A Screen for Mutations That Suppress the Phenotype of Drosophila armadillo, the ß-Catenin Homolog

Corresponding author: Mark Peifer, Biology, CB#3280, University of North Carolina, Chapel Hill, NC 27599-3280., peifer{at}unc.edu (E-mail)

Communicating editor: K. ANDERSON

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX A APPENDIX B LITERATURE CITED

During development signaling pathways coordinate cell fates and regulate the choice between cell survival or programmed cell death. The well-conserved Wingless/Wnt pathway is required for many developmental decisions in all animals. One transducer of the Wingless/Wnt signal is Armadillo/ß-catenin. Drosophila Armadillo not only transduces Wingless signal, but also acts in cell-cell adhesion via its role in the epithelial adherens junction. While many components of both the Wingless/Wnt signaling pathway and adherens junctions are known, both processes are complex, suggesting that unknown components influence signaling and junctions. We carried out a genetic modifier screen to identify some of these components by screening for mutations that can suppress the armadillo mutant phenotype. We identified 12 regions of the genome that have this property. From these regions and from additional candidate genes tested we identified four genes that suppress arm: dTCF, puckered, head involution defective (hid), and Dpresenilin. We further investigated the interaction with hid, a known regulator of programmed cell death. Our data suggest that Wg signaling modulates Hid activity and that Hid regulates programmed cell death in a dose-sensitive fashion.

THE development of a fertilized egg into a multicellular organism requires coordination of many processes. Each cell must choose the proper cell fate and must also assume its place as part of an organized tissue. In addition, apoptosis (programmed cell death; PCD) plays an important role in shaping an organism by eliminating unneeded cells. One conserved pathway that directs cell fate decisions in many animals is the Wingless (Wg)/Wnt signal transduction pathway (proteins listed as X/Y represent nomenclature in Drosophila/mammals). Loss-of-function mutations in this pathway are lethal, while inappropriate activation can be oncogenic. Wg/Wnt signals are transduced by homologous components in Drosophila, Xenopus, and mammals (reviewed in POLAKIS 1999 ). During normal development, most cells do not receive Wg/Wnt signals. In these cells the pathway is kept off through the actions of several proteins, including Zestewhite3/GSK3ß, the tumor-suppressor adenomatous polyposis coli, and axin, which work in conjunction to target Armadillo (Arm)/ß-catenin (ßcat) for degradation via the proteasome. Arm/ßcat is thus the pivotal component in the pathway. When Wg/Wnt is absent, cytoplasmic levels of Arm/ßcat are very low. However, Wg/Wnt signal relieves the destruction of Arm/ßcat. Arm/ßcat accumulates, translocates into the nucleus, and binds dTCF/TCF, forming a bipartite transcription factor that turns on Wg/Wnt-responsive genes.

The components of the Wg pathway are encoded by a subset of the segment polarity genes, mutations that affect cell fate in the embryonic epidermis. In normal fly embryos, anterior cells of each segment secrete denticles, while posterior cells secrete naked cuticle. Wg signal directs cells to choose posterior fates and thus secrete naked cuticle. In an embryo mutant for wg or other positively acting components of the Wg pathway, cell fates are altered such that all surviving cells secrete denticles. It is important to note, however, that in a wg mutant many epidermal cells fail to survive to secrete cuticle, instead undergoing PCD. Embryos mutant for genes in either the Wg or the Hedgehog pathways have elevated levels of epidermal PCD (MARTINEZ ARIAS 1985 ; KLINGENSMITH et al. 1989 ; PAZDERA et al. 1998 ).

Arm's role in Wg signaling is not its only function. The earliest requirement for Arm is in cell adhesion (COX et al. 1996 ). Arm/ßcat is an essential component of epithelial cell-cell adherens junctions (reviewed in PROVOST and RIMM 1999 ). The core components of this junction are classic cadherins, transmembrane proteins that mediate homotypic adhesion between neighboring cells. Arm/ßcat binds to the cadherin cytoplasmic tail. -Catenin then binds to Arm/ßcat, linking the actin cytoskeleton to adherens junctions. In Drosophila, Arm helps assemble adherens junctions very early during embryogenesis. This is initiated by maternal Arm, which is supplemented by zygotic Arm once transcription begins. If the embryo lacks maternal and zygotic Arm, it does not form proper adherens junctions, and cells of the cellularized blastoderm cannot form epithelia (COX et al. 1996 ). In addition to the essential role that Arm/ßcat and adherens junctions play in embryogenesis, loss-of-function mutations in the cadherin-catenin system contribute to tumorigenicity, as tumor cells must alter their adhesive properties to metastasize.

While the roles of Arm/ßcat in Wg/Wnt signaling and adherens junctions have become clearer, many questions remain concerning both processes. In addition, biochemical approaches identified many other proteins that bind ßcat, perhaps implicating it in other functions: for example, Arm/ßcat binds the epidermal growth factor (EGF) receptor at the cell surface (HOSCHUETZKY et al. 1994 ), the actin-binding protein fascin in the cortex (TAO et al. 1996 ), Presenilin proteins, presumably in the endoplasmic reticulum (ER) (ZHOU et al. 1997 ; YU et al. 1998 ), and the transcription factor Teashirt (GALLET et al. 1998 ). One strategy to identify novel proteins involved in cell adhesion and Wg signaling and simultaneously to search for biological functions of the interaction of Arm with other partners is to look for mutations that interact genetically with arm.

In designing such a genetic screen, we took advantage of Arm's dual roles in signaling and adhesion. It has been suggested that cells may use this coupling, allowing one process to regulate the other via competition for a limited pool of Arm. Although in wild-type Drosophila embryos more than enough Arm is synthesized to fulfill its roles in both signaling and adhesion, one can manipulate the pool of Arm to make signaling and adhesion competitive. For example, if one expresses excess cadherin, it titrates out all the Arm, leaving none available for Wg signaling and resulting in a segment polarity phenotype (SANSON et al. 1996 ). We utilized this balance between Arm assembled into adherens junctions and that remaining for Wg signaling to create a sensitized genetic background. We reduced the amount of available Arm until adhesion and Wg signaling became competitive by using a zygotic arm mutant that retains wild-type maternal Arm, sufficing for Arm's role in adherens junctions (COX et al. 1996 ). With most wild-type maternal Arm assembled in adherens junctions, the embryo drops below the critical threshold of Arm necessary for Wg signaling, resulting in segment polarity defects. Such an embryo is very sensitive to slight changes in arm dose; for example, doubling the maternal Arm substantially suppresses the segment polarity phenotype (WIESCHAUS and NOELL 1986 ). Thus this represents a sensitized background well suited for a modifier screen. Mutations in genes that affect adherens junction assembly, which negatively regulate Wg signaling or encode other proteins that bind the limited supply of maternal Arm, could all potentially suppress the segment polarity phenotype of arm. We previously demonstrated the feasibility of this idea, showing that reduction in DE-cadherin suppressed arm's segment polarity phenotype (COX et al. 1996 ).

