A Screen for Identifying Genes Interacting With Armadillo, the Drosophila Homolog of β-Catenin
Sarah Greaves, Bénédicte Sanson, Phoebe White, Jean-Paul Vincent


Drosophila Armadillo is a multifunctional protein implicated in both cell adhesion, as a catenin, and cell signaling, as part of the Wingless signal transduction pathway. We have generated viable fly stocks with alterations in the level of Armadillo available for signaling. Flies from one stock overexpress Armadillo and, as a result, have increased vein material and bristles in the wings. Flies from the other stock have reduced cytoplasmic Armadillo following overexpression of the intracellular domain of DE-cadherin. These flies display a wing-notching phenotype typical of wingless mutations. Both misexpression phenotypes can be dominantly modified by removing one copy of genes known to encode members of the wingless pathway. Here we describe the identification of further mutations that dominantly modify the Armadillo misexpression phenotypes. These mutations are in genes encoding three different functions: establishment and maintenance of adherens junctions, cell cycle control, and Egfr signaling.

WINGLESS (WG) is the founding member of the WNT family of secreted glycoproteins, a class of molecules involved in signaling during development. Several components that link the WG signal from the membrane to the nucleus are now known. These include Dishevelled (DSH), a PDZ domain protein that has been shown to associate with the plasma membrane (Yanagawaet al. 1995), Shaggy/ZW-3 (SGG/ZW3), the homolog of the vertebrate serine/threonine kinase GSK-3β (Simpson 1990; Siegfriedet al. 1992), Armadillo (ARM), the homolog of β-catenin, and finally the nuclear factor Pangolin, also known as dTCF (Brunneret al. 1997; Rieseet al. 1997; van de Weteringet al. 1997). Importantly for this article, WG signaling represses the activity of SGG/ZW3 which itself negatively regulates cytoplasmic ARM levels (Peiferet al. 1994). In the current model, SGG/ZW3 (or GSK-3β in vertebrates) keeps ARM phosphorylated at specific sites and this targets it for degradation by the ubiquitin pathway, thus rendering ARM (β-catenin in vertebrates) unstable and maintaining low cytoplasmic levels (Aberleet al. 1997; Jiang and Struhl 1998; Theodosiouet al. 1998). Therefore, inhibition of SGG/ZW3 by WG signaling leads to increased ARM levels and to the formation of an ARM/Pangolin complex, which modulates the transcription of target genes.

Considering that ARM was first identified as a member of the WG pathway, it is intriguing that it turned out to be the homolog of β-catenin, an adherens junction protein (Peifer et al. 1991, 1992). ARM and β-catenin have a highly conserved core region of 13 42-amino-acid repeats, which interacts with the intracellular domain of the cell adhesion molecule E-cadherin in both Drosophila and vertebrates (Ozawaet al. 1990; Peifer 1993; Odaet al. 1994). The N-terminal domain is also conserved and is needed for binding to α-catenin (Odaet al. 1993; Hoschuetzkyet al. 1994). Since α-catenin binds to actin, ARM (or β-catenin) establishes a bridge between cadherins and the actin cytoskeleton, a bridge that is needed for proper cell adhesion (Kemler 1993; Coxet al. 1996). As expected from its homology with β-catenin, ARM has a clear role in cell adhesion (Coxet al. 1996; Godt and Tepass 1998; Whiteet al. 1998) and this role can be separated from its signaling function (Orsulic and Peifer 1996; Sansonet al. 1996).

β-Catenin itself has been shown to interact, directly or indirectly, with other proteins. Proteins shown to interact directly include the junction protein ZO-1 (Rajasekaranet al. 1996), the APC tumor suppressor gene product (Rubinfieldet al. 1993), the epidermal growth factor receptor (Egfr) (Hoschuetzkyet al. 1994), and the actin bundling protein fascin (Taoet al. 1996). Other proteins are found in a complex with β-catenin even though they do not interact directly. One example is p120 (Daniel and Reynolds 1995), a tyrosine kinase that is phosphorylated following Src-mediated cell transformation. All these interactions suggest that ARM could be at the crossroad of a variety of cell biological activities, although genetic evidence for this suggestion is still scant.

To understand better how the different functions of ARM are integrated, we used Drosophila genetics to identify mutations that modify ARM's activity. Using the Gal4/UAS system (Brand and Perrimon 1993), we have altered the levels of signaling ARM: high levels were obtained by direct overexpression of wild-type ARM (Whiteet al. 1998). Conversely, overexpression of the intracellular domain of DE-cadherin leads to insufficient accumulation of signaling ARM (Sansonet al. 1996). The resulting phenotypes were studied in the eyes and wings. We found that removing one copy of genes encoding known WG pathway members modifies these phenotypes. Following a dominant modifier screen other interactors were identified. Those can be ordered in three classes: mutants in genes encoding components of the egfr pathway, cell cycle regulators, and regulators/components of adherens junctions.


Armadillo misexpression: UAS-CADHintra and UAS-ARM transgenes have already been described elsewhere (Sansonet al. 1996; Whiteet al. 1998). Engrailed-GAL4 (Yoffeet al. 1995) is used to drive expression in the wing (see also Sansonet al. 1996; Whiteet al. 1998) and GMR-GAL4 is used for expression in the eye (Moses and Rubin 1991). The genotypes of the stocks producing dominant phenotypes are as follows: misexpression in the wing: UAS-CADHintra10, en-GAL4/CyO (ARMunder10); UAS-CADHintra5, en-GAL4/CyO (ARMunder5); UAS-ARM2, en-GAL4/CyO (ARMover2). misexpression in the eye: UAS-CADHintra4/5, GMR-GAL4/CyO (ARMunder4/5); GMR-GAL4/CyO; UAS-ARM18 (ARMover18).

