Abstract
The tissue polarity genes control the polarity of hairs, bristles and ommatidia in the adult epidermis of Drosophila. We report here the identification of a new tissue polarity gene named starry night (stan). Mutations in this essential gene alter the polarity of cuticular structures in all regions of the adult body. The detailed polarity phenotype of stan on the wing suggested that it is most likely a component of the frizzled (fz) pathway. Consistent with this hypothesis, stan appears to be downstream of and required for fz function. We molecularly cloned stan and found that it encodes a huge protocadherin containing nine cadherin motifs, four EGF-like motifs, two laminin G motifs, and seven transmembrane domains. This suggests that Stan functions in signal reception, perhaps together with Fz.
INTRODUCTION
The adult cuticle of Drosophila is decorated with a large number of cuticular structures, such as hairs and bristles. These structures are polarized with respect to the plane of the epithelia and typically share a common polarity. For example, the wing is decorated with a large number of distally pointing hairs. Mutations in the tissue polarity genes disrupt the normal precise alignment of these structures (Gubb and Garcia- Bellido, 1982; Adler, 1992; Eaton, 1997). Most of the tissue polarity genes identified so far appear to be part of the frizzled (fz) signaling/signal transduction pathway, which controls the polarity of several types of cuticular structures (Adler, 1992; Gubb, 1993). Hair polarity is controlled by regulating the subcellular location for the initiation of the prehair – the cytoskeletal-mediated outgrowth that gives rise to the cuticular hair (Wong and Adler, 1993). In contrast, sensory bristle polarity is regulated by genes controlling the orientation of the cell divisions that give rise to the four cells of the bristle sensory organ (Reddy and Rodrigues, 1999; Gho et al., 1999). Ommatidia polarity is controlled by regulating the R3/R4 cell fate decision and the subsequent rotation of the developing ommatidia (Wherli and Tomlinson, 1998; Reifegerste and Moses, 1999). It is thought that there is a general tissue polarity intercellular signaling/signal transduction pathway that regulates specific downstream effector genes and proteins in the cells that give rise to the different cuticular structures (Adler, 1992).
The cadherins were originally identified as proteins that mediate Ca2+-dependent cell adhesion. These proteins typically contain four or five extracellular cadherin motifs, a single transmembrane domain and a cytoplasmic tail that contains a binding site for β-catenin (reviewed in Geiger and Ayalon, 1992). Multiple tissue type-specific cadherin proteins are found in all animals. Epithelial cells typically contain E- cadherin, which is localized to the adherens junction and is essential for the maintenance of epithelial tissue structure. The ability of classical cadherins to mediate cell-cell adhesion requires their being linked to the actin cytoskeleton via the catenins (Nagafuchi and Takeichi, 1988). Since β-catenin (armadillo in Drosophila) is also a key member of the wingless/Wnt signal transduction pathway, cadherins can, at least indirectly, affect Wnt signaling (Peifer et al., 1991). The realization that fz encodes Wnt receptors provides a clue for potential connections between cadherins and catenins in the context of tissue polarity formation (Bhanot et al., 1996). However, it does not appear that armadillo or shotgun (the Drosophila E-cadherin gene) plays important roles in the development of wing tissue polarity (Peifer et al., 1991; Uemura et al., 1996; Tepass et al., 1996).
