studies(Fig. 4) are consistent with a maternal innexin 2 complement, which probably accounts for the variable phenotype of the mutants. To ascertain whether lethality was caused by the P-lacW insertion, the P element was remobilised to induce precise excision. The stocks were tested for loss of P-lacW insertion by sequencing of the genomic innexin 2 locus and for survival of hemizygous males. The embryonic defects and the larval feeding phenotype of the kropfmutants could be rescued by inducing low level of ubiquitous innexin 2 expression in the background of homozygous kropf^P16mutants using UAS-innexin 2 (R. B. and M. H., unpublished) and daughterless-Gal4 or armadillo-Gal 4 driver lines. For mutant analysis, we used kropf^P16, if not stated otherwise.

Genomic DNA flanking the P-element insertions was recovered by plasmid rescue and inverse PCR. The precise location of the P-element insertion was determined by sequencing the rescued DNA and the amplified fragments. The P-element insertion sites are kropf^P16 +236, kropf^P34 -11, kropf^P36 +197 and kropf^P188 +188, with respect to the transcription start site at position +1 of the innexin 2 gene(Stebbings et al., 2000).

To monitor larval feeding behaviour, the animals were allowed to grow on grape agar plates and we fed them yeast that was dyed with Carmine Red (Sigma)(Pankratz and Hoch, 1995). Phenotypes were scored at various times under the dissecting microscope.

The following antibodies were used: anti-Wingless (1:20 gift of S. Cohen),anti-Dve (1:1000; gift of F. Matsuzaki), anti-Fkh (1:100, gift of H. Jäckle) and mouse anti-β-Gal (1:1000; Developmental Studies Hybridoma Bank). For immunostaining, embryos were collected and processed following standard procedures (Fußet al., 2001), incubated overnight at 4°C in primary antibody mix containing preabsorbed antibody. Biotinylated secondary antibodies,Vectastain and DAB staining reagents were used according to manufacturer's recommendations (Vector Laboratories). For fluorescent microscopy, secondary antibodies Alexa Fluor⁴⁸⁸ and Alexa Fluor⁵⁴⁶ from MoBiTec were used at dilutions of 1:200 and 1:400, respectively.

For in situ hybridisation, digoxigenin- or fluorescein-labelled RNA antisense probes were generated by in vitro transcription of a C-terminal and full-length innexin 2 fragment from the innexin 2 cDNA clone LD clone 11658 (Berkeley Genome Project), according to the manufacturer's instructions (Roche, Mannheim). Fluorescent detection of innexin 2transcripts, co-immunostained with anti-Discs lost antibody, which marks the apical sides of epithelial cells (Bhat et al., 1999), was performed as described previously(Hughes et al., 1996;Hughes and Krause, 1999). Rabbit anti-Discs lost and mouse anti-Fluorescein antibodies were used at dilutions of 1:750 and 1:2000, respectively. As secondary antibodies, we used Cy2-conjugated anti rabbit antibody (1: 200; Dianova) and Cy3-conjugated anti mouse antibody (1:400; Dianova). Images were obtained using a Leica DM IRB laser-scanning confocal microscope. Each fluorochrome was scanned individually to avoid crosstalk between channels. Images were subsequently combined using Adobe Photoshop 5.1.

Total RNA from Drosophila Schneider cells that were either transfected with the pIB expression vector alone or with pIBArm, which contains the Armadillo cDNA, was isolated using the Qiagen RNeasy Mini Kit,according the manufacturer's instructions (Qiagen). For RT-PCR analysis, 2μg of DNase I-treated total RNA was reverse transcribed by using 25 U of avian murine virus reverse transcriptase (Roche, Mannheim) with an innexin 2-specific antisense primer. The reaction mixture contained total RNA,5×reverse transcriptase buffer provided by the manufacturer, 35 U of RNase inhibitor (Pharmacia Biotech) and 0.2 mM of deoxynucleoside triphosphates (dNTP) in a final volume of 30 μl. The mixture was heated to 65°C for 5 minutes. AMV reverse transcriptase was added, and the mixture was incubated at 42°C for 60 minutes. After the completion of the first-strand cDNA synthesis, one-third of the reaction volume was amplified directly by using 5 U of Taq polymerase and 5 mM of innexin 2-specific antisense (5′-TTAGGCGTCGAAGGGCCGCTT-3′) and sense primers (5′-CTGAGCATCATGTCGGGAATATCGC-3′). PCR was performed in a final volume of 50 μl by using the Roche `High Fidelity PCR-kit'. The predicted PCR product was 273 bp. The predicted size for the actin amplification product (5′-AGCCAGCAGTCGTCTAATCCAG-3′ and 5′-CAGCAACTTTCTTCGTCACACAT-3′), which was used as an internal control, is 210 bp.