Armadillo misexpression screen: Flies from ARMunder10 and ARMover2 stocks were initially crossed to wild-type Canton-S flies to characterize each dominant phenotype. WG pathway genes were then tested for their capacity to either suppress or enhance these phenotypes. The 166 stocks of the Bloomington deficiency kit were then crossed to both ARMover2 and ARMunder10, and progeny scored for any modification. Interactions with both were screened in parallel and modified wings were mounted. Deficiencies containing modifiers of either phenotype were crossed to wild type. Some modifiers of ARMunder10 were subsequently crossed to ARMunder5.

Map positions of modifying deficiencies were compared to known Drosophila genes. Nulls or strong alleles in potential interactors were obtained and crossed to the misexpression stocks.

The following mutant alleles were used:

Group 1, Wingless pathway members: armXK22 and armXM19, strong and weak alleles, respectively (Wieschauset al. 1984); wgCX4, amorph (Baker 1987); dshV26, amorph (Greeret al. 1983) and dshb117, null mutation (N. Perrimon, personal communication); zw3M11, amorph (J. Eeken, cited in Perrimon and Smouse 1989); nkd7E, presumed amorph (Jurgenset al. 1984).

Group 2, cell adhesion molecules: shgIG29, class IV allele (Nusslein-Volhardet al. 1984; see also Tepasset al. 1996); FtG-rv, amorph (Mohr 1923); ds55, amorph (Clarket al. 1995); crb8F105, amorph (Jurgenset al. 1984); sdtEH, amorph (Wieschauset al. 1984); dlgM52, amorph (R. Vodker, cited in Perrimon 1988).

Group 3, EGF pathway components: flbIP02, amorph (Nusslein-Volhardet al. 1984); aos006, amorph (Freemanet al. 1992); veM4, strong loss of function (Jurgenset al. 1984).

Group 4, cell cycle components: Dmcdc2B47, loss of function (Sternet al. 1993); Df(3L)pbl[x1], deficiency removing pbl (Hime and Saint 1992); stg7B, strong allele (Jurgenset al. 1984); cycA183, strong allele (Lehner and O'Farrell 1989)

Group 5, wing-specific interactors: enCX1, amorph (B. Holmgren cited in Heemskerket al. 1991); nubA6, strong allele (M. Ng, personal communication)

Group 6, noninteracting genes: CadNM19, amorph (Iwaiet al. 1997); kelDE1, null allele (Xue and Cooley 1993).

Group 7, interacting P elements: l(2)00632 (A. Spradling, cited in Perrimonet al. 1996); P1532, insertion in twins/abnormal anaphase resolution (A. Spradling from Berkeley Drosophila Genome Project); P1100, insertion in fasciclin3 (Bellenet al. 1989); P1736, insertion in l(3)00761 (A. Spradling from Berkeley Drosophila Genome Center, cited in Maixneret al. 1998); P368 (A. Spradling from Bloomington Drosophila Stock Center); P277 (A. Spradling from Berkeley Drosophila Genome Project).

Wing phenotype analysis: Wings were mounted in Euparal and incubated at 65° overnight.


Decreasing the signaling function of Armadillo in the wing: Overexpressing the intracellular domain of DE-cadherin (CADHintra) leads to a “loss-of-ARM” phenotype (Sansonet al. 1996). At high levels of uniform expression embryos die with a phenotype resembling that of wg or arm mutants, while restricted expression leads to defects in adults' appendages. For example, when engrailed-Gal4 (en-GAL4) is used to drive expression of a UAS-CADHintra transgene, flies show deletions of the wing margin as seen in hypomorphic wg mutants (Sansonet al. 1996; see also Figure 1A). Concomitant expression of wild-type ARM rescues these phenotypes, suggesting that CADHintra titrates cytoplasmic ARM. Moreover, a reduction in ARM activity (using armXK22 and armXM19 alleles in the heterozygous condition) strongly enhances these phenotypes to such an extent that pupal lethality frequently follows (Sansonet al. 1996).

Since CADHintra has a dominant negative effect on cell adhesion (Kinter 1992), the truncation of the wing margin in principle could be a consequence of a partial loss of adhesion. This does not appear to be the case since the wing truncation caused by CADHintra can be rescued by concomitant expression of an ARM protein incapable of supporting cell adhesion (Figure 1B). To do this experiment we used flies carrying an UAS-ARMΔα transgene (Whiteet al. 1998). ARMΔα has a deletion of the α-catenin binding site, which is absolutely required for ARM's adhesion function (Orsulic and Peifer 1996; Whiteet al. 1998). In a converse experiment, we also assayed the rescuing activity of human Plakoglobin, which has adhesion activity comparable to that of ARM but no signaling activity (Whiteet al. 1998). Human Plakoglobin does not rescue the CADHintra phenotype, showing that the margin loss does not follow from an adhesion defect (Figure 1C). These experiments confirm that overexpression of CADHintra leads to a loss of signaling ARM and we will describe the ensuing phenotype as an ARM underexpression phenotype (ARMunder).

