Genetic Modifier Screens in Drosophila Demonstrate a Role for Rho1 Signaling in Ecdysone-Triggered Imaginal Disc Morphogenesis

Corresponding author: Carl S. Thummel, University of Utah, 15 N. 2030 East, Rm. 5100, Salt Lake City, UT 84112-5331., carl.thummel{at}genetics.utah.edu (E-mail)

Communicating editor: T. SCHÜPBACH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Drosophila adult leg development provides an ideal model system for characterizing the molecular mechanisms of hormone-triggered morphogenesis. A pulse of the steroid hormone ecdysone at the onset of metamorphosis triggers the rapid transformation of a flat leg imaginal disc into an immature adult leg, largely through coordinated changes in cell shape. In an effort to identify links between the ecdysone signal and the cytoskeletal changes required for leg morphogenesis, we performed two large-scale genetic screens for dominant enhancers of the malformed leg phenotype associated with a mutation in the ecdysone-inducible broad early gene (br¹). From a screen of >750 independent deficiency and candidate mutation stocks, we identified 17 loci on the autosomes that interact strongly with br¹. In a complementary screen of 112,000 F₁ progeny of EMS-treated br¹ animals, we recovered 26 mutations that enhance the br¹ leg phenotype [E(br) mutations]. Rho1, stubbloid, blistered (DSRF), and cytoplasmic Tropomyosin were identified from these screens as br¹-interacting genes. Our findings suggest that ecdysone exerts its effects on leg morphogenesis through a Rho1 signaling cascade, a proposal that is supported by genetic interaction studies between the E(br) mutations and mutations in the Rho1 signaling pathway. In addition, several E(br) mutations produce unexpected defects in midembryonic morphogenetic movements. Coupled with recent evidence implicating ecdysone signaling in these embryonic morphogenetic events, our results suggest that a common ecdysone-dependent, Rho1-mediated regulatory pathway controls morphogenesis during the two major transitions in the life cycle, embryogenesis and metamorphosis.

MORPHOGENETIC movements define the body plan of metazoan animals. Gastrulation, neural tube formation, limb development, and organogenesis all depend on precisely timed, coordinated cell shape changes and cell rearrangements. In certain developmental contexts, endocrine signals provide temporal cues and also aid in the proper coordination of these morphogenetic events. For example, estrogen is required for mammary epithelial growth and ductal morphogenesis (BOCCHINFUSO et al. 2000 ), whereas thyroid hormone coordinates the massive tissue rearrangements that occur during amphibian metamorphosis (TATA 1999 ). Despite the importance of endocrine signaling in developmental programs, however, the mechanisms by which hormonal signals are transduced to the cellular machinery required for morphogenesis remain largely undefined.

Development of the adult leg in Drosophila provides an ideal model system for characterizing the molecular mechanisms of hormone-triggered morphogenesis. In Drosophila, the adult legs are derived from imaginal discs that are specified during embryogenesis and set aside as discrete clusters of diploid cells that undergo extensive proliferation and patterning during larval development. At the end of the third larval instar each of the six leg imaginal discs consists of a single-layered columnar epithelium that is covered and apposed by a squamous peripodial epithelium. Transformation of this disc epithelium into an immature adult leg is triggered by a pulse of 20-hydroxyecdysone (hereafter referred to as ecdysone), the steroid hormone that directs the major developmental transitions in the Drosophila life cycle (MANDARON 1970 ; FRISTROM et al. 1973 ; RIDDIFORD 1993 ). In response to the late larval ecdysone pulse, the leg imaginal discs elongate in the proximal-distal axis as the animal pupariates and initiates metamorphosis (FRISTROM and FRISTROM 1993 ; VON KALM et al. 1995 ). The leg imaginal discs evert rapidly at 5 hr after puparium formation, bringing them to the outside of the puparium (ROBERTSON 1936 ; WARD et al. 2003 ). The proximal regions of the discs then fuse with other thoracic and cephalic discs to contribute to the formation of a rudimentary adult fly. Remarkably, leg elongation and eversion can be recapitulated in cultured discs that are exposed to physiologically relevant levels of ecdysone, demonstrating a key role for the hormone in coordinating these morphogenetic events (MARTIN and SCHNEIDER 1978 ).

Ecdysone exerts its effects primarily at the level of gene regulation (ASHBURNER et al. 1974 ). Ecdysone binds to its receptor, a heterodimer of the Ecdysone receptor (EcR) and Ultraspiracle, directly inducing the transcription of primary-response early genes (RIDDIFORD et al. 2000 ). Some of these early genes encode transcription factors that regulate large batteries of secondary-response late genes, thought to direct the appropriate spatial and temporal biological responses to the hormone (THUMMEL 1996 ). One such ecdysone-inducible early gene is the Broad-Complex (BR-C), which encodes a family of zinc-finger transcription factors (DIBELLO et al. 1991 ). Genetic studies have defined three distinct genetic functions for BR-C, of which the broad (br) function is essential for leg imaginal disc morphogenesis (BELYAEVA et al. 1980 ; KISS et al. 1988 ). Specifically, the leg discs in amorphic br⁵ mutant prepupae fail to elongate or evert and appear to arrest at a stage similar to that of a wild-type disc at puparium formation, although the animal continues to develop and makes an apparent attempt to pupate at 18 hr after puparium formation (KISS et al. 1988 ; WARD et al. 2003 ). In contrast, mutations of the hypomorphic br¹ allele display only a weakly penetrant malformed leg phenotype in adult flies (KISS et al. 1988 ).

A number of molecular, biochemical, and genetic approaches have been employed to characterize leg disc morphogenesis. These studies have revealed that elongation and eversion of the leg imaginal discs occur in the absence of cell proliferation, largely in response to changes in cell shape (GRAVES and SCHUBIGER 1982 ; CONDIC et al. 1991 ). These directed cell shape changes appear to contribute to the elongation of the disc in the proximal-distal axis while affecting a contraction in the circumferential dimension, effectively converting the flat imaginal disc into a rudimentary adult leg over the course of several hours (VON KALM et al. 1995 ). A central role for the actin cytoskeleton in driving leg morphogenesis is supported by earlier work by FRISTROM and FRISTROM 1975 demonstrating that ecdysone-induced elongation and eversion is reversibly inhibited by cytochalasin B. Several studies have also implicated an important role for proteases in imaginal disc morphogenesis during prepupal development (POODRY and SCHNEIDERMAN 1971 ; FEKETE et al. 1975 ; PINO-HEISS and SCHUBIGER 1989 ; BIRR et al. 1990 ; FESSLER et al. 1993 ).

A genetic approach for investigating imaginal disc morphogenesis was employed by BEATON et al. 1988 , who took advantage of the sensitized genetic background provided by the hypomorphic BR-C allele, br¹. Screening through a collection of recessive mutations that produce leg defects, they identified Stubble/stubbloid (Sb/sbd), which encodes an apparent type II transmembrane serine protease (APPEL et al. 1993 ), as a br¹-interacting gene. Subsequently, GOTWALS and FRISTROM 1991 conducted a small-scale screen of 19,000 randomly mutagenized F₁ animals to identify dominant modifiers of the malformed leg phenotype associated with this mutation. They identified one allele each of Sb/sbd and zipper (zip), which encodes nonmuscle myosin heavy chain. The identification of a chemomechanical motor protein that facilitates contraction of the actin cytoskeleton and an extracellular protease demonstrated the utility of their screen and supported the notion that events at the actin cytoskeleton are critically important for leg morphogenesis.

