A Screen to Identify Drosophila Genes Required for Integrin-Mediated Adhesion

Corresponding author: Nicholas H. Brown, Wellcome/CRC Institute, Tennis Court Rd., Cambridge CB2 1QR UK., nb117{at}mole.bio.cam.ac.uk (E-mail).

Communicating editor: T. SCHÜPBACH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Drosophila integrins have essential adhesive roles during development, including adhesion between the two wing surfaces. Most position-specific integrin mutations cause lethality, and clones of homozygous mutant cells in the wing do not adhere to the apposing surface, causing blisters. We have used FLP-FRT induced mitotic recombination to generate clones of randomly induced mutations in the F₁ generation and screened for mutations that cause wing blisters. This phenotype is highly selective, since only 14 lethal complementation groups were identified in screens of the five major chromosome arms. Of the loci identified, 3 are PS integrin genes, 2 are blistered and bloated, and the remaining 9 appear to be newly characterized loci. All 11 nonintegrin loci are required on both sides of the wing, in contrast to integrin subunit genes. Mutations in 8 loci only disrupt adhesion in the wing, similar to integrin mutations, while mutations in the 3 other loci cause additional wing defects. Mutations in 4 loci, like the strongest integrin mutations, cause a "tail-up" embryonic lethal phenotype, and mutant alleles of 1 of these loci strongly enhance an integrin mutation. Thus several of these loci are good candidates for genes encoding cytoplasmic proteins required for integrin function.

MULTICELLULAR organisms require cell and tissue adhesion mechanisms to mediate both static and dynamic aspects of morphogenesis and development. Integrins are cell surface receptors with essential roles in adhesion in both invertebrates and vertebrates, linking cells to their extracellular environments (for review see HYNES 1992 ). Each integrin is an /ß heterodimer of two large transmembrane proteins. In vertebrates at least 22 integrin heterodimers have been identified, which have been shown to have roles in adhesion of cells to the extracellular matrix, in adhesion between cells, and in signal transduction across the cell membrane (CLARK and BRUGGE 1995 ; GUMBINER 1996 ). Some nonintegrin proteins that appear to be part of integrin-mediated-adhesion processes have been identified in vertebrates using biochemical methods, including extracellular ligands of integrins, intracellular proteins that connect the integrins to the cytoskeleton, and proteins that may have a role in signal transduction, such as focal adhesion kinase (see HYNES 1992 ; CLARK and BRUGGE 1995 for review). The requirement for some of these proteins in integrin-mediated processes has been addressed genetically by generating mutations in mice by homologous recombination (e.g., FURATA et al. 1995 ), but the interpretations are hindered by the relative inaccessibility of the developing embryo, the resorption of the mutant embryos at an early developmental stage, and the potential for genetic redundancy (HYNES 1996 ). We wished to complement these approaches by using the powerful genetic tools of Drosophila to determine how many genes are essential for a particular integrin-mediated process.

Drosophila appears to have a smaller number of integrin genes than vertebrates, with five subunits identified so far (reviewed in BROWN 1993 ; GOTWALS et al. 1994A ). The first Drosophila integrins were identified in a series of monoclonal antibody screens for cell surface antigens with a restricted distribution in larval imaginal discs and were thus named the position specific (PS) integrins (WILCOX et al. 1981 ; BROWER et al. 1984 ). The myospheroid (mys) gene encodes the ß_PS integrin subunit (MACKRELL et al. 1988 ; LEPTIN et al. 1989 ), the multiple edematous wing (mew) gene encodes the _PS1 integrin subunit (WEHRLI et al. 1993 ; BROWER et al. 1995 ), and the inflated (if) gene encodes the _PS2 subunit (BOGAERT et al. 1987 ; BROWER and JAFFE 1989 ; WILCOX et al. 1989 ; BRABANT and BROWER 1993 ; BROWN 1994 ). These subunits form the PS1 (_PS1ß_PS) and the PS2 (_PS2ß_PS) integrins. The two Drosophila integrins PS1 and PS2 are essential for the development of both the larva and the adult and show complementary patterns of expression in particular tissues. For example, in embryos, the PS2 integrin is expressed in the muscles (BOGAERT et al. 1987 ) while the PS1 integrin is expressed in the epidermal cells and the gut endoderm (LEPTIN et al. 1989 ; WEHRLI et al. 1993 ). Their expression in these tissues is essential for adhesion between these tissue layers (WRIGHT 1960 ; NEWMAN and WRIGHT 1981 ; LEPTIN et al. 1989 ; BRABANT and BROWER 1993 ; BROWN 1994 ; BROWER et al. 1995 ; ROOTE and ZUSMAN 1995 ). Thus, null mutations of the mys gene cause embryonic lethality and cause a phenotype in which the somatic muscles detach from the epidermis, and the visceral muscles detach from the gut endoderm, which undergoes abnormal morphogenesis. In addition, dorsal closure fails, resulting in a dorsal hole in the epidermis. These phenotypes are enhanced by the removal of maternal mys activity (WIESCHAUS and NOELL 1986 ; LEPTIN et al. 1989 ; ROOTE and ZUSMAN 1995 ). A similar situation exists in the imaginal wing disc, where the PS1 integrin is expressed in the cells that will form the dorsal region of the wing, while the PS2 integrin is expressed in the cells of the future ventral region of the wing, and they both are required for the adhesion of these two surfaces to each other (WILCOX et al. 1981 ; BROWER et al. 1984 ; BROWER and JAFFE 1989 ; WILCOX et al. 1989 ; ZUSMAN et al. 1990 ; BRABANT and BROWER 1993 ; BROWER et al. 1995 ). These results show that one of the major functions of the PS integrins during development is to mediate adhesion between cell layers.

A full understanding of the mechanisms underlying integrin function will require the characterization of the other proteins that are needed. A few Drosophila homologues of those cytoskeletal proteins thought to be part of integrin-mediated processes in vertebrates have been identified, such as -actinin and vinculin, but when these were tested, genetic analysis shows they are not essential for integrin-mediated adhesion (FYRBERG et al. 1990 ; ALATORTSEV et al. 1997 ). However, two extracellular matrix components have been identified in the fly that can account for some PS integrin functions. Laminin serves as ligand for the PS1 integrin in cell-culture experiments (GOTWALS et al. 1994B ), and recent analysis suggests that PS1 binds to laminin in the developing embryo as well (PROKOP et al. 1998 ). However, laminin A does not seem to be essential for PS1 integrin function in the wing (HENCHCLIFFE et al. 1993 ). The novel extracellular matrix protein tiggrin has been identified as a potential PS2 integrin ligand (FOGERTY et al. 1994 ). The phenotype caused by tiggrin mutations is consistent with this in the embryo, but tiggrin is not essential for adhesion in the wing (BUNCH et al. 1998 ). Thus few Drosophila proteins have been found so far that by genetic evidence have a role in integrin function.

