View metadata, citation and similar papers at core.ac.uk brought to you by CORE

provided by Elsevier - Publisher Connector Cell, Vol. 98, 585–595, September 3, 1999, Copyright 1999 by Cell Press Flamingo, a Seven-Pass Transmembrane Cadherin, Regulates Planar Cell Polarity under the Control of

Tadao Usui,1 Yasuyuki Shima,1 Yuko Shimada,1 1998). In the wing, each epithelial cell is typically hexago- Shinji Hirano,1,6 Robert W. Burgess,4,7 nal in shape and localizes an assembly of actin bundles Thomas L. Schwarz,4 Masatoshi Takeichi,2 to its distal-most vertex, producing a single prehair that and Tadashi Uemura2,3,5 extends away from the cell; the resulting adult hair points 1 Department of Biophysics distally (Mitchell et al., 1983; Eaton et al., 1996; Turner Faculty of Science and Adler, 1998; see Figure 2B). Thus, cells acquire 2 Department of Cell and Developmental Biology proximal-distal (P-D) polarity. Mutations of most tissue Graduate School of Biostudies polarity genes disrupt the oriented patterns of at least Kyoto University one of the above three classes of structures without Kyoto 606-8502 much affecting each developmental field. Japan Among known tissue polarity genes, frizzled (fz)is 3 PRESTO hypothesized to be near the top of the regulatory path- Japan Science and Technology Corporation way, and it encodes a seven-pass transmembrane re- Osaka 565-0082 ceptor for an as yet unidentified ligand (Vinson et al., Japan 1989; Bhanot et al., 1996). Although how Fz activates 4 Department of Molecular/Cellular Physiology the polarity pathway is not demonstrated biochemically, Stanford University School of Medicine one model is presented in which an active form of Fz Palo Alto, California 94305-5426 evokes distinct downstream cascades (Krasnow et al., 1995; Adler et al., 1997). One pathway reflects a cell- autonomous function of Fz, and it is mediated by Dishev- Summary eled (Dsh), which also transduces the Wingless (Wg) signal. Further downstream components used in polariz- We identified a seven-pass transmembrane receptor ing wing cells are distinct from those in the Wg pathway of the cadherin superfamily, designated Flamingo (Axelrod et al., 1998; Boutros et al., 1998; reviewed by (Fmi), localized at cell–cell boundaries in the Drosoph- Shulman et al., 1998). Fz and Dsh are required to restrict ila wing. In the absence of Fmi, planar polarity was initiation of the prehair formation to the distal edge of distorted. Before morphological polarization of wing the cell; in fz or dsh mutants, a prehair emerges near cells along the proximal-distal (P-D) axis, Fmi was re- the cell center, resulting in hair outgrowth in an abnormal distributed predominantly to proximal and distal cell direction (Wong and Adler, 1993). edges. This biased localization of Fmi appears to be Besides the role in interpreting an unknown polarizing driven by an imbalance of the activity of Frizzled (Fz) signal inside a cell, mosaic studies show that Fz also across the proximal/distal cell boundary. These re- acts non–cell autonomously. When fz null mutant clones are created, hair polarity is affected not only inside the sults, together with phenotypes caused by ectopic ex- mutant clones but also in fzϩ cells distal to but not pression of fz and fmi, suggest that cells acquire the proximal to the clones (Vinson and Adler, 1987). This P-D polarity by way of the Fz-dependent boundary directional nonautonomous property has led to a hy- localization of Fmi. pothesis that the polarizing signal, or the ligand of Fz, may be propagated from one cell to its distal neighbor Introduction (Adler et al., 1997). Although the total number of tissue polarity genes In many organs, epithelial cells are polarized not only identified has been increasing, it has not been clearly along the apicobasal axis but also within a plane. Acqui- demonstrated how their gene products are orchestrated sition of the latter polarity, known as planar cell polarity to initiate reorganization of the cytoskeleton to form prehairs. One major reason is that little is known about (PCP) or tissue polarity, is considered to be crucial for subcellular compartments where those proteins exert specialized cellular functions (reviewed by Eaton, 1997). their functions. Here we report a novel polarity gene Genetic programming of PCP has been most thoroughly that encodes a of the cadherin studied in Drosophila melanogaster adults, which are superfamily. This protein is localized in specific domains decorated by parallel and unidirectional arrays of hairs of the plasma membranes in polarizing cells and con- on appendages, sensory bristles on the body, and pho- trols PCP. A cDNA fragment of this gene was isolated toreceptor clusters in a dorsal or ventral eye hemisphere in our previous hunt for Drosophila genes of the cadherin (reviewed by Adler, 1992; Gubb, 1998; Shulman et al., superfamily (Iwai et al., 1997), and we named the gene flamingo (fmi) after the molecular appearance of the 5 To whom correspondence should be addressed (e-mail: tuemura@ protein, which is a seven-pass transmembrane receptor take.biophys.kyoto-u.ac.jp). with long extracellular and intracellular segments. 6 Present address: Institute for Developmental Research, Aichi Hu- Members of the cadherin superfamily are defined by man Service Center, 713-8 Kamiya-cho, Kasugai-City, Aichi 480- extracellular tandem arrays of peculiar repeat sequences, 0392, Japan. 7 Present address: Department of Anatomy and Neurobiology, which are considered to be homophilic binding modules. Washington University School of Medicine, St. Louis, Missouri This superfamily consists of two subfamilies: the classic 63110. type and the nonclassic type (reviewed by Takeichi, 1995; Cell 586