In this study we expand upon the original Gotwals and Fristrom screen to identify links between the ecdysone signal and the cytoskeletal machinery that drives leg morphogenesis. Two independent screens were undertaken for dominant enhancers of the malformed leg phenotype associated with the br¹ mutation, using either a collection of autosomal chromosomal deficiencies or random methanesulfonic acid ethyl ester (EMS)-generated mutants. From these screens we identified 17 loci on the autosomes that interact with br¹ and we isolated 26 EMS-induced br¹-interacting mutations. Included in this collection of br¹-interacting genes are those encoding the small GTPase Rho1, the Drosophila serum response factor (SRF) transcription factor Blistered (Bs), a cytoplasmic isoform of Tropomyosin1 (cTm), and the Sb/sbd protease. These results imply an important role for Rho1 signaling in leg disc morphogenesis, a notion that we support by genetic interaction studies between the EMS-generated mutations and previously characterized mutations in the Rho1 signaling pathway. In addition, we observe defects in midembryonic morphogenetic movements in animals bearing some of the EMS-induced mutations, suggesting that common regulatory mechanisms drive morphogenesis at different stages in the life cycle.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Drosophila stocks:
All Drosophila stocks were maintained on corn meal/yeast/molasses/agar media in a room maintained at a constant temperature of 21°. The deficiency and P-element-insertion stocks used in this study were obtained from the Bloomington Drosophila Stock Center at Indiana University (Bloomington, IN). The zip^E(br), Rho^J3.8, and Rho^E3.10 stocks were obtained from S. Halsell (James Madison University; GOTWALS and FRISTROM 1991 ; HALSELL et al. 2000 ). The RhoGEF2^11-3 and zip^33-1 stocks were obtained from L. von Kalm (University of Central Florida; BAYER et al. 2003 , this issue). The cTm^eg9 and cTm^er4 stocks were obtained from D. Kiehart (Duke University; ERDELYI et al. 1995 ). Unless otherwise stated, genetic experiments were conducted in a room controlled at a constant temperature of 21° because the genetic interactions observed between br¹ and both Sb/sbd and zip are cold sensitive (BEATON et al. 1988 ; GOTWALS and FRISTROM 1991 ).

EMS mutagenesis and screening:
Twenty cohorts, each consisting of 30 3- to 5-day-old br¹ males, were treated with 25 mM EMS. Each cohort was mated to 30 br¹ virgin females. Mutagenized males were subsequently mated to a second set of br¹ virgin females to produce two broods of progeny. All progeny were maintained in bottles at 21°. In the F₁ generation, all flies showing malformed legs were backcrossed to br¹ males or females as appropriate. In the F₂ and subsequent generations, sibling flies showing malformed legs were mated inter se in an effort to remove unlinked second-site mutations by free recombination. Starting at the F₅ generation, inter se crosses producing 10% or more malformed progeny were mapped and balanced using the mapping stocks br¹;Sco/Cyo and br¹;T(2;3)ap^Xa/TM6B. The br¹ mutation was subsequently outcrossed leaving balanced Enhancer of br [E(br)] stocks. Complementation tests were conducted on E(br) stocks that mapped to the same chromosome.

Deficiency screen:
Dominant genetic interaction tests with br¹ were performed by mating five to seven br¹ virgin females to five to seven deficiency- or specific mutation-bearing heterozygous males in vials containing standard Drosophila medium. After 3 days the adults were transferred to fresh vials and, subsequently, to a third vial after two additional days. Newly eclosing F₁ flies were separated by genotype and examined for malformed legs each day for a total of 10 days. Second-site noncomplementation (SSNC) tests with br⁵ were performed in a similar manner, except that y br⁵/Binsn females were used and the crosses were maintained in an incubator at 25°.

EMS screen:
Dominant genetic interaction tests between br¹ and the E(br) mutations were performed at 21° as described above, using males of the genotype br¹/Y;E(br)/CyO or br¹/Y;E(br)/TM6B. Two vials were scored for each E(br) line in the experiment reported in Table 3. SSNC tests between br⁵ and the E(br) mutations were performed at 25° as described above using y br⁵/Binsn virgin females and w¹¹¹⁸/Y;E(br)/CyO or w¹¹¹⁸/Y;E(br)/TM6B males. SSNC tests between E(br) mutations and Rho1 pathway mutations were performed by mating five to seven w¹¹¹⁸;E(br)/CyO or w¹¹¹⁸;E(br)/TM6B virgin females to five to seven heterozygous mutant males. The crosses were maintained at 21° and three vials for each cross were examined in the manner described above. In some cases the reciprocal cross was also performed.

Complementation tests were performed by mating five to seven w¹¹¹⁸;E(br)/CyO or w¹¹¹⁸;E(br)/TM6B virgin females to five to seven heterozygous mutant males. Crosses were maintained in an incubator at 25°. The adults were transferred to fresh vials after 3 days. Newly eclosing F₁ flies were separated into genotypic classes and counted each day for a total of 7 days. In some cases the reciprocal cross was also performed.

Meiotic mapping of the E(br) mutations was performed by mating br¹;E(br)/CyO or br¹;E(br)/TM6B virgin females to br¹/Y;b pr c px sp or br¹/Y;ru th st ri roe p e ca males, respectively. In the F₁ generation, virgin females of the genotypes br¹;E(br)/b pr c px sp or br¹;E(br)/ru th st ri roe p e ca were mated to br¹/Y;b pr c px sp or br¹/Y;ru th st ri roe p e ca males, respectively. In the F₂ generation, all animals bearing malformed legs [and therefore likely containing the E(br) mutation] were examined for the presence of recessive markers. Recombination distances were calculated between the E(br) mutation and each recessive marker, and the map position of the E(br) mutation was determined using values from the two or three closest recessive markers. From 150 to 400 informative recombination events were scored for each E(br) mutation.

Lethal-phase analysis:
Embryonic lethality was determined by collecting 0- to 2-hr embryos from E(br)/Cyo, P{w⁺, ActGFP} or E(br)/TM6B, P{w⁺, UbiGFP} stocks. The embryos were aged at 25° for 15 hr and dechorionated in 50% bleach, and homozygous E(br) embryos were identified and separated on the basis of absence of green fluorescent protein (GFP) expression. The homozygous mutant embryos were allowed to develop at 25° until 48 hr after egg laying, at which point the dead embryos were counted and mounted in Hoyer's medium (ASHBURNER 1989 ) on microscope slides. Embryonic lethality was calculated as (number of dead embryos/number of total mutant embryos) x 100. Thirty to 60 E(br) mutant embryos were tested in each experiment and all E(br) stocks were tested in triplicate. The mean and standard error (SE) of embryonic lethality were calculated for each E(br) stock. Cuticle preparations were examined for terminal phenotypes. Larval lethality was determined by collecting non-GFP-expressing first instar larvae derived from E(br)/Cyo, P{w⁺,ActGFP} or E(br)/TM6B, P{w⁺,UbiGFP} stocks. Seventy-five to 100 mutant larvae were placed into vials containing standard Drosophila medium that had been lightly tilled and overlaid with fresh yeast paste. The larvae were aged for 7–10 days at 25°, at which point the number of pupae were counted. Larval lethality was calculated as [(number of total larvae - number of pupae)/number of total larvae] x 100. Experiments were done in triplicate and the mean and SE of larval lethality were calculated for each E(br) stock.

