dishevelled is required during wingless signaling to establish both cell polarity and cell identity

The dishevelled gene (dsh) is required for at least two functions in Drosophila development, namely, establishment of coherent arrays of polarized cells (Fahmy and Fahmy, 1959; Gubb and Garcia-Bellido, 1982) and establishment of embryonic segments (Perrimon and Mahowald, 1987). Polarity, as used in this communication, refers not to apical basal polarity but to the planar polarity exhibited by many cells of an epithelium. Planar polarity is evident in cells producing bristles and hairs which, in Drosophila, are cytoskeletal extensions of individual cells (Mitchell et al., 1983) and thus reflect the polarity of the cells that extend them (Piepho, 1955). In every region of every body segment, bristles and hairs are arranged in coherent arrays of defined polarity (Piepho, 1955; Gubb and Garcia-Bellido, 1982). Establishment of this polarity appears to involve a cell signaling event requiring at least the dsh, frizzled (fz) and spiny legs (sple) genes (Fahmy and Fahmy, 1959; Gubb and Garcia-Bellido, 1982; Adler et al., 1990; Vinson et al., 1989; Wong and Adler, 1993).

Complete loss of dsh (both maternal and zygotic) also produces a segment polarity defect indistinguishable from that seen in wingless (wg) embryos (Perrimon and Mahowald, 1987). Other genes that affect the same domain in each segment include the wingless, dishevelled, armadillo (arm), porcupine (porc), gooseberry, hedgehog, cubitis interruptus^D and fused genes (reviewed by Peifer and Bejsovec, 1992; Hooper and Scott, 1992). The wingless gene encodes a secreted protein (WG) (Rijsewijk et al., 1987; van den Heuvel et al., 1989), which provides a key signal during establishment of embryonic segment polarity (Nusslein-Volhard and Wieschaus, 1980; van den Heuvel et al., 1989; Bejsovec and Martinez Arias, 1991; Gonzalez et al., 1991; Peifer et al., 1991) and in patterning adult structures (Morata and Lawrence, 1977; Baker, 1988a; Struhl and Basler, 1993). Since the embryonic phenotypes of wingless, armadillo, dishevelled and porcupine are indistinguishable, it is postulated that these genes may be involved in wingless signal transduction in the embryo (Wieschaus et al., 1984; Perrimon et al., 1989; Klingensmith et al., 1989; Peifer and Bejsovec, 1992; Peifer et al., 1991). Although porcupine has not been described molecularly, activity of the porc gene is non-autonomous in mosaics and mutations prevent release of wingless from the wingless- producing cells; thus, porc acts ‘upstream’ of wingless (Klingensmith and Perrimon, personal communication). The arm gene encodes a β-catenin homologue and is expressed in all cells within the embryonic segment and expressed widely in discs (Peifer and Wieschaus, 1990; McCrea et al., 1991; Peifer et al., 1992) (Morata and Lawrence, 1977; Rijsewijk et al., 1987; Baker, 1988a; Struhl and Basler, 1993). It is required in a cell autonomous manner for response to the wingless signal during adult patterning and thus is placed downstream of the wingless signal (Peifer et al., 1991).

The extent to which pattern formation and the emergence of tissue polarity share genetic elements is not clear, nor is the nature of the patterning information transmitted by the WG signal although the latter topic has received considerable discussion (Wilkins and Gubb, 1991; Hooper and Scott, 1992; Bryant, 1993; Cohen, 1993). We have cloned and sequenced the dsh gene and investigated the developmental requirements of dsh function. Clones of dsh, dsh; wg double heterozygotes, and wg/+ heterozygotes in combination with a reduced function dsh^w transgene, all cause adult pattern abnormalities that mimic those seen in wingless mutations. The patterning abnormalities can be understood in terms of a polar coordinate system of pattern formation where loss of dsh (and/or wg) function leads to loss or mis-specification of circumferential positional values followed by regulative growth (French et al., 1976; Bryant et al., 1981). We propose that dsh encodes an intracellular protein essential for interpretation of the wingless signal(s) and that this signal is required to establish cell polarity and cell identity.

Characteristics of the various genetic markers used are described in Lindsley and Zimm, 1992. Adult flies, ras v dsh¹dy/Df(1) GA118, were prepared for SEM by dehydration in alcohol followed by critical point drying and gold palladium shadowing in the UCI EM facility. Organization of photoreceptors was determined by cutting heads off of flies and mounting the heads on a spot of Vaseline for viewing in the compound microscope by antidromic illumination (Franceschini and Kirschfeld, 1971; Franceschini, 1975). Adult cuticle elements were dissected in 70% EtOH, dehydrated briefly in isopropanol and placed in Gary’s magic mountant (Struhl and Basler, 1993) and photographed using a Nikon Optiphot.