The strength of the ARMunder phenotype depends on the individual UAS-CADHintra transgene used and therefore a phenotypic series can be obtained. We selected two UAS-CADHintra stocks that lead to phenotypes of distinctly different strengths in combination with engrailed-GAL4; UAS-CADHintra10 shows a weak phenotype, while UAS-CADHintra5 shows an intermediate phenotype (Figure 2, compare A with D and B). These two UAS-CADHintra transgenes were recombined with en-GAL4 to generate stable stocks, which we call ARMunder10 and ARMunder5. Although the assessment of phenotype strength can be subjective, two criteria allow unambiguous distinction between the phenotypes of ARMunder10 and ARMunder5. While ARMunder10 flies always retain the posterior crossvein and a tuft of marginal bristles between the tips of veins IV and V, these structures are consistently absent in ARMunder5 flies. These criteria will become relevant when we discuss the suppression of the ARMunder phenotype by various mutants later in the article.

Figure 1.

CADHintra overexpression phenotypes in the wing result from blocking ARM signaling and not adhesion. (A) CADHintra overexpression with en-GAL4 leads to posterior wing notching, mimicking Wingless hypomorphs. (B) Concomitant overexpression of ArmadilloΔα, a construct that lacks the ability to bind α-catenin but is still competent for signaling, rescues this notching phenotype to wild type. The rescue is similar to the rescue by wild-type ARM (Sansonet al. 1996). (C) Concomitant overexpression of vertebrate Plakoglobin, an ARM homolog that has no signaling activity, does not rescue the notching phenotype.

Mutations in WG pathway genes were tested for their capacity to either suppress or enhance the ARMunder10 and ARMunder5 phenotype. wgCX4 in the heterozygous condition enhances the ARMunder10 phenotype, although enhancement is not as strong as by armXK22 or armXM19 (Figure 2E). Enhancement by wgCX4 is classified as weak and scored as “+” (Table 1). dsh alleles were also tested; dshb117 weakly enhances ARMunder10 (data not shown), as seen with wgCX4. Conversely, the zw3M11 mutation suppresses ARMunder10 to a near wild-type wing (Figure 2F). This suppression is strong and is scored as “−−” (Table 1). zw3M11 also suppresses the stronger ARMunder5 phenotype (Figure 2C). The signs of the interactions of ARMunder with wgCX4 and zw3M11 are as predicted from the known roles of these gene products in the WG pathway.

Increasing Armadillo levels in the wing: Overactivation of the WG pathway in the prospective wing blade induces the formation of ectopic bristles: bristles normally found at the wing margin (where the WG signaling pathway is most active in third instar discs) form within the blade. This is true for sensory bristles in the anterior compartment (anterior triple row bristles) as well as for uninnervated bristles in the posterior compartment. This phenotype has been observed when overexpressing WG (Zeccaet al. 1996), ARM (Whiteet al. 1998), or DSH (Axelrodet al. 1996), and also in sgg/zw3 mutant clones (Simpson 1990).

In this study we exploited the phenotype obtained when ARM is overexpressed. When a weak UAS-ARM insertion (UAS-ARM2) is driven by en-GAL4, adult flies survive and their wings have ectopic bristles within the posterior wing blade (Whiteet al. 1998; see also Figure 3A). These ectopic bristles tend to be found near the margin as was shown previously when DSH is overexpressed (Axelrodet al. 1996). Since WG is produced at the margin, it is conceivable that the cells producing ectopic bristles do so in response to WG and that increased levels of ARM make these cells more responsive. In addition to inducing ectopic bristles, ARM overexpression also leads to disturbed venation (Figure 3A). We have not attempted to study the basis of the vein phenotype, although an effect of sgg/zw3 loss of function on veins has been documented (Ripollet al. 1988). The weak UAS-ARM2 line was recombined with en-GAL4 to obtain a stock displaying the “ectopic bristles” phenotype constitutively. Since this construct overexpresses ARM and generates a phenotype reminiscent of WG pathway overactivation, we called this stock “ARMover.” As with ARMunder, the ARMover phenotype is dominantly modified by removing one copy of the WG pathway members but in the opposite direction. We found that sgg/zw3 enhances ARMover (Figure 3C), leading to more ectopic bristles. Interactions with ARMover could be quantified by counting ectopic bristles. ARMover2 wings have a mean average of 9 ± 3 (n = 100) ectopic bristles in the wing blade. Enhancement by zw3M11/+ leads to wings with an average of 29 ± 5 (n = 100) ectopic bristles in the blade and was considered a strong interactor (labeled “++” in Table 1). Interactions leading to 15 ± 3 ectopic bristles were considered weaker and were labeled “+” (see below). In contrast, wg, dsh, and arm mutations dominantly suppress the formation of both ectopic bristles and vein material. arm strongly suppresses ARMover with suppression scored as “−−” (85% of modified wings have no ectopic bristles remaining). Suppression by wg and dsh is not as strong as arm (average bristle number is 4 ± 2, n = 50) and is scored as “−” (Figure 3B and not shown). Intriguingly, in the suppressed ARMover wings, not only are the ectopic bristles suppressed, but endogenous bristles are lost from the margin (see Figure 3B).

Figure 2.