We used the sensitized background of a zygotic arm mutant to carry out a modifier screen, looking for changes in the segment polarity phenotype. We screened through deficiencies covering >80% of the second, third, and fourth chromosomes, searching for regions of the genome containing a gene or genes that, when heterozygous deficient, suppress the cuticle phenotype of arm. We found 12 such regions and identified four genes with this property. One interactor is the PCD-promoting gene head involution defective (hid). Our data suggest that Hid acts as a dose-sensitive regulator of PCD in the ventral epidermis of segment polarity mutants.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX A APPENDIX B LITERATURE CITED

Fly stocks:
References for mutants used were the following: arm^XP33, arm^XM19, and zw3^m1-1arm^XM19 (COX et al. 1996 ; PEIFER et al. 1994 ); hid⁰⁵⁰¹⁴ (GRETHER et al. 1995 ); hid^WR+X1 (ABBOTT and LENGYEL 1991 ); Df(3)H99 (WHITE et al. 1994 ); wg^IG22 (NUSSLEIN-VOLHARD and WIESCHAUS 1980 ); UAS-p35 (HAY et al. 1994 ); other mutations, http://flybase.bio.indiana.edu/. The deficiency kits were from the Bloomington Drosophila Stock Center, the P-lethals from Bloomington or the Berkeley Drosophila Genome Project (BDGP), and the Dpresenilin alleles from D. Curtis.

Cuticle preparations and counting:
Cuticle preparations were as in WIESCHAUS and NUSSLEIN-VOLHARD 1986 . Care was taken to be consistent in cuticle preparations, as differences in baking and pressing alter cuticle appearance. If the first cross suggested an interaction, the cross was repeated. Each candidate interacting region was tested in two or more separate crosses, with 200 cuticles scored per cross. Percentage of suppression equaled the number of cuticles in the least severe classes divided by the total number of cuticles scored.

Terminal transferase dUTP nick end labeling (TUNEL), phalloidin and antibody staining:
TUNEL was done using reagents from Boehringer Mannheim (Indianapolis). Embryos were dechorionated in 50% bleach, fixed in 1:1 4% formaldehyde:heptane for 30 min, hand devitellinized, rinsed once in TdT reaction buffer (2.5 mM CoCl₂, 1x transferase buffer), and reacted in TdT reaction mix (50 units terminal transferase, 2:1 10 µM final concentration of dUTP:dUTP-biotin in reaction buffer) for 3 hr at 37°. After washing three times for 10 min in PBS + 0.1% Triton X-100 (PBT), the end-labeling was first amplified using the Vectastain kit (Vector Labs, Burlingame, CA) as recommended by the manufacturer, amplified with Cy3tyramide (New England Nuclear, Boston), and washed three times for 10 min in PBT. BODIPY, phalloidin (Molecular Probes, Eugene, OR) was added during the avidin-biotin reaction of the first amplification. Antiphosphotyrosine labeling was as in COX et al. 1996 .

Phosphohistone H3 staining:
The 2- to 7-hr-old embryos were dechorionated in 50% bleach, fixed in 1:1 5% formaldehyde:heptane for 20 min, blocked (50 mM Tris pH 7.4, 150 mM NaCl, 0.5% NP-40, 5 mg/ml BSA) at 4° for 2 hr and stained overnight at 4° with 1:1000 antiphosphohistone H3 (Upstate Biotechnology, Lake Placid, NY) and 1:500 anti-ß-gal (Boehringer Mannheim, Indianapolis). Secondary antibodies were from Molecular Probes. Pictures of the ventral epidermis and dorsal germband were taken, mitotic figures (stained for phosphohistone H3) counted, and means and standard deviations calculated.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX A APPENDIX B LITERATURE CITED

Strategy for the screen for modifiers:
arm^XP33 encodes a carboxy-terminally truncated Arm protein that cannot function in Wg signaling and has almost no function in adherens junctions (COX et al. 1996 ). In an arm^XP33 zygotic mutant, maternal wild-type Arm provides sufficient function for adherens junctions. However, as nearly all maternal Arm is recruited into junctions (COX et al. 1996 ), little Arm remains to transduce Wg signal, resulting in a strong segment polarity phenotype (Fig 1A vs. B). We reasoned that if one elevated the level or function of the limiting pool of maternal Arm, this should suppress the defect in Wg signaling. We hypothesized that this could occur by freeing maternal Arm from junctions, reducing the effectiveness of negative regulation of Arm's role in Wg signaling or by reducing the level of a distinct Arm-binding protein. Of course in genetic screens modifiers are also found that operate by unexpected mechanisms.

View larger version (126K): In this window In a new window Download PPT slide   Figure 1. Reducing dosage of DE-cadherin or Zw3 suppresses arm's embryonic phenotype. (A) Wild-type cuticle. The cuticle is closed dorsally and the ventral surface has alternating belts of denticles and naked cuticle. (B) armXP33/Y. armXP33 mutant cuticles are shorter than wild type, with a lawn of ventral denticles, no naked cuticle, and incomplete dorsal closure. (C) armXP33/Y; Df(3) E2/+. When the dose of DE-cadherin is reduced by half, the arm phenotype is suppressed. The cuticle is longer, the lawn of denticles is less dense, and the dorsal closure defect is rescued. (D) armXM19/Y, a somewhat weaker allele. (E) armXM19 zw3M1-1/Y. When zygotic zw3 is removed (leaving only maternal zw3), the phenotype of armXM19 is suppressed.

The feasibility of this hypothesis was supported by two observations. We previously found that heterozygosity for a chromosomal deficiency removing DE-cadherin, Df(3R)E2, suppresses the embryonic phenotype of arm^XP33—the cuticle is longer, the dorsal closure defect is substantially reduced, and denticle diversity is partially restored (Fig 1B vs. C; COX et al. 1996 ). We presume that reducing the gene dose of DE-cadherin by half creates an embryo with fewer Arm/Cadherin complexes. Although this has no apparent effect on cell-cell adhesion, wild-type maternal Arm is freed up to function in Wg signaling, leading to a suppressed phenotype. A similar suppression of arm was seen by removing the zygotic contribution of one of Arm's negative regulators, Zw3 (Fig 1D vs. E; we tested this on arm^XM19, a less severe allele).

A modifier screen for Arm interactors:
These examples demonstrated that a 50% reduction in the dose of certain genes suppresses arm. We thus screened for dose-sensitive modifiers. Rather than examining single genes one by one by mutagenesis, we evaluated large regions of the genome simultaneously by making animals heterozygous for chromosomal deficiencies that remove many genes. We obtained the "deficiency kits" for three of the four chromosomes from the Bloomington Drosophila Stock Center. These kits are designed to delete as much of the chromosome as possible using the fewest stocks; 70–80% of the euchromatin was covered by this collection of Deficiencies when we obtained them. We extended our analysis by obtaining additional Deficiencies that either covered regions not covered in the kit or overlapped interacting Deficiencies. We estimate we covered >80% of the autosomes. We have not examined the X chromosome thus far, as arm is on the X and the screen would require recombination of arm onto each deficiency. To carry out the screen, we crossed virgin arm^XP33 females to males heterozygous for each Deficiency and prepared cuticles from the dead embryonic progeny (Fig 2A). One-quarter of the progeny are arm^XP33/Y (since arm is X-linked), and these die due to loss of arm function. Half of these embryos will be hemizygous for genes deleted by the deficiency and could potentially have a modified segment polarity phenotype.