Since dsh is the second gene distal to the discs-large-1 gene (dlg-1) (Lefevre, 1981; Zhimulev et al., 1981; Geer et al., 1983; Voelker et al., 1985), a chromosome walk was conducted from the discs-large gene region (Woods and Bryant, 1989) by screening a genomic library in lambda dash (Stratagene) (a kind gift of D. Woods) using standard methods (Maniatis et al., 1978, 1982). A restriction map was generated and the hopscotch gene (hop) (Perrimon and Mahowald, 1986), which maps between dlg-1and dsh, was located by mapping a p-element insert in the hop^air allele (Watson et al., 1991; J.L. Marsh, unpublished observations). Transcribed regions were first identified by probing Southern blots of cloned genomic DNA with hydrolyzed, kinased RNA (Cox et al., 1984) from various embryonic stages and late 3rd instar larval RNA (L3) (not shown). Individual transcripts were mapped by cloning cDNAs from embryonic cDNA libraries using the genomic clones as probes (Poole et al., 1985; Brown and Kafatos, 1988). Four transcripts, which are present in early embryos and are distal to hop, were identified. Northern blots were performed to determine whether multiple transcripts were evident and to determine the size of the transcripts relative to the cDNA clones. At the level of northern blotting, the longest cDNA from the dsh gene (identified below) was similar in size to the single transcript detected.

Genomic fragments containing one or more of the transcripts (see Fig. 7A) were cloned into either the Carnegie 20B [ry⁺], PW8 [w⁺] or CasPer [w⁺] transformation vectors, transformants recovered and linkage determined by segregation of markers. Complementation tests demonstrated that constructs that contained the 2.5 kb transcript (see Fig. 7A) rescued both the viable dsh and two lethal dsh alleles thus identifying the dsh gene. Construction of transformation plasmids was as follows: For Car 7E, a 7.5 kb genomic SalI fragment was ligated to SalI cut Carnegie 20B with the left-most SalI site adjacent to the rosy gene. The left-most SalI site shown in Fig. 7A in parentheses is an artificial linker site from the phage and Car 20B is a derivative of Car 20 (Rubin and Spradling, 1983) in which the HpaI site has been replaced by XbaI. To eliminate one of the transcription units, Car 7E was cut with XbaI and religated to delete the Xba→Sal region thus leaving only one transcription unit intact (Car 7EX). For construct pW8 7EXB, the same SalI fragment was cloned into the XhoI site of pW8 (Klemenz et al., 1987) and then recut at the BamHI site of pW8 and the natural XhoI site in the genomic DNA, the ends filled in and blunt end ligated. This truncation contains only the 2 kb transcription unit intact. CasPer W was constructed by ligating an EcoRI; BglII genomic fragment into BamHI/EcoRI cut CasPer (Pirrotta, 1986). Relative to the other vectors, we experienced a low frequency of insertion using the pW8 constructs.

Helper plasmid π25.7wc (Rubin and Spradling, 1983) and test plasmid were prepared as described (Rubin and Spradling, 1982; Spradling and Rubin, 1982) and coinjected into either white mutant embryos (CasPer and pW8 constructs), or rosy⁵⁰⁶ mutant embryos (Car 20 constructs). Surviving adults were mated to appropriately marked strains and linkage determined by segregation analysis from balancer chromosomes. Transformants on the second or third chromosomes were tested for complementation of dsh mutations by crossing dsh^x/FM7c females to +/Y; transformant/CyO or /TM3 depending upon the linkage of the transformant (x=v26, VA153 or 1). The presence of non-FM7 non-CyO or non-FM7, non-TM3 males indicated that a particular construct complemented dsh.

The DNA sequence was determined by sequencing selected subclones and a series of nested deletions generated from both ends of the cDNA clones using the Erase-a-Base digestion kit from Promega. Sequencing was performed by the dideoxy chain termination method (Sanger et al., 1977) adapted for use with fluorescently labeled primers. All subclone boundaries were crossed and both strands were sequenced. Actual sequence was determined using the ALF sequencing system (Pharmacia). Sequence output from the computer algorithm was checked by visual inspection of the raw data output. Analysis of the DNA sequence was performed using the Align and MacVector programs (IBI/Kodak). Homology searches of the SwissProt +PIR +GenPept +GPUpdate non-redundant databases (Bilofsky and Burks, 1988) were performed using the BLAST algorithm (Altschul et al., 1990).