CADHintra underexpression wing phenotypes are modified by a 50% reduction in the dosage of genes encoding known members of the Wingless pathway. (A) A wild-type Drosophila wing consists of five longitudinal veins (I–V) and two cross-veins. (B) Extensive loss of wing material is seen when a strong CAD-Hintra transgene is overepressed with en-GAL4 (UAS-CADHintra5, en-GAL4; ARMun-der5). Note that margin bristles between veins IV and V are completely absent. (C) The ARMunder5 phenotype is partially suppressed in sgg/zw3 heterozygotes. In particular, some margin bristles between veins IV and V are restored (arrowhead). (D) Lowering ARM levels with a weaker CADHintra insert (UAS-CADHintra10, en-GAL4; ARMunder10) induces subtler wing notching. Some margin bristles are lost between veins LIV–LV, as indicated by the arrowhead. (E) Removing one copy of wingless (wgCX4) enhances ARMunder10-induced notches and loss of margin bristles. This enhancement is classified as “+”. (F) Removing one copy of sgg/zw3 (zw3M11) suppresses the notches of ARMunder10 to a near wild-type wing. This suppression is classified as “−−”.

Deficiency screen to identify enhancers and suppressors of ARMunder and ARMover: Having validated the use of these stocks to detect mutations that alter the level of ARM (and perhaps the ability to transduce the WG signal), we decided to search systematically for dominant modifiers of ARMover and ARMunder phenotypes. In the first instance, we screened only for interactions with one ARMover stock (UAS-ARM2, en-GAL4) and one ARMunder stock (UAS-CADHintra10, en-GAL4). We chose stocks giving the weakest phenotypes to ensure that all possible interactions were initially detected. The screen was conducted using the Bloomington Stock Center deficiency kit. This kit comprises 166 fly stocks, each deficient for regions on chromosomes X, II, or III and collectively uncovering 70% of the genome. The kit does not include deficiencies on the fourth chromosome. Each deficiency stock was crossed blind to both ARMover and ARMunder. The progeny were scored for viability and suppression or enhancement of the wing phenotypes.

View this table:

Interaction of known Drosophila genes with ARMover and ARMunder

Deficiencies that appeared to interact were then crossed to wild type to assess whether they caused a dominant wing phenotype. Two classes of nonspecific interactions were observed.

  1. Wing-notching phenotype. Several deficiencies that appeared initially to enhance the ARMunder-notching phenotype were found to also cause wing notching when crossed to ARMover. This was the case for 16 deficiencies and the effect could be attributed to a mutation in one of the following genes: vestigal, scalloped, Notch, or deltex. Mutations in vestigal or Notch also lead to wing notches when crossed to wild type.

    Figure 3.

    ARM overexpression wing phenotype is modified by a 50% reduction in genes encoding known members of the WG pathway. (A) Overexpression of cytoplasmic ARM (UAS-ARM2, en-GAL4) induces excess veins and bristles (two are indicated by arrowheads) within the posterior of the wing blade. The extra veins concentrate around the posterior crossvein. This particular combination causes a mean average of nine ectopic bristles per wing. This weak phenotype allows ready detection of enhancement. (B) Removing one copy of wg (wgCX4) suppresses the vein abnormalities in the posterior of the wing and the number of extra bristles. Interestingly, margin bristles are now lost between veins LIV–LV (as shown by arrowhead). (C) Removing one copy of sgg/zw3 (zw3M11) enhances ARMover, increasing the number of extra bristles present within the wing blade (two are highlighted with arrowheads). Unexpectedly, vein abnormalities are suppressed. The suppression by wg is recorded as “−−”, and the enhancement by zw3 as “++”. As the cellular reason for the venation phenotype is not understood, its suppression by both wgCX4 and zw3M11 was not investigated further.

  2. Vein disruption and wing-blistering phenotypes. Four deficiencies lead to blistering and/or vein disruption in the heterozygous condition. These phenotypes interfered with our ability to count ectopic bristles and we decided to exclude them from our analysis. These deficiencies uncover Plexate, Blistered, and wingblister, and mutations in all three genes cause similar phenotypes when in trans with ARMover. Mutations in either Plexate or Blistered also generate wing blisters when crossed to wild type.

The deficiencies described above were eliminated from further analysis. Other interacting deficiencies were discarded for the following two reasons:

  1. Inadequacy of the ARMover vein phenotype. A total of eight deficiencies appeared to modify dominantly the vein phenotype of ARMover without affecting the number of ectopic bristles. Again, this effect could be seen with individual mutations uncovered by these deficiencies. For example, mutations in net, thickvein, saxophone, and elbow all enhanced the vein phenotype of ARMover without affecting bristle number. Because the bristle phenotype of excess wg signaling is well documented (e.g., Axelrodet al. 1996), we decided to keep only deficiencies that modify this aspect of the phenotype.

  2. Conflicting interactions. We expected the interactions of any given deficiency with ARMover to be opposite to that seen with ARMunder. When interactions with both stocks were detected, this expectation was largely borne out. However, four deficiencies suppressed the phenotype of both ARMover and ARMunder. One possibility is that these deficiencies uncover genes encoding factors necessary for transcriptional activation by GAL4 and that halving gene dosage reduces the efficiency of the GAL4 system, leading to the suppression of both misexpression phenotypes. These deficiencies map in two chromosomal locations: 8B-9C and 42E-44C.