View larger version (16K): In this window In a new window Download PPT slide   Figure 2. Modifier screen strategy. (A) Cross scheme. armXP33/FM7 virgin females were crossed to males carrying a single deficiency on chromosome 2, 3, or 4. One-quarter of the progeny are armXP33/Y and die with a segment polarity phenotype. Half of these are heterozygous for the deficiency. Cuticles were prepared from dead embryonic progeny of each cross. If there is an interaction, these cuticles show suppression of arm. (B) The armXP33 cuticle phenotype varies about a mean (black bars). Removing one copy of DE-cadherin shifts the distribution, greatly increasing the fraction with the least severe phenotypes (light bars). Our scoring scheme for phenotypic severity was based on cuticle size, strength of the segment polarity phenotype, and degree of dorsal closure. We counted 200 embryos per cross. Percentage of suppression = the number of progeny in the least severe classes (left of the dashed line) divided by the total number of progeny scored.

To determine if there was an interaction, we grouped cuticles into phenotypic classes. arm^XP33 mutants exhibit a segment polarity phenotype, with all surviving cells adopting anterior fates and secreting denticles. However, arm^XP33 mutants show a range of severities; the phenotypes vary about a mean (Fig 2B). In embryos with the most severe phenotype (like that of the zygotic null), the cuticle is much shorter than the wild type and is open dorsally. In less severe embryos, dorsal closure is partially complete, and the embryonic cuticle is longer. Most mutant embryos fall into these classes. At a very low frequency (0.5%), arm^XP33/Y cuticles have the least severe phenotype: these are nearly wild type in length, have greater denticle diversity, and are dorsally closed (they retain an anterior hole). If one does a similar analysis of arm^XP33/Y; Df-DE-cadherin/+ embryos, as an example of suppression, one finds that the phenotypic distribution is strongly shifted toward the less severe end (Fig 2B)—in this example, 33% of the cuticles fall in the least severe classes (embryos to the left of the dotted line in Fig 2B). On the basis of this, we focused on the frequency of embryos in the least severe classes. To score whether a Deficiency suppressed the arm^XP33 phenotype, we prepared cuticles from the dead embryos, scored their phenotypes, and calculated the percentage of cuticles in the least severe classes; if this was at least six times the frequency in the control (i.e., 3%), we scored this as an interaction.

By these criteria, 32 deficiencies interacted with arm^XP33 (Table 1); a representative suppressed cuticle is shown in Fig 3B. Table 1 and Table 2, Fig 4, and APPENDIXES A and B summarize the screen, showing which regions were covered by deficiencies and which regions interacted. In all cases the suppression was qualitatively similar; embryos in the least severe class showed an increase in cuticle length, improvement in dorsal closure, and an increase in denticle diversity. The fraction of cuticles in the least severe phenotypic class ranged from 3 to 40% (each number is an average of two to three independent crosses; Table 1 and Table 2). We retested each interacting stock—all reliably interacted although in some cases the percentage of suppression varied. Of the 32 stocks that interacted, we arbitrarily made a cutoff between "weak" and "strong" interactions at the level of 6% of the embryos in the least severe phenotypic classes. Eighteen Deficiencies were thus classified as weak interactors, with 3–5.9% of the cuticles in the least severe category (Table 1; APPENDIXES A and B). Although this degree of suppression was reproducible, there were enough regions that suppressed arm^XP33 more robustly that weakly interacting regions were not investigated further. We noted in passing that six stocks had hemizygous dominant cuticle phenotypes other than effects on segment polarity (Table 2B); one was also one of the strong interactors.

View larger version (164K): In this window In a new window Download PPT slide   Figure 3. Suppression by Deficiencies from the screen and Dpresenilin. (A) armXP33/Y. (B) armXP33/Y; Df(2R)PC4/+. Example of the suppression observed in the screen. (C) armXP33/Y; Dpresenilin10/+. Heterozygosity for Dpresenilin significantly suppressed the armXP33 phenotype. (D) Putative armXP33/Y; Dpresenilin10/Df(3L)ri-79c embryo. A slightly more robust suppression was seen in progeny of flies heterozygous for Dpresenilin10 crossed to flies heterozygous for Df(3L)ri-79c. (E) armYD35/Y. The zygotic null allele. (F) armYD35/Y; Dpresenilin10/+. Heterozygosity for Dpresenilin also suppressed armYD35.

View larger version (18K): In this window In a new window Download PPT slide   Figure 4. Schematic summary of the screen. We estimate that 80% of the euchromatin of chromosomes 2, 3, and 4 were covered. Regions covered are represented by the black portions of the chromosomes; white portions are regions for which we were unable to find deficiencies. Black boxes below chromosomes represent regions containing putative suppressor(s) defined by the overlap between two or more deficiencies, or where an interacting gene was defined. White boxes represent regions defined by a single interacting Deficiency (see APPENDIX A for details).

Fourteen deficiency stocks were "strong" interactors; with 6–40% of the cuticles in the least severe classes (Table 1 and Table 2; Fig 4). Two of these deficiencies, Df(3L)W10 and Df(3L)Cat, overlap, suggesting that the gene responsible for that interaction lay in the overlapping region (75B8;C1-2) and reducing the number of interacting genomic regions to 13. While two other interacting deficiencies, Df(2L)spd and Df(2L)TE29, should not quite overlap based on their reported cytology, they fail to complement one another, strongly suggesting that they do in fact overlap, reducing the number of interacting regions to 12. We analyzed other deficiencies in the regions of the strongly interacting Deficiencies, allowing us in most cases to further pinpoint the interacting region (see APPENDIXES A and B for details). In four cases, smaller interacting Deficiencies were identified. In nine cases, overlap of the original deficiency with other Deficiencies that either interacted or did not interact allowed us to further define the interacting region.

None of the deficiencies tested resulted in any obvious enhancement of the arm^XP33 phenotype, either producing defects in epithelial integrity or enhancing the segment polarity defects. The wild-type maternal contribution of Arm appears to completely provide adherens junction function, so reducing levels of components required for adherens junction function by 50% apparently does not affect epithelial integrity in arm^XP33 mutants. In fact, when Müller and Wieschaus examined embryos homozygous for large deficiencies, they found no regions that were zygotically essential for adherens junction assembly and few that had a strong effect on junction function (MULLER and WIESCHAUS 1996 ). We realized in retrospect that the severity of the arm^XP33 segment polarity phenotype made it unlikely one could reliably recognize an enhancer of this defect.

One possible confounding factor was that mutations on the Balancer chromosomes with which the Deficiency chromosomes were heterozygous could have been the true cause of the phenotypic suppression. We think this is quite unlikely, as only a small number of Balancer chromosomes were used and none showed a consistent effect on the arm phenotype. A second potential problem is that second site mutations on the Deficiency chromosomes could in principle be responsible for certain observed interactions. This is highly unlikely for the seven strongly interacting regions that are defined by either two or more interacting Deficiencies or by a Deficiency and an identified gene (Fig 4). For the other five strongly interacting regions, some may be due to linked mutations outside the Deficiency interval, although given the overall frequency at which interactions were detected, we think this is unlikely to be the case for all.