In a screen of the Göttingen P-element insertion lines(Peter et al., 2002) for reporter gene expression patterns in the developing gastrointestinal tract, we isolated four lines: l(1)G0016, l(1)G0034, l(1)G0036 and l(1)G0188, in which the P element had inserted in a novel locus which we named kropf. The P lines showed a highly similar and dynamic reporter gene expression pattern, especially at the foregut/midgut and the midgut/hindgut junctions (Fig. 1). It is known that morphogenesis in both regions of the gut is under the control of similar regulatory networks involving hedgehog,wingless and decapentaplegic signalling activities(Pankratz and Hoch, 1995;Hoch and Pankratz, 1996). From early blastoderm stage onwards, until stage 10(Campos-Ortega and Hartenstein,1997), β-Gal expression occurs rather uniform in all embryonic tissues (Fig. 1A). The ubiquitous expression pattern resolves with the beginning of germband retraction into localised expression in the ectodermal fore- and hindgut, the endodermal anterior and posterior midgut and the excretory Malpighian tubule primordia (Fig. 1B). High levels of reporter gene expression are maintained in these tissues(Fig. 1C,D), and expression occurs in late stages of embryogenesis predominantly in the proventriculus and in the small intestine of the hindgut, which abuts the endodermal posterior midgut (Fig. 1E,F). To clone the gene affected in the kropf locus, we isolated genomic DNA flanking the P insertion sites by plasmid rescue and used the sequencing information to perform a standard nucleotide blast search. We found that the P elements of the lines l(1)G0016, l(1)G0036 and l(1)G0188were inserted in the untranslated leader region of the innexin 2 gene(Fig. 1G), which encodes a member of the Innexin family of gap junction channel proteins(Curtin et al., 1999;Stebbings et al., 2000). In situ hybridisation analysis using riboprobes for innexin 2 and orthopedia, a gene that is expressed in the hindgut and the anal pads(Simeone et al., 1994) and which serves as an internal staining control, was performed for the kropf P mutants l(1)G0016 and l(1)G0036. It showed that these P alleles were innexin 2 transcript null mutations(Fig. 1F). Mobilisation of the P element was performed in all four kropf P lines and resulted in independent excision lines that were homozygous viable and regained the innexin mRNA expression profile(Fig. 1J). And finally,experiments in which the innexin 2 cDNA was cloned into an UAS vector and ubiquitously expressed in epithelial cells of kropf mutant embryos using Daughterless or armadillo-gal4 driver lines led to a rescue of the embryonic and the larval feeding phenotypes of the kropf mutants (see Materials and Methods). We therefore conclude that the transcription unit affected by the kropf P mutations corresponds to the innexin 2 transcription unit.

The four kropf^P alleles all showed very similar mutant phenotypes. About two-thirds of the mutant animals were embryonic lethals;however, about one-third of the mutants died in the first instar larval stage(see Materials and Methods). As the P lines showed expression patterns in the developing proventriculus, we tested kropf^P larvae for their ability to feed dye-coloured yeast. We had shown previously that malformation of the proventriculus results in a feeding defect causing larval lethality (Hoch and Pankratz, 1995). Using this feeding assay, we found that in kropf mutant larvae, the food could not be efficiently transported into the midgut, resulting in an engorged oesophagus. Ectodermal cells of the foregut which in wild-type larvae migrate into the endodermal proventricular pouch tissue to form the multi-layered proventriculus organ(Fig. 2A, magnification in C),fail to move inwards in the mutant animals and are stuck on top of the endoderm cells (Fig. 2B,magnification in D). We observe lethality of the mutant larvae in the first instar stage, indicating that the feeding defect may result in death by starvation. For further mutant analysis, we used the l(1)G0016(kropf ^P16) allele.