After eliminating the nonspecific interactors (32 deficiencies), we were left with a total of 59 interacting deficiencies (out of a starting total of 166). Among those, 58 modified the ectopic bristle phenotype of ARMover and 44 modified the notching phenotype of ARMunder. A majority (33) enhanced ARMover and suppressed ARMunder (like zw3M11) while only 4 showed the converse (enhancement of ARMunder and suppression of ARMover, like dshb117). We think this difference is inherent to the nature of the starting phenotypes. Suppression of the wing notches caused by UAS-CADHintra is easier to score than enhancement. Enhancers are detectable (wgCX4 is an example) but not as readily. The same is true for the ARMover phenotype: an increase in bristle number is easier to detect than a decrease, especially when the starting number of extra bristles in the wing blade is already low. Some deficiencies interact significantly with only one stock: 15 enhance ARMover without modifying ARMunder and 7 only suppress ARMunder.

As expected, a whole range of interaction strengths was found with the modifying deficiencies. The interaction with ARMover was relatively easy to quantify since the number of ectopic bristles can be counted (see above). In the case of insufficient ARM activity, strong suppressors (“−−”) were also easily scored since they suppressed ARMunder to a near wild-type phenotype. This is exemplified by zw3M11 (see Figure 2F). In contrast, weak suppressors were sometimes ambiguous and the interaction was therefore verified with the stronger ARMunder5 stock. Recall that flies from this stock always lack the posterior cross-vein and the tuft of bristles between the tips of veins IV and V. We considered a weak interaction (“−”) significant if it yielded the restoration of these two structures. Few deficiencies enhanced ARMunder and when they did so, enhancement was weak. Enhancement of ARMunder10 was considered significant if margin bristles between veins IV and V, which are intact in the ARMunder10 stock, were deleted in the presence of the deficiency.

To identify the gene responsible for the interactions identified with individual or overlapping deficiencies, nulls or strong mutants of candidate genes from each area were obtained. Interactions with ARMover and ARMunder were tested for all these alleles. This approach identified 19 interacting mutants (corresponding to 31 deficiencies from the kit). Among those were engrailed and nubbin. Although these genes are related to wg function, they are not thought to modulate ARM's activity. It is likely that interaction with these genes is indirect and follows from their role in wing patterning. For example, nubbin is thought to function exclusively in the wing where it downregulates wg expression (Neumann and Cohen 1998). Decreasing nubbin dosage might lead to additional wg expression and thus explain the genetic interactions we detected. To exclude these potential wing-specific interactions, we decided to establish a secondary assay in the eye.

Interactions with ARM loss- and gain-of-function phenotypes in the eye: To generate phenotypes in the eye, we used the GMR-GAL4 driver (Moses and Rubin 1991) in combination with UAS lines, which we knew from previous work had a strong activity: UAS-CADHintra4/5 for ARMunder and UAS-ARM18 for ARMover (Sansonet al. 1996; P. White, unpublished observations). A wild-type eye consists of regularly spaced ommatidia separated by inter-ommatidial bristles (Figure 4A). Overexpression of ARM (GMR-GAL4; UAS-ARM18) leads to rough eyes and the loss of interommatidial bristles (Figure 4B). A similar bristle loss is seen with WG overexpression (Cadigan and Nusse 1996). As shown by Cadigan and Nusse (1996), ectopic WG suppresses the formation of interommatidial bristles by repressing achaete expression in the eye. These authors also showed that repression requires the classical WG pathway (including ARM). Intriguingly, WG has the opposite effect in the wing since it positively regulates achaete there and acts to promote bristle formation (Cousoet al. 1994). Underexpression of ARM (UAS-ARM4/5, GMR-GAL4) also leads to rough eyes. In this case extra interommatidial bristles are generated (Figure 4C), and two bristles can often be seen originating from a single socket.

By analogy with the wing assay, we recombined GMR-GAL4 with UAS-CADHintra4/5 and crossed the balanced stock with various mutants. The eyes of flies lacking both balancers were examined (using those carrying the balancer from the mutant stock as a control). To assay interactions with excess ARM, we used the UAS-ARM18 transgene, which is on the third chromosome and our test stock comprised flies of the following genotype: GMR-GAL4/Cyo; UAS-ARM18. These flies were crossed to various mutants, and once again the eye phenotype was assessed in flies lacking balancer chromosomes.

We found both eye phenotypes to be sensitive to a reduction in the dosage of known WG pathway members. Although WG is only expressed along the anterior lateral margins of the eye field (Treisman and Rubin 1995), interactions are detected throughout the eye: wgCX4 suppresses ARMover18 to wild type and enhances ARMunder4/5 (not shown). dshb117 also suppresses ARMover18 to wild type (Figure 4D) while zw3M11 suppresses ARMunder4/5 to wild type (Figure 4E). Conversely, dshb117 enhances ARMunder4/5 and zw3M11 enhances ARMover18 (data not shown). Interactions of all genes are as expected from the known roles of their products in the WG pathway and are the same as those seen with the ARM misexpression phenotypes in the wing. However, neither engrailed nor nubbin showed an interaction with the ARM misexpression phenotypes in the eye. This confirmed our suspicion that the original interactions with these mutants were wing specific. They were therefore discarded from subsequent analysis.