In addition to the classical cadherins, there is a diverse group of proteins called protocadherins that have been found in Drosophila, C. elegans and mammals. These proteins are often substantially larger than classical cadherins. For example, the fat (ft) protocadherin contains more than 30 cadherin domains (Mahoney et al., 1991). Relatively little is known about the function of the protocadherins. Results of previous cell biological experiments suggest that protocadherins only weakly promote cell adhesion. The most interesting genetic data comes from the analysis of the dachsous (ds) and ft genes of Drosophila, which have recently been shown to encode protocadherins (Clark et al., 1995; Mahoney et al., 1991). Mutations in ds result in malformed flies that often die as pharate adults or as newly eclosed adults. The legs and wings are particularly abnormal in ds flies. Both are shorter and fatter than normal, and each also displays a number of specific defects including disrupted tissue polarity (Adler et al., 1998). Mutations in ft are pupal lethals. When clones of ft are generated in imaginal tissue, they result in tumorous outgrowths. Hence, ft appears to be a tumor suppressor gene in flies (Mahoney et al., 1991). Interestingly both ds and ft were among the genes identified in a large screen for new tissue polarity mutants. The underlying biochemical mechanisms by which mutations in these genes produce a tissue polarity phenotype are unknown. At the genetic level, it was shown that ds caused an altered anatomical direction of fz signaling and enhanced fz domineering nonautonomy (Adler et al., 1998).
The starry night (stan) gene was identified in the same mutant screen that led to the identification of ds and ft as tissue polarity genes. Mutations in stan produce a wing tissue polarity phenotype that is similar to that of fz and dishevelled (dsh; Krasnow et al., 1995), both in terms of the mutant wing polarity pattern and the small number of wing cells that produce more than the normal one hair. Genetic analyses in the present study have shown that stan likely plays a role in the fz signaling/signal transduction pathway that controls tissue polarity. Mutations that inactivate stan were found to spread across about 40 kb of genomic DNA. The stan gene produces a longer than 12 kb mRNA that encodes a protocadherin. This protein is homologous in general predicted structure to proteins identified in mammals (rMEGF2, mCelsr1) and C. elegans (CeCelsr). They contain a series of 8-9 cadherin motifs, 4-7 EGF-like motifs, two laminin G motifs, and a region of seven transmembrane domains that is reminiscent of trimeric G- coupled receptors (Hadjantonakis et al., 1997; Nakayama et al., 1998). The unusual structure of the Stan protein suggests it could function in both cell adhesion and cell signaling pathways.
MATERIALS AND METHODS
Fly culture and strains
Flies were grown on standard media at 25°C, unless stated otherwise. Many mutants were obtained from the stock center at the Indiana University. A chromosome carrying Df(2R)17 was kindly provided by Dr R. Burgess (Stanford). Several different mutations that cytogenetically removed stan were used in our studies (Df(2R)E3363, Df(2R)stan1, Df(2R)stan2, Df(2R)stan9, Df(2R)stan14, Df(2R)stan38). For simplicity, they are all abbreviated as Df-stan. Alleles of stan were isolated independently in Charlottesville and Cambridge. In Charlottesville, the original stan alleles were identified in an F1 FLP/FRT screen for new wing tissue polarity mutations, while the original allele identified in Cambridge was found in an unrelated mutant screen. Additional stan alleles were isolated in both locations by F1 screens where we used a lack of complementation of viable alleles for screening. The alleles of stan were isolated by EMS, γ-ray, X-ray and P element mutagenesis. The gene was named starry night (stan) after the swirling brush strokes in the eponymous painting by Van Gogh.
Cytological procedures
To examine the process of hair morphogenesis, pupal wings were dissected in 4% paraformaldehyde/PBS, stained with a fluorescent- phalloidin which binds to F-actin (Wong and Adler, 1993) and examined by confocal microscopy (Molecular Dynamics).
Generation of genetic mosaics
Mosaic experiments were carried out using the FLP/FRT system to generate clones (Xu and Rubin, 1993). Larva of w, hs-flp; FRT42 pwn stan/FRT42 were heat shocked at 38°C for 30 minutes to induce the recombination at the FRT site. The stan clones could be recognized by the recessive mutation pawn (pwn) that resulted in cells forming thin wispy hairs and short deformed bristles. Clones homozygous for fz were marked with the autonomously acting trichome morphology marker, starburst (strb).
Scoring of mutant wings
Wings from relevant flies were mounted in Euparal (Asco labs) and examined under bright-field microscopy. The analysis of wing phenotypes is described in detail elsewhere (Wong and Adler, 1993; Krasnow and Adler, 1994; Adler et al., 1994). A distal to proximal gradient of fz expression was induced by the ‘waxing’ procedure (Adler et al., 1997).