Finding interactors by testing candidate genes:
Our first approach to identify the gene(s) within each Deficiency responsible for the interaction was to test candidate genes in each region. We considered as candidate genes those with a mutant phenotype indicating an effect on cell fate choice in the ventral epidermis, genes known to act in Wg signaling, and genes known to affect cell-cell junctions or the actin cytoskeleton. We identified one interactor by this candidate gene approach and ruled out many other candidates by two methods: testing complementation between a candidate and the interacting deficiency and checking directly whether the candidate could suppress arm.

We tested four candidate genes that are part of the Wg signal transduction pathway or that affect segment polarity: dTCF, cubitis interruptis, naked, and wg. Removing one copy of the fourth chromosome gave a very strong interaction. In examining candidates on the fourth chromosome, we found that mutations in the gene encoding the DNA-binding protein dTCF, which is required for Wg signaling, suppress arm^XP33. This was a surprise and revealed a previously unexpected role for dTCF as a repressor as well as an activator of Wg-responsive genes (CAVALLO et al. 1998 ). However, while null alleles of dTCF interact strongly, they do not suppress arm^XP33 to the same degree as removing the entire fourth chromosome. Thus, there may be a second suppressor on the fourth chromosome. cubitis interruptis, a gene involved in hedgehog signaling, was ruled out as this suppressor by testing a null allele for interaction (CAVALLO et al. 1998 ). naked is a known negative regulator of Wg signaling and maps near Df(3L)Cat. However, it complemented this Deficiency and was thus ruled out. Two deficiencies, Df(2L)TE29 and Df(2L)spd, are in the vicinity of wg. While wg is a positively acting component of the pathway, our experience with dTCF made us cautious in ruling it out without a test. We found that: (1) wg complements Df(2L)TE29 and (2) a wg null does not suppress arm. This ruled wg out, although DWnt4, which maps near wg (GRABA et al. 1995 ), remains a candidate. Finally, we tested alleles of two segment polarity genes that fell outside regions included in the Deficiencies in the kit: hedgehog and teashirt, which encodes a transcription factor that physically and functionally interacts with Arm (GALLET et al. 1998 ). Neither suppressed arm^XP33.

We also tested several genes with roles in cell-cell adhesion or cytoskeletal function. One was DE-cadherin (shotgun), which we already knew could suppress arm. Df(2)017 was suggested by its cytology to remove DE-cadherin, but both an allele of DE-cadherin and the small deficiency Df(2)E2 that removes DE-cadherin (UEMURA et al. 1996 ) complement Df(2)017. Thus this interaction is due to a different gene. Three other genes that regulate the cytoskeleton, enabled (ena; GERTLER et al. 1995 ), quail (MAHAJAN-MIKLOS and COOLEY 1994 ), and scraps (SCHUPBACH and WIESCHAUS 1989 ), map to regions covered by interacting Deficiencies (56B5, 36C2-11, and 43E7, respectively). ena is an actin cytoskeleton regulator, quail encodes a vinculin-like protein thought to associate with actin, and scraps is required for the cytoskeletal events of cellularization. ena was included in interacting deficiency Df(2R)PC4 by complementation (we did not test quail and scraps by complementation). However, when we tested alleles of all three genes, none suppressed arm^XP33. 18-wheeler, a putative cell-adhesion molecule (ELDON et al. 1994 ) that maps in or near Df(2R)017, also did not suppress arm^XP33.

As a partial test of the effectiveness and completeness of the screen, we also tested a series of additional candidate genes, some of which fell outside Deficiencies in the kit and others of which were probably included in these Deficiencies but which we expected might physically or functionally interact with Arm. The vast majority did not show an interaction. We tested a variety of genes encoding components of other signal transduction pathways that pattern the dorsal or ventral epidermis: (1) the Hedgehog pathway, hedgehog^Ac; (2) the Dpp pathway, decapentaplegic^e87 and screw^I1; (3) the EGF receptor (EGFR) and other receptor tyrosine kinase pathways, spitz^2A14, vein^147-2, argos²⁵⁷, Egfr^C18, ras85D^e1B, rolled^C18, yan^XE-12, and torso¹; and (4) the Jun N-terminal kinase pathway, basket and Djun¹. Of these, only spitz^2A14 interacted, and even in this case, only 3.8% fell into the weakest phenotypic categories, just above our cutoff for a weak interaction. We also tested five genes affecting the cytoskeleton or cuticle integrity: krotzkopf verkehrt¹, myoblast city^C1, shroud¹, steamer duck^3R-17, and scraps⁸. None interacted. Finally, we tested one candidate among proteins that interact with mammalian ß-catenin but for which the function of this interaction is not known. This was Drosophila presenilin, homolog of the mammalian presenilin family of transmembrane proteins (reviewed in HAASS and DE STROOPER 1999 ). Mammalian Presenilins bind mammalian ß-catenin (ZHOU et al. 1997 ; MURAYAMA et al. 1998 ; YU et al. 1998 ; LEVESQUE et al. 1999 ). Further, misexpression and other experiments suggest that mammalian Presenilins may regulate Wnt signaling (MURAYAMA et al. 1998 ; ZHANG et al. 1998 ; KANG et al. 1999 ; NISHIMURA et al. 1999A ). In contrast to the other candidates tested, D. presenilin showed a very strong interaction. Heterozygosity for D-presenilin strongly suppressed arm^XP33 (14.6% with weakest phenotypes; Fig 3A vs. C) and also suppressed the zygotic null arm allele, arm^YD35 (Fig 3E vs. F). A surprise from these results was that although Dpresenilin was removed by two of the Deficiencies tested, Df(3L)ri-79c and Df(3L)rdgC-co2, neither showed a significant interaction (percentage of suppressions = 1.8 and 1.5%, respectively). This suggests that some interactions are sensitive to genetic background, and thus not all potential haplo-insufficient interactors were identified in our screen.

P-element lethals that interacted:
Our second approach to identifying genes responsible for an interaction was to use the collection of P-element-induced lethal mutations (hereafter called P-lethals) characterized by the Berkeley Drosophila Genome Project. These lethals are caused by P-element transposon insertions and are thus molecularly tagged, facilitating cloning. The available P-lethals are estimated to hit ~25% of essential genes (SPRADLING et al. 1999 ). One caveat to using these mutations to uncover a dose-sensitive suppressor is that there is no guarantee that the P-lethal will be a null allele, as is a Deficiency. P-transposons tend to insert either in the 5' untranslated region or in introns, thus creating mutations that often are not null in phenotype and thus do not, when heterozygous, reduce gene function by 50%.

We obtained the P-lethals available from the Bloomington Stock Center (81 stocks) and the Kiss collection (73 stocks) in each of the interacting regions and tested their ability to suppress arm^XP33. A list of the P-lethal stocks tested is in a data supplement at http://www.genetics.org/cgi/content/full/155/4/1725/DC1. Of the P-lethals tested, we found two that suppressed arm^XP33. One of these, l(3)A251.1, mapped to region 84E. By examination of its homozygous phenotype and subsequent complementation tests, we learned that this is an allele of puckered (MARTIN-BLANCO et al. 1998 ). A detailed examination of the biology underlying this interaction is presented in MCEWEN et al. 2000 .