To identify these other components of integrin-mediated adhesion, we wished to screen for mutations that cause a phenotype similar to that produced by integrin mutations. Almost all mutant alleles within the PS integrin subunit genes cause lethality in the embryo or first-instar larva (BUNCH et al. 1992 ; BROWER et al. 1995 ; BLOOR and BROWN 1998 ). However, to screen for new genes we took advantage of an adult phenotype: the blisters produced in the absence of integrin-mediated adhesion between the two surfaces of the adult wing. Some mutations that cause viable wing blister phenotypes have already been identified in the fly, including vesiculated, wing blister, blistery, blistered, and bloated (LINDSLEY and ZIMM 1992 ; GELBART et al. 1997 ). The fact that mutations in some of these genes just cause adult viable phenotypes suggests that they may only have a role in mediating adhesion in the wing. Alternatively these genes may be required throughout development, but the viable mutations recovered so far may be rare hypomorphic alleles, and amorphic mutations would cause lethality (as is the case for the viable mys, if and mew alleles; BUNCH et al. 1992 ; BLOOR and BROWN 1998 ; our unpublished results). Therefore, we wanted to use a method that would allow us to screen for recessive mutations that produce wing blisters and may be homozygous lethal. To accomplish this we used the FLP-recognition-target (FRT) system to induce mitotic recombination at a high enough frequency to enable us to screen the F₁ generation for mutations that, like those in integrin genes, cause wing blisters when homozygous clones are produced in the wings of heterozygous individuals. This system relies on the fact that the FLPase enzyme, which is expressed under the control of a heat-shock promoter, mediates mitotic recombination between FRT sequences (GOLIC and LINDQUIST 1989 ; GOLIC 1991 ). FRTs have been inserted near the centromeres of all the major chromosome arms (XU and RUBIN 1993 ), so that for each arm, recombination at the appropriate site in heterozygotes will produce daughter cells that are homozygous for most of that arm. Thus, this efficient technique can be used to perform mosaic analysis of more than 95% of Drosophila genes.

Our screen identified 14 lethal complementation groups and 11 loci with single alleles. We recovered mutations in the PS integrin genes mys, mew, and if, as expected, and in the two previously identified genes blistered and bloated. Of the complementation groups, 9 represent novel loci, 6 of which appear to correspond to the loci isolated in a similar screen performed by PROUT et al. 1997 . To help identify those loci that are most likely to be directly involved in integrin-mediated adhesion, we examined their mutant phenotypes and ability to genetically interact with integin mutations. Some of the loci have mutant alleles that cause stronger phenotypes than the integrin mutations and thus must be required for other functions in addition to PS integrin adhesion, while mutations in other loci cause a weaker phenotype, suggesting that they are required only for a subset of PS integrin functions.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Mutagenesis and F₁ screens:
All crosses in this study were grown on the Tübingen fly food (STEWARD and NUSSLEIN-VOLHARD 1986 ) at 25° unless otherwise indicated. The mutations and balancer chromosomes used in these experiments are described in LINDSLEY and ZIMM 1992 and the flybase in GELBART et al. 1997 . The FLP-FRT strains used for the screens were those described in CHOU and PERRIMON 1992 and XU and RUBIN 1993 with various markers recombined onto these chromosomes. Screens were carried out for each chromosome arm using the following stocks: (1) y w f P[mini w⁺, FRT]^9-2 and w P[mini w⁺, FRT]^9-2; Ly ry / MKRS, P[ry⁺, hsFLP]⁹⁹ for the X chromosome; (2) w P[ry⁺, hsFLP]; P[ry⁺, hs-neo, FRT]40A and w P[ry⁺, hsFLP]; dp P[ry⁺, hs-neo, FRT]40A for 2L; (3) w P[ry⁺, hsFLP]; P[ry⁺, hs-neo, FRT]42D, P[ry⁺, y⁺]44B and w P[ry⁺, hsFLP]; P[ry⁺, hs-neo, FRT]42D sp for 2R; (4) w P[ry⁺, FLP]; th st P[ry⁺, hs-neo, FRT]80B for 3L; and (5) w P[ry⁺, FLP]; P[ry⁺, hs-neo, FRT]82B and w P[ry⁺, FLP]; P[ry⁺, hs-neo, FRT]82B e^S for 3R. For each screen the relevant chromosome was isogenic. The designs of the X chromosome and the 2R screens are shown in Figure 2. The 2L and 3R chromosome arm screens and the 2R screen were carried out in a similar way, but the 3L screen was designed with sublines as was the X screen. In each case males were mutagenized at 2–3 days old with 4000 rads of X rays and allowed to mate for 4 days before removal. To make mutant clones in the adult wings, FLP-mediated mitotic recombination was induced in the larval imaginal discs by two 2-hr heat shocks, each achieved by placing bottles containing second-instar larvae in a 38° waterbath within a 37° incubator so that the food was submerged in the water. The flies were allowed to recover for 2 hr at room temperature between the heat shocks. Adult flies with wing blisters were identified using the dissecting microscope and were retested. For the X chromosome and 3L screens it was necessary to establish stocks prior to the retest (see Figure 2), because the X screen had to be done in females, so meiotic recombination can occur between the new mutations and markers on the chromosome, and the 3L screen used an FRT chromosome that does not contain a marker found on any of the common third chromosome balancers. In the other screens the F₁ flies were crossed to stocks containing a marker that is also present on a balancer chromosome (dp, sp, and e for 2L, 2R, and 3R, respectively) so that we could immediately retest the mutants and then establish a stock of the mutant (unmarked) chromosome (see Figure 2). Stocks were established in two generations so that the w hsFLP chromosome was removed from the stock.

View larger version (92K): In this window In a new window Download PPT slide   Figure 1. Identification of mutations in genes required for adhesion between the two surfaces of the wing. (A) Wing blister produced by a clone of cells mutant for an allele of inflated, which encodes an integrin subunit. (B) Loci identified in the screen (see also Table 1). Each line indicates a chromosome, and the complementation groups isolated in the screens are indicated. Newly identified loci are depicted above each line, while previously characterized genes are underneath. The loci bee and puri have been only roughly mapped by recombination, while the others have been confirmed with deficiencies or, for pomp, a duplication.

View larger version (24K): In this window In a new window Download PPT slide   Figure 2. FLP-FRT screens for wing bubble mutations. Design of the screens for isolating mutations that cause wing blisters when homozygous clones are produced in the wing.