Embryonic and adult specimen preparations:
The devitellinized embryonic cuticles shown in Fig 4 were prepared by collecting unhatched embryos from E(br)/CyO or E(br)/TM6B stocks 48 hr after egg laying at 25°. The embryos were dechorionated in 50% bleach and devitellinized in a 1:1 mixture of heptane: 90% MeOH, 50 mM EGTA, pH 8.0. The embryos were then mounted in One-Step mounting medium (2:1:1 glacial acetic acid:CMCP10:85% lactic acid) on microscope slides. Adult leg cuticles were prepared by dissecting leg pairs from the third thoracic segment of w¹¹¹⁸, br¹;E(br)/CyO or br¹;E(br)/TM6B males in PBS, clearing them overnight in 10% KOH, and mounting them in Euporal on microscope slides. Images of the embryonic and adult leg cuticles were captured on either a Cool Snap or a SensiCamQE high performance digital CCD camera mounted on a Zeiss Axiophot microscope. Images of the dorsal thoraces from sbd^E(br)228 and sbd^E(br)228/+ animals were captured on a Cool Snap digital CCD camera mounted on a Leica stereomicroscope. Indirect immunofluorescence analysis of E(br)165 embryos was performed by collecting embryos from E(br)165/CyO parents at 25° for 2 hr, aging the embryos for an additional 19 hr, and then fixing and staining the embryos as described (FEHON et al. 1991 ). Anti-Coracle mAb C615-16B was used at a dilution of 1:250. Optical sections were captured with a Bio-Rad (Richmond, CA) MRC1024 confocal laser mounted on a Zeiss Axioplan microscope. All digital images were cropped and adjusted for brightness and contrast in Adobe Photoshop.

View larger version (40K): In this window In a new window Download PPT slide   Figure 1. Chromosomal deficiencies that enhance the br1 malformed leg phenotype. Shown are 133 deficiencies from the primary screen that were tested for dominant enhancement of the br1 malformed leg phenotype. Open boxes indicate the chromosomal extent of deficiencies that show >20% malformed legs when heterozygous in a br1 background (br1/Y;Df/+). Noninteracting deficiencies are represented by solid boxes. Polytene chromosome images are from LEFEVRE 1976 .

View larger version (91K): In this window In a new window Download PPT slide   Figure 2. The bristle phenotype and molting defects associated with sbd mutations. Brightfield photomicrographs are shown of (A) the dorsal thorax from an sbdE(br)228/+ heterozygous adult and (B) an sbdE(br)228 homozygous adult. Note the short, thick, and barb-ended scutellar bristles on the sbdE(br)228 thorax relative to the long and thin wild-type bristles on the sbdE(br)228/+ thorax (arrows). Depicted below are brightfield photomicrographs of cuticle preparations showing the anterior regions of (C) wild-type and (D) sbdE(br)623 mutant third instar larvae. The mutant larva has retained an extra set of mouth hooks and head skeleton (arrowheads) along with some attached cuticle from the previous molt (arrow). sbdE(br)623 mutants also show an unusual sclerotization of the anterior epidermis.

View larger version (36K): In this window In a new window Download PPT slide   Figure 3. Representative leg phenotypes of E(br) mutants. Brightfield photomicrographs are shown depicting cuticle preparations of adult legs from the third thoracic segment. (A) A w1118/Y leg showing normal morphology. (B) A br1/Y; sbdE(br)623/+ leg showing a short, fat femur and tibia. Fourteen of the E(br) mutations show this interaction phenotype. (C) A br1/Y;E(br)155/+ leg showing bulbous and bent tarsal segments. E(br)165 also shows this interaction phenotype. (D) A br1/Y;E(br)121/+ leg showing a bent tibia. Four of the E(br) mutations show this interaction phenotype. Femur (fe), tibia (ti), and tarsal segments 1–5 (ta) are labeled.

View larger version (87K): In this window In a new window Download PPT slide   Figure 4. Defective midembryonic morphogenetic events in E(br) mutants. Brightfield photomicrographs of cuticle preparations from (A) wild type, (B) RhoE(br)233, (C) RhoE(br)246, (D) sbdE(br)623, and (E) E(br)155 are shown. (F) Confocal optical section of an E(br)165 mutant embryo stained with an antibody against the septate junction protein Coracle. All animals are shown with anterior to the left and dorsal up. (B and C) Rho1 mutants show completely penetrant embryonic lethality characterized by a dorsal anterior open phenotype, indicative of a defect in head involution. Many Rho1 mutants also show exaggerated curvature of the ventral surface caused by a mild defect in germband retraction. (D) More than 20% of sbdE(br)623 mutants die as embryos, of which 19% display dorsal anterior holes. (E) E(br)155 mutant animals show nearly completely penetrant embryonic lethality with 10% of the embryos showing a dorsal anterior hole and 50% showing a dorsal hole indicative of a defect in dorsal closure (not shown). (F) The head is completely involuted but the dorsal surface is not closed in E(br)165 mutants as indicated by the sharp boundary of Coracle staining near the dorsal surface (arrowheads) and the extruded hindgut (arrow).

RNA isolation and Northern blot analysis:
Progeny from a cross of y br⁵/Binsn x Binsn/Y were staged on standard Drosophila media supplemented with 0.1% bromophenol blue as described in ANDRES and THUMMEL 1994 . Total RNA was isolated by direct phenol extraction from leg imaginal discs dissected from staged y br⁵/Y and Binsn/Y males. Approximately 9 µg of total RNA per sample was separated by formaldehyde agarose gel electrophoresis and transferred to a nylon membrane. The membrane was hybridized and stripped as described by KARIM and THUMMEL 1991 . Generation of probe fragments for BR-C (core), ImpE3, and rp49 is described in ANDRES and THUMMEL 1994 and for Sb in D'AVINO and THUMMEL 1998 . A probe to detect Rho1 was generated by PCR amplification of a 581-bp fragment from a late third instar larval random-primed cDNA collection (T. KOZLOVA, personal communication) using the primer set 5'-AACTTCCAATGACGACGATTCGC-3' and 5'-GCAAAAGGCATCTGGTCTTCTTCC-3'. A probe to detect bs was generated by PCR amplification of a 429-bp fragment from a late third instar larval random-primed cDNA collection (T. KOZLOVA, personal communication) using the primer set 5'-CGTTGAGTGTTTTCTGTGTGG-3' and 5'-CTGGGAGGCGTGCTGTGGG-3'. A probe to detect Rho kinase was generated by PCR amplification of a 569-bp fragment from cDNA LD36258 (Research Genetics, Birmingham, AL) using the primer set 5'-CGAAATAAAATAAGTGCAACGCGC-3' and 5'-CATTGCTGGACACCACTTGGCC-3'. A probe to detect RhoGEF2 was generated by PCR amplification of a 588-bp fragment from cDNA SD04476 (Research Genetics) using the primer set 5'-CGTCGTGTGCGTGTTGATGGCG-3' and 5'-GACGGGCCTGCAGATGTCGC-3'. Specific probes were labeled by random priming of gel-purified fragments (Stratagene, La Jolla, CA).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Screening autosomal deficiency stocks for genetic interactions with br:
As a first step toward identifying loci that function with br to direct leg morphogenesis, we screened through an ordered collection of chromosomal deficiency stocks from the second and third chromosomes. We screened for deficiencies that exert a dominant increase in the penetrance of the malformed leg phenotype associated with the weakly hypomorphic br¹ allele. In control experiments conducted prior to the screen, we found that br¹ animals (hemizygous males and homozygous br¹ females), maintained at 21°, display malformed legs at a low frequency of 0.4% (n = 3629). For the purpose of this screen, we arbitrarily considered an interaction to be significant if 20% or more of the br¹/Y;Df/+ animals displayed at least one malformed second or third leg, representing a 50-fold increase over the br¹ background.