To assess the autonomy of the dsh requirement in cell polarity, clones were produced using two lethal alleles of dsh both with and without forked (f) as a marker (i.e., y w dsh^VA153/f^36a, y w dsh^v26/f^36a, y w dsh^v26/w⁺, y w dsh^VA153/w⁺). In one experiment, irradiated animals were heterozygous for the bristle marker forked and only twin spots, where patches of forked bristles could be found adjacent to patches of yellow tissue, were scored. In the second experiment, forked was eliminated in order to better score putative polarity defects in normal cells when adjacent to mutant cells. The w⁺ insertion just proximal to dsh in the second set of animals is immaterial in this experiment. Although there is no independent check for distal recombinants in the second experiment, the frequency of such recombinants has been determined to be less than 14% (Becker, 1976). For clone induction, eggs were collected from the appropriate cross and aged to 24–48 or 70–74 hours and irradiated for 1.5 or 3 minutes respectively at 750 rad/minute in a gammator with a ¹³⁷Cs source. The irradiated animals were placed in fresh bottles of food and analyzed for clones upon eclosion. For analysis of patterning defects, sibs from the y w dsh^v26/w⁺ and y w dsh^VA153/w⁺ animals were analyzed.

The original dsh¹ mutation was described as a viable mutation with deranged hairs on the thorax, divergent and blistered wings, and ellipsoid eyes (Fahmy and Fahmy, 1959; Lindsley and Zimm, 1992). To document the extent of the dsh requirement in polarized cells, we examined the effect of the dsh¹ mutation on planar cell polarity using light microscopy and SEM. All cells in which polarity is evident are affected by dsh mutations (examples of the thorax, abdomen and legs are shown in Fig. 1). Note the abnormal polarity of hairs as well as bristles on the thorax in Fig. 1A-C. In the sternites (Fig. 1D,E), complete reversals of bristles are seen and hairs are also affected. On the leg, bracts normally arise on the proximal side of each bristle by a polarized induction from the bristle cells (Tobler et al., 1973). The dsh¹ mutation leads to the induction of bracts in the wrong position relative to the bristle (Fig. 1F) (Held et al., 1986). In addition, there are extra, mirrorimage duplications of tarsal joints in the legs (Fig. 1G). The rough eye phenotype is seen by SEM to actually reflect abnormal facet packing and incorrect placement of the sensilla (Fig. 2A,B). In addition, polarity of cuticular hairs adjacent to the eye is disrupted.

We used antidromic illumination (Franceschini and Kirschfeld, 1971; Franceschini, 1975) to examine the internal organization of the eyes of dsh mutant flies. In dsh mutants, (i.e., Df (1)GA112/dsh¹), the photoreceptors are correctly organized into the characteristic trapezoidal array and all pattern elements are present, but the photoreceptor clusters are misoriented reflecting an abnormal polarity of the whole ommatidium (Fig. 2C,D). Two classes of abnormalities are evident. Some ommatidia are rotated from their normal orientation but otherwise exhibit normal handedness. Others are reversed in either the A/P or D/V axis (relative to the axes of the adult) with varying degrees of rotation from these axes (Fig. 2D). In mutants (i.e., Df (1)GA112/dsh¹), about 41% of the ommatidia show polarity reversals while approximately 10% of the ommatidia are rotated but exhibit the correct handedness. By examining eyes at different planes of focus (not shown), it is evident that the location of the bristle sensilla in the anterior equatorial vertices is altered in a manner consistent with the polarity reversals and rotations (note placement of sensilla in Fig. 2B). We do not observe incorrect placement or orientation of individual photoreceptor cells within an ommatidium. Thus, the orientation defects appear to affect ommatidia as a whole.

Since only a single viable allele of dsh exists, it was important to determine whether the cell polarity phenotype is a unique property of the dsh viable allele versus the lethal alleles. X-ray induced somatic recombination was used to generate mosaics of the dsh lethal alleles dsh^v26 and dsh^VA153, marked with yellow in two separate experiments. Clones of both lethal alleles were recovered in all body segments and in both cases cells exhibited polarity defects indistinguishable from those seen in dsh¹ homozygotes (e.g., Fig. 3A). Thus, the cell polarity defect is not a novel effect of a single unusual allele and lethal alleles of dsh are not cell lethals. Clones of all sizes were observed from a few bristles to almost half the notum (Fig. 3A). Even in the smallest clones, dsh mutant bristles adopted abnormal orientations despite being intimately surrounded by normal cells. Along the edges of large clones and around the perimeter of smaller mutant patches of tissue, correctly oriented normal bristles were located immediately adjacent to incorrectly oriented mutant bristles. Thus, normal cells are unable to correctly orient dsh mutant cells and mutant cells do not adversely affect neighboring normal cells indicating that dsh functions autonomously in establishing cell polarity.