Complementation tests and genetic mapping:
Complementation tests of mutations recovered in the screen were based on the failure to complement the lethality of the new loci or the phenotype of viable alleles of previously characterized loci. X chromosome mutations were initially tested for complementation to alleles of the three X chromosome PS integrin genes, through use of the following: the viable myospheroid allele mys^NJ42 (DE LA POMPA et al. 1989 ); if^B4, a null inflated lethal allele (BROWN 1994 ); and mew^A1-5, a viable multiple edematous wing allele (J. W. BLOOR and N. H. BROWN, unpublished results). The viable X-linked allele recovered from the screen fails to complement the vesiculated allele vs¹. The remaining three X chromosome lethal mutations were initially mapped by recombination through the use of y cv v f car and found to map to the same region. The duplication Dp(1;2)v⁺65b was found to cover two of the alleles, allowing for the appropriate complementation tests and deficiency mapping to be done, which demonstrated that all three were alleles of a single locus. Once the autosomal mutations were assigned to complementation groups they were mapped using the Bloomington Stock Center deficiency kits, and at least two alleles from each complementation group were used to confirm the results. Complementation groups that did not map to any of the deficiencies were mapped by recombination using the second chromosome markers al dp b c px sp. Complementation tests with the loci isolated by PROUT et al. 1997 were performed with two alleles from each screen, if possible.

Determination of the lethal phase of mutant alleles:
We determined the stage of development during which each mutant allele causes lethality by outcrossing males from balanced mutant stock to Oregon R virgins and crossing the heterozygous progeny (virgin females or males, as appropriate) to other alleles or deficiencies in small cages. Embryos were collected on apple juice agar plates and the proportions of unhatched embryos and dying larvae and pupae were calculated.

Germline clones of mutations:
Germline clones of one mutant allele of each locus were produced using the FLP-recombinase system and the dominant female sterile mutation ovo^D1 according to the methods described in CHOU and PERRIMON 1992 , CHOU and PERRIMON 1996 . For the mutations on the X and 2R it was necessary to recombine on the appropriate FRT site for use with the ovo^D1 stocks. The following chromosomes were used in these tests: w ovo^D1 v²⁴ P[mini w⁺, FRT]¹⁰¹; MKRS, FLP⁹⁹/ + for the X chromosome; P[mini w⁺, ovo^D1]^2L-13X13 P[ry⁺, hs-neo; FRT]40A/ CyO for the 2L chromosome arm; P[mini w⁺, FRT]^2R-G13 P[mini w⁺, ovo^D1]^2R-32X9/ CyO for the 2R chromosome arm; and P[ry⁺, hs-neo; FRT]82B P[mini w⁺, ovo^D1]^3R-C13a31n9/ TM3, Sb for the 3R chromosome arm.

Cuticle preparations and antibody staining of embryos:
To examine embryonic phenotypes, mutants were first outcrossed to Oregon R flies to remove the balancer chromosome. Flies were set up in cages and the eggs collected on apple juice agar plates. Eggs were collected overnight, aged a further 24–36 hr and cuticle preparations made as described in WIESCHAUS and NUSSLEIN-VOLHARD 1986 . Antibody staining of embryos was done using standard procedures. The primary antibodies used were rabbit anti-muscle myosin heavy chain (KIEHART and FEGHALI 1986 ) at a concentration of 1:250, mouse anti-ß_PS integrin subunit monoclonal CF6G11 (BROWER et al. 1984 ) at a concentration of 1:500, and rat anti-_PS2 subunit monoclonal hc/2 (BOGAERT et al. 1987 ) and mouse anti-Fasciclin III monoclonal DA1B615 (BROWER et al. 1980 ) at concentrations of 1:5. The secondary antibodies used were the anti-mouse Ig, horseradish peroxidase-linked, whole antibody from sheep NA931 (Amersham, Little Chalfont, UK) and the anti-rabbit Ig, horseradish peroxidase-linked (Jackson ImmunoResearch, West Grove, PA) at concentrations of 1:250. For the _PS2 monoclonal a biotinylated anti-rat antibody was followed by vectastain elite (Vectalabs, Burlingame, CA). Stained embryos were photographed with a Zeiss (Thornwood, NY) Axiophot microscope. The images were scanned with a Nikon (Garden City, NY) Coolscan, assembled with Adobe Photoshop software (Adobe Systems, Mountain View, CA), and labeled with FreeHand 5.5 (Macromedia, San Francisco, CA).

Generation of marked mutant clones in the wing:
For most of the loci we crossed alleles to virgin females from stocks that will produce clones marked with forked and M+ in a M/+ backround following X-ray irradiation to induce clones: for 2L f^36a; P[f⁺]30 M(2)z/CyO; for 2R f^36a; M(2)l P[f⁺]52 / CyO; and for 3R f^36a; P[f⁺]98 M(3)w/TM3, Sb (kind gifts of Paloma Martin and Antonio Garcia Bellido). For papillote we crossed on f^36a onto the pot^P14 allele. Mitotic recombination was induced with X rays (1000 rad) and, by irradiating second-instar larvae, we produced clones that are restricted to the dorsal or ventral compartments of the wing imaginal discs. For the pomp allele we crossed males to the stock P[ry⁺, hsFLP] f^36a; ck P[f⁺]30 P[ry⁺, hs-neo, FRT]40A/CyO (RODRIGUEZ and BASLER 1997 ) and generated clones with a 15 min heat shock at 37°. Wings were dissected in ethanol and mounted in a 1:1 glycerol:lactic-acid solution and examined for clones of cells with forked trichomes using bright-field microscopy.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Isolation of mutations in genes required for adhesion between the two surfaces of the wing:
With the aim of identifying genes involved in integrin-mediated adhesion, we screened for mutations that cause a wing blister when clones of cells homozygous for the mutation are produced in the wings of heterozygous individuals (Figure 1). We were initially concerned that this phenotype might not be selective enough if it occurred in response to a wide variety of defects that are not linked to the cell-adhesion process. If this were the case then we would end up with mutations in too many loci. Our assumption was that if the number of loci isolated in the screen were small, then there would be a greater probability that all of the loci were involved in the same process: integrin-mediated adhesion between the two surfaces of the wing. A second, related question was how efficient the screen would be at recovering mutations in loci required for adhesion, which would determine how many mutant chromosomes should be screened. To be able to test these factors we initially screened the X chromosome because the three integrin genes on this chromosome could serve as controls for the screen. If we recovered multiple alleles of each integrin gene, then this would indicate that the screening procedure is efficient at recovering mutations in relevant genes, and the number of loci that we recover would indicate the selectivity of the screen.

Using the genetic scheme outlined in Figure 2A, we mutagenized with X rays males containing an FRT site near the centrosome (18A) and crossed them to females homozygous for the same FRT site and containing a source of the FLP recombinase. We decided to use X rays rather than EMS to avoid an F₁ generation mosaic for the mutagenized chromosome (independent of the FLP-FRT-induced mosaicism) and to increase our chances of recovering mutations that would be detectable as molecular aberrations. The F₁ larva were heat shocked to induce mitotic clones, and 20,000 F₁ adult females of the appropriate genotype were screened, yielding 80 with wing blisters similar to that shown in Figure 1A. From these 80 initial females we recovered 16 recessive lethal mutations that produce wing blisters in clones and 1 allele of the viable wing blister locus vesiculated. Of the lethal mutations 13 are integrin alleles (Table 1) and the remaining 3 lethal mutations map to a single locus. Thus this approach appears to be very successful; from a screen of only 20,000 F₁ individuals we have recovered multiple alleles in each of 4 loci on the X chromosome, 3 of which are the known integrin genes, and 1 of which is a new locus. Because the X chromosome is approximately one fifth of the genome, if this distribution of genes were representative, we would expect to recover only 16 additional loci by screening the autosomes, indicating that the screen is suitably selective.