Classifying the genes within interacting chromosomal regions: Following our various criteria, we were left with mutations in 17 genes (26 deficiencies) that interacted with ARMover and/or ARMunder in a similar direction, as did the original deficiency. Interaction strength varied from deficiency to point mutation, suggesting that several genes in the original deficiencies could have contributed to, or modified, the interaction. Only for 7 of the 17 genes were interactions identical between the point mutation and the corresponding starting deficiency. The 17 genes were sorted into four groups described below (see also Table 1).

Group 1, wingless pathway genes: Four known members of the pathway were identified: wg, dsh, zw3, and nkd. All interacted in the direction expected (wg, dsh, and sgg/zw3 had already been tested prior to the screen). naked (nkd) was also identified as a suppressor of ARMunder and an enhancer of ARMover; however, these interactions are much weaker than those seen for zw3M11 (Table 1).

Figure 4.

ARM misexpression eye phenotypes are modified by a 50% reduction in genes encoding known WG pathway members. (A) A wild-type eye consists of regularly spaced ommatidia with a patterned array of interommatidial bristles. (B) Overexpression of cytoplasmic ARM with the GMR driver (GMR-GAL4; UAS-ARM18; ARMover18) leads to eyes that are rough and contain empty bristle sockets (arrowheads in inset). (C) Underexpression of ARM (UAS-CADH intra4/5, GMR-GAL4; ARMunder4/5) also creates eyes that are rough but have extra bristles. Often two bristles can be seen coming from the same socket (arrowhead in inset). (D) Removing one copy of dishevelled (dshb117) suppresses the ARMover18 rough eye to wild type, restoring the regular ommatidial pattern. This suppression is scored as “−−”. (E) Removing one copy of sgg/zw3 (zw3M11) suppresses the ARMunder4/5 eye phenotype to wild type. This suppression is also scored as “−−”. Interactions by both are consistent with those seen in the wing.

Group 2, genes required for cell adhesion: This group includes shotgun (which encodes DE-cadherin) as expected. Also uncovered were fat (ft) and dachsous (ds). These two genes encode nonclassical cadherin characterized by a huge extracellular domain containing up to 35 cadherin repeats and a bipartite ARM binding site (Clarket al. 1995). Interactions with these two mutants are similar to those observed with shotgun (DE-cadherin), the only difference being that ft interacts more weakly than shg with ARMover. In addition to genes encoding cadherins (classical and nonclassical), we observed interactions with some of the genes known to be essential (directly or indirectly) for the assembly or maintenance of adherens junctions—stardust (sdt), discs-large (dlg), and crumbs (crb). These interacted in the same direction as shg; however, the suppression of ARMunder was always weaker and only dlgM52 enhanced ARMover to the same extent as zw3M11 (Table 1).

Group 3, EGF pathway genes: Interactions were observed with some, but not all, members of the EGF pathway. Those identified were Egfr, veinlet/rhomboid (ve/rho), and argos (aos). All enhance ARMover, increasing the number of ectopic bristles in the wing blade with veM4 being the strongest interactor (Table 1). None showed an interaction with the ARMunder stocks.

Group 4, cell cycle genes: Four cell cycle genes, cdc2, string/cdc25, pebble (pbl), and cyclinA, were uncovered by the screen. string enhances ARMover and weakly suppresses ARMunder (Table 1). The other three, cdc2, pbl, and cycA, all modify ARMover. These modifications are weak and variable and more work is needed to assess their significance.

The candidate gene approach did not always yield an interaction. For example, two obvious candidates uncovered by an interacting deficiency on chromosome II, Kelch and cadherin-N, did not modify either ARM misexpression phenotype (the deficiency modifies both). The lack of interaction with cadherin-N is particularly surprising considering the high degree of homology of its intracellular domain with that of E-cadherin (Iwaiet al. 1997 and see discussion).

Figure 5.

l(2)00632 modifies the ARM misexpression phenotypes in wings. (A) ARMover wing showing ectopic bristles in the blade (UAS-ARM2 en-GAL4; see Figure 3A). (B) Wing from a fly carrying the same transgenes as in A and also heterozygous for l(2)00-632. When one copy of l(2)00-632 is mutated, the number of ectopic bristles increases. (C) Wing from an ARMunder fly. Phenotype is characterized by the loss of blade and margin material (UAS-CADHintra10 en-GAL4; see Figure 2D). (D) Wing from a fly carrying the same transgenes as in C and also heterozygous for l(2)00632. Margin and blade tissue is restored.

Ultimately, there remained 28 deficiencies without a mutant that could account for the interaction (this includes the Kelch/cadherin-N deficiency). These map to 10 overlapping chromosomal regions. Testing smaller deficiencies further refined the chromosomal location of each interactor. Once a small interacting region was identified, P-element-induced mutations were obtained and tested for an interaction with ARMover and ARMunder. This approach identified six P-element-induced mutations that modify the ARM wing phenotypes in a manner similar to the original deficiencies. All enhance ARMover and suppress ARMunder, both in the eye and the wing. An example of one such P-element-induced mutant is l(2)00632. Enhancement of ARMover and suppression of ARMunder in the wing are shown in Figure 5. Qualitatively similar interactions are seen in the eye (Figure 6). Note that these interactions are similar to those with sgg/zw3, which encodes a negative regulator of Armadillo levels. Thus the wild-type product of l(2)00632 contributes negatively to Wingless signaling, possibly by lowering the level of signaling by Armadillo. So far, we know that another interacting P insertion mutates fasciclin3 and yet another disrupts twins (Gomeset al. 1993; Uemuraet al. 1993). The remaining four are in genes that are as yet not cloned. Summary information on these P mutations and corresponding interactions are listed in Table 2.