Molecular biology
P1 clones were generously obtained from Dr G. Rubin (UC Berkeley) and Dr E. Nitasaka (Kyushu University). P1 DNA was prepared by the alkaline lysis method described previously (Hartl et al., 1994) with slight modifications. DNA and RNA isolation, Southern analysis, northern analysis, genomic DNA and cDNA library screening was performed by standard methods (Sambrook et al., 1989). Genomic walking with λ clones was performed as described previously (Park et al., 1996). DNA sequencing was primarily done in the Core Center of the University of Virginia.
Seven RT-PCR were performed to obtain the full stan transcript. Reverse transcription was carried out using SuperScript Preamplification System (GIBCO-BRL). The resulting cDNA was subject to polymerase chain reaction that consisted of 35 cycles of denaturation for 30 seconds at 94°C, annealing for 1 minute at 55°C, and extension for 2 minutes at 72°C. For RT1 (1962 bp), the primer sets were 5′-ACT TGC GAG TGT GAT TCC-3′ and 5′-ATT TAG GCG AAT GTG TGG-3′; for RT2 (1918 bp), 5′-AGC TCC TTG GCG ATT CTC-3′ and 5′-CGG CAG CAC AAC ACT TTC-3′; for RT3 (1897 bp), 5′-ACG TAA CGA TGC CCA CAT-3′ and 5′-CTC TTT CTG GCG TGT TCG-3′; for RT4 (1598 bp), 5′-GTG TAC CAC CCA AAT CGG C-3′ and 5′-GGC TTC GTT AAT TCC ATC CG-3′; for RT5 (1861 bp), 5′-ACG CCT CCT ATG CCA TTC-3′ and 5′-AGC GAC TGG GCA ATG GTA-3′; for RT6 (1891 bp), 5′-GGC CTG CTC CTC GTG CT-3′ and 5′-TCA CAT CAC CGG ACA TCG-3′; for RT7 (1194 bp), 5′-TAG TCG TCG TGG CGT AG-3′ and 5′-CCT GTG TCT CAT CTC ATT GTC G-3′.
In situ hybridization
The 4,513 bp stan cDNA was digested with ApaI and SacI, and the resulting 645 bp fragment was subcloned into pGEM-T-Easy vector (Promega). Sense and antisense RNA probes were prepared using the Dig RNA labelling kit according to the manufacturer’s instruction (Boehringer Mannheim). In situ hybridization was performed as described previously (Tautz and Pfeiffle, 1989) with minor modification.
Sequence analysis
Genomic DNA sequence was analyzed for the presence of a putative open reading frame or exons using computer programs such as WebGeneMarker and BCM gene finder (Smith et al., 1996). For RT- PCR, optimal primers were designed using Vector NTI (Informax) and Oligo 5.0 (National Biosciences). The PROSITE database was utilized for motif search of the Stan protein (Bairoch et al., 1997). The presence of putative signal peptide was determined using SignalP (Nielsen et al., 1997). Putative transmembrane domains were predicted using SOSUI (Hirokawa et al., 1998) and PHD (Rost et al., 1995) algorithms.
RESULTS
Isolation of starry night (stan) mutations
Mutations in the stan gene cause a tissue polarity phenotype over much of the epidermis. Epidermal hairs, sensory bristles and ommatidia are all affected by stan mutations, thus in this way stan resembles fz, dsh, Van Gogh(Vang)/strabismus (Taylor et al., 1998; Wolff and Rubin, 1998), and prickle (Gubb et al., 1999). We found the stan gene to be highly mutable. In one experiment, we simultaneously screened for mutations in Vang, kojak (koj) and stan after γ-ray mutagenesis. We isolated five times more stan (21 mutations) than Vang (4 mutations) mutations and seven times more stan than koj (3 mutations) mutations (P. N. Adler, unpublished). This suggested that stan was a large gene. The stan gene was mapped to position 62 on the second chromosome by meiotic mapping, and localized to 47B3-5 on the basis of several alleles that contained breakpoints in this interval and deficiencies that failed to complement stan point mutants (data not shown).