We screened 30,000–50,000 F₁ adults for each of the major autosomal arms (the design of one of the screens is shown in Figure 2B) and recovered 59 mutations that cause blisters in clones, all of which are on chromosomes containing a lethal mutation (Table 2). Through complementation testing we found that we had recovered lethal mutations in 21 genes (Table 2). Because the mutagenesis can result in the mutation of more than one gene on the chromosome, it is difficult to be sure that the homozygous phenotype of a single mutant chromosome is due to the mutation that causes the wing blister phenotype. However, the chance that two alleles of a particular wing blister locus also both have an allele of a second unlinked lethal mutation is low. Therefore we have restricted our further analysis to the 10 autosomal complementation groups, which each contain at least two alleles, and to the new locus on the X chromosome (Table 1 and Table 2). The locations of these complementation groups were mapped with deficiencies or by mitotic recombination, which showed that 2 of them correspond to previously described loci: bloated (blo; WADDINGTON 1939 , WADDINGTON 1940 ) and blistered (bs; FRISTROM et al. 1994 ). The remaining novel loci were named on the basis of the wing blister phenotype [papillote (pot), bladderwrack (bad), pompholyx (pomp), kopupu (kop), puri (puri), sac (sac), gonfle (gon)] or their embryonic phenotype [beerbelly (bee) and scorpion (sci); see below]. These loci are shown schematically in Figure 1B.

All of the new genes are required on both sides of the wing:
The two integrins PS1 and PS2 are each required only on one side of the developing wing, with PS1 on the dorsal side and PS2 on the ventral side (BRABANT and BROWER 1993 ; BROWER et al. 1995 ). If one of the new genes we have isolated were involved only in the function of one of the integrins, then we would expect that it would also be required only on one side of the wing. Conversely, if it were required for the function of both integrins then it would be required on both sides of the wing. We tested each of the new genes by making mutant clones marked with the trichome marker forked (see MATERIALS AND METHODS). One allele from each complementation group was tested and mutant clones were scored for the presence or absence of blisters. Blisters were found to be associated with both dorsal and ventral clones for all of the new genes (an example of a marked mutant clone is shown in Figure 3). Therefore, the new genes are required on both sides of the wing to mediate adhesion between the dorsal and ventral epithelial cell layers and may be required for the function of both integrins.

View larger version (156K): In this window In a new window Download PPT slide   Figure 3. Clones of cells mutant for most of the loci isolated just cause a wing blister, but bloated mutant clones cause additional defects. (A) Wild-type wing. (B) Close up of a wing containing a homozygous clone of kopV104 (the outline of the clone marked by f bristles is shown with a dotted line), showing that the blister is larger than the clone and that the venation and polarity of the wing are unaffected. (C and D) Two wings, each containing a clone of bloV99. In C, a blo mutant clone (marked with a dotted line) causes three effects: a blister, changes in the shape of the wing, and a vesicle of cells in the center of the wing. In D, a close-up shows that the vesicles produced in wings containing blo mutant clones can alter the venation pattern of wild-type portions of the wing.

Phenotypes of marked wing clones:
We also examined the marked mutant clones for any additional defects besides the loss of adhesion between the two wing surfaces, such as vein abnormalities, overgrowth of clones, defects in trichome polarity, or changes in the normal shape or pattern of the wing. Consistent with previous work showing that bs is required for intervein cell fates (FRISTROM et al. 1994 ; MONTAGNE et al. 1996 ), clones of one of the new bs alleles show a transformation toward vein tissue (not shown). Also consistent with the previous description of the viable blo¹ allele (WADDINGTON 1939 , WADDINGTON 1940 ), clones of the blo^V99 allele show a more complex phenotype than just wing blisters. One defect caused by blo mutations is the production of clusters of wing epithelial cells that are found between the two sides of the wing in both intervein and vein regions (Figure 3). The cells within the vesicles have secreted cuticle, contain trichomes, and can consist of either darkly pigmented cells like vein cells or unpigmented cells like intervein cells. Some of the vesicles appear to induce ectopic vein outgrowths from nearby wild-type veins (Figure 3D). The vesicles of cells can end up fairly distant from the clone of blo mutant cells (not shown), but at least some of the trichomes within the vesicles appear to have the forked phenotype, suggesting that the cells in the vesicles have originated from the marked mutant clones in the wing epithelium (not shown). An additional phenotype is found in the wings of blo¹/Df, blo¹/blo^V99, and blo¹/blo^V134, which have a distorted shape with an elongation of the distal anterior region and a truncation of the most distal region, somewhat similar to the wing phenotype caused by certain dumpy mutations (data not shown). Clones of blo^V99 also produce distortions in the shape of the wing (Figure 3C). Even though the clones occur on just one side of the wing, the wing is distorted on both surfaces and beyond the area of the clone. Therefore, mutations in blo cause nonautonomous defects in the pattern of the wing. The wing blister phenotype is independent of the wing distortion phenotype, as some mutant clones produce just one or the other phenotype (not shown). However, larger numbers of mutant clones need to be examined before it is clear whether size and/or position causes a particular wing phenotype. These results suggest that the product of the bloated gene is involved in multiple aspects of cell-cell interactions in the wing: adhesion between cells within each epithelial layer, growth control in the wing blade, and adhesion between the two surfaces of the wing blade.

Clones of the mutation pomp^E16 also produce vesicles between the two wing surfaces (data not shown), as seen in the case of clones of blo^V99, but pomp mutant clones do not cause any distortions in the shape of the wing, and there is no other obvious link between pomp and blo. Mutations in the other eight genes isolated in this screen cause a specific defect only in the process of adhesion between the two wing surfaces. The clones of cells mutant for any one of these loci do not disrupt the normal pattern of veins or trichome polarity and do not show any overgrowth or disruption to the overall wing shape and pattern (one example is shown in Figure 3B). This is also true for integrin mutant clones, suggesting that these eight new loci could be genes that specifically function in integrin-mediated processes.

Dominant enhancement of viable integrin mutations by mutations in the new wing blister loci:
Mutations in genes involved in PS integrin-mediated adhesion might be expected to show genetic interactions with PS integrin mutations. One particularly strong genetic interaction has been described between integrin mutations: the antimorphic mys^XR04 allele is lethal over amorphic if alleles (WILCOX 1990 ; BUNCH et al. 1992 ). We tested whether any of the new mutations also show this strong interaction and found that none of them do, including the mew mutant alleles (data not shown). Thus the pseudoallelism of mys^XR04 with mutations in other loci appears to be specific for inflated.

