Abstract
The common subunit of the PS antigens of Drosophila is homologous with vertebrate integrins and is encoded by the lethal(1)myospheroid gene. We have generated flies mosaic for wild-type and mutant alleles of lethal-(1) myospheroid using adult gynandromorphs and radiation-induced somatic crossing over. The defects observed in the gynandromorphs demonstrate widespread requirements for PS integrins during development especially in ventrally derived structures, which also show strong expression of PS β integrin. Smaller lethal(1)myo-spheroid clones induced during larval development result in blister and vein defects in the wings and aberrant development of photoreceptor cells, demonstrating roles for PS integrins during development of both wings and eyes. PS integrins are required for the close apposition of the dorsal and ventral wing epithelia and for the proper arrangement of photoreceptor cells. However, many other adhesive and morphogenetic processes proceed normally in the absence of integrins containing the fl subunit encoded by lethal(1)myospheroid.
Introduction
The sequence of morphogenetic events observed during the formation of a multicellular organism often requires changes in cell adhesion, shape, migration and proliferation. In order to understand such basic developmental mechanisms, it is important to identify the molecules that mediate them. The integrins, a family of transmembrane glycoprotein receptors, have been associated with these functions. Integrins are heterodimers consisting of noncovalently associated crand subunits. Their extracellular domains bind to adhesive molecules such as fibronectin, laminin and collagen, and their intracellular domains interact with the cytoskeleton (see Hynes, 1987; Buck and Horwitz, 1987; Ruoslahti and Pierschbacher, 1987; and Ruoslahti, 1988 for reviews). The combination of a specific a subunit with a specific subunit has been shown to be important in determining the affinity for specific ligands. As many as 11 distinct α subunits and 5 distinct β subunits have been identified in vertebrate cells (Hynes, 1987; Ruoslahti and Pierschbacher, 1987; Sonnenberg et al. 1988; Cheresh et al. 1989; Hemler et al. 1989; Kajiji et al. 1989).
The Drosophila position-specific (PS) antigens are integrin subunits. PSI and PS2 are asubunits (Leptin et al. 1987; Bogaert et al. 1987) each of which binds to a common β subunit known as PS3 or PS α (Wilcox et al. 1984; Leptin et al. 1987, 1989). PS integrins are present during most of embryonic development, but are concentrated in specific embryonic tissues. PS2 α mainly found in mesodermal derivatives, while PSlcr is in ectodermal and endodermal derivatives and PS β is found in all three germ layers (Bogaert et al. 1987; Leptin et al. 1989).
During later stages, the PS integrins localize to specific regions of the eye-antennae, wing and leg disk epithelia, as well as to other larval and adult tissues (Wilcox et al. 1981; Brower et al. 1984, 1985; and Wilcox and Leptin, 1985). In the wing and eye-antennae imaginai discs, PSl α and PS2 α are found in complementary patterns, which suggest they may cooperate in their functions. In the wing disc, PSlaand PS2trare expressed on the dorsal and ventral surfaces respectively, while PS/3 is found throughout the disc (Brower et al. 1985). This distribution suggests that integrins may be involved in bringing together or maintaining the apposition of the dorsal and ventral surfaces of the wing at metamorphosis. This event is thought to be necessary for the proper shaping and organization of the wing as well as the normal patterning of wing crossveins (Waddington, 1940; Garcia-Bellido, 1977). Development of the eye is accompanied by an intriguing shift in PS integrin expression. During the third larval instar, a morphogenetic furrow travels across the eye disc; undifferentiated epithelium ahead of this furrow develops after passage of the furrow into the organized pattern of ommatidia characteristic of the adult eye (Ready et al. 1976). PSl α and PS2 α are expressed on opposite sides of this morphogenetic furrow (Brower et al. 1985) suggesting roles for integrins in the normal development of the eye disc.
The advantages of genetic analysis in Drosophila make it possible to study the functions of integrins in vivo. Genomic DNA clones of the lethal(1)myospheroid locus have been isolated (Digan et al. 1986). Sequence analysis of cDNA clones corresponding with this gene has shown that it encodes a protein homologous to vertebrate ft integrins (MacKrell et al. 1988) and studies of PS antigen expression in l(1)mys embryos show that this gene encodes the PS/3 subunit (Leptin et al. 1989). The l(1)mys mutation has allowed an initial examination of the role of fi integrins in Drosophila development. Embryos hemizygous or homozygous for loss of function l(1)mys mutations can develop approximately normally through a large portion of embryonic development. The first defects in mutant embryos are observed when the first muscle contractions normally occur. During this time the somatic muscles pull away from their attachment sites. Also, distortion of visceral mesoderm and gut epithelia, herniation of brain tissue and abnormal dorsal closure are observed in these embryos. These abnormalities result in embryonic death (Wright, 1960; Newman and Wright, 1981). In addition to the previously mentioned defects, removal of both maternal and zygotic 1(1)mys expression, resulting in a complete absence of integrin, causes abnormal germband retraction (Wieschaus and Noell, 1986; Leptin et al. 1989).
Because of the embryonic lethal phenotype of l(1)mys, it has not been possible to study the functions of integrins in later stages of Drosophila development. Cell adhesion undoubtedly plays important roles in later morphogenesis and the presence of PS integrins in imaginai discs (Brower et al. 1985) suggests that they function there. In order to address questions of possible functions of integrins in later stages of development we have used Drosophila genetic methods to generate mosaic animals that contain patches of wild-type and l(1)mys tissue. In this way, the functions of integrins in particular regions can be examined in the context of an organism with normal integrin expression elsewhere.
In the experiments described below, we produced adult l(1)mys gynandromorphs and l(1)mys somatic clones and have examined the requirements for integrins during development. The location of these requirements in the developing embryo is compared with l(1)mys transcription patterns and with the distribution of ft integrin in the developing embryo. We show widespread requirements for l(1)mys expression in the developing embryo and demonstrate roles for integrins in the development of both wings and eyes.