Uemura, 1998). The classic-type molecules, simply called with an fmi cDNA construct and in lysates of whole cadherins, interact with cytoplasmic molecules, catenins, embryos (arrowhead in Figure 1C). In embryos homozy- functioning as cell–cell adhesion molecules, whereas pro- gous for fmiE59, which has a nonsense mutation in the teins of the nonclassic type do not interact with catenins; ectodomain (Figure 1A), this 300 kDa band was missing, Fmi belongs to the latter subfamily. Fmi is localized in and instead we detected a faint band of roughly 200 plasma membranes, and its ectodomains interact ho- kDa (sharper arrowhead in Figure 1D), which is consis- mophilically in vitro; this type of binding appears to also tent with the predicted size of the truncated polypeptide. take place in vivo. During a restricted interval prior to This finding corroborates our view that the 300 kDa prehair outgrowth, Fmi distribution was polarized along fragment was derived from fmi. A 400 kDa band was the P-D axis; Fmi molecules were present predominantly sometimes detectable (arrows in Figures 1C and 1D), at proximal and distal cell boundaries (P/D boundaries) and we speculate that this is a precursor and processed rather than at anterior and posterior ones. We provide to convert into the 300 kDa fragment (arrowhead in Fig- evidence that the fz gene is required to generate this ure 1A; see details in Experimental Procedures). biased pattern and demonstrate how Fz is involved in To address whether Fmi has an intercellular adhesion making Fmi distribution polarized within the wing cell. activity, we expressed fmi in Drosophila S2 cells that In addition, results of gradient expression of fmi and fz exhibit a very weak self-aggregating property (Figure 1E). provided a clue to explain how Fmi molecules at the Cell aggregates were formed by transfection of the S2 P/D boundary discriminate between the proximal and cells with a cDNA construct that encoded the full-length distal cell edges. Fmi protein, but not by transfection with ⌬EX construct, producing a molecule without most of its extracellular Results subdomains except for the signal peptide. When green fluorescent protein (GFP)-labeled cells that produced Flamingo Is a Seven-Pass Transmembrane the full-length protein and unlabeled nontransfectants Receptor of the Cadherin Superfamily were mixed prior to aggregation, only the labeled cells The predicted translational product of fmi has 3575 were incorporated into cell clusters, showing that Fmi amino acids. The Fmi ectodomain has nine cadherin mediated homophilic cell adhesion (Figure 1E). repeats, three cysteine-rich domains (Cys-rich), and two laminin A globular domains (LmA-G; Figure 1A). A com- fmi Belongs to the “Core” Group of Tissue bination of Cys-rich and LmA-G domains is a common Polarity Genes feature of many invertebrate members and some verte- In both embryos and imaginal tissues, fmi is broadly brate molecules of the cadherin superfamily (for exam- expressed in epithelia and the nervous system, but we ple, see Miller and McClay, 1997; Costa et al., 1998; have confined ourselves to the detailed analysis of the Nakayama et al., 1998, and references therein). In con- role of Fmi in wing epithelia in this article. We isolated trast to those of classic-type cadherins, the carboxy- mutations of the fmi gene and identified base substitu- terminal intracellular tail of Fmi does not possess ca- tions in the protein coding sequences of two genetically E45 E59 tenin-binding sequences, nor does it bind to catenins null alleles, fmi and fmi (Figure 1A), neither of which gave a signal in immunostained epithelia (see Figure (data not shown). Thus, Fmi can be classified into the 3K). The fmi null mutations were embryonic lethal, and nonclassic-type subfamily. the mutant embryos showed local disconnection of lon- Compared with functionally characterized members gitudinal fascicles in the central nervous system of this superfamily, Fmi is structurally unusual, as it is (T. Usui and T. Uemura, unpublished observation). This predicted to be a seven-pass transmembrane (7TM) pro- lethality and the axonal defect could be rescued by fmi tein. Sequences of the 7TM region show similarity to cDNA expression in the nervous system of the mutant, those of one particular family of G protein–coupled re- fmiE45/fmiE59, indicating that the role of Fmi in axon out- ceptors, and the first such protein to be isolated is a growth is essential for viability (see details in Experimen- receptor for a peptide hormone, secretin (Ishihara et al., tal Procedures). 1991; asterisks in Figure 1B). Many molecules of this In the rescued animals, epithelia and imaginal photo- family have been shown to increase receptor cells were devoid of fmi expression, and those the intracellular levels of cAMP and/or inositol phos- adults exhibited defects in planar cell polarity in three phates upon ligand binding (reviewed by Bockaert and classes of structures: ommatidia, sensory bristles, and Pin, 1999). Whether Fmi is coupled to G proteins remains wing hairs (Figure 2C). This finding classifies fmi into to be demonstrated. the core group of tissue polarity genes that includes fz The fmi gene is conserved across species: a Caeno- and dsh (reviewed by Shulman et al., 1998). In region rhabditis elegans counterpart gene is present in cosmid D of the wing of the fmiE45/fmiE59 escaper, hairs were clones F15B9 and W07G4 (GenBank), and two paralogs deflected from the P-D axis and oriented toward the of mammalian receptors were reported, that is, mouse posterior wing margin (Figure 2C). This pattern alteration Celsr1 (Hadjantonakis et al., 1998) and rat MEGF2 (Na- is similar to that reported for the same region of fz mutant kayama et al., 1998). In addition, we isolated mouse wings (reviewed by Adler, 1992). As reported in null mu- cDNA clones that encode the entire protein of a third tants in all the known tissue polarity genes, most hairs paralog, mouse Flamingo1 (mFmi1; Figures 1A and 1B). of the fmi mutant did not point in random directions but possessed an abnormal, nondistal polarity that is similar Fmi Has a Homophilic Cell Adhesion Activity to their neighbors. This “stream” pattern suggests that an Antibodies to an extracellular subregion of Fmi recog- fmi-independent system is working to align adjacent nized a 300 kDa band in lysates of S2 cells transfected cells. Role of Flamingo in Planar Cell Polarity 587

Figure 1. Flamingo (Fmi) is a Seven-Pass Transmembrane Protein of the Cadherin Superfamily (A) Schematic representations of the Fmi protein and its mouse homolog, mouse Flamingo1 (mFmi1). The Fmi ectodomain includes nine cadherin repeats, three cysteine-rich (Cys-rich) regions, and two laminin A globular (LmA-G) domains. A subdomain between the last cadherin repeat and the first Cys-rich region is conserved between Fmi and mammalian homologs and is designated Flamingo box. Some but not all cysteine residues in Cys-rich regions match a consensus of the EGF-like motif. Also indicated are missense and nonsense mutations in fmiE45 and fmiE59, respectively. A hypothetical internal cleavage site is indicated by the arrowhead (see details in Experimental Procedures). (B) Sequence alignments of transmembrane domains of Fmi and three mammalian counterparts: mFmi1, mouse Celsr1 (mCelsr1), and rat MEGF2 (rMEGF2). Identical amino acid residues between Fmi and the other proteins are shown in black boxes. Asterisks represent conserved residues between these proteins and a G protein–coupled receptor, rat secretin receptor. The seven transmembrane domains are indicated by lines above each block. (C and D) Major bands of 300 kDa fragment (arrowheads) were detected in extracts of the wild-type embryos (embryo and WT) and of S2 cells transfected with a cDNA construct (S2/Fmi) but not in lysates of untransfected S2 cells (S2). The arrows point at a probable precursor (400 kDa). Embryos homozygous for fmiE59 (E59) produced a 200 kDa fragment (sharp arrowhead in [D]). (E) S2 cells producing the ⌬EX protein, which lacks most of the ectodomain, and the full-length Fmi were labeled by GFP and mixed with untransfected cells prior to aggregation assays (S2/⌬EXϩS2 and S2/FmiϩS2, respectively). Production of the full-length protein induced aggregation of S2 cells, and those aggregates contained only Fmi-producing cells (S2/FmiϩS2), showing that Fmi mediates homophilic cell interaction.

Polarity disruption of the above structures was repro- phenotype of fz mutations (Wong and Adler, 1993). In duced in homozygous fmiE59 clones, and prehairs emerged the core group, several mutations in fz and strabismus/ near the cell center or at wrong locations along the cell Van Gogh (stbm/Vang) show non–cell autonomy in spec- periphery (data not shown), which is again similar to a ifying hair polarity, and those nonautonomous effects Cell 588

3B). At 30 hr APF, the polarized pattern became most prominent; signals ran zigzag orthogonal to the P-D axis, indicating that Fmi molecules were predominantly local- ized at the P/D boundaries (arrowheads in Figure 3C; see also Figure 6E) rather than at the anterior/posterior boundaries (A/P boundaries; arrows in Figure 3C). In every typical hexagonal cell aligned parallel to the P-D axis, we could easily confirm the emergence of a prehair at its distal cell vertex where two Fmi-rich distal bound- aries met (Figure 3D). The zigzag pattern was seen throughout the dorsal and ventral surfaces of the wing. This biased localization of Fmi provided a striking con- trast to the honeycomb distribution of an epithelial clas- sic-type cadherin, DE-cadherin (Figure 3F; Oda et al., 1994). Along the apicobasal cell axis at the cell–cell junction, Fmi was present apically and its distribution was at least partially overlapped with that of DE-cad- herin concentrated at the adherens junction (Figures 3G and 3H; Uemura et al., 1996; Hough et al., 1997). Once prehairs had emerged and initiated outgrowth, the distribution started to be depolarized, although a temporal coordination between the extent of prehair length and that of depolarization of Fmi was not strictly fixed in this transition phase (30–36 hr APF). The staining pattern became almost nonpolar by 36 hr APF, when prehairs were shifting toward cell centers (Figure 3E).