To test for dominant genetic interactions with mutations from the other complementation groups, we next used the PS integrin viable hypomorphic alleles mys^NJ42, mew^A1-5, and if³, which all give rise to wing blister phenotypes. Flies hemizygous for both mys^NJ42 and if³ have large blisters in almost 100% of the wings demonstrating that the phenotype of each allele can be significantly enhanced (WILCOX et al. 1989 ; WILCOX 1990 ). Males hemizygous for the mys^NJ42 allele are unable to jump due to defects in the muscles (DE LA POMPA et al. 1989 ), and have held-out wings and a low frequency of wing blisters (~2%), which is enhanced when the allele is placed over a deficiency for mys (WILCOX et al. 1989 ; ZUSMAN et al. 1993 ). The mew^A1-5 allele is a new temperature-sensitive viable allele (J. W. BLOOR and N. H. BROWN, unpublished results) that at 28° causes a high frequency of wing blisters in hemizygous males (~65%), which is reduced to ~10% at 25° (Table 3). Males hemizygous for the viable if³ mutation have blisters in ~50% of the wings at 25°, which is reduced at 28° (BROWER and JAFFE 1989 ; WILCOX et al. 1989 ). We examined whether mutant alleles of the new loci are able to dominantly enhance the hemizygous phenotypes caused by these integrin mutations. We tested two mutant alleles from each locus (except pot) and examined the phenotypes at 25° and 28° (Table 3).

We found that both mutant alleles of three loci, bs, puri and sci, enhance the if³ mutation. Mutations in four loci, bee, gon, kop, and pomp, did not enhance the phenotype caused by any of the integrin mutations. One mutant allele of the remaining loci, bad, blo, pot and sac, enhanced the phenotype caused by one or more integrin mutation. In all combinations, when enhancement was observed it was an increase in the frequency of the blisters rather than their size (not shown), and no clear examples of suppression were observed. Since each pair of mutant alleles was produced in a stock isogenic for the relevant chromosome and the pairs were balanced with the same stocks, the differing ability of the two alleles of some of the loci in enhancing integrin mutations is most likely due to either different strengths of the two alleles or second-site mutations. Consistent with the former, the bs alleles give complex results: both enhance if, but one enhances mew while the other enhances mys. The enhancing activity of bs has been described previously (WILCOX 1990 ) and is likely due to some partial transformation of the intervein cells toward a nonadhesive vein cell fate, because in homozyous clones of bs the intervein cells are transformed to vein (FRISTROM et al. 1994 ; MONTAGNE et al. 1996 ). The difference in the behavior of these two mutant alleles suggests that there may be additional complexities to bs function. As mentioned, blo appears to have diverse activities in the wing, and, although the mutant cells are not transformed into vein tissue, ectopic vein tissue is produced. In contrast, mutant clones of the other five enhancing loci, bad, pot, puri, sac and sci, do not show any transformation from intervein to vein. The consistent enhancement of if³ by both sci alleles suggests that the sci gene product is the most likely to act directly in integrin-mediated adhesion.

Characterization of mutant phenotypes of the complementation groups:
If the new loci were generally required for integrin function, then we would expect their homozygous mutant phenotypes to share features with the phenotypes caused by PS integrin mutations. We first examined the lethal phase of each of the mutant alleles by scoring hatching of embryos and larval or pupal lethality (Table 1). For five loci, all of the mutations we recovered cause embryonic lethality (pot, bad, kop, puri, sci); for three loci, both mutant alleles cause larval lethality (pomp, blo, gon); and three loci have both embryonic lethal and larval lethal alleles (bee, bs, sac). In this latter category we may have recovered some amorphic and some hypomorphic alleles in each locus, or the embryonic lethality may be due to a mutation in a second gene. In the case of bee, the two embryonic lethal alleles are also embryonic lethal when transheterozygous, showing that the embryonic lethal is closely linked to bee.

We initially analyzed the cuticle phenotypes of the embryonic lethal mutations (using transheterozygous combinations of mutant alleles and mutant alleles over a deficiency, when available) to determine the overall pattern of the epidermis. For the integrin genes, embryos mutant for if and mew have normal cuticles (BRABANT and BROWER 1993 ; BROWN 1994 ; BROWER et al. 1995 ), and, as discussed previously, embryos mutant for mys have a dorsal hole and, in the absence of the maternal component, a failure in germ-band retraction. Mutations in three of the loci, pot, puri, and sac, do not disrupt the normal pattern of the cuticle, although pot mutant embryos have a faint cuticle (not shown), and we have categorized this group as class II loci (Table 1 and Table 4). Mutations in four loci (bad, kop, bee and sci) cause dramatic defects (Figure 4), which include a dorsal open or tail-up phenotype, resembling embryos mutant for mys, and we called these class I loci. Contrary to previous results suggesting that dorsal closure always occurs in mys mutant embryos and then tears open (WRIGHT 1960 ), we have found that dorsal closure frequently fails (our unpublished results), as it does in the class I loci. The mutations in the remaining loci cause larval lethality, and these loci we termed class III loci. We then assayed the effect of mutations in class I and class II loci on the expression of the PS integrins, using an antibody against the ß_PS subunit encoded by mys (data not shown) and an antibody against the _PS2 subunit (examples are shown in Figure 4, S–U). None of the mutations were found to substantially alter the level of PS integrin expression. Analysis of the muscles, gut, and epidermis by antibody staining of stage-16 embryos revealed defects caused by mutations in the class I loci, but failed to reveal any developmental defects in embryos mutant for pot, puri and sac (data not shown) and so the reason mutations in these class II loci cause embryonic lethality remains unclear.

View larger version (161K): In this window In a new window Download PPT slide   Figure 4. Embryonic phenotypes caused by mutations in class I loci. In A–R the phenotypes caused by mutations in mys and the four class I loci are compared to wild-type embryos: by examining the pattern of the cuticle (first column), by staining the epidermis with anti-fasIII antibody (second column), and by staining the muscles with anti-muscle myosin antibody (third column). The bottom row (S–U) shows embryos stained with an anti-PS2 integrin antibody. The cuticles are from dead 24- to 36-hr-old embryos, while the antibody stained embryos are all stage 16. In all panels anterior is left and ventral is up. Most panels show a lateral view, but B, H, and N, are dorsolateral views to show the dorsal midline. Genotypes of mutant embryos: (D–F) mysXG43/Y; (G–I) badE44/Df(2L)319; (J–L) beeV27/beeV42; (M and N) kopV104/Df(2R)CX; (O) kopV167/Df(2R)CX; (P) sciT1/sciT90; (Q and R) sciT1/Df(3R)2-2; (T) beeV27/beeV42; and (U) sciT1/sciT90.