The apoptosis-promoting gene head involution defective is a dose-sensitive suppressor of arm:
The second P-lethal that interacted with arm was l(3)05014, which maps to 75C1-2 and gave as strong a suppression as either of the interacting deficiencies in this region. l(3)05014 is an allele of head involution defective (GRETHER et al. 1995 ). This allele is likely to be a null, as the P-element is inserted early in the protein-coding region. Null mutations in hid are embryonic lethal with defects in head involution during embryonic development (ABBOTT and LENGYEL 1991 ; Fig 5A vs. B), although occasional escapers survive to adulthood. More recently, it was revealed that hid mutations affect programmed cell death.

View larger version (88K): In this window In a new window Download PPT slide   Figure 5. Decreasing cell death suppresses arm. Wild type (A), hid05014 homozygotes (B), and Df(3L)H99 homozygotes (C) have very similar cuticle phenotypes. hid05014 and Df(3L)H99 homozygotes have head defects, but their segment polarity is normal. (D) armXP33/Y (E) armXP33/Y; hid05014/+. Removing one copy of hid suppresses armXP33. (F) armXP33/Y; armGAL4/UAS-p35. Expression of the baculovirus antiapoptotic protein p35 in an armXP33 mutant background also suppresses arm.

The machinery that triggers PCD in Drosophila has been the subject of intense investigation (reviewed in ABRAMS 1999 ). This work was initiated by a screen for genomic regions required for PCD (WHITE et al. 1994 ). When chromosomal region 75C1-2 is deleted, in embryos homozygous for the small deficiency Df(3)H99, essentially all PCD in the embryo is eliminated (WHITE et al. 1994 ). Subsequent analysis revealed that this chromosomal region contains three genes involved in PCD: hid, reaper, and grim (reviewed in ABRAMS 1999 ). Ectopic expression of any of these will trigger PCD. However, loss-of-function mutations are only available for hid. In hid mutants, a subset of the cells that normally undergo PCD do not do so (GRETHER et al. 1995 ), resulting in defects in head development. In embryos homozygous for Df(3L)H99, which thus lack hid, reaper, and grim, all PCD is abolished (WHITE et al. 1994 ); these embryos have slightly stronger defects in head development (Fig 5C).

hid plays an important role in PCD. Ectopic expression of hid is sufficient to induce PCD in the eye, and this is completely suppressed by the baculovirus caspase inhibitor p35, suggesting hid acts upstream of caspases (GRETHER et al. 1995 ). hid has no clear homologs in other organisms; however, Hid overexpression triggers PCD in mammalian cells. Hid, Reaper, and Grim all share a short region of weak sequence similarity near their N termini. In Hid, this region is required for initiating cell death in mammalian tissue culture, while Hid's C terminus is required for localization to mitochondria, an organelle involved in PCD (HAINING et al. 1999 ). Recent work supports the idea that Hid functions by blocking interaction between Inhibitor-of-apoptosis (IAP) family caspase inhibitors and caspases (VUCIC et al. 1998 ; WANG et al. 1999 ).

Heterozygosity for hid⁰⁵⁰¹⁴ suppresses arm^XP33 (Fig 5D vs. E), as well as the zygotic null allele arm^YD35 (data not shown). Heterozygosity for an X-ray-induced loss-of-function allele, hid^WR+X1, causes the same degree of suppression, further supporting the idea that hid is the gene responsible for the interaction. In addition, we generated revertants of the P element in hid⁰⁵⁰¹⁴ by mobilizing the P element and screening for viable stocks that lost the genetic marker carried by the P element. These revertant chromosomes fail to suppress arm^XP33 (data not shown). Further reducing hid levels by making embryos homozygous for hid⁰⁵⁰¹⁴ does not increase the degree of suppression of arm^XP33. Likewise, either heterozygosity or homozygosity for the small deficiency Df(3L)H99, which removes hid, grim, and reaper, suppresses arm^XP33 to the same degree as removal of one copy of hid.

The suppression by hid can be mimicked by blocking PCD:
PCD is elevated in segment polarity mutants (MARTINEZ ARIAS 1985 ; KLINGENSMITH et al. 1989 ; PAZDERA et al. 1998 ). The dramatically shortened cuticle secreted by an arm mutant is presumably caused, at least in part, by loss of ventral epidermal cells via PCD. The suppression of arm^XP33 by hid⁰⁵⁰¹⁴ could thus be due to Hid's role in PCD; alternately, it could be due to an unknown function of Hid. To test if the arm suppression results from an effect on PCD, we reduced embryonic PCD by expressing the baculovirus antiapoptotic protein p35, which acts as a caspase inhibitor; p35 suppresses the PCD triggered by hid overexpression in the fly eye (GRETHER et al. 1995 ). We found that arm^XP33 mutant embryos in which we ubiquitously expressed p35 using the GAL4-UAS system (BRAND and PERRIMON 1993 ) had a suppressed phenotype (Fig 5F vs. D). The suppression by p35 was similar in degree to that resulting from hid heterozygosity (Fig 5E). This suggests that decreased PCD in the embryo can suppress arm and supported the idea that the interaction between hid and arm was due to hid's role in PCD.

hid suppresses wg in a highly dose-sensitive fashion:
We next tested whether the effect was arm specific or whether reduction in PCD would suppress the phenotype caused by other reductions in Wg signaling. To do so, we examined whether reduction in PCD suppressed a null allele of wg, wg^IG22. wg^IG22 mutant cuticles have a lawn of uniform, large denticles covering the ventral epidermis (NUSSLEIN-VOLHARD and WIESCHAUS 1980 ; Fig 6A), and they are much smaller than wild type, but unlike arm mutant cuticles they are usually closed dorsally. We tested whether heterozygosity or homozygosity for either hid or for Df(3L)H99 suppressed wg^IG22.

View larger version (139K): In this window In a new window Download PPT slide   Figure 6. Reducing PCD suppresses wg in a dose-sensitive fashion. (A–C) Cuticle preps. (D–F) Embryos labeled with phalloidin, which recognizes f-actin and thus highlights denticles. These embryos were also labeled via TUNEL (data not shown) to confirm their genotypes. (A, F) wg null mutants completely lack segment polarity and have only large, thick denticles. (B) wgIG22; hid0501. (C, E) wgIG22; Df(3L)H99. (D) Wild type. In the double mutants, denticle number greatly increases. (G–K) Embryos stained with antiphosphotyrosine antibody to outline cells. (G) Wild type, showing reiterated groups of narrow cells, which will secrete denticles, and less narrow cells, which will secrete naked cuticle. There are 12 rows of cells per segment. (H) wg single mutants have fewer, larger cells. All cells are cuboidal, and there are about 8 rows of cells per segment. (I) wg; Df(3L)H99 double mutants have many more ventral epidermal cells than wg single mutants—there are 12 to 14 rows of cells per segment. Cells in the double mutant are much smaller. (J) Lateral view of a wild-type embryo during germband retraction. (K) Df(3L)H99 at the same stage, revealing an increased number of cells compared to wild type. Excess cells form a lateral fold and ectopic folds near the maxillary and labial segments and toward the posterior. (L) Wild-type cuticle. (M) armXP33. (N) armXP33; UAS-dsh/VP16::armGAL4. Expressing high levels of dsh in armXP33 results in an increase in denticle number and reduction in denticle size, similar to that in wg; Df(3L)H99 double mutants.