Combinations of wingless alleles that die as pupae produce a characteristic set of abnormalities including leg defects, loss of head structures and duplicated nota in place of wings (Sharma and Chopra, 1976; Morata and Lawrence, 1977; Baker, 1988b; Peifer et al., 1991). Leg defects in wg pupae can be formally described by the polar coordinate model as including both convergent and divergent duplications (Girton, 1982) (Fig. 4C,D). Similar leg abnormalities are observed when dsh clones are induced by irradiation (e.g., Fig. 4G,H). The ectopic leg associated with the dsh clone in Fig. 4G converges distally as seen by the symmetry of duplicated ventral elements, (except where the dsh clone occupies the location of a putative ventral element). The ectopic leg in Fig. 4H appears to be diverging as it is asymmetric and forms dorsal elements (e.g., the claw); however, as it extends distally, the dsh clone occupies the complete circumference and prevents ventral structures from forming. It is worth noting that, unlike previous studies of pattern regulation which always involved marked normal cells that were fully capable of responding to patterning signals (Girton, 1982), the dsh clones shown here are different in that the mutant clonal tissue is unable to respond to patterning cues during outgrowth and is incapable of producing ventral pattern elements. The location of dsh clones is shown on the fate map of the leg disc in Fig. 5. Clones that result in patterning defects are always located anteroventrally or ventrally on the fate map and mutant cells are often included in the abnormal leg but never in the normal one. Clones of dsh in the posterior and dorsal regions of the leg do not result in patterning defects. Clones of dsh also exhibit defects when they occur in the dorsal-medial region of the head, a region that exhibits defects in wg mutants (Peifer et al., 1991). Defects associated with dsh clones in the head and antenna include loss, ectopic location and duplication of pattern elements (not shown).

To explore the possibility that dsh and wg participate in a common patterning signal in imaginal tissue, we tested for genetic interactions between dsh and wg alleles. Phenotypes observed in animals that are heterozygous for more than one recessive mutation (i.e., synthetic phenotypes) often indicate that the two genes function in a common pathway (e.g., Simon et al., 1991). Several dsh and wg alleles were crossed to produce F₁ daughters which are doubly heterozygous for dsh and wg (Table 1). These flies were examined for pattern abnormalities and scored for lethality. Controls included scoring the sibs and scoring flies from the individual mutants used. A small percentage of doubly heterozygous flies exhibited abnormalities that mimic those seen in wg pharate adults including converging duplications (compare Fig. 4E to C) and bifurcations of leg segments (compare Fig. 4F to D). The abnormalities also mimic those seen in dsh clones (Fig. 4G,H), and those observed in temperature-sensitive wg heteroallelic heterozygotes exposed to restrictive temperature during late embryogenesis (Sharma and Chopra, 1976; Morata and Lawrence, 1977; Baker, 1988b; Peifer et al., 1991; Couso et al., 1993). The rate of spontaneous defects observed in the wg alleles used is rare (i.e., ∼0.35%) while, in the double heterozygotes, the nature of defects observed is more severe and the frequency is increased approximately 10-fold (e.g., to 4.4%) (Table 1). The synthetic interaction appears to be allele specific since the frequency of defects is significantly higher than controls when wg^IG is in combination with either dsh allele but the frequency of defects with wg^CX4 is only elevated in combination with dsh^v26. Thus, dsh and wg interact to give patterning defects similar to those seen in wg alone.

A transformant line carrying a dsh transgene with 184 bp of 5′ flanking and 240 bp of 3′ flanking DNA exhibits reduced dsh activity (P[dsh^w], described in Fig. 7A) such that flies carrying lethal alleles of dsh and one copy of the transgene survive to adulthood but they exhibit cell polarity defects (e.g., dsh^VA153/dsh^VA153; P[dsh^w/+]). 8% of these flies also have missing wings and duplicated nota and other defects typical of wg mutants (Fig. 6A,B). If these flies are also made heterozygous for wg (i.e., dsh/dsh; wg/+;P[dsh^w/+]), the frequency of wingless-like defects increases to 100% (Fig. 6C). The defects include missing and reduced wings, duplicated nota, head defects and leg abnormalities. The fact that reduced wingless exacerbates the defects seen in the reduced dsh background from 8% to 100% provides further evidence that dsh and wg interact in a common pathway to specify cell fate.

Examination of wg^CX4/wg^P pharate adults revealed a small fraction of animals with misoriented bristles on the notum, head and abdomen (Fig. 3B). The polarity defects could be observed through the pupal case or through the pupal cuticle after dissection and, thus, were not the consequence of mechanical disturbance. The polarity defects seen in these wg pharate adults are indistinguishable from those seen in dsh pharate adults (Fig. 3B,C). Further, some dsh/+; wg/+ double heterozygotes described above exhibited defects of the distal wing blade and those animals also exhibited bristle polarity defects but only on one side of the notum; namely, the same side as the wing defect. The concordance of the developmental defect in the wing blade and the polarity defect in the same disc derivative suggests a common cause for the two defects. Thus, loss of wg function alone or simultaneous reduction of dsh and wg activities can lead to polarity defects similar to those seen in dsh.