Of the class I loci, mutations in scorpion (sci) and beerbelly (bee) cause the most consistent developmental defects. Embryos mutant for sci have a very consistent tail-up phenotype (90% of the cuticles from unhatched embryos; Figure 4, P–R), both in sci transheteryzogotes and in hemizygous embryos, sci/Df, (Figure 4P and Figure U vs. Q and R; data not shown), showing that the two mutant alleles are amorphic. In these mutant embryos the germ band fails to retract normally, the embryos remain dorsally open, and the head fails to involute (Figure 4Q). The pattern of the muscles is severely disrupted (Figure 4R), but the muscles appear to remain attached and the PS2 integrin is correctly localized to the ends of the muscles (Figure 4U). The disruption of the muscle pattern is more severe than in mys mutant embryos (Figure 4, D–F), even in the absence of the maternal component (data not shown), and the defective head involution phenotype is also not found in mys mutant embryos. Embryos transheterozygous for the two strongest bee mutant alleles also have fully penetrant defects (95% of cuticles; Figure 4, J–L). The embryos undergo germ-band retraction but are unable to complete dorsal closure or head involution (Figure 4K), and the gut does not become normally constricted and therefore frequently extrudes through the dorsal hole as a large sphere (Figure 4L). Mutations at this locus also cause a defect in muscle fusion since many unfused myoblasts are observed at this late stage (Figure 4L and Figure T); however, some of the ventral muscles are formed and attached and the PS2 integrin is localized properly to the ends (Figure 4T).

Both bladderwrack (bad) and kopupu (kop) mutant embryos have more variable developmental defects. Mutations at both loci cause much stronger phenotypes over a deficiency than when transheterozygous, suggesting that we have not recovered amorphic alleles of these loci. However, embryos homozyogus for deficiencies of these genes also have variable phenotypes (data not shown), suggesting that even amorphic alleles would give variable defects. For both loci, the transheterozygous combinations of mutant alleles cause embryonic lethality but with relatively mild defects. Thus, the kop mutant embryos in general look normal, but a few have a tail-up phenotype. Hemizygous mutant embryos, kop/Df(2R)CX, have a range of phenotypes from apparently normal to the strongest phenotype, in which the embryo is much shorter, fails to involute the head, and has very abnormal germ-band retraction and dorsal closure and a disrupted pattern of muscles (Figure 4, M–O). In addition, in some kop mutant embryos parts of the abdominal segments appear to be missing, leading to missing or fused denticle bands in the cuticle preparations (not shown). The embryos transheterozygous for two bad mutant alleles have a low penetrance tail-up phenotype and a higher penetrance of an anterior open phenotype (data not shown). When these mutant alleles are hemizygous over Df(2L)319, the frequency of the tail up phenotype in the mutant embryos is much higher (60%, a frequency similar to that caused by the homozygous deficiency) and the abnormal head involution and dorsal closure is much clearer (Figure 4G and Figure H). The muscle pattern looks relatively normal although the muscles appear thinner and the pattern is mildly disrupted (Figure 4I). The localization of the PS2 integrin to the ends of the muscles appears normal (not shown).

In summary, the seven novel loci result in a range of phenotypes, none of which exactly mimics the phenotypes caused by integrin mutations, although the phenotypes of the four class I loci share some similarities. Thus, bee and sci mutations cause phenotypes that are reminiscent of the mys mutant phenotype, but more extensive in their disruption of morphogenetic events. Mutations in the loci bad and kop cause some integrin-like phenotypes when mutant alleles are hemizygous, and, therefore, mutations in all four of these class I loci cause phenotypes similar to those caused by mys mutations, defects in germ-band retraction, and dorsal closure. They differ, however, in also causing defects in head involution and in not disrupting the attachment of the muscles. Mutations in the three class II loci, pot, puri, and sac, cause embryonic lethality but do not cause any dramatic defects in embryonic morphogenesis. Mutations in the remaining four class III loci, bs, blo, gon, and pomp, cause larval lethal phenotypes.

One possible reason some of these loci might have relatively mild mutant phenotypes is that they have a substantial maternal component that suffices for much of embryonic development. To test this we made germline clones of one mutant allele from each of the new loci using the FLP-FRT system combined with the dominant female sterile mutation ovo^D1 (CHOU and PERRIMON 1992 , CHOU and PERRIMON 1996 ) and analyzed the lethality and cuticle phenotypes of embryos derived from the germline clones of each mutation, which thus lacked any maternal contribution. We did not see substantial enhancement of the mutant phenotypes of any of the loci, and all mutations were fully rescued by a paternal wild-type allele of the gene (data not shown). This differs from the mutant phenotype of mys, which is enhanced by the absence of the maternal component, but is similar to mutations in if and mew, which are not affected by the absence of maternal gene activity.

Allelism with other wing blister loci:
While this manuscript was in preparation, a similar FLP-FRT screen for mutations that cause wing blisters was reported (PROUT et al. 1997 ). To determine the extent of overlap between the two screeens we performed simple complementation tests with their mutations, which were isolated from the autosomes. On 2L we found that bladderwrack (we recovered 3 alleles) appears to be allelic to cassowary (they recovered 5 alleles), and pompholyx (2 alleles) appears to be allelic to pygoscelis (6 alleles; previously called penguin). The single alleles on this arm, 2L-A and 2L-F, of PROUT et al. 1997 each fail to complement 1 of the 4 single mutant alleles that we isolated, resulting in the identification of two additional complementation groups, which we have renamed bubblewing (bub, replacing 2L-A) and blisterwing (blis, replacing 2L-F). Our balancing strategy meant that we would not have retained mutations in dumpy as they did and we did not recover alleles of Ostrich. There is also extensive overlap on 2R between the two screens. Mutations in many of the loci on 2R were recovered with a similar frequency in the two screens: both isolated blistered alleles (4 vs. 2 alleles); bloated (2 alleles) appears to be allelic to kitikete (3 alleles); gonfle (2 alleles) appears to be allelic to auk (3 alleles); and sac (2 alleles), to moa (4 alleles). A single mutant allele of ours appears to be allelic to piopio (2 alleles). Mutations in other loci on 2L were recovered at quite different rates: one of our single-mutant alleles appears to be allelic to kiwi (17 alleles); kopupu (19 alleles) appears to be allelic to kakapo (5 alleles); and beerbelly (9 alleles), to takehe (1 allele). PROUT et al. 1997 did not recover any alleles of puri (4 alleles). We do not have any alleles of mastermind (2 alleles), xenecid (1 allele), 2R-F (1 allele), or 2R-L (1 allele). It is possible that 1 of the other 5 single-mutant alleles that we recovered could have been allelic, but they have been lost. There is no overlap between the two screens on the third chromosome; scorpion alleles complement the two single alleles PROUT et al. 1997 recovered on 3R that had not been mapped, 3R-B and 3R-C, and the other two loci they recovered map to different positions. The phenotype of 3R-B is very similar to that caused by mutations in the previously identified locus blistery (GLASS 1934 ), and we found that it fails to complement blistery¹. The relative rates at which each group recovered mutations on each autosomal arm are similar: 2L 9/24 (ours/theirs); 2R 48/42; 3L 0/3; 3R 2/8. This could reflect the relative density of wing blister genes on the different arms, but may also reflect the efficiency of recovering mutations that cause blisters in clones. The high degree of overlap between the two screens on the second chromosome emphasizes the selectivity of the screen and suggests that we may be close to identifying all of the loci on this chromosome that can be mutated to give this phenotype. The small number of mutations isolated on the third chromsome and lack of overlap suggests that additional genes are likely to exist. Nonetheless, these screens have provided a large set of genes required for adhesion between the two wing surfaces.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Identification of 14 complementation groups essential for the adhesion of the two surfaces of the adult wing:
The aim of our screen was to identify genes that are involved in integrin-mediated adhesion. We screened for a phenotype caused by mutations in each of the three integrin subunits _PS1, _PS2, and ß_PS: namely a failure in the adhesion of the two wing surfaces. In this screen we identified 14 recessive complementation groups of mutations that cause wing blisters in wing clones and, when homozyogus, cause lethality during embryonic or larval stages. These included mutations in the three integrin genes, as expected, and mutations in two other previously identified wing blister genes: blistered, which encodes the Drosophila serum response factor (MONTAGNE et al. 1996 ); and bloated (WADDINGTON 1939 , WADDINGTON 1940 ), which has not yet been cloned. The remaining 9 complementation groups do not appear to be allelic to any mutations that cause wing blisters and were identified prior to the development of the FLP-FRT methodology for efficiently making mosaics; thus they may represent novel loci. Of the 9, 6 were also identified in the similar screen performed by PROUT et al. 1997 . All of the new loci are required on both sides of the wing and therefore appear to be required for both PS1 and PS2 integrin-mediated adhesion.