hid modified wg^IG22 in a dose-sensitive fashion, but the nature of the phenotypic modification was different from that seen with arm. There was not any pronounced improvement in the wg segment polarity defect; in wg^IG22; hid⁰⁵⁰¹⁴ (Fig 6B) or wg^IG22; Df(3L)H99 (Fig 6C) double mutants, all cells still secrete a uniform lawn of denticles, and the cuticle of the wg^IG22; Df(3L)H99 double mutant remains much smaller than that secreted by a wild-type embryo, contrasting with the increase in cuticle size in arm; Df(3L)H99 double mutants. However, we found a striking effect of hid dose on the number and size of the denticles on the ventral epidermis. The number of denticles is more than doubled in wg^IG22; Df(3L)H99 compared to wg^IG22 alone, and the denticles secreted by the double mutant are much smaller than those in the wg null (Fig 6A vs. B and C; the change in denticle size may be less meaningful, as denticle size is also somewhat reduced in Df(3L)H99 homozygotes that are wild type for arm and wg; Fig 5C).

wg^IG22 is less sensitive to reduction in hid dose than arm, and thus the effect on wg^IG22 is additive. Removal of one copy of hid in a wg^IG22 background has only a subtle effect on cuticle pattern (data not shown), while removal of both copies of hid has a stronger effect (Fig 6B vs. A). Removing one copy of the region covered by Df(3L)H99 has a greater effect than removing both copies of hid (data not shown), suggesting that removing all three cell death genes results in a more pronounced interaction. The effect on cuticle pattern is thus most pronounced in wg^IG22; Df(3L)H99 double mutants (Fig 6C), which have many more, much smaller denticles than does a wg^IG22 single mutant (Fig 6A). To confirm this, we labeled embryos with phalloidin to visualize the filamentous actin in denticles and with TUNEL to identify embryos with cells undergoing PCD. Embryos homozygous for Df(3L)H99 do not label with TUNEL, as they have no PCD, allowing us to unambiguously identify double mutants. The results matched the cuticle data: wg^IG22 mutants showed the characteristic lawn of denticles (Fig 6F), while wg^IG22; Df(3L)H99 double mutants (embryos without cells undergoing PCD as measured by TUNEL) had many more much smaller denticles (Fig 6E). In the course of this analysis, we also observed that Df(3L)H99 mutants have significantly more epidermal tissue in the head (as was previously observed by GRETHER et al. 1995 ) and in the lateral epidermis (Fig 6J vs. K), consistent with the idea that wild-type embryos reduce the number of epidermal cells via PCD. We also examined whether reduction in PCD suppressed a weaker wg heteroallelic combination, wg^PE4/Df(2)DE. We saw no noticeable suppression of the segment polarity phenotype and no noticeable increase in denticle number caused by either heterozygosity or homozygosity for Df(3L)H99 (data not shown). This may not be surprising as this weaker wg phenotype likely primarily reflects changes in cell fate without significant ectopic cell death, as the cuticle is nearly wild type in length.

Blocking cell death in wg^IG22 increases cell number but decreases cell size:
The novel phenotype of wg^IG22; Df(3L)H99 double mutants could have several causes. Extra denticles could result if individual cells secreted more denticles; alternately, they could result from an increased number of cells. To distinguish these possibilities, we examined the cell morphology of wild type, wg^IG22, and wg^IG22; Df(3L)H99 double-mutant embryos, using antibodies to phosphotyrosine to outline ventral epidermal cells and to label developing denticles. In wild-type embryos (Fig 6G), ventral epidermal cells form a reiterated pattern of denticle-secreting cells, which are very narrow in the anterior/posterior (A/P) axis, and naked cuticle-secreting cells, which are much less narrow. There are, on average, 12 rows of cells per segment. In contrast, in wg mutants there are only 8 rows of cells per segment (Fig 6H; the segment boundary was determined by comparison of the denticle pattern in cuticles to the phosphotyrosine pattern). In the wg^IG22; Df(3L)H99 double mutant, cell number is greatly increased relative to the wg^IG22 single mutant. The double mutant has 12 to 14 rows of cells (Fig 6I), equaling or exceeding the number of cell rows in the wild type. Thus eliminating PCD in a wg^IG22 mutant embryo increases cell number, as expected.

Blocking cell death in a wg^IG22 mutant also had a second, unexpected consequence—cell size was significantly decreased. As mentioned above, in wild-type embryos anterior denticle-secreting cells are narrowed in the A/P axis, while posterior naked cuticle-secreting cells are not. In contrast, in a wg mutant all cells are both uniformly cuboidal (Fig 6H) and significantly larger than denticle-producing cells of a wild-type embryo. This increase in size likely reflects an increase in cell volume, because in optical cross sections wg^IG22 and wild-type cells were the same height (data not shown). In contrast, cells of wg^IG22; Df(3L)H99 double mutants are much smaller than those in wg single mutants (Fig 6I). Ventral cells of double mutants do resemble wg^IG22 single mutants in several ways; most cells are cuboidal, the cells create a pattern of block-like pseudosegments (though with more rows of cells than in wg single mutants), and all cells secrete denticles. We do not have a good explanation for the qualitative difference in the effect of hid on the arm and wg^IG22 phenotypes. We observed one other situation where manipulating Wg signaling resulted in an increased number of smaller denticles. Overexpression of dsh using the GAL4-UAS system in an arm^XP33 mutant gives rise to a cuticle with many very small denticles, but with the length of an arm^XP33 single mutant (Fig 6M vs. N). Dsh is a positive effector of Wg signaling mapping upstream of Arm in the Wg pathway; we imagine that Dsh overexpression slightly augments the residual Wg signaling in an arm zygotic mutant.

Our comparison of wg^IG22 and wg^IG22; hid suggests that an increase in PCD contributes to the reduced number of ventral epidermal cells in wg, consistent with previous observations (PAZDERA et al. 1998 ). This might also explain the increased cell size in wg mutants, as epidermal cells have been observed engulfing dying neighbors (PAZDERA et al. 1998 ), thus potentially increasing their size. However, since Wnts are mitogens in certain cell types (e.g., DICKINSON et al. 1994 ; NEUMANN and COHEN 1996 ), we also considered an alternate explanation that could explain both reduction in cell number and increase in cell size in wg^IG22 mutants: a failure to complete the normal program of cell division. Since no growth occurs within the embryo, as cells divide they are reduced in size, and thus if wg^IG22 mutant cells failed to complete one round of mitosis, they would be twice as large.