Materials and methods
Mutant strains and definition of PS β integrin
The lethal(1)myospheroid alleles used in these experiments, l(1)mysXG43 and l(1)mysXBS7, were induced by EMS (Wies chaus et al. 1984). Late embryos homozygous for 1(1 )mysXC43 or l(1)mysXBS7 do not produce immunologically detectable PS3 P subunit (Leptin et al. 1989) and behave as loss-of-function alleles in various complementation tests (Zusman, unpublished results). Mutant chromosomes carrying l(1)mysXC43, l(1)mys*B87 or the marker mutations, yellow, white or forked, alone or in various combinations, were kept over the balancer Fm7a (Merriam and Duffy, 1972). For a detailed description of the marker mutations, see Lindsley and Grell (1968). PS/3 integrin will refer to the P subunit which is encoded by the l(1)mys gene.
Clones for lethal(1)myospheroid
A genomic phage encompassing the lethal(1)myospheroid gene was generously provided by Dr S. Haynes (Digan et al. 1986) and used to isolate cDNA clones from a λ gtll library prepared from 0-16 h embryos (a generous gift of Drs K. Zinn and C. Goodman). Genomic and cDNA clones were analyzed by restriction mapping, subcloning and sequencing (Patel-King and Hynes, unpublished data). Sequences match those previously described by MacKrell et al. (1988).
RNA isolation, Northern blot analysis
RNA was isolated from staged wild-type embryos using a phenol/chloroform extraction method (Ayme and Tissieres, 1985). Total RNA from each of the indicated stages was electrophoresed in 1.0% agarose/formaldehyde gels containing 50 mM Hepes pH 7.8 and IIDM EDTA, and RNA sizes were determined by comparison with an RNA ladder (Bethesda Research Laboratories). Separated RNAs were transferred to nylon membranes (Zetaprobe, Biorad.) and hybridized as described in DeSimone and Hynes (1988) with a [32P]-labeled 3.3 kb l(1)mys cDNA probe covering the entire coding region. Blots were washed in 2 ×SSC/0.1% SDS at 65 °C for 1h prior to autoradiography. Equal loading and transfer of the RNA samples was checked by ethidium bromide staining of the gel and the blotted nylon filter.
In situ hybridization to sectioned embryos
Embryos were collected from females on agar plates coated with yeast paste and were dechorionated, fixed, dehydrated and embedded in paraffin as described in Ingham et al. (1985). 7 micron sections were prepared and processed as described in ffrench-Constant and Hynes (1988) using either 35S-labeled antisense or sense RNA transcripts of a 1.1 kb. l(1)mys cDNA probe from the coding region as hybridization probes.
Preparation of embryos for whole mounts and antibody staining
0 –16 h embryos were collected from Oregon R females at 25°C and were dechorionated, fixed and prepared for antibody reaction as described by Zusman and Wieschaus (1987), except that fresh paraformaldehyde was used for fixation. Embryos were mounted in a 3:1 solution of methyl salicylate and Canada Balsam, and were viewed under bright-field optics on a Zeiss Axiophot microscope.
The β-specific antiserum reported by Marcantonio and Hynes (1988) was used in this study to examine the distribution of p integrin on Drosophila embryos. This antiserum was raised against a synthetic peptide corresponding to the COOH-terminal end of chicken β1 integrin. Previous studies have shown that this antiserum recognizes α βintegrin complexes in Drosophila cells (Marcantonio and Hynes, 1988). Staining performed in the absence of primary serum or with immune serum in the presence of competitor peptide gave very little signal.
Production of adult gynandromorphs
l(1)mys gynandromorphs were collected from crosses in which males carrying the unstable ring X chromosome ln(1)wvc (Hinton, 1955; Hall et al. 1976) were mated toyl(1)myslFm7a females. The ring X males also had a Y-linked duplication of the Notch region to increase survival (y+w+N*Y), although this is not essential to the experiment. Random loss of the ring X chromosome during early development produces patches of cells, which are y l(1)mys/-, that is, are male cells with the yellow cuticle marker and are lacking a wild-type l(1)mys gene. l(1)mys gynandromorphs in which both maternal and zygotic l(1)mys+ expression were removed from mutant tissue were produced by crossing y l(1)mys/OvoD females containing homozygous y 1(1 )myrC43f germ cells to ring X males. Ovo13 is a dominant female sterile that prevents egg formation (Jimenez and Campos-Ortega, 1982; Garcia-Bellidoand Robbins, 1983; Perrimon et al. 1984). When l(1)mys/OvoD female larvae are irradiated to induce mitotic recombination, some germ cells become homozygous for l(J)mys and lose the OvoD gene. Germline clones were identified using the procedures described by Zusman and Wieschaus (1985). A y w f/Fm7a or y w f/OvoD control cross was run simultaneously with each experiment using ring X individuals from the same bottles used in the experimental crosses.
All adult gynandromorphs were examined under a dissecting microscope or prepared for microscopic examination using the methods described by Szabad et al. (1979). The genotype of 20 landmark structures was recorded and the left and right sides of each individual were scored independently. A ‘maleness average’ score was determined for each landmark structure as the number of times a structure was mutant divided by the number of times that structure was scored. Significant differences between values were detected using x2 contingency tables (P<0.05).
Mitotic clones
Mitotic recombination was induced by gamma irradiation (1500 rads) of y l(1)mysXC43/OvoD, y Ifljmys097/OvoD, w 1(1 )mysXC43/ OvoD or y w//OvoD larvae, between 48 and 66 h old. The OvoD chromosome was used in these experiments, since flies containing germline clones could also be used to generate gynandromorphs (see above). Wings from flies containing chromosomes marked with yellow and forked were embedded in Faure’s mountant and examined under bright-held optics on a 2Leiss Axiophot microscope. Heads containing w l(1)mys clones were removed, submerged in immersion oil and examined under antidromic illumination (Francescini, 1975) or were fixed and embedded in JB4 (Polysciences Inc,) as described in Mahowald et al. (1979). 4 micron sections were cut using a Leitz 1516 microtome, stained with methylene blue, and viewed under bright-held or Nomarski optics.