The proventriculus is an organ that functions as a valve to regulate food passage from the foregut into the midgut in Drosophila larvae(Snodgrass, 1935;Strasburger, 1932). It develops during early embryogenesis at the junction of the ectodermal foregut and the endodermal anterior midgut primordia(Skaer, 1993;Fuß and Hoch, 1998;Nakagoshi et al., 1998). There is initially an outward buckling of the ectoderm, in a region that is referred to as the `keyhole' structure (Pankratz and Hoch, 1995). This area will undergo further outward movement,then fold back on itself and move inwards into the endodermal proventricular pouch to form the mature, multi-layered proventriculus(King, 1988). Wingless is functionally required for cell migration of the proventricular cells and is initially expressed in wild-type embryos in a domain that includes the region from which the keyhole will form (Pankratz and Hoch, 1995; Fuß and Hoch, 1998) (Fig. 3A, magnification in C). With the onset of cell migration during keyhole formation, Wingless expression is either lost or repressed in the centre of its expression domain resulting into two domains that are split(Fig. 3B, magnification in D). One domain is at the anterior border of the keyhole in the ectodermal foregut cells and the other in endodermal cells posterior to the keyhole. In kropf P mutants, initial Wingless expression in the keyhole cells is normal(Fig. 3E, magnification in G). However, the Wingless expression in the central region persists(Fig. 3F, magnification in H). The cells that would normally form the keyhole stay fixed within the tightly organised epithelial foregut tube in the mutant until late stages of development, and the migration of the ectodermal cells into the endodermal pouch seems not to occur. Instead, these cells stay on top of the endoderm, as can be demonstrated using by anti-Fkh/anti-Dve double staining(Fig. 3I,J). We conclude from these results that innexin 2 is required during early stages of proventriculus development for keyhole formation. In kropf mutants,the keyhole defect prevents the correct folding and inward movement of the proventriculus tissue layers most probably resulting in the feeding defect observed later in the mutant larvae.

In order to analyse the role of innexin 2 during development further, we studied the mRNA expression pattern of the gene using an innexin 2 antisense probe (see Materials and Methods). During oogenesis, innexin 2 transcripts are found in the nurse cells, border cells and in a cortical localisation in the oocyte(Fig. 4A), consistent with a maternal complement that could explain the variability of the kropfmutant phenotype. innexin 2 mRNA is ubiquitously distributed during the early embryonic blastoderm stage (Fig. 4B). A segmental expression pattern is found during germ band extension stage (Fig. 4C)(Stebbings et al., 2000) and we find expression in the stomodeal and proctodeal invaginations, which are the primordia of the foregut and the hindgut. During keyhole formation at the boundary of the foregut and the anterior midgut, innexin 2 expression persists in the ectodermal and endodermal tissue regions of the proventriculus primordium (Fig. 4D). During the invagination of the ectodermal cells into the endodermal pouch, the expression pattern of innexin 2 is restricted to the endodermal part of the proventriculus (not shown). Fluorescent detection of innexin 2transcripts (Materials and Methods) and co-immunostaining with anti-Discs lost antibody, which marks the apical sides of epithelial cells(Bhat et al., 1999)demonstrates that innexin 2 mRNA is localised to the apical sides of the proventricular cells (Fig. 4E-G). In summary, our mRNA expression studies are compatible with a functional role of innexin 2 during proventriculus development.

Morphogenetic cell movements of the proventriculus epithelium and of the hindgut depend on the activities of the genes hedgehog, wingless and decapentaplegic (Pankratz and Hoch, 1995; Hoch and Pankratz,1996). Through the restricted expression of these genes in local domains, they define signalling centres at the foregut/midgut and the midgut/hindgut boundaries. Wingless and Hedgehog were shown to control cell migration of epithelial gut cells whereas Decapentaplegic was shown to prevent morphogenesis in the gut (Pankratz and Hoch, 1995; Hoch and Pankratz,1996). In order to analyse whether innexin 2transcription is dependent on these signalling cascades, we studied the transcriptional regulation of the gene by crossing the P lines into various mutant backgrounds or by determining the innexin 2 mRNA profile (the nuclear β-Gal expression pattern reflects the mRNA expression profile of innexin 2 during proventriculus development). Whereas in hedgehog and decapentaplegic mutant embryos, innexin 2 or reporter gene expression was not affected, they were absent in the proventriculus rudiment and reduced in the small intestine region of amorphic wingless mutants (Fig. 5A,B). Because the gut morphology in general is strongly affected in wingless mutants, we also performed a gain-of-function analysis using the UAS-Gal4 system (Brand and Perrimon, 1993). Ectopic expression of Wingless in the visceral mesoderm using the twi-Gal4 driver in combination with the UAS-wingless effector line or ectopic expression in the endoderm using the 14-3fkh-Gal4 driver in combination with the UAS-winglesseffector, resulted in ectopic expression of innexin 2 or β-Gal expression in the anterior and posterior regions of the gut(Fig. 5C,D). Both, the expression domains in the proventriculus and in the small intestine region that lies at the complementary endoderm/ectoderm boundary in the posterior region, dramatically expand (Fig. 5D). It is known that upon ectopic Wingless expression, endogenous decapentaplegic expression is ectopically activated and covers the region of parasegments 2-7 of the visceral mesoderm(Yu et al., 1996). However,upon ubiquitous expression of Decapentaplegic in the visceral mesoderm using UAS-decapentaplegic transgenes, in combination with twist-Gal4, we could not find a similar alteration of the innexin 2 expression domains.