Out of an initial collection of 154 autosomal deficiency stocks, we tested 133 stocks for genetic interactions with br¹. The remaining deficiency stocks could not be tested either because of unmarked duplications that prevented unambiguous identification of progeny classes or due to the presence of Sb/sbd alleles on the deficiency chromosome that would obscure a possible genetic interaction with br¹ (BEATON et al. 1988 ). Of the 133 deficiency stocks screened, 43 reproducibly enhance the malformed leg phenotype associated with br¹ (Fig 1, open boxes). To confirm these interactions and refine the genomic regions containing the putative br¹-interacting loci, we tested >175 additional deficiency stocks that were predicted to overlap with br¹-interacting deficiency stocks identified in the primary screen. Overall, we found 64 deficiency stocks that enhance the br¹ leg phenotype to >20% (Table 1). We also tested >425 P-element-insertion stocks and individual mutations in candidate genes in an attempt to identify single loci that could account for the br¹ interaction detected with the deficiency stocks. From these studies we found 17 br¹-interacting loci and identified mutations in three genes that act as dominant enhancers of the malformed leg phenotype of br¹.

Rho1:
Five overlapping deficiency stocks, Df(2R)Jp1, Df(2R)Jp4, Df(2R)Jp5, Df(2R)Jp7, and Df(2R)Jp8, enhance the br¹ malformed leg phenotype to frequencies ranging from 23 to 78% (Table 1). Each of these deficiencies is predicted to remove sequences in the 52F cytological region. We tested five P-element-insertion stocks from this interval and found one, l(2)k02107, that also strongly enhances br¹ (Table 2). Plasmid rescue of genomic DNA adjacent to the P-element-insertion site revealed that l(2)k02107 is an insertion into an intron of the Rho1 gene, confirming a subsequent report that this P-element mutation is a bona fide allele of Rho1 (MAGIE et al. 1999 ). To verify the dominant genetic interaction between Rho1 and br¹, we tested Rho1^J3.8 and Rho1^E3.10 and found that both alleles also strongly enhance the malformed leg phenotype of br¹ (Table 2). Interestingly, HALSELL et al. 2000 found that whereas Rho1 fails to complement Df(2R)Jp4 and Df(2R)Jp8, it fully complements Df(2R)Jp1. Our finding of a strong dominant genetic interaction between br¹ and Df(2R)Jp1 therefore suggests the presence of another br¹-interacting locus in the cytogenetic interval 51D3-52F9. A potential candidate gene mapping to this region is myosin light chain kinase. A direct test of this candidate, however, awaits the identification of a specific mutation in this gene.

blistered:
Df(2R)Px2 (60C05;60D09-10) is a very strong dominant enhancer of the br¹ malformed leg phenotype, producing malformed legs at a frequency of 50% (Table 1). Crosses with an overlapping deficiency, Df(2R)Px1 (60B08-10;60D01-02), produced malformed legs at a frequency of 19% (n = 112) in the br¹ genetic background, mapping the br¹-interacting locus to the 60C5-D1 interval. Mutations in blistered (bs), which encodes the Drosophila SRF transcription factor, fail to complement these two deficiencies, and bs² was previously shown to enhance the malformed leg phenotype of br¹, suggesting that bs might contribute to the interaction seen with these deficiencies (GOTWALS and FRISTROM 1991 ; AFFOLTER et al. 1994 ). To confirm these results, we tested three alleles of bs for a dominant genetic interaction with br¹ and found that bs^k03267, a P-element-insertion allele, displays a frequency of 15% malformed legs, whereas the P-element-insertion allele bs^k03267 displays a weaker interaction of 9% (Table 2). The genetic interactions with both alleles are cold sensitive, showing a dramatic increase in the percentage of animals displaying malformed legs at 18° (Table 2). bs^ba fails to interact with br¹ at either temperature (Table 2). The allele specificity and cold sensitivity of these genetic interactions are consistent with earlier observations (GOTWALS and FRISTROM 1991 ). We conclude that bs is a dominant enhancer of br¹ and that its absence in Df(2R)Px2 contributes to the genetic interaction we observe between this deficiency and br¹.

cytoskeletal Tropomyosin:
Df(3R)ea (88E07-13;89A01) enhances the malformed leg phenotype of br¹ with a frequency of 62% (Table 2), a particularly strong enhancement. We were unable to refine the interval containing the br¹-interacting locus using available deficiencies and therefore tested 20 P-element-insertion stocks that map within the interval from 88E1 to 89A9. We found one stock, l(3)2299, that acts as a dominant enhancer of br¹ and also fails to complement Df(3R)ea (Table 2; TETZLAFF et al. 1996 ). l(3)2299 results from a P-element insertion into the twentieth codon of a cytoskeletal-specific exon of Tm1 (TETZLAFF et al. 1996 ). To confirm this interaction, we tested a maternal-effect mutation in cytoskeletal Tropomyosin (cTm; Tm1^eg9; ERDELYI et al. 1995 ) and found that it weakly enhances br¹ (Table 2), although a lethal excision allele generated from this mutation, cTm^er4 (ERDELYI et al. 1995 ), does not enhance the br¹ malformed leg phenotype (Table 2). In Drosophila there are two tandem tropomyosin genes (Tm1 and Tm2) that produce several muscle-specific isoforms and one cytoskeletal-specific isoform due to alternative splicing. We therefore tested a hypomorphic allele of Tm2 that specifically affects jump and indirect flight muscles and found that it does not act as a dominant enhancer of the malformed leg phenotype of br¹ (Table 2). We conclude that br¹ interacts with cTm (Tm1) and that this interaction contributes to the observed dominant enhancement of br¹ by Df(3R)ea.

Other interacting loci defined by deficiencies:
In addition to the genetic interactions of br¹ with Rho1, bs, and cTm, we predict that at least seven loci on the second chromosome and at least seven loci on the third chromosome harbor br¹-interacting genes (Table 1). This prediction is based on the finding of two or more overlapping br¹-interacting deficiency stocks that were derived from different parental chromosomes (Table 1). In five cases, each of the overlapping deficiency stocks produce an interaction phenotype at a frequency of >40%, >100-fold over the br¹ background. Of these, one interval is particularly noteworthy. Df(3R)D6, Df(3R)D7, and Df(3R)p712 compose a set of three overlapping deficiencies that remove genomic sequences from 84D04 to 84F02. Efforts to identify a br¹-interacting gene within this interval have thus far failed, although we have tested 36 deficiency stocks, 16 P-element-insertion stocks, and 21 representative EMS alleles derived from a saturation mutagenesis screen of this region (BAKER et al. 1991 ). Of the remaining candidate genes within this interval, one gene in particular stands out, ImpE3. This gene was originally isolated in a molecular screen for genes that encode ecdysone-inducible cell-surface or secreted imaginal disc proteins (NATZLE et al. 1986 ). Subsequent experiments confirmed that ImpE3 is induced by ecdysone and strongly expressed in imaginal discs (MOORE et al. 1990 ). At present there are no mutant alleles of ImpE3 to test for a genetic interaction with br¹, but Northern blot analysis of ImpE3 expression in br mutant imaginal discs has demonstrated a regulatory interaction between these two genes (see below).

br⁵ second-site noncomplementation screen with deficiency stocks:
We conducted SSNC tests between a collection of 133 autosomal deficiency stocks and br⁵, an amorphic allele of br, to determine whether a stronger allele might identify novel br-interacting genes. In pilot studies we were unable to detect malformed legs in br⁵/+ females and therefore set an arbitrary threshold for significant interaction at 5% malformed legs in the br⁵/+;Df/+ genotypic class. Surprisingly, we found only one deficiency that specifically and reproducibly interacts with br⁵ in this SSNC assay. Df(3R)DG2 (89E01-F04;92D03-06) displays malformed legs at a frequency of 11% (n = 62) when heterozygous in a br⁵/+ genetic background. Df(3R)DG2 also shows a very strong dominant genetic interaction with br¹ (Table 1).