Fmi Molecules Are Most Likely Distributed on Both Sides of the P/D Boundary Strong Fmi signals at P/D boundaries may be explained by a concentration of Fmi at either the proximal or distal Figure 2. fmi Mutations Alter Planar Polarity in the Wing edge of each cell (Figure 3I); alternatively, the protein (A) Different regions of the adult wing are indicated, and arrows could be abundant at both edges (Figure 3J). To distin- represent wing hair polarity in individual regions on the dorsal sur- guish between these possibilities, we generated mutant face (Wong and Adler, 1993). clones using an almost protein-null allele, fmiE59, and (B and C) Region D of the wild-type (B) and the fmi null mutant that escaped embryonic lethality by transgene expression ([C]; see examined whether Fmi was localized at interfaces be- ϩ E59/ details in Experimental Procedures). In the fmi mutant wing, hairs tween fmi cells (fmi ϩ or ϩ/ϩ) and mutant cells directed posteriorly. (fmiE59/fmiE59) along clone borders. Along the borders of (D) The fmiE59 mutant clones, which were marked by a hair shape more than 50 clones observed, Fmi was missing at all E59 marker, pawn (pwn), are outlined with a dotted line. fmi is essen- interfaces between fmiϩ and fmiϪ cells, no matter where tially cell autonomous on hair polarity. A few wild-type cells along the interface was positioned (Figure 3K). Assuming that a clone border made twin hairs (arrow). Only 4 mutant clones out of 104 examined were in direct contact with 1–3 wild-type cells with Fmi molecules bind to each other in a homophilic fashion twin hairs. Distal is to the right and anterior is at the top in all panels. in vivo as in vitro, the most plausible interpretation of this mosaic analysis would be that both of the two ap- posed cells need to produce Fmi to localize this receptor range over ten cells when clones are large (Vinson and at the interface and that retention of Fmi at P/D bound- Adler, 1987; Taylor et al., 1998; Wolff and Rubin, 1998). aries is achieved through homophilic interaction be- In contrast, the fmi null mutation behaved essentially in tween its ectodomains. If our interpretation is correct, a cell-autonomous way (Figure 2D). it follows that Fmi is present at both proximal and distal edges in normal epithelial cells, as drawn in Figure 3J. Transient Polarization of Subcellular Localization of Fmi along the P-D Axis Abnormal Distribution of Fmi in the Absence of Fz Hairs in adult wings are derived from prehairs that As an attempt to pursue functional relationships be- emerge 30–36 hr after puparium formation (hr APF) at tween fmi and previously discovered tissue polarity 25ЊC. To gain an insight into where Fmi functions, we genes, we studied whether Fmi distribution is altered in stained wing epithelia for Fmi at various pupal stages those polarity mutants, particularly in fz complete loss- and found dynamic transitions in the subcellular distri- of-function mutants (Figures 4A and 4B). In the total bution of Fmi. absence of Fz protein (fzD21/fzK21; Park et al., 1994a), Fmi In 18 hr APF wings, immunofluorescence signals of was not redistributed toward the P/D boundaries at 24 Fmi were present almost entirely at cell-to-cell bound- or 30 hr APF. At the onset of prehair formation (30 hr aries (Figure 3A). However, at 24 hr APF, Fmi became APF), bright staining at cell boundaries was greatly re- redistributed; its localization looked biased toward the duced in length, and the fragmented signals were not P/D cell boundaries in many cells (arrowheads in Figure necessarily restricted to the P/D boundaries (compare Role of Flamingo in Planar Cell Polarity 589

Figure 3. Transient Polarization of Subcellu- lar Localization of Fmi (A–H) Immuofluorescence pictures of pupal wings before or during prehair development. Distal is to the right and anterior is at the top in all photographs except for (G) and (H), which are images of vertical sections. (A) At 18 hr APF, Fmi is localized at almost all cell– cell boundaries. (B and C) In the 24 hr APF wing (B), Fmi tended to be enriched at proxi- mal/distal (P/D) cell boundaries (arrowheads) rather than anterior/posterior boundaries (arrows), and this biased localization became most prominent at 30 hr APF, just before the onset of prehair morphogenesis (C). The po- larized pattern was seen in all epithelial cells, irrespective of their size and shape. For ex- ample, an asterisk indicates a pentagonal cell that is the smallest in this panel. (D) A 34 hr APF wing stained with dye-conjugated phal- loidin (green) and Fmi antibodies (red; overlap with green and yellow). Prehairs had emerged from distal vertexes of cells. (E) Fmi localiza- tion became almost nonpolar at 36 hr APF. (F) A 30 hr APF wing stained for DE-cadherin. All images were of region C except for (E), an image of region E. (G and H) Images of vertical sections of the 30 hr APF wing that was dou- bly stained for Fmi ([G] and red in [H]) and DE-cadherin (green in [H]). Arrows indicate apical-free surfaces of the epithelial cells. At this stage, the cells are about 12 ␮m in height along the apicobasal axis. (I and J) Two models to explain the Fmi zigzag pattern at 30 hr APF. For simplicity, Fmi mole- cules, which assembled at cell boundaries, are illustrated as if they were undercoat pro- teins (thick blue lines). (K) Asterisks indicate cells homozygous for fmiE59. Fmi was missing at all interfaces be- tween fmi-expressing and -nonexpressing cells. Bar, 5 ␮m for (A–H) and (K).

Figures 3C and 4A), indicating that generation of the Apposition of fz-Expressing and Nearly normal Fmi pattern is strongly dependent on Fz. Resid- Nonexpressing Cells Resulted ual boundary signals became even less prominent at in Accumulation of Fmi later stages, leaving only fine dots both in the cytoplasm at the Cell Interface and along cell borders (compare Figures 3E and 4B). In the preceding experiment, we studied Fmi distribution Along the apicobasal cell axis, these intracellular parti- in pupal wings where all the cells lost fz expression. To cles were present from near the apical surface to the dissect how Fz-dependent intercellular communication basolateral level. controls Fmi localization, we adopted two approaches dsh1 is a genetic null allele for planar polarity (Perrimon to juxtapose cells with different fz expression levels and and Mahowald, 1987). Wings with this mutation also investigated how those conditions affect Fmi distribu- showed a decrease in intensity of Fmi staining at cell tion. One approach was to generate fz mutant clones (Figure 5), and the other was to express fz in a gradient boundaries, and the distribution looked much less polar- fashion (Figure 6). ized than that in wild type (Figure 4C). We stained wings We made clones of cells homozygous for a strong fz of other tissue polarity mutants and found that mutations allele, fzR52, which produces a truncated polypeptide at of all the genes did not necessarily disrupt the Fmi distri- a very low level (Jones et al., 1996), and we stained for bution (T. U. et al., unpublished data). For example, in Fmi. Strikingly, the fz mutant cells appeared to decide mutant cells of the multiple wing hair (mwh) gene, which where to localize Fmi molecules in a neighbor-depen- is currently considered to be further downstream in the dent manner (Figure 5B). Along clone borders, Fmi was tissue polarity pathway (Wong and Adler, 1993), Fmi accumulated at almost all the interfaces between fzϩ molecules were present predominantly at P/D bound- (fzR52/ϩ or ϩ/ϩ) and fzR52/fzR52 cells whether the interface aries (Figure 4D). Therefore, Fz and one downstream was a P/D cell boundary or not (Figure 5B). In contrast, component, Dsh, are necessary to accomplish the nor- Fmi was never localized intensely at boundaries be- mal distribution of Fmi. tween outermost mutant cells (yellow dots in Figure 5B; Cell 590