In our collection of 49 stan alleles, there is a great deal of phenotypic variation (a detailed description of the stan alleles and their genetic interactions with other tissue polarity genes will be presented elsewhere). Most stan mutations are recessive lethals. In this study, we made extensive use of stan3, which is homozygous and hemizygous viable. Flies with this allele show a relatively strong polarity phenotype in the adult cuticle (Fig. 1B,D). The phenotype of stan3/stan3 wings is only slightly weaker than stan3/Df-stan wings (compare Fig. 1G and H), thus by this criteria stan3 is a strong hypomorphic allele for the wing tissue polarity phenotype. Since stan3 is healthy, the mutation is presumably functional for the stan vital functions. It is possible that stan3 is analogous to the dsh1 mutation, which is defective for tissue polarity, but functional for the essential functions of this gene (Krasnow et al., 1995).
The stan wing phenotype
We have used viable stan mutants (particularly stan3/stan3 and stan3/Df-stan) to examine the stan wing phenotype. As is the case for other tissue polarity mutants, stan mutants do not show a complete loss or randomization of hair polarity across the wing. Rather, they show a stereotypic abnormal polarity pattern (Fig. 1B,G,H). As noted previously, the polarity patterns that result from mutations in many tissue polarity genes are quite similar, albeit not identical. We call this pattern the fz/inturned (in) pattern after two of the best-studied genes in fz pathway (Fig. 1F). The stan mutant wings also have this general pattern. A second criterion that we have used to characterize tissue polarity mutants is the frequency of wing cells that form more than the normal one hair. Wings homozygous for stan3 have relatively few multiple hair cells (an average of 1.03 hairs/cell in our typical test region). This is similar to wings mutant for fz (1.02 hairs/cell), dsh (1.01 hairs/cell) and pk (1.02 hairs/cell). In contrast, it is much lower than that seen in wings mutant for inturned (in) (1.82 hairs/cell), fuzzy (fy) (1.92 hairs/cell) and multiple wing hair (mwh) (3.94 hairs/cell). We have also examined the process of hair morphogenesis in stan pupal wings. In wild-type wings, the prehairs that develop into the adult cuticular hairs are formed in the vicinity of the distal- most vertex of the cell and extend away from the cell in a distal direction (Fig. 1C). In stan mutants, many cells formed prehairs at a relatively central location on the apical cell surface (Fig. 1D). Other cells formed prehairs at abnormal locations along the cell periphery. This is similar to what we have found for mutations in the fz-like genes (Wong and Adler, 1993). Thus, based on several phenotypic criteria we place stan in the fz-like group of genes.
Cell autonomy of wing clones
Several recessive stan alleles were isolated due to a wing hair polarity phenotype seen in a wing in an FLP/FRT based F1 screen. Thus, it was clear that the presence of wild-type neighboring cells would not rescue all of the mutant cells in a clone. Several of the tissue polarity genes display domineering nonautonomy in wing clones – that is the presence of mutant cells in a clone alters the development of wild-type cells that are near the clone. To see if stan clones also displayed domineering nonautonomy, we generated mosaic wings where stan clones were marked with the hair marker pwn. Several different alleles were tested including the putative null allele stan24, the recessive lethal allele stan21 and the recessive viable allele stan3, and similar results were obtained. In all cases, the majority of clones behaved cell autonomously (53/60 for pwn stan21 clones, 72/83 pwn stan24 clones and 19/22 pwn stan3 clones) (Fig. 2A). Further, the extent of domineering nonautonomy for those clones scored as showing it was typically much weaker than we have seen with fz or Vang (Fig. 2B). We conclude from these experiments that stan principally functions cell autonomously.