In this article, we describe two types of sensitized Drosophila stocks that can be used to detect subtle changes in ARM levels and, more generally, signaling by WG. By design, our sensitized flies interact in opposite ways to any given change in WG signaling. Flies of one stock overexpress ARM (ARMover) and, as a result, have ectopic bristles in the wing blade, a well-characterized consequence of overactivating the WG pathway. Flies of the other stock have depressed levels of cytoplasmic ARM (ARMunder) because of overexpressed CADHintra. Again, the resulting phenotype, loss of the wing margin and blade material, is a well-studied effect of partial loss of WG function. We confirmed that these stocks are sensitive to the dosage of genes involved in WG signaling. For example, in zw3M11/+ flies, the ARM-over phenotype is enhanced (more ectopic bristles) while the ARMunder phenotype is suppressed (wing margin restored). As a general rule, it can be difficult to assess the significance of a dominant interaction and we have found the requirement for opposite interactions with ARMover and ARMunder to be one useful criterion for specificity. However, this was not the sole criterion. Some mutants interact solely with either ARMunder or ARMover, and if this interaction was strong, we considered it to be significant.

We also generated flies over- or underexpressing ARM in the eye. Both have phenotypes that can be modified by mutations in members of the WG pathway (wg, dsh, and zw3) in the same manner as in the wing. This provided us with an additional criterion to identify specific interactors.

Both loss- and gain-of-function experiments have implicated the WG pathway (though not necessarily WG) in ommatidial polarity (Treisman and Rubin 1995; Cadigan and Nusse 1996; Tomlinsonet al. 1997; Herberlein et al. 1998). An effect on interommatidial bristle formation has also been shown: overexpressed wg leads to suppression of interommatidial bristle development. Such widespread effects of WG throughout the eye field are surprising, considering its restricted domain of expression (along the anterior eye margin). However, our results indicate that all eye cells are indeed sensitive to changes in WG dosage and that they are therefore probably all under the influence of WG during normal development.

Figure 6.

l(2)00632 modifies the ARM misexpression phenotypes in the eye. (A) Armadillo overexpression phenotype in the eye (GMR-GAL4 UAS-ARM18; as shown in Figure 4B). (B) Eye from a fly carrying GMR-GAL4 UAS-ARM18 and also heterozygous for l(2)00632. The rough eye phenotype is enhanced. (C) Eye from a fly carrying UAS-CADHintra4/5 GMR-GAL4 (Armadillo underexpression as shown in Figure 4C). (D) Eye of a fly carrying the same transgenes as in C and also heterozygous for l(2)00632. The rough eye phenotype is completely suppressed.

Screening with deficiencies had the advantage that, in a relatively short time, we could scan a large proportion of the genome. However, the drawback of using large deficiencies is that conflicting genetic interactions with more than one locus might arise. Indeed, we found that overlapping deficiencies and mutants from a given chromosomal location rarely interacted to the same extent. Nevertheless, in many cases, we were able to narrow down an interacting deficiency to a single mutant that itself interacted significantly. Our screen was by no means exhaustive. First, the deficiency kit uncovers only 70% of the genome. And second, we cannot exclude that opposite interactors present in a given deficiency could mask each other. A pilot screen for EMS-induced dominant modifiers suggests that more interactors are to be discovered. It should also be pointed out that our stocks provide a useful assay to test genetic interaction with WG signaling. For example, CREB binding protein (CBP; encoded by the gene nej) was initially suggested from biochemical experiments to regulate WG signaling and the ARMunder stock was used to add genetic evidence (Waltzer and Bienz 1998).

View this table:

Modification of six P elements with ARMover and ARMunder wing and eye phenotypes

Among the interactors identified, we found naked (nkd), a mutant that has long been associated with excess WG activity (for a review, see Perrimon 1994). The embryonic phenotype of nkd mutants is characterized by an excess of naked cuticle, just like that of sgg/zw3 mutants or embryos overexpressing WG. In the case of sgg/zw3, this phenotype clearly follows from overactivation of the pathway—irrespective of whether endogenous wg is present or not. In contrast, wg/nkd double mutants resemble the wg single mutant (Bejsovec and Wieschaus 1993), suggesting that nkd is upstream of wg. More precisely, since nkd mutants have enlarged stripes of Engrailed [and concomitant HH (Hh)] expression, nkd has been proposed to be a negative regulator of Engrailed expression. Broader hh expression in nkd embryos (as a result of widened engrailed expression) is thought to induce ectopic stripes of wg expression and this would cause the naked cuticle phenotype (Dougan and DiNardo 1992). However, in wing imaginal discs, wg expression is not controlled by engrailed or hh (Cohen 1993) and therefore our finding that nkd modifies the ARMover and ARMunder phenotypes in the wing implies a more widespread role of nkd in WG signaling. Maybe absence of nkd function renders cells more responsive to WG. This would explain why endogenous WG is required for the nkd phenotype to arise. It would also be consistent with the genetic interactions we detected in the wing. Note that, so far, no function has been ascribed to nkd in disc development.