The dsh gene was cloned and identified by transformation complementation (see Fig. 7A and Methods). A dsh cDNA clone that is the same length as the mRNA as measured by northern blots (not shown) and thus likely to be near full length was sequenced. The transcription unit represented by this cDNA is completely contained within the fragments that rescue dsh mutants (Fig. 7A). The sequence revealed a single long open reading frame beginning with the first AUG in the sequence (Figs 7B, 8). The cDNA contains 250 bp of 5′ UT and 529 bp of 3′ UT and terminates in a 42 nt poly(A) tail which is preceded by a poly(A) addition signal (AATAAA). In situ hybridization reveals widespread expression of dsh throughout development. As expected from the almost ubiquitous requirement for dsh function throughout development, dsh mRNA is found in egg chambers of the ovary and essentially ubiquitously throughout embryogenesis and in discs (not shown). Expression is not seen in salivary glands, muscles or ventral ganglia but is observed in brain lobes.

Conceptual translation of the dsh cDNA predicts a 68.9×10³M_r protein with a pI of 5.9. No hydrophobic regions were found that might correspond to either a secretory leader sequence or a membrane spanning or anchoring region (Kyte and Doolittle, 1982). Thus, the dsh protein exhibits the structural characteristics of an intracellular protein. The amino terminal half of the protein contains a stretch of 34 glutamine (Q) residues interrupted by 2 histidines and 5 other amino acids near the end of the stretch (Figs 7B, 8). The carboxy terminal half of the protein contains a string of 10 contiguous glycine (G) residues. The 530 bp 3′ untranslated region following the ORF is extremely A/T rich (70%) and repetitive. The motif TAA is repeated 15 times broken only by 3 TTA triplets and at position 2456, 50 of the next 56 bp are AT repeats. The 3′ untranslated region contains multiple stop codons in all frames.

A BLAST search (Altschul et al., 1990) identified no proteins with extensive structural homology to dsh; thus, dsh appears to encode a novel protein. However, seven proteins, most of which are found or predicted to be found in association with cell junctions or junctional complexes, showed significant similarity in a shared motif (Fig. 7C) originally called the GLGF repeat (Cho et al., 1992) and now referred to as the DHR repeat (Discs - large Homology Region) (Bryant et al., 1993). Proteins containing this motif include the rat postsynaptic density protein (PSD-95) (Cho et al., 1992), the discs large tumor suppressor protein (DLG (Woods and Bryant, 1991), nitric oxide synthetase from rat brain (NOS) (Bredt et al., 1991), the human tight junction protein ZO-1 (Willott et al., 1992), erythrocyte membrane protein p55 (Ruff et al., 1991), an intracellular protein tyrosine phosphatase (PTP-meg) (Gu et al., 1991), and the putative Friedriech ataxia gene, x11 (Duclos et al., 1993). PSD-95 and DLG each have three copies of this motif, while DSH and the other proteins contain a single copy. The dsh DHR/GLGF motif, aligned with all the DHR/GLGF motifs in Fig. 7C, is most closely related to the first two repeats of PSD-95 and DLG and the single repeat of ZO-1 exhibiting 41 or 40 similar amino acids out of 80 with these 5 repeats.

In this report, we present evidence that (1) planar cell polarity requires the widespread action of the dishevelled gene, (2) a dsh-mediated wg signal is required for both cell fate choice and establishment of planar cell polarity and (3) dishevelled is required autonomously for response to the wingless signal. We also report the cloning and sequence of the dishevelled gene which coupled with autonomy studies suggests that (4) dishevelled encodes an intracellular protein necessary for reception of the wingless signal. We argue that this signal specifies some of the circumferential positional values predicted by the polar coordinate model.

In dsh¹ mutants, bristle and hair polarity is disrupted in all cells in all body segments (Fig. 1). In addition, two polarized structures, which arise by inductive events (bracts and ommatidia), exhibit abnormal polarity in dsh mutants. Bracts on the legs are induced in inappropriate locations rather than proximal to the bristle that induces them (Fig. 1F) (Tobler et al., 1973; Held et al., 1986). The photoreceptor cells of the ommatidia are neurons that elaborate rhabdomeres. Since the polarity of individual photoreceptor cells is always consistent with the other members of that ommatidium and dsh mutations cause misorientation of entire ommatidia (Fig. 2D), it appears that ommatidia are independently developing units and that the orientation of the unit may be controlled by a primary organizing cell. R8 is the first cell of emerging photoreceptor clusters to exhibit differentiation and by implication is suggested to be the ‘founder’ cell (Tomlinson, 1987, 1988; Banerjee and Zipursky, 1990). Mosaic studies are underway to determine whether dsh function is required in a particular organizing cell, possibly R8. The independent development of individual ommatidia suggests that the ommatidia represent tertiary morphogenetic fields in Drosophila (the primary fields being segments and secondary fields being discs (Williams et al., 1993)). The range of polarized structures affected indicates that dsh is required both for polarity of individual cells as well as for polarized interactions between cells.