cell boundaries in some fzϩ cells (arrows in Figure 5E). Deformation of the Fmi pattern was minimal in areas proximal to the clone (left half of Figure 5C; see also Figure 5D). Whether prehair morphogenesis always occurred in the vicinities of such Fmi-rich A/P boundaries was diffi- cult to determine because Fmi localization was being depolarized once prehairs started outgrowth. Neverthe- less, the directional nonautonomous effect of the fz mu- tation on Fmi localization was observed in 20 out of 20 clones examined that were well isolated from one another, suggesting a tight coupling of Fmi mislocaliza- tion with incorrect placement of prehair sites. This ob- servation is consistent with the hypothesis that assem- bled Fmi molecules play an important role in initiating prehair morphogenesis.

Fz Gradient along the A-P Axis Provided Sufficient Information to Localize Fmi at A/P Cell Boundaries Besides the clonal analysis, we took advantage of the GAL4-UAS system (Brand and Perrimon, 1993) to ap- Figure 4. Distribution of Fmi in Tissue Polarity Mutants pose cells with different levels of fz expression as de- Immunostaining pictures for Fmi of pupal wings of tissue polarity scribed previously (Adler et al., 1997) and examined how mutants fzK21/fzD21 (A and B), dsh1 (C), and mwh1 (D). Images of region the distribution of Fmi was altered. The patched (ptc)- D at 30 hr APF (A, C, and D) or at 36 hr APF (B) are shown. GAL4 driver was used to generate a short-range gradient (A and B) Absence of Fz protein made boundary signals of Fmi of fz expression along the A-P axis within an anterior fragmentary and less polarized, and cytoplasmic staining was more compartment of region C (Figure 6A), and this ectopic evident when compared to normal patterns (Figures 3C and 3E, respectively). expression made the wing hairs point from high to low (C) Distribution of Fmi was altered similarly in dsh1. levels of Fz (Adler et al., 1997; Figures 6B and 6C). In (D) The Fmi zigzag pattern appeared to be preserved in mwh1. Distal contrast to the normal zigzag patterns of Fmi at P/D cell is to the right and anterior is at the top in all panels. Bar, 5 ␮m. boundaries (Figure 6E), Fmi was concentrated primarily at A/P cell boundaries in the presence of the Fz gradient along the A-P axis (Figure 6F). This result showed that see also Figures 5D and 5E); in other words, those outer- under this experimental condition, juxtaposition of cells most mutant cells always restricted the distribution of with different Fz levels is sufficient to accumulate Fmi Fmi to contact sites with fzϩ cells. These results indi- at the interface. cated that every cell has a system to monitor fz expres- It was previously shown that effects of fz overexpres- sion levels across each boundary and that if there is an sion are suppressed by mutations in the dsh gene, which imbalance across a certain boundary, the cell deposits acts downstream of fz (Krasnow et al., 1995; Adler et Fmi molecules preferentially at this particular cell–cell al., 1997); this result suggests that fz overexpression contact site. mimics activation of the signaling pathway. Given that Inner mutant cells that did not contact the clone bor- levels of fz gene expression correlate with those of Fz ders displayed fuzzy Fmi signals both at interfaces be- protein activity, the results of the gradient expression tween themselves and in the cytoplasm, and the distri- and the clonal analyses allowed us to draw two conclu- bution of the boundary signals was not polarized (Figure sions. First, a cell is able to communicate with its neigh- 5B; see also Figure 5C). These abnormal patterns were bors to monitor the amplitude of Fz activity across each reminiscent of those in wings of the fz null mutant (Figure cell–cell boundary. Second, enrichment of Fmi at one 4A). This observation of the fz clone confirmed that the boundary is a hallmark of a difference of Fz activity in fz gene is necessary to concentrate Fmi to cell–cell between the two juxtaposed cells. boundaries and, in addition, to bias the distribution of Fmi toward P/D boundaries. Opposing Directions of Polarization by Gradient R52 It is known that most fz mutations, including fz , Expression of fz and fmi show non–cell autonomy in orienting wing hairs with Results described so far allowed us to build a model perfect penetrance (Jones et al., 1996). In areas distal showing (1) that in the wild-type wing, Fz activity is made to but not proximal to fz mutant clones, hairs of fzϩ cells unequal across every P/D cell boundary, (2) that this swirl toward the clone; this is due to initiation of prehair difference causes bilateral assembly of Fmi molecules formation at abnormal subcellular sites, for example, at at the P/D boundary, and (3) that those assembled Fmi posterior and proximal cell vertexes (Adler et al., 1997). molecules play an important role in initiating prehair We found that the misselection of prehair sites appeared formation in the vicinity of the P/D boundary. However, to be prepatterned by mislocalization of Fmi (Figures bilateral distribution of Fmi per se does not explain how 5C–5E). The Fmi zigzag pattern was distorted in areas the distal edge, not the proximal one, is selected to distal to the clone; many zigzags ran obliquely rather reorganize the cytoskeleton to form a prehair. An experi- than orthogonally to the P-D axis (arrowheads in Figure ment of fmi overexpression, described next, provided 5C), and at a subcellular level, Fmi was present at A/P an insight into how Fmi breaks cellular symmetry. Role of Flamingo in Planar Cell Polarity 591

Figure 5. How Subcellular Localization of Fmi Was Affected within, along the Border of, and outside an fz Clone Clones of cells homozygous for a strong al- lele, fzR52, were generated by the Flp-FRT technique. Distal is to the right and anterior is at the top in all panels. This pupal wing was fixed at 30 hr APF and stained for N-Myc to detect the fz clones (A) and for Fmi (B–E). (A) A clone of fzR52/fzR52 cells was recognized by the absence of N-Myc staining, whereas cells in its twin spot had two copies of the wild-type fz gene (ϩ/ϩ) as well as the N-Myc marker gene and thus were stained brightly. (B) High-power image of Fmi staining of the boxed area in (A). Note that intense Fmi sig- nals were detected at almost all interfaces between outermost fz mutant cells (yellow dots) and wild-type cells (fzR52/ϩ in this panel); in contrast, Fmi was hardly localized at con- tact sites between the outermost mutant cells. (C–E) A large region containing the fz mutant clone and its wild-type twin (ϩ/ϩ), which were outlined with yellow and green dots, respec- tively (C). In an area distal to the clone, many Fmi zigzags ran obliquely rather than orthog- onally to the P-D axis (a string of blue arrow- heads). (D and E) High-power images of areas high- lighted in (C). In cells proximal to the fz clone, Fmi molecules were predominantly present at P/D boundaries (D), whereas Fmi was present ectopically at A/P boundaries in many cells distal to the clone (arrows in [E]). Besides twin clones like (C), where the ϩ/ϩ clone was positioned distal to the mutant clone, we ob- served Fmi patterns in cases where ϩ/ϩ twin spots were located proximal to mutant clones. Irrespective of the relative position of the clones of the two genotypes, Fmi distribu- tion was affected in areas distal to but not proximal to mutant clones (data not shown). Similarly, it was previously reported that the location of ϩ/ϩ cells was not important for producing the nonautonomous effect on hair polarity (Taylor et al., 1998). Bar, 20 ␮m for (A); 5 ␮m for (B), (D), and (E); 10 ␮m for (C).