Our screen identified several genes involved in the assembly or maintenance of adherens junctions. shotgun (encoding DE-cadherin) itself is not very illuminating since it is expected that the phenotype caused by excess intracellular domain of cadherin will be suppressed by decreasing endogenous cadherin levels. Still, this interaction shows that the level of overexpression afforded by the Gal4p system is within physiological levels. Interaction with fat and dachsous suggests that these two nonclassical cadherins interact (maybe directly) with ARM. Initial analysis of the intracellular domain of Fat and Dachsous failed to identify an ARM/β-catenin binding site homologous to that found in E-cadherin (Mahoneyet al. 1991). However, subsequent sequence examination suggested the existence of a bipartite site (Clarket al. 1995). Genetic interactions with fat and dachsous suggest strongly that this proposed site is functional, and thus removing one copy of the fat or dachsous gene would release additional ARM to the cytoplasm and make it available for use in WG transduction. Interactions with fat and dachsous in the eye confirm the ability of these genes to modify cytoplasmic ARM levels. It also indicates that these genes are expressed in the eye and may be functional there.

Cadherin-N (CadN) binds to ARM (Iwaiet al. 1997). Therefore the failure of CadN to interact in our screen suggests that CadN may not be expressed in significant levels in the posterior compartment of wing imaginal discs or in eye precursors. In contrast, crumbs (crb) and stardust (sdt) do interact. The proteins encoded by these genes are not thought to participate in junctional complexes per se. Rather, they control the biogenesis of the junctions (Graweet al. 1995). We suggest that decreasing the activity of crb or sdt has a quantitative effect on the number or size of adherens junctions and this would lead to more ARM being released from the membrane and made available for WG signaling.

The interaction with discs-large (dlg) is somewhat surprising since DLG is presumed to act in septate junctions. DLG localizes at septate junctions once adherens junction contacts are established (Woods et al. 1996, 1997). However, the vertebrate homolog of DLG, ZO-1, has been shown to interact with β-catenin (Rajasekaranet al. 1996). The genetic interaction of dlg with ARMover and ARMunder implies either that DLG binds ARM in vivo, or that altering DLG levels affects septate junctions, which in turn are needed for the stability of adherens junctions. In support for the latter alternative the dlgM52 mutation leads to disruption not only of septate but also adherens junctions (Woodset al. 1997). In fact, in dlgM52 mutants ARM no longer localizes to the membrane (Woodset al. 1997) and this could lead to increased ARM in the cytoplasm.

We initially dismissed the interactions with genes encoding components of the EGF pathway because argos and Egfr, which have opposite effects on the EGF pathway, interacted similarly with ARMunder and ARMover. However, recent work has demonstrated an antagonism between the WG and EGF pathways in the embryonic epidermis (O'Keefeet al. 1997; Szutset al. 1997). This antagonism is probably not universal since another embryonic function of WG, the maintenance of Engrailed expression, is not affected by EGF signaling (Schejter and Shilo 1989). However, the interactions that we uncovered in the wing suggest that the EGF-WG antagonism may not be limited to cuticle patterning. It is noteworthy that, in the wing, we only saw an interaction with ARMover (which involves ectopic bristles). It may thus be that WG and EGF only compete at places where specialized cuticular structure are formed, although, while denticles are negatively regulated by WG, bristles are made in response to WG signaling. Also, we have no explanation as to why argos, a negative regulator of EGF signaling should interact in the same manner as Egfr.

Genes encoding cell cycle components also interact with our ARM misexpression lines and this was initially unexpected by us. However, recent work by others has established a link between WG and the cell cycle (Johnston and Edgar 1998). It was shown that, at the wing margin of the anterior compartment, wg suppresses string (stg) transcription via induction of the proneural genes achete (ac) and scute (sc). This is followed by G2 arrest and the formation of specialized sensory bristles. In the posterior compartment no such G2 arrest takes place, and therefore there is no simple explanation as to why stg should enhance the number of ectopic noninnervated bristles induced by excess wg signaling. This and the finding that stg suppresses the ARMunder phenotype implies that somehow loss of stg activity potentiates WG signaling. We are currently investigating this further.

Work in both fruit flies and vertebrates has hinted at the central role played by ARM/β-catenin in many cellular functions, most notably, WG signaling, cell adhesion, and, more recently, EGF signaling and the cell cycle. We expect that the characterization of the P-induced mutations identified during the screen described here will broaden our perspective on ARM function. We are encouraged by the fact that one such P has been inserted in a gene that has previously been implicated in cell adhesion (fasciclin3; Snowet al. 1989). The interaction with twins (tws), a regulatory subunit of protein phosphatase 2A (PP2A), is also promising. A vertebrate regulatory subunit of PP2A has recently been shown to regulate β-catenin activity (Seelinget al. 1999). It will be interesting, therefore, to determine if ARM, APC, or Axin are substrates of PP2A. We hope that further experiments with tws and fas3, as well as the characterization of the remaining four mutants, will help us understand how diverse functions like cell adhesion and cell cycle and signaling might be integrated by the usage of one common component, ARM.


We are grateful to all those who contributed fly stocks, especially to the Bloomington Stock Center. Thanks also to Tanita Casci and Peter Thomason for comments on the manuscript and Steven Marygold for discussions. S.G. and P.W. were supported by a Medical Research Council Studentship. B.S. was supported by a Marie Curie European Union Fellowship.


  • Communicating editor: K. Anderson

  • Received May 7, 1999.
  • Accepted August 10, 1999.


View Abstract