When we examined the homozygous phenotypes caused by the mutations, we found that we could separate them into three groups. Members of the first group, consisting of mutations in bad/cass, bee/tak, kop/kak, and sci, cause embryonic lethality and one prominant phenotype that is also caused by mutations in mys: a failure in germ-band retraction. Mutations in the second group also cause embryonic lethality, although tissue morphogenesis appears to occur normally, and this group includes the loci pot, puri, and sac/moa. The third group, consisting of pomp/pyg, blo/kit, bs, and gon/auk, have mutant alleles that cause larval lethality and also do not cause obvious defects in tissue morphogenesis. These latter two classes are more similar to the phenotype caused by mutations in mew, which are largely larval lethal although they cause a clear defect in gut morphogenesis. Mutations in the first group cause additional phenotypes, such as defects in head involution, which are not caused by embryonic lethal if and mys mutations, suggesting that the products of this group are required for other functions in addition to PS integrin-mediated adhesion.

How many wing blister loci are directly involved in PS integrin function?
The isolation of mutations in the gene blistered in our screen shows that the products of the loci recovered in this screen may be involved in the specification of intervein cell fate. The product of the blistered gene is the Drosophila homologue of serum response factor transcription factor, which is required to promote intervein cell development (MONTAGNE et al. 1996 ). In bs mutant wings, intervein tissue is converted toward vein, which normally does not show adhesion between dorsal and ventral surfaces of the wing, and this results in blisters (FRISTROM et al. 1994 ; MONTAGNE et al. 1996 ). Genetic interactions between mutant alleles of bs and mys and if (WILCOX 1990 ; FRISTROM et al. 1994 ) and mew (this study) cause an increase in the penetrance of intervein blisters, but do not enhance vein defects. The dominant genetic enhancement of integrin mutations by bs may occur through a partial transformation to vein fate in the absence of one copy of bs, which results in a reduction in adhesion in the transformed interveins, possibly due to a reduction of integrin expression. In pupal wings, PS integrin expression is normally absent from the veins (FRISTROM et al. 1993 ). This suggests that blistered is upstream of the integrins, and the maintenance of integrin expression is one important aspect of intervein fate. None of the mutations in the other genes identified in our screen appears to be causing blisters due to a similar change of fate from intervein to vein, because they did not show a detectable transformation in the mutant clones.

Mutations in two genes, bloated/kitkete and pompholyx/pygocelis, cause defects that suggest that these genes encode proteins involved in cell-cell adhesion within each layer of the disc as well as adhesion between dorsal and ventral layers. In wings containing clones of cells homozyogous for mutations in either of these loci, vesicles of cells are observed that have delaminated from the wing surfaces, a phenotype that was initially observed to be caused by the first, viable blo mutant allele (WADDINGTON 1939 , WADDINGTON 1940 ). This suggests that the adhesion between cells is weak during pupal development and occasionally clumps of cells become dissociated from the epithelium. The only protein that is known to play a role in epithelial cell adhesion is the cadherin encoded by the shotgun gene (TEPASS et al. 1996 ; UEMURA et al. 1996 ), which has not yet been examined in mutant wing clones. The vesicles of cells produced in the blo mutant wings can induce ectopic veins, but, in general, blo mutant clones are not transformed into a vein fate. Clones of lethal blo alleles also cause changes in the shape of the wing, suggesting that its product may link adhesion, patterning, and/or growth in the wing.

This leaves eight genes for which mutant clones only cause a loss in adhesion between the two surfaces of the wing. It is possible that some of these gene products are involved in an as yet uncharacterized pathway that is required for adhesion of the two wing surfaces, but does not involve the PS integrins. However, the simplest interpretation is that the products of each of these genes are components of integrin-mediated adhesion.

Possible functions of the products of the wing blister loci in integrin-mediated adhesion:
The link between the two sheets of cuticle that form the mature wing is constructed during puparation (FRISTROM et al. 1993 ). At the apical surface of each epidermal cell, the plasma membrane is linked to the cuticle by an apical hemiadherens junction. This junction serves as an organizing center for the microtubules that extend from the apical to basal surfaces, which with their associated actin filaments form transalar arrays (see FRISTROM et al. 1993 for references). The PS integrins are localized at the basal surface, where we think that they link to the opposite layer of cells via components of the extracellular matrix. The genes identified in this screen could encode any part of this link, including components of the apical junction as well as those more directly connected to integrins at the basal surface. Integrin-mediated adhesion between the two basal surfaces of the wing could potentially require proteins that have a wide variety of functions: (1) extracellular ligands, which may be secreted proteins or other transmembrane proteins; (2) intracellular proteins that link the integrins to the cytoskeleton; (3) intracellular proteins that signal to activate integrins into a high affinity state; and (4) intracellular proteins involved in transmitting signals that are essential for adhesion. The screen described here can only lead to the isolation of some of these proteins. For example, mutations in genes encoding secreted extracellular ligands are likely to cause nonautonomous phenotypes: a mutant clone of cells would be rescued by secretion of the ligand from the surrounding wild-type cells. The screen would also not identify genes that encode products that are involved not only in integrin adhesion but also a process that is essential for cell survival or division, because mutant clones of cells would not be produced. We did not recover loci in a previously described gene with a wing blister phenotype, wing blister (LINDSLEY and ZIMM 1992 ), suggesting that this gene falls into one of the above classes.