To further substantiate the conclusion drawn from our genetic experiments that innexin 2 transcription is regulated in response to Wingless signalling, we performed transfection experiments in Drosophilatissue culture S2 cells (see Materials and Methods). Total RNA was isolated from Schneider S2 cells that were either transfected with an expression vector alone or with the same vector containing the cDNA of Armadillo, theβ-catenin homologue that has been shown to serve as the transducer of the Wingless signalling cascade (Peifer and Polakis, 2000). We monitored innexin 2 mRNA levels by performing a RT-PCR analysis using actin mRNA levels as an internal control(see Materials and Methods). As shown inFig. 5E, innexin 2mRNA levels are increased approximately five times in response to Armadillo when compared with the control reaction, whereas the actin mRNA levels remain constant in both situations. We conclude from our genetic lack- and gain-of-function studies, and from our tissue culture experiments that innexin 2 is a target gene of the Wingless signalling cascade that is induced in wild-type embryos at ectoderm/endoderm boundaries during gut organogenesis (Fig. 5F).

The regulatory networks controlling the assembly of gap junction channels to make cells competent for signalling during patterning and organogenesis are largely unknown in both vertebrates and invertebrates. We have investigated the role and the regulation of the Drosophila innexin 2 gene, which encodes a core component of invertebrate gap junction proteins, during proventriculus development in the Drosophila embryo.

Our results indicate that kropf mutant larvae which are devoid of innexin 2 expression, display a feeding defect resulting from a failure of proventriculus fomation in the embryo: the ectodermal proventriculus cells fail to move into the endodermal pouch and are stuck instead on top of the endoderm tissue; a multi-layered organ does not form(Fig. 2). Our mutant analysis using Wingless and Fkh/Dve as markers to visualise the ectodermal/endodermal boundaries in the foregut strongly suggests that this late invagination defect originates from a failure of cells to form the keyhole structure at much earlier stages (Fig. 3). The keyhole is the first visible morphological structure that is formed by an outbuckling of primordial cells of the proventriculus, in a region of the caudal foregut primordium that is free of visceral mesoderm(Pankratz and Hoch, 1995). innexin 2 mRNA is expressed in the primordial cells prior to the initiation of keyhole formation (Fig. 4). The analysis of Wingless expression in the proventriculus primordium of kropf mutants suggests that it is not the specification of the keyhole cells that is affected in the mutant. At the time when the keyhole forms, cell proliferation in the foregut has already been completed(Campos-Ortega and Hartenstein,1997). Furthermore, we could not observe cell death of keyhole cells in kropf mutants using TUNEL analysis (data not shown). We therefore favour a model that suggests that gap junctional communication may be important for the cellular processes involved in the outbuckling, the folding and the subsequent migration of the keyhole cells, as has been observed previously for the migration of neural crest cells in vertebrates(Lo et al., 1999;Huang et al., 1998).

Epistasis experiments indicate that innexin 2 transcription is abolished in the proventriculus rudiment that is left in winglessmutants and that expression of the gene is dramatically expanded upon ectopic expression of Wingless in the gain-of-function experiment(Fig. 5). Furthermore Armadillo, the signal transducer of the Wingless signalling pathway is able to induce innexin 2 transcription in tissue culture cells, as show by RT-PCR experiments (Fig. 5E). In wild-type embryos, both innexin 2 and wingless are co-expressed during early and late stages of proventriculus development(Fig. 3A,C, compare toFig. 4D). In summary, these data provide strong evidence that the gap junction channel gene innexin 2 is induced as a Wingless target gene during keyhole formation in the proventriculus primordium. Recently, defective proventriculus(dve), a novel homeobox gene, has been identified as another target gene of Wingless signalling in the proventriculus endoderm(Fuß and Hoch, 1998;Nakagoshi et al., 1998). dve mutants display, however, a late invagination defect caused by a malformation of the proventriculus endoderm(Fuß and Hoch, 1998;Nakagoshi et al., 1998). This suggests that Wingless has the potential to activate target genes that act during different phases of proventriculus development. innexin 2expression is, however, not only regulated by Wingless in the proventriculus primordium but also at the junction of the midgut and the hindgut. This region, which is called the small intestine(Hoch and Pankratz, 1996;Snodgras, 1935;Strasburger, 1932), is characterised by extensive morphogenetic cell movements and thus displays many similarities to the keyhole in the anterior region. Wingless-dependent gap junctional communication may also be required in this area of the gastrointestinal tract, although we have not observed gross morphological changes in this region in kropf mutants.