In mosaics, we frequently found abnormally oriented mutant bristle cells adjacent to correctly oriented normal cells indicating that dsh⁺ activity in a nearby cell cannot serve to orient correctly a mutant cell. This analysis is limited by the fact that the yellow marker cannot be detected in hairs. Thus, normal and mutant cells along clone edges are separated by a variable number of hair-producing cells of unknown genotype. The closest apposition of normal and misoriented bristles that we observed has about four cells separating a normal and misoriented mutant bristle. Since we cannot determine which, if any, misoriented bristle cell are actually in direct contact with a normal cell, we cannot rule out the possibility of very short range interactions over 2 or 3 cell diameters. However, if dsh were capable of influencing neighboring cells, we would expect to see a ring of correctly oriented yellow bristles around the perimeter of dsh clones and this is not seen. Thus, within the limits of this technique, we conclude that dsh is required autonomously for the reception or interpretation of a cell polarity signal.

Loss of dsh function (in mutants or clones) does not specify a new axis or orientation but rather it removes an apparent bias leaving the cell ‘undecided’ as to which orientation to adopt. This is similar to other handed situations such as left/right handedness bias in mammals in which mutation removes a bias toward right handed but does not ‘specify’ left handedness (Annett, 1978, 1979). An alternative view is that all polarity reversals are the consequence of local pattern disruptions followed by regulation giving reversed tissues (a micro scale version of what happens with duplicated, reversed legs). Although the fine structure regulation of pattern in discs is not understood, this alternative seems unlikely since polarity disruptions are apparently not associated with extra, symmetrical or misplaced pattern elements as might be expected from a regulative cause of polarity disruptions.

Significantly, two observations implicate wingless in the establishment of cell polarity. Foremost is the observation that a small percentage of wg^P/wg^CX4 pharate adults exhibit bristle polarity defects identical to those observed in dsh pupae and adults (Fig. 3B). Similar polarity defects in wingless mutant combinations can be seen in several figures of other publications, although polarity was not addressed (Morata and Lawrence, 1977; Baker, 1988a,b; Couso et al., 1993). Secondly, we also observed polarity defects on the nota of dsh;wg double heterozygotes. In each case, the polarity defects were restricted to one half of the notum and were associated with a defective wing blade. The concordance of wing and polarity defects to derivatives of a single disc under conditions of reduced wg signaling also suggests that a wg signal influences cell polarity.

Although acquisition of cell polarity and cell fate share some genes in common, other genes are clearly unique to each process. For example, unlike dsh, the sple and fz genes are required only for cell polarity since complete loss of function of either gene allows a viable animal with polarity defects but no patterning defects (Gubb and Garcia-Bellido, 1982; Adler et al., 1990). The fz gene encodes a novel protein with seven transmembrane domains reminiscent of G-protein-linked membrane receptors (Vinson et al., 1989) while the molecular nature of the sple gene is unknown. It is possible that WG interacts with two sets of signal reception/transduction molecules and that the two sets share some common elements (e.g., dsh) but otherwise lead to separate events within the cell. Indeed, the fz gene appears to carry out two functions, one being a cell autonomous transduction of an intercellular signal to the actin cytoskeleton and the other being a directionally non-autonomous transmission of an intercellular polarity signal (Vinson and Adler, 1987; Wong and Adler, 1993). Interaction with genes such as fz and sple may provide the distinction between the cell polarity and positional value functions mediated by dsh.

Mutations that reduce or eliminate wg function cause loss of structures from regions of the discs that extend beyond the regions of wg expression. The same structures are also affected by mutations that eliminate dsh or arm function (Peifer et al., 1991). Expression of wingless in discs is highly localized while expression of dsh (this report) and arm (Peifer et al., 1991) is widespread. In the leg disc, wg is expressed in an anterior/ventral wedge (Baker, 1988b; Peifer et al., 1991; Couso et al., 1993; Struhl and Basler, 1993) adjacent to the engrailed expression domain, which occupies the posterior compartment of the leg disc (Fig. 5) (Hannah-Alava, 1958; Steiner, 1976; Brower, 1986; Hama et al., 1990). wingless is not expressed in wing discs until the second larval instar when it initially appears as a wedge that refines to a focus of wg expression in the anlage of the notum and a stripe along the margin of the wing blade (Couso et al., 1993). In the eye antennal disc, a peak of wg expression occurs in the region fated to develop the dorsal portion of the head, and in a wedge in the antennal region (Baker, 1988b; Peifer et al., 1991; Couso et al., 1993; Struhl and Basler, 1993). The defective structures seen in dsh (this report) and arm mutants (Peifer et al., 1991) originate from regions in or near regions of high wingless expression in discs and are concordant with structures eliminated by wg mutations.

One of the earliest regions of wg expression in discs is the anterior/ventral wedge in leg discs (Couso et al., 1993). In dsh; wg double heterozygotes, anterior ventral structures of the leg are lost and dorsal structures are duplicated (Note the duplicated preapical bristles, Fig. 4E,F). Other defects include ectopic outgrowths (e.g., Fig. 4F), which exhibit the same circumferential restriction as those seen in wg mutants and dsh clones and mimic the abnormalities seen in wg pharate adults.