Results
Expression of the lethal(1)myospheroid gene
The lethal(1)myospheroid gene is expressed both maternally and zygotically (Digan et al. 1986; Wieschaus and Noell, 1986). To investigate further the pattern of l(1)mys expression during embryonic development, Northern blots of RNA from various embryonic stages were probed with a 3.3 kb l(1)mys cDNA. These blots reveal two major mRNAs of sizes 3.8 and 4.4 kb in 0 –4 h embryos. These two mRNA species could represent the existence of multiple transcripts or alternative splicing. During subsequent stages of development, we observed only the 4.4 kb mRNA. The level of l(1)mys RNA drops significantly in 4 –8 h embryos (extended germbands) and again becomes abundant in 8-14 h embryos (shortened germbands, dorsal closure; Fig. 1). l(1)mys RNA continues to be observed throughout embryonic development.
Northern blot of embryonic RNAs probed with a 3.3 kb 32P-labeled l(1)mys cDNA. Each lane contains 10 μg total RNA. The number at the top of each lane represents the age of embryos (hours after fertilization) from which RNA was extracted. RNA sizes are given in kb and were determined by comparison with the migration of marker RNAs.
Northern blot of embryonic RNAs probed with a 3.3 kb 32P-labeled l(1)mys cDNA. Each lane contains 10 μg total RNA. The number at the top of each lane represents the age of embryos (hours after fertilization) from which RNA was extracted. RNA sizes are given in kb and were determined by comparison with the migration of marker RNAs.
To examine the spatial distribution of 1(1 )mys mRNA during embryonic development, we hybridized 35S-labeled l(1)mys antisense and sense RNAs to paraffin sections of 0 –24 h embryos. Hybridization was seen with the antisense probe at all stages examined, while none was seen in control experiments using the sense probe. Fig. 2 shows that, during the first 9 nuclear divisions of development (see Foe and Alberts, 1983 for review of nuclear divisions), l(1)mys RNA is uniformly distributed throughout the embryo (Fig. 2A, B). As the nuclei migrate to the periphery of the embryo (nuclear division 12-13), the transcripts also move to the periphery of the embryo (Fig. 2C, D). By the cellular blastoderm stage, l(1)mys transcripts are localized to all the newly formed cells. As in the Northern blot results, the in situ hybridization experiments also demonstrate a temporary decrease in RNA levels when the germband is fully extended (Fig. 2E, F). Abundant RNA levels are again observed during germband retraction (Fig. 2G, H) and also during later stages (data not shown). Thus l(1)mys transcripts were found in most, if not all, embryonic tissues.
Localization of l(1)mys transcripts in embryos. Bright-field (left) and corresponding dark-field (right) photographs were taken after autoradiography. Autoradiography and exposure for all embryos was 10 days. The montage shows appropriately staged embryos from a single experiment photographed and enlarged at the same magnification so that RNA levels at particular stages can be compared. Orientation of embryos is anterior left and dorsal up. (A,B) syncytial blastoderm (stage 2); (C,D) cellular blastoderm (stage 5); (E,F) gastrulation with a fully extended germband (stage 10); (G,H) embryo with shortened germband (stage 13). See Campos-Ortega and Hartenstein (1985) for a description of the stages.
Localization of l(1)mys transcripts in embryos. Bright-field (left) and corresponding dark-field (right) photographs were taken after autoradiography. Autoradiography and exposure for all embryos was 10 days. The montage shows appropriately staged embryos from a single experiment photographed and enlarged at the same magnification so that RNA levels at particular stages can be compared. Orientation of embryos is anterior left and dorsal up. (A,B) syncytial blastoderm (stage 2); (C,D) cellular blastoderm (stage 5); (E,F) gastrulation with a fully extended germband (stage 10); (G,H) embryo with shortened germband (stage 13). See Campos-Ortega and Hartenstein (1985) for a description of the stages.
The distribution of βintegrin in the developing embryo
Antiserum against a synthetic vertebrate β integrin peptide (Marcantonio and Hynes, 1988) was used to examine the location of β integrin protein in developing embryos. 0 –16 h embryos were fixed, incubated with the antibodies, followed by anti-rabbit antibody complexed to HRP and examined in whole-mount preparations. The staining patterns described below were found only when embryos were incubated with immune serum and were not observed when embryos were incubated with preimmune serum or with immune serum plus the peptide used in the production of antisera (data not shown).
The first observable β integrin appears during the early stages of gastrulation. In wild-type embryos, gastrulation begins with the formation of the ventral furrow, which brings the future mesoderm into the interior of the embryo. The resulting ventral tube of presumptive mesoderm and the immediately surrounding ectoderm comprise the germband. Also, during this stage of development the posterior and anterior midgut primordia become apparent. The earliest definitive staining for β integrin is seen about 20 min after the onset of gastrulation (Fig. 3A). A dispersed staining pattern is especially obvious in the germband and posterior and anterior midgut primordia. As the germband extends, β integrin appears to become concentrated in the area of the germband between mesoderm and ectoderm (Fig. 3B). Previous studies of sections of embryos stained with PSβ antibodies suggest that this staining pattern reflects a large concentration of integrin on opposing basal surfaces of these two epithelial sheets (Leptin et al. 1989). integrin is also observed during this stage around the stomodeum and proctodeum, the primordia of the larval hindgut and foregut, respectively (Fig. 3C). As the germband shortens, β integrin becomes concentrated in the round balls of somatic mesoderm (Fig. 3D). During this and subsequent stages, it is also concentrated around the foregut, hindgut and midgut (Fig. 3E, 3F). Little or no staining, however, was observed in the lateral hypoderm. As previously described by Leptin et al. (1989), we observed f integrin in approximately 14 h old embryos localizing to the attachment sites of the body wall muscles (not shown). Little staining was observed in the central nervous system. These data indicate widespread expression of β integrin in the embryo.