br¹ dominant genetic interaction screen of EMS-treated animals:
There are two significant limitations when chromosomal deficiencies are used for a genetic interaction screen: incomplete coverage of the genome and the requirement for detecting an interaction with an amorphic allele. In an effort to overcome these limitations, we conducted an F₁ screen of EMS-treated animals. We mutagenized br¹ males, mated them to br¹ virgin females, and screened through 112,000 F₁ progeny for flies displaying the malformed leg phenotype. Malformed progeny were backcrossed to br¹ animals to generate stocks. These stocks were then maintained for several generations as inter se crosses, selecting for animals with malformed legs to remove unlinked second-site mutations by free recombination. We kept those stocks in which the inter se crosses produced 10% or more malformed progeny at the F₅ generation. From this screen we identified 26 mutations that map to a single chromosome and reproducibly enhance the br¹ leg phenotype. Two E(br) mutations map to the X chromosome, 9 map to the second chromosome, and 15 map to the third chromosome. We have analyzed 20 of these E(br) lines in detail (Table 3 and Table 4). Lethal complementation analyses revealed one complementation group consisting of six members that maps to the third chromosome, one complementation group with two members that maps to the second chromosome, and 12 E(br) mutations that complement every other E(br) mutation. Six of these map to the second chromosome and 6 map to the third chromosome.

Complementation analyses identify Rho1, sbd, and bs alleles as E(br) mutations:
As a first step toward identifying the br¹-interacting genes disrupted by the EMS mutations, we performed a large-scale complementation analysis between each of the E(br) lines and a panel of deficiencies and specific mutations. This set of stocks included representative br¹-interacting deficiencies from each of the intervals identified in the deficiency screen, as well as mutations in bs, Rho1, RhoGEF2, sbd, cTm, and zip. From these analyses we determined that the large complementation group on the third chromosome is allelic to Sb/sbd on the basis of the following four observations. First, E(br)20, E(br)48, E(br)448, E(br)536, and E(br)623 compose a single complementation group, and E(br)228 partially fails to complement these mutations for lethality. Second, E(br)20, E(br)536, and E(br)623 fail to complement sbd⁴⁵. Third, complementation tests between these same three E(br) lines and sbd¹⁰⁵ produce viable trans-heterozygous adults that show a stubble bristle phenotype. Fourth, E(br)228 is partially viable, producing homozygous mutant adults at 20% of the expected frequency, all of which show a completely penetrant stubble bristle phenotype (Fig 2B). All of the sbd^E(br) mutations show normal bristle morphology when heterozygous, classifying all six alleles as sbd with respect to the bristle phenotype (Fig 2A). Similar complementation analyses demonstrated that E(br)233 and E(br)246 are allelic to Rho1. Both mutations fail to complement one another, E(br)246 fails to complement Df(2R)Jp8 (52F05-09; 52F10-53A01), and both mutations fail to complement Rho1^k02107, Rho1^J3.8, and Rho1^E3.10. Finally, E(br)292 is allelic to bs on the basis of its failure to complement Df(2R)Px2 (60C05;60D09-10) and two bs alleles, bs^k03267 and bs^k07909.

Meiotic mapping of the E(br) mutations:
Our complementation studies revealed one instance of noncomplementation between an E(br) mutation and a br¹-interacting deficiency. E(br)420 fails to complement Df(3R)RD31 (89E02;90D) and Df(3R)DG2 (89E01-F04;91B01-02). It does, however, fully complement Df(3R)C4 (89E03-04;90A01-07), suggesting that the E(br)420 mutation is located in 89E02-04 or 90A-D. To verify this mapping, we performed meiotic mapping experiments with br¹;ru th st ri roe p e ca. We conducted these crosses in the br¹ genetic background to specifically map the br¹-interacting mutation rather than a lethal lesion that might be linked but not causative of the interaction (see MATERIALS AND METHODS). Recombination distance from thread (3-[43.2] on the recombination map) places E(br)420 at 59.9 on the recombination map, which approximately corresponds to 89F (FLYBASE 1999 ). Similarly, recombination distance from scarlet (3-[44]) places E(br)420 at 60.7 and from radius incompletus (3-[46.8]) at 60.8 on the recombination map, both of which correspond roughly to 90D. The meiotic mapping data therefore support the complementation data and place E(br)420 in 89E02-04 or 90A-D (Table 3).

In addition to E(br)420, we used meiotic mapping to locate the br¹-interacting mutations in E(br)24, E(br)65, E(br)121, E(br)155, and E(br)165 (Table 3). The mapping of E(br)121 to 3-[41-43] (corresponding to 70C-71C) is supported by the failure of E(br)121 to complement Df(3L)fz-CALS (70C02-06;70E01). We are currently attempting to confirm the meiotic mapping of the other four E(br) mutations through complementation studies and are beginning the initial mapping of the remaining five E(br) mutations.

Characterization of the E(br) mutations:
We undertook preliminary characterization of the E(br) mutations as a first step toward understanding the function of the affected genes. To this end we have: (1) determined the penetrance of the dominant genetic interaction with br¹ (Table 3), (2) tested for SSNC with an amorphic allele of br (Table 3), (3) conducted thorough lethal-phase and terminal phenotypic analyses of the E(br) mutations (Table 4), (4) conducted SSNC analyses with representative mutations in the Rho1 pathway (Table 5), and (5) performed Northern blot analysis of br¹-interacting genes on RNA collected from staged wild-type and br⁵ mutant imaginal discs (Fig 5). The results of these studies are reported below.

View larger version (88K): In this window In a new window Download PPT slide   Figure 5. ImpE3 transcription, but not that of Sb/sbd, is dependent upon br function. Total RNA isolated from collections of staged Binsn/Y (control) and br5/Y leg imaginal discs was fractionated by formaldehyde agarose gel electrophoresis and analyzed by Northern blot hybridization. The time in hours relative to puparium formation is depicted at the top. Hybridization to detect BR-C mRNA isoforms (BR-C Z1 isoform is indicated by an arrow; BR-C Z2, -Z3, and -Z4 isoforms are indicated by arrowheads) was used to follow changes in ecdysone titer. Hybridization to detect rp49 mRNA was used as a control for loading and transfer. The br5/Y -4-hr RNA sample is underloaded relative to the other samples.

Phenotypic analyses of interactions between br¹ and the E(br) mutations:
Individual E(br) mutations vary in their ability to enhance the br¹ malformed leg phenotype, ranging from 5% for sbd^E(br)228 to 100% for E(br)165 (Table 3). Eight E(br) mutations enhance br¹ with a penetrance of <20%, which is the threshold used for the deficiency screen. These results are, however, significant because the E(br) mutations were generated in a uniform genetic background, unlike the wide range of genetic backgrounds found in the deficiency stocks. Therefore, for example, the 5% malformations seen in br¹/Y; sbd^E(br)228/+ animals represent a 12-fold increase over br¹/Y that can be attributed primarily to the mutation. It is also noteworthy that the poorest interacting mutation is an allele of Sb/sbd, a known br¹-interacting gene (BEATON et al. 1988 ).