This screen should have succesfully identified genes that encode cytoplasmic proteins that are essential and fairly specific for integrin functions. Thus we anticipate that the eight loci identified in this screen that appear to be specifically involved in adhesion of the two surfaces will encode proteins involved in several processes. The most likely types are proteins that link the PS integrins to the cytoskeleton and proteins essential for activating integrins to a high affinity state. If integrin signaling should prove to be an essential part of adhesion in the wing, then we would expect to isolate mutations in the genes encoding this signaling pathway. Finally, we may recover proteins that are transmembrane ligands for the PS integrins or proteins that link the apical surface of the wing to the cuticle. Therefore, the next goal in this study is to clone these genes to allow a molecular characterization of their roles in adhesion between the two surfaces of the wing.

	ACKNOWLEDGMENTS

We thank John Overton for excellent technical assistance, Danny Brower for sending us the mew^M6 allele, Jose de Celis for advice and stocks for making marked wing clones, Mary Prout and Jim Fristrom for sending alleles from their screen, and Andrea Knox for testing allelism with blistery. We thank the Bloomington Stock Centre, Umea Stock Centre, and the Tübingen Stock Centre for fly stocks, and Sarah Bray, S. Gregory, M. Martin-Bermudo, and R. Smith for critical comments on the manuscript. This work was supported by a studentship from the Biological and Biotechnological Sciences Research Council to E.P.W. and a Wellcome Trust Senior Research Fellowship to N.H.B.

Manuscript received January 21, 1998; Accepted for publication July 13, 1998.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

ALATORTSEV, V. E., I. A. KRAMEROVA, M. V. FROLOV, S. A. LAVROV, and E. D. WESTPHAL, 1997  Vinculin gene is non-essential in Drosophila melanogaster. FEBS Lett. 413:197-201[Medline].

BLOOR, J. W. and N. H. BROWN, 1998  Genetic Analysis of the Drosophila _PS2 integrin subunit reveals discrete adhesive, morphogenetic and sarcomeric functions. Genetics 148:1127-1142[Abstract/Free Full Text].

BOGAERT, T., N. H. BROWN, and M. WILCOX, 1987  The Drosophila PS2 antigen is an invertebrate integrin that, like the fibronectin receptor, becomes localized to muscle attachments. Cell 51:929-940[Medline].

BRABANT, M. C. and D. L. BROWER, 1993  PS2 Integrin Requirements in Drosophila Embryo and Wing Morphogenesis. Dev. Biol. 157:49-59[Medline].

BROWER, D. L. and S. M. JAFFE, 1989  Requirement for integrins during Drosophila wing development. Nature 342:285-287[Medline].

BROWER, D. L., R. J. SMITH, and M. WILCOX, 1980  A monoclonal antibody specific for diploid epithelial cells in Drosophila. Nature 285:403-405[Medline].

BROWER, D. L., M. WILCOX, M. PIOVANT, R. J. SMITH, and L. A. REGER, 1984  Related cell-surface antigens expressed with positional specificity in Drosophila imaginal discs. Proc. Natl. Acad. Sci. USA. 81:7485-7489[Abstract/Free Full Text].

BROWER, D. L., T. A. BUNCH, L. MUKAI, T. E. ADAMSON, and M. WEHRLI et al., 1995  Nonequivalent requirements for PS1 and PS2 integrin at cell attachments in Drosophila: genetic anlysis of the _PS1 integrin subunit. Development. 121:1311-1320[Abstract].

BROWN, N. H., 1993  Integrins hold Drosophila together. Bioessays 15:383-390[Medline].

BROWN, N. H., 1994  Null mutations in the _PS2 and ß_PS integrin subunit genes have distinct phenotypes. Development 120:1221-1231[Abstract].

BUNCH, T. A., R. SALATINO, M. C. ENGLESJERD, L. MUKAI, and R. F. WEST et al., 1992  Characterization of mutant alleles of myospheroid, the gene encoding the ß subunit of the Drosophila PS integrins. Genetics 132:519-528[Abstract].

BUNCH, T. A., M. W. GRANER, L. I. FESSLER, J. H. FESSLER, and K. D. SCHNEIDER et al., 1998  The PS2 integrin ligand tiggrin is required for proper muscle function in Drosophila. Development 125:1679-1689[Abstract].

CHOU, T. B. and N. PERRIMON, 1992  Use of a yeast site-specific recombinase to produce female germline chimeras in Drosophila. Genetics 131:643-653[Abstract].

CHOU, T. B. and N. PERRIMON, 1996  The autosomal FLP-DFS technique for generating germline mosaics in Drosophila melanogaster. Genetics 144:1673-1679[Abstract].

CLARK, E. A. and J. S. BRUGGE, 1995  Integrins and signal transduction: the road taken. Science 268:233-239[Abstract/Free Full Text].

DE LA POMPA, J. L., J. R. GARCIA, and A. FERRUS, 1989  Genetic analysis of muscle development in Drosophila melanogaster. Dev. Biol. 131:439-454[Medline].

FOGERTY, F. J., L. I. FESSLER, T. A. BUNCH, Y. YARON, and C. G. PARKER et al., 1994  Tiggrin, a novel Drosophila extracellular matrixprotein that functions as a ligand for Drosophila _PS2ß_PS integrins. Development 120:1747-1758[Abstract].

FRISTROM, D., M. WILCOX, and J. FRISTROM, 1993  The distribution of PS integrins, laminin A and F-actin during key stages in Drosophila wing development. Development 117:509-523[Abstract].

FRISTROM, D., P. GOTWALS, S. EATON, T. KORNBERG, and M. STURTEVANT et al., 1994  blistered: a gene required for vein/intervein formation in wings of Drosophila. Development 120:2661-2671[Abstract/Free Full Text].

FURATA, Y., D. ILIC, S. KANAZAWA, N. TAKEDA, and T. YAMAMOTO et al., 1995  Mesodermal defect in late-phase of gastrulation by a targeted mutation in focal adhesion kinase, FAK. Oncogene 11:1989-1995[Medline].

FYRBERG, E., M. KELLY, E. BALL, C. FYRBERG, and M. C. REEDY, 1990  Molecular genetics of Drosophila alpha-actinin: mutant alleles disrupt Z disc integrity and muscle insertions. J. Cell Biol. 110:1999-2011[Abstract/Free Full Text].

GELBART, W. M., M. CROSBY, B. MATTHEWS, W. P. RINDONE, and J. CHILLEMI e