stan suppresses the domineering non-autonomy of fz clones
As an in vivo assay for fz pathway function, we used the domineering non-autonomy of fz clones. To do this, we induced fzR52strb clones in stan3 wings. In a wild-type wing, more than 80% of fz clones show distal domineering non-autonomy. That is, cells distal (and in part anterior/posterior) but not proximal to the clone show altered polarity that extends to cells that do not border the clone (Fig. 2G,H). We scored 54 fzR52strb clones in regions of stan3 wings, where the polarity was consistent enough for us to be able to score the clones for domineering non-autonomy. Forty two clones behaved cell autonomously (Fig. 2C,D) and only 12 clones showed evidence of domineering non-autonomy. Further, the extent of domineering non-autonomy in these 12 clones was modest (Fig. 2E,F). Thus, stan appears to be a suppressor of the domineering non- autonomy of fz. That there remains some fz domineering non- autonomy in stan3 wings may reflect stan3 not being a null allele. The ability of a stan mutation to suppress this fz phenotype argues that stan is downstream of fz and required for the cell non-autonomous function of the fz pathway.
stan is required for the ability of a gradient of fz expression to repolarize wing hairs
As a second in vivo assay for fz pathway function, we used the ability of a gradient of fz expression, with its highest point near the distal tip of the wing, to reverse the normal distal polarity of wing hairs (Adler et al., 1997). This result argues that cells can ‘sense’ the fz activity of neighboring cells and respond to this information. The production of a region of reversed polarity is likely to require both cell non-autonomous (e.g., a fz-dependent intercellular signal) and cell autonomous functions (e.g., transduction of the fz-dependent signal). We found that stan3 completely blocked the ability of a gradient of fz expression to reorganize wing hair polarity (Table 1). Hence we conclude that stan functions downstream of fz and is required either for the cells to be able to sense the fz activity of neighboring cells or to respond to this information.
stan does not block the late fz gain of function
The overexpression of fz just prior to prehair initiation causes the formation of large numbers of multiple hair cells that are a phenocopy of the in-like mutations (Krasnow and Adler, 1994). We have previously used this fz gain-of-function phenotype as a test to identify genes that are downstream of and required for the transduction of the fz signal (Krasnow et al., 1995). It was shown that the function of the dsh gene, which is thought to function downstream of fz (Klingensmith et al., 1994; Theisen et al., 1994), was indeed required for this phenocopy (Krasnow et al., 1995). However, the function of several other tissue polarity genes, pk, ds and Vang, was not required (Krasnow et al., 1995; Adler et al., 1998; Taylor et al., 1998). To determine if stan was required for the transduction of the fz signal, we constructed stan; hs-fz flies and induced fz expression just prior to prehair initiation. We found that the stan3 did not block the ability of fz overexpression to induce cells to form multiple hairs (Table 2). Rather, it appeared to slightly enhance the ability of fz overexpression to induce multiple hair cells.
Mapping of the stan locus
We initially mapped stan to 47B2-4 based on the cytogenetic analysis of two independent stan inversions, stan10 and stan35 (data not shown). P1 clones covering this region were obtained and used to screen a Drosophila genomic λ phage library. The clones were assembled into a contig covering about 76 kb and a restriction map was constructed (Fig. 3). RFLPs associated with stan10 and stan35 were identified by genomic Southern blot analysis, and these were found to be separated by approximately 40 kb, consistent with the large genetic target size of stan. We also identified RFLPs associated with two cytologically normal alleles (stan15 and stan24) induced by treatment with ionizing radiation. Both mapped in the GC5 region relatively close to the RFLP associated with stan35.