We thus assessed the pattern of cell division in wg^IG22. Ventral epidermal cells divide three times after the blastoderm stage and arrest in G1 of the 17th embryonic cell cycle (EDGAR and O'FARRELL 1989 ). Cell divisions can be visualized by pulse labeling with BrdU to detect replicating nuclei or with an antibody that specifically recognizes a phosphoisoform of histone H3 that only occurs during mitosis (Fig 7; SU et al. 1998 ). Condensed mitotic chromosomes can be easily identified in fixed tissues with this antibody (Fig 7C and Fig D). We analyzed embryos at a time (stage 12) when cells of the ventral epidermis normally complete their 16th cell cycle, to see if wg^IG22 mutants fail to undergo this last cell division. BrdU labeling indicated that ventral epidermal cells replicate during S phase 16 in wg^IG22 mutants (data not shown). In addition, mitotic figures are as readily apparent in the ventral epidermis of stage 12 wg^IG22 mutant embryos as they are in wild type (Fig 7A and Fig B). Thus, lack of Wg activity does not cause inappropriate cell cycle arrest. To compare the mitotic index between wild type and wg^IG22, we counted the total number of phosphohistone H3 positive nuclei of the ventral surface of individual embryos (arrow in Fig 7C and Fig D). There was no significant difference between the average number of mitotic cells in wild-type (175 ± 75.7; n = 7) vs. wg^IG22 mutant (232 ± 44.2; n = 9) embryos. The variance in the absolute numbers of mitotic cells from embryo to embryo can be attributed to at least two factors: First, the precise age of each embryo scored differs slightly, as stage 12 spans ~2.5 hr of development at 25°, and second, there is some cell cycle asynchrony among individual cells of a particular epidermal region that enter mitosis "together" (i.e., mitotic domains). We conclude that wg^IG22 mutant embryos complete the normal number of cell divisions in the ventral epidermis, supporting the idea that the reduction in cell number in this region of late stage wg^IG22 embryos is primarily a consequence of elevated levels of PCD.

View larger version (104K): In this window In a new window Download PPT slide   Figure 7. wg mutants enter mitosis 16 in the ventral epidermis. Ventral views of representative wgIG22 (B, D) and phenotypically wild-type sibling (A, C) embryos stained with antiphosphohistone H3 antibodies. The magnification in C and D is twice that of A and B. Arrows in C and D indicate condensed chromosomes of cells that were counted. Only mitotic figures in the epidermis were included in the counts; out of focus staining (e.g., arrowhead) is from mitotic cells in the underlying central nervous system. Embryos were from a wgIG22/CyO ftz-lacZ stock, allowing unambiguous identification of wg mutants by lack of ß-galactosidase expression. wg mutants were also apparent due to defects in the normal segmental pattern of S phase and mitosis.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX A APPENDIX B LITERATURE CITED

Drosophila Arm and its human homolog ßcat are multifunctional proteins that play roles in cell-cell adherens junctions and in the transduction of Wg/Wnt signals. In both roles, Arm/ßcat acts as a scaffold upon which a multiprotein complex is assembled. In addition to these well-documented roles, Arm/ßcat associates with other proteins, such as the EGF receptor (HOSCHUETZKY et al. 1994 ), fascin (TAO et al. 1996 ), and Presenilin 1 (ZHOU et al. 1997 ). The biochemical function of these complexes is unknown.

We desire to learn more about the known roles of Arm in adherens junctions and in Wg signaling and also to begin to learn what Arm might do with its other partners. Genetics offers the opportunity to look for proteins that are functionally linked with Arm without assumptions as to their identity or biochemical role. Our initial goal in the screen was to identify novel proteins essential for adherens junction assembly or structure. However, as in all genetic screens, we had cast our net much wider. In the four cases where we proceeded from Deficiency to single gene, none encode new junctional proteins and each reveals a separate aspect of Arm biology. The fourth chromosome interactor dTCF revealed a previously unexpected role for a known component of the Wg pathway, providing evidence that dTCF not only activates Wg responsive genes but also, in the absence of Arm, represses them (CAVALLO et al. 1998 ). Characterization of the interactor in 84E, Puckered, revealed a novel role for a known protein and led to data suggesting that the JNK and Arm pathways act in parallel both in dorsal closure and in ventral patterning (MCEWEN et al. 2000 ). The interaction with Dpresenilin suggests that the biochemical interaction observed in mammalian cells has impact on Arm function in vivo. The fourth interactor, Hid, demonstrated that altering a downstream consequence of the loss of Wg signaling, programmed cell death, could suppress arm and, as is discussed below, suggested that Wg may act as a survival factor by modulating Hid activity.

We were initially concerned about the amount of labor required to screen for suppressors reducing the severity of an embryonic lethal phenotype without restoring viability (we expected that suppression to viability was unlikely). In retrospect, the screen, while labor intensive, was quite straightforward and could be applied to other embryonic lethal genes with a clear cuticle phenotype (arm's position on the X chromosome eased the effort). Use of the Deficiency kit reduced the number of stocks screened, although having completed the screen we now believe one could carry out such a screen using individual mutagenized lines. Others also recently screened for suppressors or enhancers of embryonic lethal phenotypes, suggesting that this approach may be widely applicable (RAFTERY et al. 1995 ; HUDSON et al. 1998 ; A. BEJSOVEC, personal communication).

Our screen had several limitations that affected the spectrum of genes identified. First, for a gene to be identified, it had to affect the arm phenotype in a dose-sensitive way. Second, the effect on arm had to be consistent and substantial. Our arbitrary cutoff for degree of interaction likely eliminated genes in the desired categories in which mutations did not sufficiently suppress arm. For example, loss-of-function mutations of Drosophila abelson or Deficiencies that remove it suppress arm, but not to a sufficient degree to be scored positive in our screen (LOUREIRO and PEIFER 1998 ). Third, due to the allele of arm we chose, we could not reliably score enhancement of the segment polarity phenotype, and likewise, potentially due to high levels of maternal Arm, we did not detect any interactor that produced defects in epithelial integrity. Fourth, since many P-element alleles are not null, our ability to move from Deficiency to single gene using these mutations was consequently limited. Finally, genetic background may obscure some interactions—for example, we saw a clear genetic interaction with two alleles of Dpresenilin but saw no significant interaction with two Deficiencies that remove it.

During preparation of this manuscript, an article appeared describing a different strategy for identifying genetic interactors with arm, which provides an interesting comparison. GREAVES et al. 1999 expressed in the posterior compartment of the wing the intracellular domain of DE-cadherin, which they previously found could sequester Arm and thus block its signaling activity (SANSON et al. 1996 ). In parallel they overexpressed Arm in the same place. Each caused a reproducible wing phenotype, which appears to reflect reduction and elevation of Wg signaling, respectively. They then screened for modifiers of these phenotypes using, as we did, the Deficiency kit. Many of the deficiencies tested thus overlapped (though not all, as we did not test the X chromosome and they did not test the fourth chromosome).

We compared the spectrum of modifying Deficiencies obtained in our screen with the 59 interacting Deficiencies identified in their screen. The Deficiencies identified were quite different, likely reflecting the distinct methods used to examine interaction and the different tissues involved. These differences illustrate the benefit of taking a variety of genetic approaches to modifier screens and emphasize that no one screen will identify all or even most potential interactors. Most interacting Deficiencies identified in their screen did not interact in our screen; for 39 of their interacting deficiencies, the percentage of suppression in our screen was <3%. Eight of their interacting Deficiencies were weak interactors in our screen [Df(2L)sc19-8, Df(2L)prd1.7, Df(2L)J32, Df(2L)H20, Df(3L)vin5, Df(3L)ZN47, Df(3R)crb87-5, and Df(3R)Hu]. Four interacting Deficiencies from their screen, Df(3L)Spd, Df(3L)Cat, Df(3R)D1-BX12, and Df(3R)p712, were strong interactors in our screen. Within two of these latter regions we identified interactors: hid from Df(3L)Cat and puc from Df(3R)p712.