As is also the case with fz, not only loss-of-function hypothetical downstream signaling cascade, whether mutations but also overexpression of fmi could disrupt they are located in plasma membranes or in cytoplasm, planar polarity. By using a number of GAL4 drivers, we it follows that wing hairs point toward cells with stronger compared the effects of overproduction of Fmi with activity of Fmi. This assumption permits a model that those of overproduction of Fz. The most striking result explains generation of distally oriented hairs in the wild- was obtained when a gradient of Fmi was induced with type wing, where neither Fz nor Fmi is distributed in a the ptc-Gal4 driver. The phenotype induced by the Fmi gradient fashion along the P-D axis (Park et al., 1994a, gradient presented a sharp contrast to that caused by 1994b; this study). In this model, the distal side of the the Fz gradient, that is, hairs were oriented toward the P/D boundary, that is, the proximal edge of every cell, highest point of Fmi levels (Figures 6C and 6D). Cells has a stronger Fmi activity in spite of the apparent sym- within the gradient tend to accumulate overproduced metrical distribution (Figure 7C; discussed later). Fmi molecules in cytoplasm (Figure 6G). This polariza- tion along the Fmi gradient required the Fmi ectodomain, Discussion as gradient production of the ⌬EX form had no effect on the hair polarity (Figure 6B). Fz Signaling Causes Assembly of Fmi Preferentially The above results strongly suggested that when cells at the P/D Boundary to Polarize Wing Cells with equivalent P-D positional information are given the None of the endogenous proteins of known tissue polar- graded expression of fmi, they perceive the difference ity genes, except for Fz, have been reported in terms in the expression level across the cell boundaries and of subcellular distribution in the pupal wing (Park et al., direct hairs toward their neighbors that produce higher 1994a; see below). We have identified a novel tissue levels of Fmi. If overproduced Fmi molecules activate a polarity gene, fmi, and shown that Fmi is localized in a Cell 592

Figure 6. Experiments of Gradient Expres- sion of fz and fmi (A) ptc-Gal4 drives gradient expression of tar- get genes along the A-P axis in region C, which was visualized by GFP fluorescence. A peak of the gradient expression is repre- sented by the arrowhead. We expressed tar- get genes that encode the ⌬EX protein, which lacks most of the ectodomain of Fmi (B), Fz (C and F), the full-length Fmi (D and G), and GFP (E). (B–D and G) Shown are magnified views of areas that roughly correspond to the box in (A). Hair polarity is indicated by arrows in (B–D). (B) Gradient production of the ⌬EX protein did not alter polarity. (C and D) In contrast to hairs pointing away from the peak of fz expression ([C]; Adler et al., 1997), hairs were oriented toward the highest point of the Fmi gradient (D). (G) The Fmi gradient was visualized by the antibody staining, and this confocal image was underexposed to ob- serve the peak of the expression (bracket). (E and F) Shown are distributions of Fmi in small areas of 30 hr APF wings, which are located within boxes in (B) and (C), respectively. In contrast to the normal localization of Fmi preferentially at P/D cell boundaries (E), Fmi molecules were concentrated primarily at A/P boundaries in the presence of the Fz gradient (F). Smaller cells are prevein cells. Distal is to the right and anterior is at the top in all panels. Bar, 5 ␮m for (E) and (F); 15 ␮m for (G). polarized fashion only for a limited period of time prior to Fmi in the dsh mutant imply a possibility that the sorting prehair outgrowth. Fmi is a seven-pass transmembrane of Fmi could be mediated by either direct or indirect receptor of the cadherin superfamily and has a homo- interaction between Fmi and Dsh. Whichever mecha- philic binding activity in vitro. This property and the result nism works, our observation suggested that this opera- of fmi mosaic analysis strongly suggest that both proximal tion is initiated after 18 hr APF. Consistently, tempera- and distal boundaries of individual cells are enriched in ture-shift experiments using a cold-sensitive fz mutation Fmi molecules. suggest that fz function is required between approxi- Under various conditions of dysfunction or experi- mately 15 hr APF at 25ЊC and the start of prehair morpho- mental manipulation of Fz signaling that altered planar genesis (Adler et al., 1994). polarity, patterns of Fmi were affected and those abnor- One obvious question that has not been rigorously mal patterns prefigured how polarity would be dis- answered is where Fz is localized within the cell. Staining turbed. For example, in fz or dsh mutants, disorientation for endogenous Fz proteins was challenging because of wing hairs was preceded by a low-level and nonpolar of its low abundance; it was sometimes possible to distribution of Fmi at cell boundaries; additionally, pre- detect Fz signals at both the apical-free surfaces and hair outgrowth in the anterior or posterior direction was apical portions of cell boundaries irrespective of the associated with ectopic Fmi accumulation at A/P cell stage of prehair development (Park et al., 1994a, 1994b). boundaries. Together with the fact that loss of fmi func- A more difficult task is visualizing the sites where Fz is tion disrupted polarity, all of these results support the activated; one reason for this difficulty is that the ligand hypothesis that Fz signaling plays a pivotal role in bias- for Fz, which is considered to be the polarizing signal, ing Fmi distribution toward P/D cell boundaries in the has not been identified. wild-type wing and that generation of this polar Fmi Apart from Fmi, the cell fate determinant Numb has pattern is prerequisite for the choice of correct intracel- been the only protein reported that displays biased dis- lular sites for prehair development, that is, genesis of tribution along a planar axis, and its localization is also distally directed hairs. Visual inspection of the Fmi pro- under the control of Fz signaling (Gho and Schweisguth, tein has revealed a previously untractable process be- 1998). Numb is a membrane-associated intracellular fore prehair morphogenesis and provided a novel tool protein and forms a crescent that overlies one of the two to dissect the molecular basis for planar polarization. spindle poles of cells that undergo asymmetric divisions (reviewed by Jan and Jan, 1998). A search for proteins What is the Molecular Mechanism to Polarize interacting with Numb led to the identification of Pon, Fmi Distribution? which colocalizes with Numb during mitosis and directs Polarization of the Fmi distribution can be explained by Numb-asymmetric localization (Lu et al., 1998). Although two mechanisms that are not mutually exclusive. One polarization and depolarization of the Fmi distribution is sorting of Fmi preferentially to P/D cell boundaries, occurred in postmitotic cells, a hunt for binding partners and the other is involved in selective retention and deg- may help to disclose the molecular machinery that regu- radation of Fmi at P/D and A/P boundaries, respectively. lates Fmi location. Fz is able to recruit a signal transducer, Dsh, from cyto- plasmic vesicles to cell–cell interfaces in a heterologous Roles of Fmi and Fz in Breaking Cellular Symmetry system (Axelrod et al., 1998). This finding may be sug- along the P-D Axis gestive of the intracellular sorting event. This Fz-depen- We observed behavior of Fmi molecules when cells with dent translocation of Dsh and aberrant localization of different fz expression levels were juxtaposed, and the Role of Flamingo in Planar Cell Polarity 593