The malformed leg phenotypes in br¹/Y;E(br)/+ males can be classified into three distinct classes, as shown in Fig 3. Fourteen of the E(br) mutant lines produce legs that have short, fat femurs and tibias that are often kinked or twisted (Fig 3B; Table 3). Included in this group of mutations are all the sbd, Rho1, and bs alleles. In two E(br) lines, the interaction phenotype is predominantly restricted to the tarsal segments, which are shorter and fatter than those of wild type (Fig 3C; Table 3). In four E(br) lines, the interaction phenotype consists entirely of a moderate to strong bend in the mid-tibia that is not associated with an ectopic joint (Fig 3D; Table 3). In all cases, the defects manifest themselves in the third pair of legs and are often unilateral. Malformations of the second pair of legs occur much less frequently and are almost always associated with extreme malformations of the third pair of legs. E(br)420 is unique in producing malformations at a high frequency in the first pair of legs. In this case, the malformed legs show defects primarily in the tibia and tarsal segments, and occasionally the first pair legs are missing middle tarsal segments (data not shown). We have not, however, quantified these phenotypes. Finally, we occasionally observe wing malformations as a br¹ interaction phenotype with the E(br) mutations. We did not quantify these phenotypes because their penetrance is low and the expressivity is more varied than that of the malformed leg phenotypes.

We also outcrossed the br¹ allele and examined w¹¹¹⁸;E(br)/+ animals for the presence of malformed legs. This study revealed that five of the E(br) mutations show a semidominant malformed leg phenotype. E(br)24, E(br)121, E(br)155, E(br)333, and E(br)444 all show >5% malformed legs under these conditions (Table 3). In each case, however, the penetrance of malformed legs in the br¹ background is at least threefold higher. Interestingly, all four of the E(br) mutant lines that show the bent tibia malformed leg phenotype are also semidominant (Fig 3D; Table 3).

SSNC analyses between E(br) mutations and br⁵:
To test the specificity of the E(br) mutations we conducted SSNC experiments with br⁵. Whereas only one deficiency from the deficiency screen showed >5% malformed legs in this assay (see above), 10 of the E(br) mutations display >5% malformed legs when heterozygous in a br⁵/w¹¹¹⁸ genetic background (Table 3). Included in this collection are two sbd alleles, both Rho1 alleles, the bs allele and five of the unidentified E(br) mutations. The percentage of malformed legs seen with E(br)155 and E(br)444 in a br⁵/w¹¹¹⁸ genetic background, however, was similar to the level of malformed legs found in w¹¹¹⁸; E(br)/+ animals, arguing against a relevant genetic interaction.

Lethal-phase and terminal phenotypic analyses of E(br) mutants:
To characterize the function of the genes disrupted by the E(br) mutations, we conducted lethal-phase studies using balancer chromosomes that express GFP to unambiguously identify homozygous mutant embryos and larvae (Table 4). These experiments revealed that both Rho1^E(br)233 and Rho1^E(br)246 show completely penetrant embryonic lethality with nearly every embryo possessing a large dorsal anterior hole with the head skeleton extruded (Fig 4B and Fig C). In addition, 5–10% of these embryos possess a second cuticular hole often positioned near the posterior pole (data not shown), and many of the dead embryos display slight curvature of the ventral surface, indicating a mild defect in germband retraction (Fig 4C). These observations are consistent with phenotypic analyses describing the zygotic loss-of-function phenotypes for Rho1 mutations (MAGIE et al. 1999 ) and suggest that both of these new Rho1 mutations are strong loss-of-function alleles.

Lethal-phase analysis of bs^E(br)292 indicated a requirement for bs throughout the life cycle (Table 4). The predominant lethal period occurs during larval stages, although we also detected significant embryonic lethality. All of the mutant late embryos and larvae examined lack tertiary branching of the tracheal system (data not shown), consistent with strong loss-of-function mutations at the bs locus (GUILLEMIN et al. 1996 ). Examination of wings dissected from rare escapers that survive to the pharate adult stage revealed a tube wing phenotype identical to that found in known bs mutants (MONTAGNE et al. 1996 ). Interestingly, during the course of four independent embryonic lethal-phase experiments with bs^E(br)292, 18% of the mutant animals showed embryonic lethality (Table 4), of which 29% of the dead embryos displayed dorsal anterior holes and several others showed aberrant head skeletons (data not shown). Likewise, bs^E(br)292/Df(2R)Px2 hemizygous animals show a similar degree of embryonic lethality and >25% of the dead embryos show head skeleton defects or dorsal anterior holes (data not shown). Since these phenotypes had not been reported previously, we also examined the strong loss-of-function bs^k03267 allele; however, we did not recover any dead embryos exhibiting these phenotypes (data not shown). Additional experiments are required to determine if the novel embryonic phenotypes we observe in bs^E(br)292 mutants represent a stronger loss-of-function bs phenotype or an antimorphic or neomorphic phenotype associated with bs or are due to a closely linked second-site lethal mutation.

Lethal-phase analyses of the sbd^E(br) mutations revealed that five of the six mutants are predominantly larval lethal, whereas sbd^E(br)228 shows some larval and pupal lethality with 21% viable adults (Table 4). sbd^E(br)228 adults and adult escapers from trans-heterozygous combinations between sbd^E(br)228 and the other sbd^E(br) alleles show the short, thick, and barbed bristles characteristic of mutations at the Sb/sbd locus (Fig 2B). Most of these adult escapers also display severely malformed legs (data not shown). Unexpectedly, we found that mutant animals from all six sbd^E(br) alleles show defects in larval molting characterized by two complete sets of head skeleton. Occasionally, some cuticle from an earlier molt was found attached to the mouth hooks of dead mutant larvae (Fig 2D). sbd^E(br)48 and sbd^E(br)623 also display a unique phenotype characterized by excessive sclerotization of the anterior-most cuticle (Fig 2D). Interestingly, 9% of the dead embryos from sbd^E(br)48 and 19% of the dead embryos from sbd^E(br)623 show a dorsal anterior hole similar to that seen with Rho1 mutations (Fig 4D). These embryonic and larval phenotypes have not been described previously, although SPILLMAN and NOTHIGER 1978 reported early larval lethality in several sbd lines.

Characterization of the unidentified E(br) mutations revealed four mutant lines that display embryonic lethality characterized by defects in the midembryonic morphogenetic processes of dorsal closure and head involution (Table 4). E(br)24, E(br)65, and E(br)155 mutants show 25, 14, and 97% embryonic lethality, respectively. In all three E(br) lines at least 10% of the dead embryos display a dorsal anterior hole similar to that found in Rho1^E(br) mutants (compare Fig 4E with 4, B and C). In addition, we found that E(br)155 and E(br)165 mutants show a high penetrance of dorsal holes (Table 4 and Fig 4F). Specifically, nearly 50% of the E(br)155 mutant embryos fail to complete dorsal closure (data not shown). Lethal-phase and phenotypic analyses of E(br)165 revealed completely penetrant embryonic lethality characterized by a naked cuticle (data not shown). Because it was not possible to assess the terminal phenotype of E(br)165 mutants using cuticle preparations, we resorted to indirect immunofluorescence analysis of late stage 17 embryos with an antibody directed against the septate junction marker Coracle (FEHON et al. 1994 ). Whereas their heterozygous siblings were resistant to the antibody due to the deposition of impermeable cuticle, the E(br)165 mutant embryos showed robust Coracle staining revealing a highly penetrant dorsal open phenotype (Fig 4F).