One dsh⁺ transgene (P[dsh^w]) rescues lethal alleles of dsh but the rescued animals exhibit polarity defects similar to dsh¹ animals. When the sole source of dsh⁺ activity comes from the transgene (i.e., dsh/dsh; P[dsh^w]/+), 8% of the flies exhibit defects that phenocopy wg defects (Fig. 6B) and arm defects (Baker, 1988b; Peifer et al., 1991). When the same genetic background is made heterozygous for a wg lethal allele (i.e., dsh/dsh; wg/+; P[dsh^w]/+), the frequency of defects increases to 100% with the same range of wingless-like defects including loss of wing, duplication of notum, defects in the legs and in the dorsocentral region of the head and antennal defects (Fig. 6C). The phenocopies of wg seen under conditions of reduced dsh or armadillo activity and the indistinguishable phenotype of dsh, wg and arm mutant embryos (Perrimon and Mahowald, 1987; Peifer et al., 1991) strongly suggest that dsh, wg, arm and perhaps other segment polarity genes act together during pattern formation in both embryos and discs.

The polar coordinate model of pattern formation proposed a set of rules for cell interactions that have proven remarkably robust in predicting tissue behavior (French et al., 1976; Bryant et al., 1981). However, a molecular basis for this model has not been described (Bryant, 1993). A specific prediction of the model is that genes that specify circumferential positional values should be required in cells all along the proximal distal axis of the leg but their activity should be restricted to a particular sector around the circumference of the leg. Previously, it has not been possible to test this prediction by clonal analysis since wg is non autonomous (therefore clones are rescued) and arm is cell lethal hence precluding locating the site of origin of clonal tissue (Peifer et al., 1991). Our finding that dsh clones lead to pattern regulation in a restricted portion of the leg disc circumference provides a unique test of the circumferential signal hypothesis.

While a complete discussion of the implications of the polar coordinate model is beyond the scope of this paper, a brief discussion of a few examples will be illustrative (Girton, 1982; Wilkins and Gubb, 1991; Held, 1993 provide additional discussion of this topic). Pattern regulation will occur when cells are confronted with neighbors of incorrect positional values. This can occur either by loss (e.g. due to cell death) or when values are mis-specified (e.g. as in a ventral clone adopting a more dorsal value; Bryant, 1975; Haynie and Bryant, 1976; Adler, 1981). The predicted outcomes are different depending upon whether the disparity at the confrontation is greater or less than half of the positional values (shortest intercalation rule). Thus, it is significant that fate map studies show that the anterior half of the leg disc contains more positional values than the posterior region (Bryant, 1980), an observation that is consistent with the effects of loss of wg signal in leg discs.

Two predictions of the polar coordinate model are a circumferentially restricted requirement and regulative growth accompanying pattern regulation. Clones of dsh that occur in the posterior compartment or the dorsal-most region of the anterior compartment of the leg disc cause no defects. Clones that occur in the anterior/ventral one third of the leg disc cause bifurcations of the leg. The location of clones that cause patterning defects overlaps the territory in which wg function is required and spans more than half of the positional values on the leg disc fate map (Fig. 5). Clones associated with pattern regulation expand distally while clones in the posterior compartment do not lead to pattern regulation and remain long and thin thus satisfying the prediction that regulative growth accompanies pattern regulation. The defects caused by dsh clones fall into two categories, diverging duplications and converging triplications (e.g., Fig. 4). Like wg defects, they can be explained by loss or mis-specification of ventral values followed by regulative growth. The two clones shown in Fig. 4G,H occurred in the anterior/ventral region of the disc and produce pattern abnormalities similar to those seen in wg mutants with only slight differences due to their clonal nature. The pattern defect associated with the dsh clone in Fig. 4G is a converging triplication which is similar to the converging duplication seen in wg mutants (Fig. 4C). A polar coordinate explanation for these legs is loss of ventral values followed by regulative growth. The reason for duplication versus triplication is that the clone affects only a patch of tissue thus leaving a point of symmetry on either side while wg mutants affect the whole disc including the perimeter of the pattern. Thus, the effects of dsh clones satisfy the two predictions of the polar coordinate model.

The question of autonomy deserves special mention. Mosaic analysis can be used to distinguish a gene involved in the sending of a signal from one involved in the reception or response to a signal. Genes involved in the sending of a signal tend to give no phenotype in clones because neighboring normal cells can provide the diffusible function. For example, clones of wg give no patterning phenotype indicating that normal cells can rescue or compensate for the mutant cells (Baker, 1988a). In contrast, clones of dsh are not rescued by adjacent normal cells suggesting that dsh functions on the receiving side of a signal. The fact that both mutant and normal tissue participate in the ectopic structures formed is expected from the regulative growth that follows from dsh cells being unable to receive the correct positional value signal. The mixed nature of the ectopic outgrowths does not imply non autonomy of dishevelled’s cellular function.