Whole mounts of wild-type embryos stained with an antiserum produced against a β integrin peptide. Antibody localization was visualized using biotinylated goat antirabbit antibodies bound to an HRP complex. Note the large concentration of β integrin in the germband (gb) and its derivatives (i.e. sm; arrows) and in the gut primordia (i.e. pr, st, amg, and pmg; arrows). Orientation of embryos is anterior left and dorsal up. (A) gastrula approximately 20 min after the onset of gastrulation; (B) gastrula with a fully extended germband; (C) embryo shortly after the onset of segmentation; (D and E) embryos with shortened germbands; (F) embryo shortly after dorsal closure, gb, germband; amg, anterior midgut; pmg, posterior midgut; st, stomodeum; pr, proctodeum; sm, somatic mesoderm; mg, midgut; hg, hindgut.
Whole mounts of wild-type embryos stained with an antiserum produced against a β integrin peptide. Antibody localization was visualized using biotinylated goat antirabbit antibodies bound to an HRP complex. Note the large concentration of β integrin in the germband (gb) and its derivatives (i.e. sm; arrows) and in the gut primordia (i.e. pr, st, amg, and pmg; arrows). Orientation of embryos is anterior left and dorsal up. (A) gastrula approximately 20 min after the onset of gastrulation; (B) gastrula with a fully extended germband; (C) embryo shortly after the onset of segmentation; (D and E) embryos with shortened germbands; (F) embryo shortly after dorsal closure, gb, germband; amg, anterior midgut; pmg, posterior midgut; st, stomodeum; pr, proctodeum; sm, somatic mesoderm; mg, midgut; hg, hindgut.
Localized requirements for zygotic and maternal gene activity
The mRNA detected at the early stages is presumably largely maternal, whereas later stages also contain embryonic transcripts. Both maternal and embryonic expression can contribute to development. The maternal mRNA is not absolutely required since lethal-β )myospheroid+ embryos derived from homozygous l(1)mys oocytes are normal (Leptin et al. 1989). Nevertheless, maternal l(1 )mys expression has been shown to produce functional β integrin that can contribute to the developing embryo. This is demonstrated by the observations that: (1) extra maternal copies of l(1)mys+ can partially rescue the phenotype resulting from loss of zygotic expression (Wieschaus and Noell, 1986); and (2) removal of the maternal contribution produces a more severe defect in embryos that are l(1)mys homozygotes (Wieschaus and Noell, 1986; Leptin et al. 1989).
We conducted two series of experiments to address the role of zygotic and maternal PS integrin expression. To determine whether the zygotic expression of 1(1 )mys is required in all regions of the embryo for normal viability, gynandromorphs with l(1)mys+ and l(1)mysXC43 or l(1)mysB87 patches of tissue were produced from ring X/ y l(1 )mys flies by random loss of the ring X chromosome containing the wild-type l(1)mys gene (Hall et al. 1976, see Materials and methods). Only 7% of the l(1)mys gynandromorphs survived to the adult stage, whereas control l(1)mys+ gynandromorphs showed normal viability. In the surviving l(1)mys mosaic animals, mutant patches were small and covered no more than 10 % of the fly, while control patches constituted an average of 50 % of the fly. Mutant tissue observed in the head, thorax and abdomen was often normal in appearance. However, wing blisters, missing legs or leg parts, tergite defects and missing pieces of cuticle were frequently observed in l(1)mys gynandromorphs, whereas such defects were rare in l(1)mys+ gynandromorphs (Table 1). The largest mutant patches constituting approximately 10 % of the fly were in the dorsally located tergites and could be normal or defective in appearance. These observations suggest that, while PS/3 integrin is not required in every cell of the developing fly for viability, it appears to influence or be necessary for the development of many adult structures including the wings, legs and tergites.
To determine localized requirements for PS/3 integrin in particular regions of the embryo, we measured the frequency with which 20 adult structures in l(1)mysXG43 and control mosaics were mutant. Since such mutant patches are made up of male (XO) cells, this is called the maleness average score. The maleness average scores calculated for structures in control mosaics give the probability that a particular structure will be mutant, if there is no lethality associated with loss of the wild-type gene; reduced scores in the l(1)mys gynandromorphs indicate lethality. Table 2 shows significant lethality associated with all scored structures in l(1)mys mosaics suggesting that most large patches of mutant tissue result in lethality. There is a particularly strong bias for lethality in those mosaics that have mutant cells in ventrally derived structures. For example, in the abdominal segments of surviving mosaics, the ventrally located stemites were found to be mutant at a significantly lower frequency than the more dorsally located tergites. In addition, marked clones lacking l(1)mys expression were not observed in the legs, the most ventrally located structures scored. Finally, no l(1)mys mosaic borders were observed to cross the ventral midline; however, they crossed the dorsal midline in 10 % of the surviving mosaics.
Calculated maleness average scores for landmarks in control and l(1)mysXG43 adult gynandromorphs

The conclusion that not every cell in the developing embryo requires the normal level of l(1)mys expression for viability is complicated by the fact that maternally derived PSβ integrin may still be present in mutant cells. To determine whether these cells can survive throughout development without either maternally or zygotically derived PSβ integrin, we performed an additional series of experiments in which both the maternal and zygotic components were eliminated. Gynandromorphs were produced by crossing ring-X males to y l(1)mysXC3f/OvoD females containing homozygous y l(1)mysXG43f germline clones (See Materials and methods). 30 adult gynandromorphs were observed with l(1)mys- patches. As in the previous experiment, there was a 95 % lethality associated with these mosaics and the same bias against mutant patches in the ventral tissue of surviving gynandromorphs was observed. Also, wing blisters, missing leg parts, missing cuticle and tergite defects were found only in l(1)mys mosaic flies. Because the level of mosaic survival and the mean maleness average score for this second series of experiments were the same as for the first series, both sets of data are combined in Table 2. The fact that the data from the two series are the same indicates that the maternal expression of l(1)mys does not appear to play a role different from that of zygotic expression.
Collectively, the combined data suggest that, although certain localized regions of the embryo can develop in the complete absence of PS/5 integrin (maternal or zygotic), expression of PSβ integrin is required in many parts of the developing fly. There is an especially large requirement for l(1)mys+ expression (PSβ integrin) in ventrally derived tissues.