SSNC analyses of E(br) mutations and mutations in Rho1 signaling genes:
The similar embryonic lethal phenotypes seen in several E(br) and Rho1 mutants raises the possibility that one or more of the genes affected by the E(br) mutations may function in a Rho1 signaling pathway. To examine this possibility, we conducted a series of SSNC experiments between the E(br) mutations and several alleles of genes known to function in Rho1 signaling, including Rho1, RhoGEF2, and zip (Table 5). The strong genetic interaction observed between the two Rho1^E(br) alleles and mutations in RhoGEF2 and zip serve as a useful control for these experiments and show that the Rho1 pathway is amenable to dose-sensitive genetic interaction studies (Table 5). In general, the sbd^E(br) alleles display strong SSNC with zip^E(br) and weak SSNC with Rho1 alleles, although there is considerable allele-specific variation. Similar findings were observed by BAYER et al. 2003 (this issue). In tests conducted with the unidentified E(br) mutations, we found that 6 of the 11 lines show >20% malformed legs in SSNC assays with at least one of the Rho1 pathway mutations (Table 5). Interestingly, all three of the unidentified E(br) mutants that display anterior open embryonic phenotypes show SSNC with Rho1. E(br)65 also shows SSNC with RhoGEF2 and zip. Finally, E(br)121 and E(br)444 show very robust SSNC with alleles of all three Rho1 pathway genes tested and, although both of these mutations are semidominant, the strength of these interactions suggests that the corresponding genes likely function in a Rho1 pathway.

Transcription profiles of br¹-interacting genes and Rho1 signaling pathway genes in wild-type and br⁵ mutants:
The central role of ecdysone signaling and br function in imaginal disc morphogenesis raises the possibility that one or more of the genes identified through our screens might be transcriptionally regulated by ecdysone and dependent on br activity. To test these possibilities, we analyzed the expression of genes identified through the br¹-interacting screens as well as additional genes in the Rho1 signaling pathway in both wild-type and br⁵ mutants. Total RNA was isolated from collections of 100 hand-dissected leg imaginal discs per time point from staged br⁵/Y mid-third instar larvae and prepupae, as well as from their Binsn/Y siblings. The expression of BR-C, an ecdysone-inducible early gene, was used as a control to follow the timing of the late larval ecdysone pulse (Fig 5). Previous studies have demonstrated that all BR-C isoforms are induced as a primary response to ecdysone in imaginal discs and that BR-C Z2, Z3, and Z4 isoforms are expressed at the beginning of the ecdysone peak, while the strongest expression of BR-C Z1 is delayed several hours coincident with a reduction in the expression of Z2, Z3, and Z4 (BAYER et al. 1996 ). The expression profile of the BR-C isoforms indicates that the discs respond to the late larval ecdysone pulse, although a slight developmental delay is detected in br⁵ mutant discs where the shift from predominantly Z2, Z3, and Z4 isoform expression to Z1 occurs at +4 rather than at +2 (Fig 5). This delay in br⁵ mutant discs was confirmed by monitoring the patterns of EcR, E74, and E75 early gene transcription (data not shown). Previous work from APPEL et al. 1993 has shown that Sb/sbd transcription is dependent on ecdysone. The expression of Sb/sbd that initiates between -4 and 0 hr and peaks from +2 to +4 hr in control leg imaginal discs supports their findings (Fig 5). Interestingly, however, Sb/sbd transcription is unaffected by the br⁵ mutation, showing only the 2-hr developmental delay described above. In similar Northern blot experiments using RNA collected from whole animals and pooled leg and wing imaginal discs, we found that Rho1, RhoGEF2, Rho kinase, and bs are expressed in imaginal discs, that the level of expression of each gene remains constant from -18 to +6 hr, and that the expression of these genes is unaffected in br⁵ mutant whole animals (data not shown).

We also examined the pattern of ImpE3 transcription in control and br⁵ leg imaginal discs in an effort to test for a possible regulatory interaction between br and ImpE3 suggested by the deficiency screen results described above. In control leg imaginal discs, ImpE3 expression begins 4 hr before puparium formation, peaks at 2–4 hr after pupariation, and begins to subside by +6 hr (Fig 5). Interestingly, this expression is substantially reduced in br⁵ leg imaginal discs, indicating that the ImpE3 expression is dependent upon br function in this tissue.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Detailed studies over the past decade have focused on understanding how the imaginal discs undergo proliferation and pattern formation during larval development. In contrast, we know little about how the mature imaginal discs are transformed into their corresponding adult structures during metamorphosis—structures that bear no physical resemblance to the imaginal discs from which they were derived. This study is aimed at addressing this topic by focusing on the ecdysone-dependent morphogenesis of the adult leg in Drosophila.

Two screens for dominant enhancers of the br¹ malformed leg phenotype:
We conducted two large-scale genetic modifier screens as a first step toward identifying novel links between the ecdysone signal and the cytoskeletal components that drive Drosophila leg morphogenesis. Both approaches took advantage of a hypomorphic mutation in the ecdysone-inducible br early gene, screening for enhancement of a rare malformed leg phenotype in adult flies. Screening through >750 stocks bearing either a chromosomal deficiency or a specific mutation, we identified nine loci on the second chromosome and eight loci on the third chromosome that interact with br¹. In a complementary F₁ screen of EMS-treated br¹ animals, we obtained 26 enhancer lines of which 20 were analyzed in detail. From these screens combined, we identified Rho1, bs, sbd, and cTm as br¹-interacting genes.

Conducting both a deficiency-based screen and a random mutagenesis screen allowed us to play the strengths of one approach off the weakness of the other. While a deficiency-based screen can quickly map br¹-interacting loci, it can be difficult to subsequently identify specific mutations to account for those interactions. On the other hand, an EMS screen rapidly generates specific br¹-interacting mutations, but significantly more effort is required to map and clone the corresponding gene. Here, we used complementation analyses between br¹-interacting deficiencies and EMS-induced mutations to identify br¹-interacting genes. Both screening strategies identified mutations in Rho1 and bs. We also found an unidentified E(br) mutation that fails to complement a br¹-interacting deficiency. Moreover, further complementation tests between the E(br) mutations and mutations in previously identified br¹-interacting genes allowed us to identify six EMS-derived mutations as new alleles of Sb/sbd.

The results of our screens indicate that we have not yet begun to approach saturation in this pathway. Although we identified six alleles of sbd and two alleles of Rho1 from the EMS screen, we also identified 12 mutations that are each represented by a single allele. In addition, we found at least 14 br¹-interacting loci by deficiency screening for which we did not recover an EMS-derived mutation. Principally, these results illustrate the complexity of imaginal disc morphogenesis and suggest that many genes are required to ensure the fidelity of this process. It is apparent that a larger screen will generate additional br¹-interacting mutations, some of which might map to genomic intervals identified through the deficiency screen.

A central role for Rho1 signaling in ecdysone-triggered leg disc morphogenesis:
The identification of Rho1 mutations as dominant enhancers of br¹, as well as the genetic interactions we observe between E(br) alleles and mutations in the Rho1 signaling pathway, indicate a central role for Rho1 in directing leg morphogenesis at the onset of metamorphosis (Fig 6). Signaling through the Rho1 small GTPase depends on a shift in the cellular equilibrium between inactive Rho-GDP and active Rho-GTP (reviewed in VAN AELST and D'SOUZA-SCHOREY 1997 ). This equilibrium is influenced by guanine nucleotide exchange factors (GEFs) that activate Rho1 by removing GDP from inactive Rho1 molecules, thereby allowing Rho1 to bind GTP. Countering this action, GTPase activating proteins (GAPs) stimulate the weak GTPase activity of Rho1. A key effector of activated Rho1 is Rho kinase, a serine/threonine kinase that regulates contractile events at the actin cytoskeleton. Rho kinase exerts its effect by phosphorylating and thereby inactivating the myosin-binding subunit of the myosin phosphatase complex. The principal substrate for myosin phosphatase is myosin regulatory light chain—a component of the actin cytoskeleton that can also be directly phosphorylated by Rho kinase. Therefore, the net effect of activating Rho kinase is to maintain the phosphorylated state of myosin regulatory light chain, which, in turn, results in the activation of the myosin heavy chain, allowing myosin complexes to move along actin filaments.