al., 1997; Taylor et al., 1998). One is the cell-by-cell signaling model, in which release and uptake of the ligand take place across each P/D boundary, the ligand inactivates Fz only at the proximal cell edge, and, thus, intracellular Fz activity becomes uneven. Alternatively, a proximal-distal concentration gradient of a diffusible ligand leads to differential activation of Fz in cells aligned along the P-D axis (Figures 7A and 7B). Although it has been discussed that both models have difficulties in explaining some data and may need to be modified, cells would monitor this preset bias of the Fz activity and may recruit Fmi molecules to the distal edge, where Fz activity is stronger. Then, the homophilic bind- ing property of Fmi likely induces the assembly of Fmi on the other side of the P/D boundary, leading to forma- tion of “Fmi zipper.” In this way, the asymmetrical Fz activity can be converted into the distribution of Fmi molecules, which is apparently symmetrical. Importantly, however, the result of gradient expression of fmi sug- gested that Fmi molecules at the proximal cell edge are more active in normal wing development (Figure 7C). One plausible mechanism to make Fmi activity unequal across the boundary may be an inhibitory action of Fz against Fmi at the distal edge of the cell. Why do cells employ a system that is so complicated and requires spatially distinct activation of the two recep- Figure 7. A Model of Interplay between Fz and Fmi in Specifying tor species instead of Fz activation alone? What is Fmi’s Planar Polarity role in breaking cellular symmetry along the P-D axis? Schematic drawing of wing cells that are aligned along the P-D axis We speculate that the initial bias of the Fz activity is too and subcellular localization and/or activity of a hypothetical ligand subtle to drive cytoskeletal reorganization and that Fmi for Fz, Fz, and Fmi. Distal is to the right. (A) Before receiving the ligand for Fz, all the cells possess a low is responsible for making this bias stronger. If active Fmi and equal level of Fz activity (short red bars on individual cells). molecules at the proximal edge downregulate activity of (B) The ligand is distributed in a proximal-distal concentration gradi- Fz molecules in the same domain of the plasma mem- ent (orange), and this leads to differential activation of Fz in cells. brane, this inhibitory effect of Fmi against Fz could en- Note that this schema represents one simplified version of a situa- hance the difference in Fz activity across the P/D bound- tion where Fz activity is slightly higher at the distal cell edge than ary (Figure 7D). The antagonistic interplay between the at the proximal one across each P/D boundary. two receptors could make the imbalance of Fz activity (C) Flamingo molecules are recruited at the P/D boundary and dis- tributed bilaterally (blue rectangles on plasma membranes). Fmi is exceed a certain threshold to initiate prehair formation postulated to be less active at the distal cell edge (open rectangle), at the distal cell vertex. Planar polarity is also reflected in due to the stronger Fz activity, and to be more active at the proximal the arrangement of photoreceptor cells in the Drosophila cell edge (closed rectangle). eye, and it was recently proposed that Fz sets up an (D) Active Fmi molecules at the proximal edge downregulate Fz initial small bias, which is amplified by the Delta-Notch activity, which enhances the difference in Fz activity across the pathway (Cooper and Bray, 1999; Fanto and Mlodzik, boundary (arrows). This makes the imbalance of Fz activity exceed 1999). Fmi is also required for polarity formation in the a certain threshold to initiate prehair formation at every distal cell vertex (green). eye; therefore, it is intriguing to investigate where Fmi is localized in ommatidia and whether Fmi is involved in the connection between Fz and Notch signaling results strongly suggested that the difference in Fz activ- pathways. ity promotes assembly of Fmi molecules at the cell–cell The model shown in Figure 7 predicts that fmi mutant contact site. From this assumption, it follows that in the clones show proximal one-cell nonautonomy, that is, wild-type wing there exists an imbalance of Fz activity fmiϩ cells, which abut on mutant cells along the proximal across every P/D cell boundary prior to the onset of prehair clone border, cannot localize Fmi molecules at distal morphogenesis. Such an activity imbalance could exist cell edges and thus are expected to display a polarity not only across the P/D boundary but also across the defect. However, the fmi null mutation behaved mostly, A/P boundary, and cells may take account of relative if not perfectly, in a cell-autonomous manner. This dis- values of the differences and target the P/D boundary, crepancy could be explained by the possibility that each where the imbalance is perhaps larger, to localize Fmi. wing cell has an fmi-independent system, which most In the experiment of gradient expression of fz, wing likely depends on physical contacts with neighbors to hairs point down the Fz slope; this result can be inter- align itself, and that this system serves as a safeguard preted to suggest that the proximal side of the P/D against such a one-cell nonautonomous effect. boundary, that is, the distal edge of every cell, has a In summary, we identified evolutionarily conserved stronger Fz activity in normal wing development (Adler seven-pass transmembrane cadherins and pursued a et al., 1997). Two models were previously proposed to role of the Drosophila homolog Fmi in regulating planar make Fz activity unequal across the boundary (Adler et cell polarity in the wing. We propose that wing cells Cell 594