Identification of the stan transcript
To identify and map gene products in the putative stan locus, northern blots were screened with the inserts of all the λ phage clones. Distinct RNA bands of 4.9 kb, >10 kb, and 3.8 kb in length were detected with the inserts from the clones GC71, GC5 and GC1 respectively. A late embryonic cDNA library (Brown and Kafatos, 1988) was then screened with probes from these clones. We isolated cDNA clones representing these three independent mRNAs and sequenced members of each cDNA class. One group (obtained after screening with GC71) was found to have 93% sequence identity to the Fw repetitive elements found in the white locus of wi flies. These sequences show an extensive sequence similarity to L1 sequences, a major family of repetitive DNA elements dispersed in the mammalian genome (Di Nocera and Casari, 1987). The second group was found to encode the Drosophila rab3 gene, which was previously mapped to 47B1-14 (Johnston et al., 1991).
The third group was represented by a 4.5 kb cDNA that hybridized to a larger than 10 kb mRNA on northern blots, suggesting it was an incomplete cDNA. Sequencing showed that it encoded part of a novel protein. A consensus poly(A) signal (AATAAA) was found and followed by a stretch of poly(A) tail, implying that the cDNA contains a true 3′ end. The cDNA mapped to restriction fragments altered by three of the stan RFLPs arguing that it was a product of the stan gene. To identify the remainder of the stan transcript, we sequenced genomic DNA encompassing about 50 kb and identified potential coding regions by computer analysis. We then used RT-PCR to test if the identified regions were indeed found in stan mRNA and they were present in the same RNA molecule as our original cDNA clone. The RT-PCR products were sequenced to confirm the predicted structure of stan. As summarized in the Fig. 3, the stan gene consists of at least four exons, and the stan mRNA is over 12 kb in length encoding a putative 3,579 amino acid protein.
stan expression
Northern blot analysis showed that the large stan mRNA was present in a number of developmental stages (Fig. 4A). It was most abundant in 6-9 hour embryos and more abundant in pupae than larvae. We used in situ hybridization to examine the expression of stan in pupal wings. We found stan mRNA to be present at relatively even levels in all regions of the pupal wing (Fig. 4B). This is consistent with the genetic experiments showing that stan mutations have a phenotype over most if not all of the wing.
The Stan protein is a protocadherin
The conceptual translation of the stan mRNA yielded a protein with several notable motifs (Figs 5, 6). When analyzed with SignalP algorithm (Nielsen et al., 1997), the amino-terminal peptide consisting of the first 29 amino acids was predicted to serve as a signal peptide. This and the presence of seven putative transmembrane domains in the carboxy terminus suggest that Stan is a type 1 membrane protein. In addition, the huge extracellular part of the protein contains nine cadherin motifs, two laminin G motifs, and four EGF motifs implying that Stan may communicate with many signaling molecules or receptors. Database searches showed that stan is closely related to mammalian genes, rMEGF2 (Nakayama et al., 1998) and mCelsr1 (Hadjantonakis et al., 1997) and a C. elegans gene, CeCelsr. Their function is largely unknown except that rMEGF2 is specifically expressed in cerebellum and olfactory bulb (Nakayama et al., 1998) and mCelsr1 is expressed in the developing central nervous system (Hadjantonakis et al., 1998). While carrying out the molecular analysis of stan, we learned that the same gene had been cloned by the Uemura group (Kyoto University) due to the cadherin domain homology (Usui et al., 1999). They have called the gene flamingo and obtained cell biological data showing that flamingo is downstream of fz.
DISCUSSION
stan interacts with fz
Several lines of evidence point to stan function being required for the function of the fz pathway in wing tissue polarity. Mutations in stan produce an abnormal polarity pattern that is typical of tissue polarity genes whose function is thought to be essential for the pathway, such as fz and dsh. We also found a number of strong genetic interactions between stan and several fz pathway genes, including an allele-specific interaction between stan and fz (data not shown). Further, the domineering non-autonomy of fz is strongly suppressed by the stan3 mutation (Fig. 2). Finally, the ability of a directed gradient of fz expression to redirect the polarity of cells on the wing is blocked by the stan3 mutation (Table 1). All of these observations argue that stan function is essential for the function of the fz pathway.