Positional confrontations that lead to pattern regulation can occur either by loss of cells or by mis-specification of positional values. Positional values can be lost through surgical removal (French et al., 1976) or genetically induced cell death (Girton, 1982). Loss of armadillo activity is cell lethal and arm clones produce pattern abnormalities similar to those produced by cell death (Peifer et al., 1991). Positional confrontation can also occur due to mis-specification of positional values. For example, ectopic expression of wingless under heat-shock control leads to mis-specification and pattern regulation (Struhl and Basler, 1993). dsh clones could induce pattern abnormalities either because dsh cells die or because dsh cells are unable to respond to the wingless signal and thus adopt a default positional value. Several observations are relevant here. Firstly, dsh cells survive in clones demonstrating that loss of dsh function is not generally cell lethal. This does not rule out the possibility that a subset of dsh cells die as a consequence of not having dsh⁺ function. Secondly, duplications show symmetry of ventral structures when the regenerated tissue is wild type but when ventral regions are mutant, those cells produce other structures suggesting that they adopt a default value (e.g., Fig. 4G). Thirdly, wg embryos secrete an excess of a single type of denticle in a repeating pattern and this pattern duplication is not accompanied by cell death or by ectopic cell proliferation (Bejsovec and Martinez Arias, 1991), suggesting that cells have adopted a default value. In contrast, many dsh clones are associated with ectopic cuticle elements inside the leg and/or dark fragments of tissue which might be melanized (e.g., Fig. 4G). This tissue could represent dying cells although dead tissue was never observed when genetically induced death was known to occur (Girton, 1982). Nevertheless, it is conceivable that dsh clones cause continued low levels of cell death and that late dying cells do not have time to be cleared thus accounting for why a pulse of cell death never left traces of dying cells while dsh clones may leave such traces. Alternatively, excess tissue might be formed if dsh cells are unable to participate in the pattern and are mechanically displaced during metamorphosis. Additional studies will be required to distinguish between death versus mis-specification as a mechanism.

The sequence of dsh suggests an intracellular protein that shares an amino acid motif with seven proteins that are localized to cell junctions (DLG; PSD-95; ZO-1) or to junctional-like complexes (p55) raising the possibility that DSH may localize to junctional complexes. DSH also has a string of 34 glutamine residues near the amino terminus. Such glutamine-rich strings are thought to be involved in protein-protein interactions (Pascal and Tjian, 1991; Su et al., 1991) which would be consistent with a protein interacting in a complex.

Recent studies indicate that cell signaling events often occur in apical regions of epithelial cells and that proteins mediating these interactions are localized in apical junctions (reviewed in Woods and Bryant, 1993, 1992; Bryant et al., 1993). The transfer of WG protein during signaling can occur in the apical part of the cell immediately basal to the adherens junctions (Gonzalez et al., 1991). The ARM protein, a β-catenin homologue, is part of a membrane-associated complex that includes a large cadherin-like glycoprotein and is localized at the adherens junction (Peifer et al., 1993). In embryonic cells that receive the WG signal, ARM protein is released from the membrane and accumulates in the cytoplasm (Peifer, personal communication). Mutations of dsh, porc and wg block the relocalization of ARM and thus function upstream of ARM (Peifer, personal communication). The intracellular nature of the dsh protein and the autonomous requirement for dsh function for patterning demonstrate that DSH acts downstream of WG. Coupled with the ARM studies, this puts DSH function between reception of the WG signal and redistribution of ARM in response to that signal in embryos. β-catenin molecules such as ARM serve as a link between the adhesive junctions of epithelial cells and the cytoskeleton (Peifer, 1993). Thus transduction of the WG signal may involve a DSH-mediated modulation of ARM and ultimately a change in the organization of the cytoskeleton. We cannot determine whether DSH interacts with the same gene products during both cell fate choice and cell polarity from this study. Our results do show that DSH is an intracellular protein that is required for the response to a wingless signal. The territories that require this signal and the consequences of interfering with it indicate that a dsh-requiring wg signal specifies circumferential positional value in leg discs and thus provides a molecular basis for a theoretical function predicted by the polar coordinate model (Bryant, 1993).

This work was supported by an NSF Research Grant DCB-8916666 to J. L. M and an NIH Research Program Project PO1 HD27173. Some of the work presented in this paper was submitted in partial fulfillment of undergraduate research projects by D. K. and A. S. and their contributions are gratefully acknowledged. J. P and M. B. were supported in part by PHS grant GM28972 to J. L. M and H. T. was supported in part by a PHS training grant 5T32 GM07311-17. We are indebted to B. Cohen and N. Perrimon for stocks and S. Bryant, M. O’Connor, M. Peifer, D. Woods and P. J. Bryant for stimulating discussions and M. Peifer and N. Perrimon for communicating results prior to publication. We gratefully acknowledge the computational support of GenBank and the UCI office of academic computing.