modulate Fz activity to redistribute Fmi molecules to- (Serotec), and mouse anti-Myc (Santa Cruz). Stained pupal wings ward the P/D cell boundary, enhance the initial imbal- were viewed with a laser scanning confocal microscope MRC1024 ance of Fz activity across the boundary through the (BioRad). To detect high-molecular-weight Fmi polypeptides, we often followed a protocol of a modified PAGE system (Bolt and antagonistic interplay between Fz and Fmi, and cause Mahoney, 1997), except that samples were run in 5% polyacrylamide cytoskeletal reorganization at the distal cell vertex. To gels. dissect a role of Fmi at the molecular level and verify this model, it would be desirable to clarify an active form Cell Culture of Fmi biochemically and visualize subcellular distribu- fmi was transiently expressed in S2 cells by use of an inducible tion of those molecules in the context of prehair devel- expression construct as described previously (Iwai et al., 1997) or opment. by cotransfection with actin5C-Gal4 (a gift from Yasushi Hiromi) and UAS-fmi. In some experiments, cells were transfected with actin5C- GAL4, UAS-fmi, and UAS-GFP.RN3 (Zernicka-Goetz et al., 1996) Experimental Procedures and mixed with untransfected cells. Molecular Cloning and Sequencing Drosophila cDNA fragments with cadherin repeats were amplified Acknowledgments as described (Iwai et al., 1997), and we subsequently isolated clones from cDNA and genomic libraries. Transmembrane domains of Fmi We thank P. Adler very much for his openness to share a number were predicted according to the application SOSUI (Hirokawa et of fly strains and discuss unpublished results. We also thank J. al., 1998). The transformation construct p[UAS-fmi] was a pUAST Axelrod, Bloomington Stock Center, J. Gurdon, S. Hayashi, Y. Hiromi, derivative with a 12.2 kb fragment containing the fmi ORF. The ⌬EX R. Niwa, and T. Tabata for reagents. We are grateful to members construct was designed so that the protein product lacks almost of A. Nose’s group for allowing us to perform microinjection in their the entire ectodomain (amino acid residues 56–2769). cDNA of a laboratory; to the Human Genome Center at the Institute of Medical mouse fmi homolog (mfmi1) was isolated by screening brain cDNA Sciences of Tokyo University for use of its databases; and to H. libraries (Clontech) with KIAA0279 (Nagase et al., 1996) and W55583 Ueda for discussion from a mathematical point of view. This work (one of I.M.A.G.E. Consortium cDNA clones; Lennon et al., 1996) as was supported by grants from PRESTO of Japan Science and Tech- probes and by 5Ј-RACE-PCR using Marathon-ready cDNA (Clon- nology Corporation (T. U.), from Grants-in-Aid for Specially Pro- tech). KIAA0279 encodes a human homolog of mouse Fmi1 and rat moted Research of the Ministry of Education, Science, and Culture MEGF3 (Nakayama et al., 1998). We determined sequences of fmi of Japan (M. T.), and by NIMH 48108 (T. L. S.). T. Usui is a recipient mutations essentially as described (Iwai et al., 1997). of a Fellowship of the Japan Society for the Promotion of Science for Junior Scientists. Drosophila Strains fmi is cytologically located at band 47B of the second chromosome. Received June 22, 1999; revised July 27, 1999. We searched for fmi mutations in a collection of 234 EMS-induced lethal strains that did not complement Df(2R)17, which eliminates the chromosomal interval 47A to 47E/F (R. W. B. and T. L. S., unpub- References lished data). In this collection, 17 strains greatly reduced the inten- sity of the Fmi signal in homozygous embryos. We found that 31 Adler, P.N. (1992). The genetic control of tissue polarity in Drosoph- lines, including these 17, belonged to the same complementation ila. Bioessays 14, 735–741. group. At least two alleles, fmiE45 and fmiE59, were genetically null, Adler, P.N., Charlton, J., Jones, K.H., and Liu, J. (1994). The cold- as homozygotes of either allele showed axonal phenotypes similar sensitive period for frizzled in the development of wing hair polarity to that of each allele over Df(2R)17. Lethality and the axonal defect ends prior to the start of hair morphogenesis. Mech. Dev. 46, of fmiE45/fmiE59 were recovered in animals of fmiE45 p[Gal4-1407]/ 101–107. fmiE59;p[UAS-fmi]/ϩ. Gal4-1407 allows expression of target genes in both neuronal precursors and in embryos (Luo et al., Adler, P.N., Krasnow, R.E., and Liu, J. (1997). Tissue polarity points 1994) but not in imaginal photoreceptor cells. The rescued adults from cells that have higher Frizzled levels towards cells that have displayed planar polarity phenotypes, whereas fmi cDNA expression lower Frizzled levels. Curr. Biol. 7, 940–949. under Gal4-1407 in the wild-type background did not disrupt polar- Axelrod, J.D., Miller, J.R., Shulman, J.M., Moon, R.T., and Perrimon, ity, supporting the idea that the polarity phenotypes in the rescued N. (1998). Differential recruitment of Dishevelled provides signaling animals were due to loss of fmi function. Other strains were fz specificity in the planar cell polarity and Wingless signaling path- mutants (fzK21 and fzD21; Park et al., 1994a), UAS-fz (Adler et al., 1997), ways. Genes Dev. 12, 2610–2622. dsh1 (Krasnow et al., 1995), mwh1 (Gubb and Garcia-Bellido, 1982), Bhanot, P., Brink, M., Samos, C.H., Hsieh, J.C., Wang, Y., Macke, and ptc-Gal4 (Hinz et al., 1994). We made recombinant chromo- J.P., Andrew, D., Nathans, J., and Nusse, R. (1996). A new member of somes of ptc-Gal4 UAS-fz and stained homozygotes for Fmi. the Frizzled family from Drosophila functions as a Wingless receptor. Nature 382, 225–230. Generating Clones Bockaert, J., and Pin, J.P. (1999). Molecular tinkering of G protein- Clones of mutant cells (fmiE59 or fzR52) were made with the Flp-FRT coupled receptors: an evolutionary success. EMBO J. 18, 1723– system (Xu and Rubin, 1993). Relevant genotypes were hsp-70- 1729. flp/ϩ;p[FRT]42D fmi/p[FRT]42D p[hs-CD2] (Rodriguez and Basler, Bolt, M.W., and Mahoney, P.A. (1997). High-efficiency blotting of 1997) and hsp-70-flp/ϩ;p[FRT]80B fz/p[FRT]80B p[N-Myc] (Adler proteins of diverse sizes following sodium dodecyl sulfate-poly- et al., 1997) for staining pupal wings and hsp-70-flp/ϩ;p[FRT]42D acrylamide gel electrophoresis. Anal. Biochem. 247, 185–192. pwn fmi/p[FRT]42D for adult wing hairs. Boutros, M., Paricio, N., Strutt, D.I., and Mlodzik, M. (1998). Dishev- elled activates JNK and discriminates between JNK pathways in Antibodies, Immunostaining, and Western Blot planar polarity and wingless signaling. Cell 94, 109–118. Fusion proteins were made in Escherichia coli and used to generate Brand, A.H., and Perrimon, N. (1993). Targeted gene expression as rat antisera to extracellular or intracellular epitopes and mouse a means of altering cell fates and generating dominant phenotypes. monoclonal antibodies to the ectodomain. Reciprocal immunopre- Development 118, 401–415. cipitation and Western blot analysis suggested that the 400 kDa precursor is internally cleaved to generate amino-terminal 300 kDa Cooper, M.T., and Bray, S.J. (1999). Frizzled regulation of Notch and carboxy-terminal 100 kDa fragments and that they form a com- signaling polarizes cell fate in the Drosophila eye. Nature 397, plex (data not shown). Antibodies used were as follows: mouse anti- 526–530. Fz (Park et al., 1994a), DCAD2 (Oda et al., 1994), mouse anti-CD2 Costa, M., Raich, W., Agbunag, C., Leung, B., Hardin, J., and Priess, Role of Flamingo in Planar Cell Polarity 595