Since stan mutations do not block the ability of the late overexpression of fz to produce an in-like phenotype (Table 2), the interaction between fz and stan is complex. It appears that fz carries out multiple (both cell autonomous and non-autonomous functions) roles in tissue polarity. The requirement for stan function for some but not all fz functions provides further evidence that fz has multiple roles in the development of tissue polarity. The two functions of fz could represent fz functioning in two distinct pathways or fz functioning at two times/locations in a single pathway. In the former model stan could be downstream of fz in one of the two pathways, and in the latter model it could function between the two times/locations where fz functions and hence be downstream of some fz functions and upstream of others. The requirement for stan function for a gradient of fz expression to redirect wing hair polarity and for the domineering nonautonomy of fz clones, but not for the late overexpression of fz to induce an in-like phenotype is not unique to stan, but is also a property of Vang (Taylor et al., 1998).
The Stan protein has multiple domains
Conceptual translation of the stan open reading frame reveals that it encodes a member of a protocadherin family found in both mammals and C. elegans. The protein has a remarkable structure that includes nine cadherin motifs, four EGF motifs, two laminin G motifs and seven transmembrane domains. The protocadherins are not nearly as well studied as classical cadherins, and relatively little is known about their in vivo function. The classical cadherins were identified as proteins that mediated Ca2+-dependent cell adhesion (reviewed by Geiger and Ayalon, 1992). E-Cadherin is present at adherens junctions in epithelial cells and is essential for the maintenance of tissue structure. Cadherin-dependent cell adhesion requires the linkage of the cadherin to the actin cytoskeleton, which is mediated by the catenins (Nagafuchi and Takeichi, 1988). Some protocadherins, such as Ds and Ft have potential catenin binding sites although these are interrupted (Clark et al., 1995; Mahoney et al., 1991). Stan does not apparently have such catenin binding sites. Ds and Ft have been suggested to have an adhesion function based on a number of phenotypes. For example, large ft clones in the wing often end up evaginating inward to form cuticular nodules. This might be a consequence of altered adhesion. Smaller ft clones as well as ds clones exhibit a marked tendency to assume an oval shape, once again consistent with the hypothesis that the clone cells have altered adhesivity (P. N. Adler, unpublished data). In contrast, we saw no such phenotypes with stan. Hence, we think it unlikely that stan mutations have a major effect on the adhesion of wing cells. However, we note that the Uemura group has obtained evidence for stan being an adhesion protein in cultured cells (Usui et al., 1999).
The large size of the Stan protein provides it with many potential sites for binding to other proteins. The presence of both cadherin domains and a region whose topology is similar to a G-coupled receptor suggest that Stan could potentially function either in cell adhesion and/or in signal reception. The failure to see any evidence of altered wing cell adhesion in stan mutants suggests that at least in tissue polarity Stan functions primarily in signaling/signal transduction. If Stan does not function as an adhesion protein, what is the reason for the cadherin domains that are conserved in both vertebrates and invertebrates? One intriguing possibility is that the cadherin domains serve to localize the Stan protein to a plasma membrane where it functions in signal reception and/or perhaps to localize other tissue polarity proteins. Alternatively, the cadherin domains could be required for stan function in other tissues or for subtle but important effects on cell adhesion in the wing that we could not detect by obvious changes in clone shape.
Acknowledgement
During this work, W. J. P. was supported by a grant (G70800- GG01570) from KRF (Korea Research Foundation) and a grant (G01130) from KOSEF (Korea Science and Engineering Foundation), and P. N. A. was supported by a grant from the NIH (GM37136). S. C. and D. G. were supported by a programme grant from the British Medical research Council to M. Ashburner and D. Gubb. W.J.P conceived this work and supervised the molecular analysis of stan. J. C. and M. J. K. carried out most of the molecular experiments together. J. H. G took a part in the initial characterization of P1 and λ clones. S. C. and D. G. independently identified stan and isolated several key alleles (stan3 and stan10), J. C. and P. N. A. independently identified stan and carried out most of the genetic experiments. W. J. P. and P. N. A were co-authors of the paper.