Even in cases where the two screens identified the same Deficiency, it is not clear that the same gene is responsible. First, in several cases different subsets of overlapping Deficiencies interacted in the two screens. Second, GREAVES et al. 1999 identified interacting genes in many of their Deficiencies. In four cases, we also examined those candidates. Two were identified as interactors in our screen as well (DE-cadherin and zw3, used in our reconstruction experiments). In contrast, one of their interactors, wg, did not interact in our screen, even though a Deficiency that removes it, Df(3L)Spd, did interact. Likewise, components of the EGFR pathway interacted in their tests but not ours. Finally, in our hands naked complements Df(3L)Cat, thus ruling it out as our interactor in that region; in this region we identified hid as the interactor.

Hid activity, PCD, and the segment polarity phenotype:
It has been known for more than a decade that PCD plays an important role in the segment polarity phenotype resulting from inactivation of either the Hedgehog or Wg pathways (MARTINEZ-ARIAS 1985 ; KLINGENSMITH et al. 1989 ). Recently, Minden and colleagues carried out a detailed analysis of this process, quantitating cell death in wg, arm, gooseberry, and naked. They found that the elevation in cell death affected particular cells (PAZDERA et al. 1998 ). Since the first reports of cell death in segment polarity mutants, the machinery that drives PCD in embryos has begun to be identified. Homozygosity for the small chromosomal Deficiency, Df(3L)H99, blocks essentially all PCD (WHITE et al. 1994 ). Within this interval, three genes play roles in PCD: grim, reaper, and hid (reviewed in ABRAMS 1999 ). Ectopic expression of any of these can trigger PCD, but loss-of-function mutations are only available for hid.

Given the role of PCD in the segment polarity phenotype, it is perhaps not surprising that elimination of PCD would alter it. Several aspects of the effect of PCD reduction were unexpected, however. First, and most striking, the phenotypes of arm and wg mutants were very sensitive to relatively small changes in the dose of hid and the other cell-death promoters. For example, while heterozygosity for hid has no known effects on normal development, it strongly suppresses arm. Further reductions in the levels of hid or the other cell-death regulators had no additional effect on arm, suggesting that reducing the Hid dose by half eliminated the relevant ectopic PCD that occurs in an arm mutant. The wg phenotype was also suppressed in a highly dose-sensitive fashion, but in a different dosage range. A 50% reduction of hid caused slight but detectable effects, a 50% reduction in all three death promoters caused greater suppression, while homozygosity for the deletion removing all three genes resulted in the strongest wg suppression.

Recent observations regarding the role of Hid in PCD in the eye may explain this. Signaling through the ras/mitogen-activated protein kinase (MAPK) pathway promotes cell survival by antagonizing Hid (BERGMANN et al. 1998 ; KURADA and WHITE 1998 ). These authors suggested that Hid serves as a rheostat, with its levels determining the probability of PCD. They further suggest that Hid activity has to exceed a threshold to trigger PCD; the accumulation of hid mRNA in cells that are not programmed to die is consistent with this (GRETHER et al. 1995 ). Our observations further support this model. Wg signaling may normally antagonize Hid, potentially by regulating its expression. In embryos where Wg signaling is attenuated, elevated Hid activity may trigger PCD when it rises above a critical threshold. A threshold model could explain why the segment polarity phenotype is so sensitive to the dose of Hid and its partners.

Another surprise was the qualitative difference in the effect of cell death reduction on wg and arm mutants. While the resulting cell number was likely increased in both double-mutant genotypes in the arm; hid double mutant, the reduction in PCD restored an almost wild-type-length cuticle, while in the wg; hid double mutant, the increase in cell number was not reflected in an increase in cuticle length. The reason for this remains mysterious. One possible explanation for this discrepancy is the difference in the degree to which Wg signal is compromised in the two situations and the embryonic stage at which this disruption occurs. In the wg null, Wg signaling is totally eliminated from the beginning of development. In contrast, perdurance of maternal Arm substantially rescues early defects in Wg signaling in arm zygotic nulls (KLINGENSMITH et al. 1989 ). arm mutants remain more normal in morphology than wg mutants through the onset of germband retraction and retain remnant denticle diversity. Thus when one eliminates PCD in an arm mutant a more normal pattern is restored. The difference in amount and timing of Wg signaling in the two backgrounds may also explain why arm mutants are affected by smaller alterations in Hid level. The remaining Wg signaling in an arm zygotic mutant may promote cell survival to some extent, meaning that a smaller reduction in Hid activity prevents ectopic PCD.

We were also surprised that reduction in cell death alleviated arm's dorsal closure defect. We previously suspected that this defect was due solely to Arm's role as a catenin. However, recent data suggest that dorsal closure is promoted by Wg signaling (MCEWEN et al. 2000 ). We now suspect that defects in Wg signaling and catenin function combine to block dorsal closure in arm mutants. Restoring either rescues the arm dorsal closure defect. However, blocking PCD alone should not restore Wg signaling or catenin function. Perhaps the excess cell death in the head region or in the amnioserosa of an arm mutant contribute to its dorsal closure defect.

Presenilins and Arm function:
While evaluating the effectiveness of our screen, we tested a variety of candidate genes, including some that mapped within noninteracting Deficiencies. Heterozygosity for one of these, Dpresenilin, strongly suppressed arm. Presenilins form a family of multipass transmembrane proteins that were first identified because missense mutations in two human Presenilins cause early onset familial Alzheimer disease (FAD; reviewed in HAASS and DE STROOPER 1999 ; NISHIMURA et al. 1999B ). The cell biological function of Presenilins and how dysfunction contributes to disease remain controversial. Genetic data in Caenorhabditis elegans and Drosophila implicate Presenilins in the function of Notch proteins, most likely via effects on protein processing. Likewise, human Presenilin mutations affect proteolytic processing of the plaque protein Aß; this may lead to pathology (reviewed in HAASS and DE STROOPER 1999 ). Recently, it was found that both ßcat and other Arm repeat proteins such as -catenin associate with Presenilins in vivo. The function of this interaction remains confusing. ZHANG et al. 1998 reported that wild-type Presenilin stabilizes ßcat and that this is abrogated by missense mutations found in FAD patients, and NISHIMURA et al. 1999A reported that presenilin missense mutant cells from FAD patients have less nuclear ßcat. These data support a role for Presenilins as positive regulators of Wnt signaling via Arm/ßcat. In contrast, both MURYAMA et al. (1998) and KANG et al. 1999 report that overexpression of wild-type Presenilin destabilizes ßcat; Kang et al. further show that ßcat is stabilized in both Presenilin1 null fibroblasts or if FAD mutants of Presenilin1 are overexpressed, while Muryama et al. demonstrate that a Wnt-responsive promoter is downregulated by Presenilin overexpression. These data support a conclusion opposite from that above, in which wild-type Presenilins negatively regulate Wnt signaling. Finally, GEORGAKOPOULOS et al. 1999 suggest that the presenilin-ßcat complex includes cadherins, in contravention of most other data. Our genetic data are most consistent with a model in which Presenilins negatively regulate Wg signaling (Fig 3) either directly or indirectly by binding Arm/ßcat or by influencing adherens junction assembly. Clearly much work remains to differentiate between the different possible mechanisms.

	APPENDIX A

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION APPENDIX A APPENDIX B LITERATURE CITED

Table 3

	APPENDIX B

TOP ABSTRA