Comparison of the data on gynandromorphs with the immunolocalization data suggests a correlation between regions of the embryo that have a large concentration of β integrin and regions that show the greatest requirement for normal levels of ) β integrin for viability. Such regions include the ventrally derived mesoderm and surrounding ectoderm and perhaps part of the gut. In contrast, we detected much less )3 integrin staining in most dorsally and laterally derived regions of the embryo, which appear to have a smaller requirement for local expression of PSβ integrin as indicated by the gynandromorph studies.
Requirements during wing development
Earlier studies have demonstrated the presence of PSβ integrin in the developing eye-antennae, leg and wing discs (Brower et al. 1985). To examine the effect of PS β integrin loss on the development of these and possibly other imaginai discs, larvae heterozygous for a l(1)mysXG43 chromosome marked with yellow (y) and forked (f) or white (w) were irradiated with gamma rays during larval stages to generate clones of cells homozygous for y l(1)mys f or w l(1)mys.
20 % of the 597 irradiated y l(1 )mysXG43f heterozygous flies examined contained wing defects, including blisters (the most common defect), folds in one or both surfaces of the wing, vein abnormalities and missing or enlarged halteres. In contrast, these defects were not observed in nonirradiated y l(1 )mys f heterozygotes, or in irradiated y w f heterozygotes, but were observed when flies containing another l(1)mys loss of function allele, Kljmys*1*87, were irradiated. These data suggest that the observed wing defects were not due to dominant affects of l(1)mys, other mutations, or irradiation damage, but rather were due specifically to the loss of PSβ integrin in a group of wing disc cells.
This hypothesis was confirmed by mounting 85 wings with detectable abnormal morphology and examining them for clones. Virtually all of these wings contained detectable mutant l(1)mys cells. However, most blisters or defects encompassed a significantly larger area of the wing than that of the associated clones. In addition, l(1)mys wing clones often extended in the proximaldistal direction and were confined to either the posterior or anterior compartment (Garcia-Bellido et al. 1973,1976), while blisters were typically oval or circular in shape and often crossed the compartment boundary.
Fig. 4 shows that the severity of the wing blisters can range from small separations of the dorsal and ventral epithelia (Fig. 4B) to large blisters (Fig. 4C) and folds causing larger, more general disruption of dorsal and ventral epithelia (Fig. 4D). The trichomes (hairs, derived from individual cells in the wing epithelium) develop normally and in apparently normal patterns in the blister, but are often more sparsely distributed than in the surrounding regions of the wing, as if the epithelium constituting the blister is stretched. Cell death and/or cell proliferation do not appear to play a major role in the blistering effect since y l(1)mys f clones were as large (average of 80 cells) and as abundant (20 % of irradiated flies) as in the control y w f clones. Wing veins included in the blisters were often abnormal, and associated with brown cuticle, while the overall vein pattern in the wing was usually maintained (Fig. 4). The differentiation of bristles within blisters remained normal. No apparent relationship was observed between the morphology of the wing and whether the clone was dorsally or ventrally located.
Requirements for β integrin in wing disc development. (A) Wild-type wing; (B–D) wings containing l(1)mys clones and increasingly severe wing blisters. Note brown cuticle around blister in B, abnormal venation in C, and folds around the wing margin in D (arrows).
Since PS1 and PS2 integrin a subunits are expressed in the mature wing imaginai disc on dorsal and ventral epithelium, respectively, we examined clones crossing the wing margin, which is derived from the boundary of PS1 and PS2 expression. Most of these clones, however, leave the pattern of marginal bristles looking normal and do not produce blisters.
These data indicate that PSβ integrin is required for the normal development of the wing and suggest that it is necessary for keeping the dorsal and ventral wing epithelia apposed and for normal wing vein development.
We did not observe any abnormalities in the legs of irrradiated y l(1)mys f heterozygotes. However, since the expected leg clones are small we would not necessarily have expected any gross defects and more detailed analysis would be required to identify subtle defects.
Requirements during eye development
The differentiation of the Drosophila eye disc into the adult compound eye occurs during the late larval and early pupal stages of development. The adult eye consists of several hundred ommatidia, each containing 8 photoreceptor cells (R1–R8) at its core and twelve surrounding accessory cells. Cross sections show a reiterated asymmetrical arrangement of 6 peripheral rhabdomeres (R1–R6) surrounding 2 central ones (R7, R8). R1–R6 span the entire thickness of the retina, while R7 and R8 are set one above the other, with R7 more apically located and R8 more basally located. Above the photoreceptor unit lies the lens system, consisting of the corneal lens, the pseudocone, 2 primary pigment cells and 4 cone cells. Around the photoreceptor cells and lens system is a ring of secondary and tertiary pigment cells, surrounding each omma-tidial unit. These pigment cells are shared by neighboring units and therefore sections of the adult compound eye show a honeycomb-like array of hexagonal ommatidia (see Tomlinson, 1988 for a recent review of adult eye structure).
15 % of the 384 heterozygous w l(1)mysXG43 irradiated flies and 14 % of the 454 heterozygous y w f irradiated flies were observed to have white eye clones constituting 5–30% of the eye (Fig. 5A). In order to examine the effect of the absence of PSj3 integrin on ommatidial shape and rhabdomere organization, eyes of 28 w l(1)mys flies and 21 y w f flies that contained clones were immersed in oil and examined under antidromic illumination (Francescini, 1975). In all eye clones examined, the hexagonal array of ommatidia appeared normal and continuous with that of neighboring wild-type ommatidia (Fig. 5B). However, more detailed examination showed that the organization and structure of the rhabdomeres in l(1 )mys ommatidia was abnormal. Unlike the regular organization of wild-type or y w f rhabdomeres (Fig. 5C), l(1)mys rhabdomeres showed a more random arrangement and irregular structure (Fig. 5D). Under antidromic illumination, l(1)mys photoreceptor cells often appeared to be larger than wild-type cells, and at times, showed fewer than 7 rhabdomeres in a particular optical plane. In addition, in 16 out of the 28 l(1)mys clones examined, no rhabdomeres were detectable under antidromic illumination. These data suggest that PSβ integrin is not required for establishment of the hexagonal pattern of ommatidia, but is required for the normal arrangement of rhabdomeres within each ommatidium. The fact that we were unable to observe rhabdomeres in certain l(1)mys clones further suggests that loss of PSβ integrin might result in areas without rhabdomeres or cause rhabdomeres to be oriented in such a way that they are no longer visible under antidromic illumination.