J.R. (1998). A putative catenin-cadherin system mediates morpho- KIAA0280) deduced by analysis of cDNA clones from cell line KG-1 genesis of the Caenorhabditis elegans embryo. J. Cell Biol. 141, and brain. DNA Res. 3, 321–329, 341–354. 297–308. Nakayama, M., Nakajima, D., Nagase, T., Nomura, N., Seki, N., and Eaton, S. (1997). Planar polarization of Drosophila and vertebrate Ohara, O. (1998). Identification of high-molecular-weight proteins epithelia. Curr. Opin. Cell Biol. 9, 860–866. with multiple EGF-like motifs by motif-trap screening. Genomics 51, Eaton, S., Wepf, R., and Simons, K. (1996). Roles for Rac1 and Cdc42 27–34. in planar polarization and hair outgrowth in the wing of Drosophila. J. Oda, H., Uemura, T., Harada, Y., Iwai, Y., and Takeichi, M. (1994). Cell Biol. 135, 1277–1289. A Drosophila homolog of cadherin associated with Armadillo and Fanto, M., and Mlodzik, M. (1999). Asymmetric Notch activation essential for embryonic cell-cell adhesion. Dev. Biol. 165, 716–726. specifies photoreceptors R3 and R4 and planar polarity in the Dro- Park, W.J., Liu, J., and Adler, P.N. (1994a). The frizzled gene of sophila eye. Nature 397, 523–526. Drosophila encodes a membrane protein with an odd number of Gho, M., and Schweisguth, F. (1998). Frizzled signaling controls transmembrane domains. Mech. Dev. 45, 127–137. orientation of asymmetric sense organ precursor cell divisions in Park, W.J., Liu, J., and Adler, P.N. (1994b). frizzled gene expression Drosophila. Nature 393, 178–181. and development of tissue polarity in the Drosophila wing. Dev. Gubb, D. (1998). Cellular polarity, mitotic synchrony and axes of Genet. 15, 383–389. symmetry during growth. Where does the information come from? Perrimon, N., and Mahowald, A.P. (1987). Multiple functions of seg- Int. J. Dev. Biol. 42, 369–377. ment polarity genes in Drosophila. Dev. Biol. 119, 587–600. Gubb, D., and Garcia-Bellido, A. (1982). A genetic analysis of the Rodriguez, I., and Basler, K. (1997). Control of compartmental affinity determination of cuticular polarity during development in Drosophila boundaries by hedgehog. Nature 389, 614–618. melanogaster. J. Embryol. Exp. Morphol. 68, 37–57. Shulman, J.M., Perrimon, N., and Axelrod, J.D. (1998). Frizzled sig- Hadjantonakis, A.K., Formstone, C.J., and Little, P.F.R. (1998). naling and the developmental control of cell polarity. Trends Genet. MCelsr1 is an evolutionarily conserved seven-pass transmembrane 14, 452–458. receptor and is expressed during mouse embryonic development. Takeichi, M. (1995). Morphogenetic roles of classic cadherins. Curr. Mech. Dev. 78, 91–95. Opin. Cell Biol. 7, 619–627. Hinz, U., Giebel, B., and Campos-Ortega, J.A. (1994). The basic- Taylor, J., Abramova, N., Charlton, J., and Adler, P.N. (1998). Van helix-loop-helix domain of Drosophila lethal of scute protein is suffi- Gogh: a new Drosophila tissue polarity gene. Genetics 150, 199–210. cient for proneural function and activates neurogenic genes. Cell Turner, C.M., and Adler, P.N. (1998). Distinct roles for the actin and 76, 77–87. microtubule cytoskeletons in the morphogenesis of epidermal hairs Hirokawa, T., Boon-Chieng, S., and Mitaku, S. (1998). SOSUI: classi- during wing development in Drosophila. Mech. Dev. 70, 181–192. fication and secondary structure prediction system for membrane Uemura, T. (1998). The cadherin superfamily at the synapse: more proteins. Bioinformatics 14, 378–379. members, more missions. Cell 93, 1095–1098. Hough, C.D., Woods, D.F., Park, S., and Bryant, P.J. (1997). Organiz- Uemura, T., Oda, H., Kraut, R., Hayashi, S., Kataoka, Y., and Takeichi, ing a functional junctional complex requires specific domains of the M. (1996). Zygotic Drosophila E-cadherin expression is required for Drosophila MAGUK Discs large. Genes Dev. 11, 3242–3253. processes of dynamic epithelial cell rearrangement in the Drosophila Ishihara, T., Nakamura, S., Kaziro, Y., Takahashi, T., Takahashi, K., embryo. Genes Dev. 10, 659–671. and Nagata, S. (1991). Molecular cloning and expression of a cDNA Vinson, C.R., and Adler, P.N. (1987). Directional non-cell autonomy encoding the secretin receptor. EMBO J. 10, 1635–1641. and the transmission of polarity information by the frizzled gene of Iwai, Y., Usui, T., Hirano, S., Steward, R., Takeichi, M., and Uemura, Drosophila. Nature 329, 549–551. T. (1997). Axon patterning requires DN-cadherin, a novel neuronal Vinson, C.R., Conover, S., and Adler, P.N. (1989). A Drosophila tissue adhesion receptor, in the Drosophila embryonic CNS. 19, polarity locus encodes a protein containing seven potential trans- 77–89. membrane domains. Nature 338, 263–264. Jan, Y.N., and Jan, L.Y. (1998). Asymmetric cell division. Nature 392, Wolff, T., and Rubin, G.M. (1998). Strabismus, a novel gene that 775–778. regulates tissue polarity and cell fate decisions in Drosophila. Devel- Jones, K.H., Liu, J., and Adler, P.N. (1996). Molecular analysis of opment 125, 1149–1159. EMS-induced frizzled mutations in Drosophila melanogaster. Genet- Wong, L.L., and Adler, P.N. (1993). Tissue polarity genes of Drosoph- ics 142, 205–215. ila regulate the subcellular location for prehair initiation in pupal Krasnow, R.E., Wong, L.L., and Adler, P.N. (1995). Dishevelled is a wing cells. J. Cell Biol. 123, 209–221. component of the frizzled signaling pathway in Drosophila. Develop- Xu, T., and Rubin, G.M. (1993). Analysis of genetic mosaics in devel- ment 121, 4095–4102. oping and adult Drosophila tissues. Development 117, 1223–1237. Lennon, G., Auffray, C., Polymeropoulos, M., and Soares, M.B. Zernicka-Goetz, M., Pines, J., Ryan, K., Siemering, K.R., Haseloff, (1996). The I.M.A.G.E. Consortium: an integrated molecular analysis J., Evans, M.J., and Gurdon, J.B. (1996). An indelible lineage marker of genomes and their expression. Genomics 33, 151–152. for Xenopus using a mutated green fluorescent protein. Develop- Lu, B., Rothenberg, M., Jan, L.Y., and Jan, Y.N. (1998). Partner ment 122, 3719–3724. of Numb colocalizes with Numb during mitosis and directs Numb asymmetric localization in Drosophila neural and muscle progeni- GenBank Accession Numbers tors. Cell 95, 225–235. Luo, L., Liao, Y.J., Jan, L.Y., and Jan, Y.N. (1994). Distinct morphoge- fmi and mfmi1 have been deposited in GenBank under accession netic functions of similar small GTPases: Drosophila Drac1 is in- numbers AB028498 and AB028499, respectively. volved in axonal outgrowth and myoblast fusion. Genes Dev. 8, 1787–1802. Miller, J.R., and McClay, D.R. (1997). Characterization of the role of cadherin in regulating cell adhesion during sea urchin development. Dev. Biol. 192, 323–339. Mitchell, H.K., Roach, J., and Petersen, N.S. (1983). The morphogen- esis of cell hairs on Drosophila wings. Dev. Biol. 95, 387–398. Nagase, T., Seki, N., Ishikawa, K., Ohira, M., Kawarabayasi, Y., Ohara, O., Tanaka, A., Kotani, H., Miyajima, N., and Nomura, N. (1996). Prediction of the coding sequences of unidentified human genes. VI. The coding sequences of 80 new genes (KIAA0201-