View larger version (60K): In this window In a new window Download PPT slide   Figure 6. Model of the signaling events that direct leg imaginal disc morphogenesis. Activation of Rho1 to the GTP-bound state plays a central role in directing the cell shape changes that drive disc morphogenesis. Possible regulatory interactions are represented by question marks. See text for details.

Over the past several years, genetic, molecular, and pharmacological perturbations of Rho1 signaling have revealed key roles for this signaling cascade in directing a variety of morphogenetic processes, including embryonic elongation in Caenorhabditis elegans and neural tube closure in the mouse (WISSMANN et al. 1997 , WISSMANN et al. 1999 ; BROUNS et al. 2000 ; WEI et al. 2001 ). In Drosophila, Rho1 signaling is required for cellularization of the blastoderm embryo, gastrulation, dorsal closure, head involution, neural development, and the establishment of planar polarity (reviewed in SETTLEMAN 2001 ). Recently, HALSELL et al. 2000 found that mutations in Rho1 enhance the malformed leg phenotype associated with heterozygous zip mutations, suggesting a role for Rho1 in imaginal disc morphogenesis. Here we confirm and extend their observations by linking several Rho1 signaling components to the genetic functions of an ecdysone-inducible transcription factor, suggesting that ecdysone activation of the Rho1 signaling pathway may drive the cell shape changes associated with leg disc morphogenesis.

Genetic studies indicate that at least five components of the Drosophila Rho1 signaling cascade are required during imaginal disc morphogenesis: Rho1, RhoGEF2, myosin phosphatase, myosin regulatory light chain (encoded by spaghetti squash or sqh), and nonmuscle myosin heavy chain (encoded by zip; Fig 6). First, we found that deficiencies that uncover the Rho1 locus as well as specific mutations in Rho1 enhance the malformed leg phenotype associated with the br¹ mutation (Table 2). We also recovered two new alleles of Rho1 as E(br) mutations from our EMS mutagenesis screen (Table 3). Second, we found strong SSNC between both Rho1^E(br) alleles and RhoGEF2^11-3 (Table 5), suggesting that RhoGEF2 is playing a pivotal role in activating Rho1 during imaginal disc morphogenesis. Consistent with this observation, BAYER et al. 2003 (this issue) detected strong SSNC between the same allele of RhoGEF2 and three additional alleles of Rho1, and HALSELL et al. 2000 reported SSNC between three independent alleles of RhoGEF2 and zip^E(br). At least 20 potential RhoGEF genes are present in the Drosophila genome (SETTLEMAN 2001 ), however, raising the possibility that other RhoGEFs may also contribute to Rho1 activation during imaginal disc morphogenesis. Third, MIZUNO et al. 2002 demonstrated that mutations in the myosin-binding subunit of myosin phosphatase can ameliorate the malformed wing phenotype associated with zip^E(br)/zip⁰²⁹⁵⁷ mutants and also reduce the penetrance of malformed wings in animals displaying SSNC between zip^E(br) and mutations in Rho1, RhoGEF2, and Rho kinase. Fourth, the zip^E(br) allele of nonmuscle myosin strongly enhances the malformed leg phenotype of br¹ (GOTWALS and FRISTROM 1991 ) and displays robust SSNC with several alleles of Rho1 and RhoGEF2 (HALSELL et al. 2000 ). Finally, malformed leg and wing phenotypes are seen in sqh mutants (EDWARDS and KIEHART 1996 ). Taken together, these studies provide strong evidence of a key role for Rho1 in directing the cell shape changes that drive imaginal disc morphogenesis (Fig 6).

In addition to direct effects on the actin cytoskeleton, Rho1 signaling can also transduce extracellular signals to the nucleus by activating SRF transcription factors. The mechanism of Rho1-dependent SRF activation is poorly understood but appears to require at least one of several Rho1-specific effector molecules, including Rho kinase, LIM kinase, and formin-homology proteins of the mDia family, in a cell-type-specific manner (GENESTE et al. 2002 ). A current model proposes that the coordinated effects of these Rho1 effector molecules is to increase F-actin assembly and reduce F-actin severing, thereby decreasing the cytoplasmic pool of G-actin, which promotes the nuclear accumulation of MAL, an SRF coactivator (SOTIROPOULOS et al. 1999 ; GENESTE et al. 2002 ; MIRALLES et al. 2003 ). Transcriptional targets of activated SRF include - and ß-actin, vinculin, and tropomyosin (GINEITIS and TREISMAN 2001 ; MACK et al. 2001 ; NAKAMURA et al. 2001 ). In this context, Rho1-dependent transcriptional activation of SRF appears to be reinforcing the direct effects that Rho1 is producing on the actin cytoskeleton. It is intriguing then that we identified mutations in bs (the Drosophila ortholog of SRF) and cTm as dominant modifiers of br¹ for leg disc morphogenesis, suggesting that a transcriptional pathway downstream of Rho1 is also important for this morphogenetic process and that cTm may be a transcriptional target of bs (Fig 6). Consistent with these ideas, HALSELL and KIEHART 1998 identified cTm in an SSNC screen with zip^E(br) using a similar malformed leg assay. We performed SSNC experiments between bs^E(br)292 and mutations in the Rho1 signaling pathway as an initial test for Rho1-dependent bs function during leg morphogenesis, but failed to detect any significant interaction (Table 5). Additional work will be required to assess the relative importance of the transcriptional effects of Rho1 signaling in imaginal disc morphogenesis.

Roles for ecdysone in directing leg disc morphogenesis:
The identification of genetic interactions between members of the Rho1 signaling pathway and the ecdysone-inducible transcription factor encoded by br provides an intriguing tie between the steroid hormone and the cellular machinery that drives morphogenesis. To investigate whether members of the Rho1 pathway might be transcriptionally regulated by ecdysone, we examined the expression of Rho1, RhoGEF2, and Rho kinase in whole animals and imaginal discs dissected from staged late larvae and prepupae of both wild type and br⁵ mutants (data not shown). This study, however, revealed no changes in transcript levels in response to the late larval ecdysone pulse, and no effects of the amorphic br⁵ mutation on their expression (data not shown). It should be noted, however, that there are many possible targets for ecdysone regulation of Rho1 activity, including multiple RhoGEFs and RhoGAPs. Ecdysone may also be responsible for inducing the expression of one or more proteins required for the appropriate subcellular localization of the Rho1 complex, a level of regulation that is thought to be critical for its activation (reviewed in SYMONS and SETTLEMAN 2000 ). Possible targets for this regulation include a transmembrane protein that anchors Rho1 signaling components to the plasma membrane or a kinase that phosphorylates a RhoGEF to promote membrane association. The use of microarrays to identify ecdysone-inducible genes in imaginal discs would provide a powerful counterpoint to our genetic screens as well as a means of identifying these possible intermediates between the ecdysone and Rho1 signaling pathways.

Intriguingly, Sb/sbd represents the only known br¹-interacting gene that is induced directly by ecdysone in imaginal discs as they undergo morphogenesis (Fig 5; APPEL et al. 1993 ). The function of the Sb/sbd type II transmembrane serine protease in this response, however, remains unknown. It has been suggested