Requirements for p integrin during eye disc development. Eyes were submerged in oil and viewed under antidromic illumination. (A) Eye containing a w l(1)mys clone. (B) Ommatidia from the clone shown in A). Plane of focus at the level of the lens. Note the normal shape and arrangement of ommatidia and bristles. (C) Enlargement of wild-type ommatidia in A. Plane of focus at the level of the rhabdomeres. (D) Enlargement of l(1)mys ommatidia in A. Plane of focus at the level of the rhabdomeres. Note the abnormal shape and size of the rhabdomeres (arrow). C and D were photographed at the same magnification and were enlarged to the same size.
Requirements for p integrin during eye disc development. Eyes were submerged in oil and viewed under antidromic illumination. (A) Eye containing a w l(1)mys clone. (B) Ommatidia from the clone shown in A). Plane of focus at the level of the lens. Note the normal shape and arrangement of ommatidia and bristles. (C) Enlargement of wild-type ommatidia in A. Plane of focus at the level of the rhabdomeres. (D) Enlargement of l(1)mys ommatidia in A. Plane of focus at the level of the rhabdomeres. Note the abnormal shape and size of the rhabdomeres (arrow). C and D were photographed at the same magnification and were enlarged to the same size.
To investigate further the nature of the defects in the l(1)mys eye clones, 15 additional eyes containing one or more l(1)mys clones were embedded in plastic and sectioned. In sections transverse relative to the ommatidia (Fig. 6), the wild-type ommatidia showed the typical trapezoidal array of rhabdomeres, whereas the l(1 )mys clones showed abnormally large and disorganized rhabdomeres. Although the number of rhabdomeres per ommatidial unit appeared to be approximately normal, the deformed shapes frequently made accurate counts difficult. In some cases, rhabdomeres appeared to be duplicated, at least in cross section (Fig. 6D). In sections perpendicular to the plane of the retina (Fig. 7), the rhabdomeres remained together as units, but they failed to span the thickness of the retina and were disorganized. This disorganization was more severe at basal positions where holes were frequently observed. However, the lamina, the first optic ganglion, was intact and appeared to have a normal morphology. The apical segments of the rhabdomeres appeared more wild-type in organization than the basal segments and the arrangement of the corneal lens cells appeared relatively normal. In both transverse and longitudinal sections, the ommatidial units closest to the boundaries of the clones often looked the most normal (Figs 6, 7), which might be expected since ommatidia are not clonal units.
Cross sections of eyes containing w l(1)mys clones. (A) Cross section of an eye showing the borders of a w l(1)mys clone (arrows). (B) Higher magnification of the boundary between wild-type (left) and mutant (right) tissue. Compare the regular size and arrangement of the wild-type rhabdomeres with the abnormally large and irregular shape of the l(1)mys rhabdomeres and their irregular .arrangement (arrows). (C) Enlargement of l(1)mys+ ommatidia illustrating the regular pattern of rhabdomeres. (D) Enlargement of w l(1)mys ommatidia showing derangement, large size and occasional duplication or folding of mutant rhabdomeres. C and D were photographed at the same magnification and were enlarged to the same size.
Cross sections of eyes containing w l(1)mys clones. (A) Cross section of an eye showing the borders of a w l(1)mys clone (arrows). (B) Higher magnification of the boundary between wild-type (left) and mutant (right) tissue. Compare the regular size and arrangement of the wild-type rhabdomeres with the abnormally large and irregular shape of the l(1)mys rhabdomeres and their irregular .arrangement (arrows). (C) Enlargement of l(1)mys+ ommatidia illustrating the regular pattern of rhabdomeres. (D) Enlargement of w l(1)mys ommatidia showing derangement, large size and occasional duplication or folding of mutant rhabdomeres. C and D were photographed at the same magnification and were enlarged to the same size.
Parasagittal sections of eyes containing w l(1)mys clones. (A) Parasagittal section of an eye showing the borders of a w l(1)mys clone (large thin arrows). Note the normal lamina of the optic ganglion in the area of the clone (small thin arrow). (B and C) Higher magnification of sections containing w l(1)mys clones. Note the disoriented and fragmented rhabdomere bundles (small thin arrows) in the w l(1)mys ommatidia. Also note the holes on the basal side of mutant ommatidia (large thin arrows), w l(1)mys lens cells usually form and appear generally normal (small thick arrows). Orientation of the sections is apical down and basal up.
Parasagittal sections of eyes containing w l(1)mys clones. (A) Parasagittal section of an eye showing the borders of a w l(1)mys clone (large thin arrows). Note the normal lamina of the optic ganglion in the area of the clone (small thin arrow). (B and C) Higher magnification of sections containing w l(1)mys clones. Note the disoriented and fragmented rhabdomere bundles (small thin arrows) in the w l(1)mys ommatidia. Also note the holes on the basal side of mutant ommatidia (large thin arrows), w l(1)mys lens cells usually form and appear generally normal (small thick arrows). Orientation of the sections is apical down and basal up.
In summary, the overall arrangement of ommatidia and the lamina is normal in l(1)mys clones, rhabdomeres are induced in normal or near normal numbers, but fail to develop their normal shape and organization within each ommatidial unit. These results indicate that PS integrins are necessary for some, but not all aspects of development of ommatidial arrays (See Discussion).