Whether the transcription factor complex that transduces the Wingless signal and which is composed of the β-catenin homologue Armadillo and Tcf/Lef DNA-binding proteins, directly interacts with the regulatory region of innexin 2 to activate its transcription, is presently unknown. However it is noteworthy that a consensus Tcf/Lef binding site is found 5 kb upstream of the putative innexin 2 promotor. Armadillo is distributed throughout the foregut, but is concentrated in specific areas of the developing proventriculus (Pankratz and Hoch, 1995). At early stages, it is strongly expressed near the keyhole region and at later stages it becomes highly concentrated in areas undergoing the most extensive cell movements. The proventriculus phenotype of armadillo mutant embryos is similar to that of wingless null embryos, strongly suggesting the requirement of armadillo during proventriculus morphogenesis. In addition, innexin 2 transcription is induced by Armadillo in tissue culture cells. It is thus possible that Armadillo and Tcf/Lef directly regulate innexin 2 transcription.

Our expression studies show that innexin 2 mRNA is localised to the apical side of cells. mRNA expression studies with other members of the innexin family indicate a similar apical localisation pattern (C. L. and M. H., unpublished). It is also noteworthy that wingless mRNA is localised at the apical sides of epithelial cells(Baker, 1987). The apical localisation has been shown to depend on discrete elements within the wingless 3`UTR (Simmonds et al., 2001). It requires cytoplasmic dynein protein that assembles the apical RNAs selectively into particles that are transported apically along microtubules (Wilkie and Davis, 2001). Redistribution of the transcripts causes a dramatic loss of Wingless signalling activity. Furthermore, the subcellular localisation of WNT transcripts also occurs in other organisms, suggesting a conserved function (Simmonds et al.,2001). Various other components of the Wingless signalling pathway, such as the Wingless transmembrane receptor Frizzled, theβ-catenin homologue Armadillo, Dishevelled and the Armadillo-modifying complex are also localised apically in the cells (for reviews, seeWodarz and Nusse, 1998;Peifer and Polakis, 2000). As we have shown that innexin 2 is a target of Wingless signalling in the proventricular cells, it is likely that apical innexin transcript localisation is an important functional feature required for Wingless-dependent gap junctional communication in the keyhole cells. Furthermore, these data suggest the existence of yet unknown factors (most probably RNA-binding proteins) that are required for apical innexin 2mRNA localisation in epithelial cells. It remains to be shown whether connexin mRNAs are also localised and, if so, what functional role the localisation plays. Electrophysiological experiments in the heterologous Xenopussystem have demonstrated that Innexin 2 forms heteromeric gap junctions with Innexin 3, another Innexin family member(Stebbings et al., 2000). Whether heteromerisation is also required for keyhole formation is not known because innexin 3 mutants have not been identified, yet.

Our results suggest that although invertebrate innexin genes share no sequence homology to the vertebrate connexin genes(Phelan et al., 1998a;Phelan and Starich, 2001),both multigene families might be regulated by an evolutionarily conserved signalling pathway, the WNT/Wingless signalling cascade. In the developing Xenopus embryo, it has been observed that Wnt1 expression leads to an enhancement of gap junctional communication in ventral cells (Olson et al.,1991; Olson and Moon, 1992; Krufka et al.,1998). Studies in the mouse have shown that ectopic expression of Wnt1 in the limb mesenchyme results in an increase of connexin 43 transcription (Meyer et al.,1997). Recent tissue culture studies using rat PC12 cell lines and cardiomyocytes have further provided molecular evidence that connexin43 is a downstream target gene of WNT1 signalling(van der Heyden et al., 1998;Ai et al., 2000). Our studies about innexin 2 regulation and the evidence from the connexin regulation in vertebrates suggest that WNT/Wingless signalling may be an evolutionarily conserved signalling pathway regulating the expression of gap junction-encoding multigene families.

We thank M. Peifer, H. Jäckle, U. Schäfer, J. Campos-Ortega and D. St Johnston for flies and antibodies, our laboratory members for helpful comments. The work was supported by a DFG grant to M. H. (FOR 425).