Discussion
Earlier studies have described the defects in embryogenesis of l(1)mys embryos (Wright, 1960; Newman and Wright, 1981; Wieschaus and Noell, 1986; Leptin et al. 1989). These defects include incomplete germband extension and contraction, aberrant attachment of visceral muscles, detachment of somatic muscles from their attachment sites and failure of dorsal closure leading to herniation. Many of these defects are more severe when the maternal component of l(1)mys is also deleted; additionally, elevation of the maternal contribution can partially restore the deficit in l(1)mys zygotes (Wieschaus and Noell, 1986). These results demonstrate that both maternal and zygotic expression of l(1)mys can participate in normal development, although the maternal component is dispensable. It is tempting to speculate that the two l(1)mys mRNA species detected in 0–4 h embryos (Fig. 1, see also Digan et al. 1986) are maternally derived and that they decay over the first 6–8 h of development and are replaced by a zygotic transcript at later times (Fig. 1).
However, we cannot at present distinguish maternal from zygotic transcripts. The l(1)mys mRNAs become segregated to all the cells of the blastoderm (Fig. 2 C,D) and mRNA is detectable in all or most tissues throughout development (Fig. 2). The PS/3 integrin encoded by the mRNA is first readily detectable at the time of gastrulation (Fig. 3) and is detectable in many tissues, including the germband and the gut throughout development. Integrins are concentrated in muscle attachment sites (Bogaert et al. 1987; Leptin el al. 1989; our unpublished results) and their absence in l(1)mys mutants is likely the cause of much of the mutant phenotype. Given the known functions of integrins in vertebrates, namely cell-matrix and cell-cell adhesion (Hynes, 1987; Ruoslahti and Pierschbacher, 1987; Buck and Horowitz, 1987), it is likely that they play a similar role in Drosophila and that, in their absence, cell attachments are defective.
The mutations we have used in the generation of mosaics appear to be null alleles of l(1)mys, which produce no PSβ integrin subunit. The two associated PS α-subunits (PS α and PS2α) fail to be processed to the cell surface in the absence of PSβ (Leptin et al. 1989). Thus, cells lacking l(1)mys+ lack both of these PS integrins as well as any others that share the same β subunit. If other subfamilies of integrins exist in Drosophila, which is not yet known, they presumably persist.
The gynandromorphs (Tables 1 and 2) contain large clones derived from cells that lose l(1)mys+ and PSβ integrin relatively early in development. As would be expected, many (93%) of these gynandromorphs die before the adult stage. Although our gynandromorph data do not provide any information on the age or phenotype of dead mosaics, the surviving mosaics provide indirect information on the requirements for PSβ integrin in different regions of the organism. The data (Table 2) indicate that the greatest requirement is in ventral structures. For instance, we never observed mutant patches on the legs and legs were frequently missing, suggesting either that development of legs has a strong requirement for PS integrins, or that adjacent ventral structures have such a requirement, or both. The selection against survival of gynandromorphs lacking PS integrins in ventral structures (Table 2) is concordant with the high level of expression of PSβ integrin in the germband and its derivatives (Fig. 3) and suggests a primary role for PS integrins in the development or maintenance of one or more of these structures. In addition, our results suggest that the abnormalities (i.e. dorsal rupture) observed along the dorsal midline of l(1)mys embryos (Wright, 1960; Newman and Wright, 1981; Wieschaus and Noell, 1986) might be secondary and result from the lack of integrins in ventrally derived cells. While some mutant patches in the l(1)mys gynandromorphs appeared to develop normally, many others gave rise to defects of varying severity, such as wing blisters and tergite defects. These results also indicate a requirement for PS integrins in the development of tergites and wings.
Somatic clones generated by radiation-induced mitotic crossing-over produce smaller mutant patches and lead to defects in the development of wings (Fig. 4) and eyes (Figs 5–7). The nature of the defects in the wings and eyes provides information as to which aspects of development require the presence of PS integrins and which do not. The very existence of wing and eye clones at the same frequency and size in both l(1 )mys and wildtype crosses indicates that PS integrins are not required for cell viability or proliferation or for maintenance of cell sheets. The defects produced by loss of PS integrins appear, rather, to arise from defects in interactions between sheets or groups of cells.
The blisters that form in the wings appear to result from a failure of apposition of the two surfaces of the wing, each of which is derived from an epithelial sheet. While these sheets appear relatively normal in the blisters, they are not properly attached to each other and the areas of detachment are larger than the areas of the clones. This suggests that a defect in attachment in the area of the clone may generate a point of weakness allowing separation of the two epithelial sheets over a large area. During development of the wing disc, the dorsal and ventral epithelial layers separate and rejoin several times (Waddington, 1940; Milner and Muir, 1987). During the separations, the two epithelial layers remain attached via long cellular processes, which contain microfilaments and microtubules (Mogensen and Tucker, 1988) and are connected at their ends by cell-cell attachments. Since PS1α is expressed in the dorsal epithelium and PS2α in the ventral epithelium (Brower et al. 1985), it seems quite likely that these two integrins participate in the attachment of the two epithelial layers either directly or via intervening extracellular matrix. Similar wing blisters have been reported by Wilcox et al. (1989) in flies heterozygous for a hypomorphic allele (l(1)mysni42) of the l(1)mys locus and for an amorphic allele (l(1)mysXG43’) and in flies homozygous for inflated, the gene for PS2a’. Flies doubly homozygous or hemizygous for mutations in l(1)mys and inflated show even more extreme wing defects. These observations are in good agreement with our results. Absence of l(1)mys+, and the PSβ integrin that it encodes, would lead to defects in interepithelial attachment, the failure to maintain the correct apposition of dorsal and ventral wing surfaces and the development of the blisters and folds (Fig. 4).
Interpretation of the eye defects is a little more complex, but previous descriptions of pattern formation during eye development provide a good basis for analyzing potential roles for PS integrins. The process of ommatidial development involves cell–cell adhesion, cell movements and cell-matrix interactions, exactly those processes in which integrins have been implicated in vertebrates (see Introduction). The defects seen in l(1 )mys somatic eye clones indicate that PS integrins are necessary for the proper organization of ommatidial units, although many aspects of eye development proceed in the absence of PS integrins.
Examination of l(1)mys somatic eye clones first allows identification of the processes for which PS integrins are not necessary. For example, the fact that lenses, and the hexagonal array of ommatidia, develop essentially normally in clones (Figs 5–7) indicates that the cone cells, and the secondary and tertiary pigment cells, develop in the normal positions. Also an apparently normal lamina, the first optic ganglion, forms in mutant l(1)mys areas. Together with the appearence of photoreceptor cells with recognizable rhabdomeres, albeit significantly distorted (Figs 5–7), these results suggest that the development of the patterned array of ommatidial units, initial differentiation of photoreceptors and the projection of their axons into the lamina all proceed along approximately normal paths without PS integrins. Thus, the ‘preclusters’ of photoreceptor precursors that appear just behind the morphogenetic furrow and form the core of the developing ommatidial units (Ready et al. 1976, 1986; Tomlinson, 1988) must appear in their normal array. That is, the clustering of these cells from the previously undifferentiated epithelial layer happens in the absence of PS integrins despite the switch of PSI a integrins in front of the furrow to PS2a integrins behind it (Brower et al. 1985). These integrins also appear to be unnecessary for the cell movements that position subsequent members of the ommatidial unit in relation to the core of precluster-derived cells. The cells that move apically along the sides of the precluster include the precursors of the cone cells (Ready et al. 1986; Tomlinson, 1988) and, as noted above, cone cells develop normally in the correct apical location. Since an apparently normal lamina forms in mutant l(1)mys areas, integrins also appear not to be absolutely required for projection of photoreceptor axons, as previous studies (Meinertzhagen, 1973; Meyerowitz and Kankel, 1978; Fischbach and Technau, 1984) have shown that proper retinal innervation is required during late larval stages for the normal development of the optic ganglia. Thus, one can conclude that PS integrins are not necessary for segregation of preclusters, for many of the apicobasal movements of the cells of the ommatidial units, for the projection of photoreceptor axons or for determination or differentiation of cone cells, photoreceptor cells, secondary and tertiary pigment cells or lamina.
The fact that the photoreceptors are disorganized within the mutant ommatidia indicates that integrins are needed for the normal patterning of these cells. The disorganization precludes the identification of individual photoreceptors by their positions within the ommatidia. Thus, while it appears that each of the mutant ommatidia contains an approximately normal number of photoreceptors (Fig. 6), our current level of analysis does not rule out deficiencies or excesses in photoreceptor number and does not allow one to define exactly which of the usual eight receptors in each unit is correctly formed, let alone correctly positioned. Since a ‘default’ pathway for R7 is to develop into cone cells (Tomlinson and Ready, 1986,1987), we also cannot rule out aberrant determination of certain photoreceptor cells.
The simplest interpretation of the current data is that precluster formation, recruitment, induction and differentiation of photoreceptors, cone cells and pigment cells all proceed relatively normally without PS integrins. Therefore, it appears that photoreceptor cells and rhabdomeres depend on the expression of PS integrins predominantly during later stages of development, perhaps as late as the pupal stages when the ommatidial units elongate. The enlarged and distorted profiles of rhabdomeres in transverse section (Figs 5 and 6) and their apparent basal retraction in the parasagittal sections (Fig. 7) suggest that the attachments of these cells at the basal surface of the retina depend on integrins. The relatively normal array at the apical end of each ommatidial unit could rely on apical cell-cell interactions among the photoreceptor cells. A more detailed analysis will be necessary to define more precisely the nature of the defects caused by the absence of specific PS integrin subunits.
The lethality and defective development of both embryonic and adult structures, consequent upon loss of PS integrins, is not unexpected given their presumed involvement in multiple cell adhesion phenomena. It is striking, however, how much development can occur in the absence of these proteins. This point is particularly clear from the analysis of eye development. The extensive development of ommatidial units and their normal overall patterning in l(1)mys eye clones must involve cell adhesion events not dependent on PS integrins. Other molecules, such as cadherins (Takeichi, 1988), immunoglobulin-related adhesion molecules such as N-CAM (Cunningham et al. 1987) or fasciclins (Patel et al. 1987; Zinn et al. 1988), and possibly other integrins with ft subunits distinct from that encoded by l(1)mys may be involved. The same could be true for early developmental processes, such as germband extension. Dual involvement of two subfamilies of integrins in a given biological process has precedent in the case of fibroblast adhesion (Singer et al. 1988; Dejana et al. 1988), as does the dual involvement of integrins and cell–cell adhesion molecules in the case of neurite outgrowth (Tomaselli et al. 1988). In both these instances, ablation of both classes of cooperating adhesive receptors is necessary to block the biological process. Further work will be necessary to investigate potential cooperation between PS integrins and other adhesive receptor systems, perhaps including other integrins.
In summary, the data discussed above demonstrate requirements for PS integrins in several parts of the early embryo and especially in ventral structures. They are also required for the normal development of the wing and the eye. The defects occurring in the absence of PS integrins can all be interpreted in terms of loss of adhesion of a cell layer and of individual cells, consistent with the suspected functions of integrins. However, other adhesions are completely normal in the absence of PS integrins and must rely on other molecules.
ACKNOWLEDGEMENTS
We thank Eric Wieschaus, Ted Wright and the stock center at Bowling Green for Drosophila stocks, Susan Haynes and Igor Dawid for the l(1)mys genomic clone, Kai Zinn and Corey Goodman for the cDNA library and Mary Lou Pardue and Jim Garbe for providing embryos and advice on RNA preparation. We also wish to thank Norbert Perrimon for his useful comments and for the use of his sectioning equipment and Linda Ambrosio for all her technical help, information and encouragement. We are grateful to Ruth Lehmann, Terry Orr-Weaver, Hermann Steller, Elizabeth George, Gene Huh, Pamela Norton and Larry Zusman for their helpful comments on this manuscript and Sue Bergin, John Cohan, Spenser Ladd and David Smouse for their photographic assistance and general support. This work was supported by the Howard Hughes Medical Institute.
References
Note added In proof
While this paper was under review, Brower, D. L. and Jaffe, S. M. reported wing blisters caused by l(1)mys somatic clones and inflated mutations. Nature342, 285-287 (1989).