The function of PS integrins in Drosophila wing morphogenesis

Integrins are cell surface receptors both for certain extracellular matrix components and serum proteins, and for surface molecules on other cells (for review see Hynes, 1987; Ruoslahti and Pierschbacher, 1987). The receptors are α β heterodimers in which both subunits are transmembrane glycoproteins. In Drosophila, genetic analysis can be used to investigate in vivo the function of integrins during development. During embryogenesis, the two Drosophila integrins, PSI and PS2 (Wilcox et al. 1981; Leptin et al. 1987; Bogaert et al. 1987), which share a common β subunit, PS/β (Brower et al. 1984; Wilcox et al. 1984), function in the attachment of both somatic and visceral muscles to ectoderm and endoderm, respectively, and in certain cell sheet migrations (Leptin et al. 1989). In these events, the PSI and PS2 integrins are usually found in complementary positions - that is, on adjacent cells or tissues - and it seems likely that they cooperate in their function. The two are also expressed in a complementary fashion, later in development, on the basolateral cell surfaces of the wing imaginai disc epithelia, which will give rise to the dorsal and ventral surfaces of the adult wing (Brower et al. 1984). The PSI integrin is expressed on the presumptive dorsal surface, PS2 on the ventral surface. At metamorphosis, the disc evaginates, bringing together the basal surfaces of these dorsal and ventral wing epithelia. Although the major veins of the wing are specified within the epithelia earlier in disc development, their final differentiation requires an inductive effect of the upper, dorsal epithelium upon the lower, while the shaping and detailed organization of the wing, including the formation of crossveins, results from interactions between the cells of the two surfaces (Garcia-Bellido, 1977). Although much of this biology was first described fifty years ago (Waddington, 1940), little is known as yet of the molecules mediating the events.

The occurrence of the PSI and PS2 integrins on the two presumptive wing surfaces suggests that they may be involved in some aspect of the attachment of the two surfaces during wing morphogenesis. We now show that viable mutations in either the PSβ or the PS2α integrin genes lead to defects in the organization of the adult wing; the mutants show incomplete attachment of the two surfaces, leading to ‘blistering’ of the wing, and also distortions of vein patterning. The two classes of mutant interact to produce much more extreme wing defects in which the wings fail either to unfold properly or to expand after hatching of the fly. Thus, PS integrins are required both for embryonic development and for later stages of Drosophila morphogenesis.

Two wildtype strains (Oregon R and Barton) of D. melano-gaster were used: identical results were obtained with each. l(l)myospheroid mutants were l(1)mys^G43 and l(1)mysR⁰⁴ (both EMS-induced; Wieschaus et al. 1984) and Df(l)sn^C128 (Gans et al. 1980). nj42 (Costello and Thomas, 1981) was obtained from Dr A. Ferrus, l(l)k27e (Falk et al. 1984) from Dr D. Falk, and if from the Mid-America Drosophila Stock Center, Bowling Green State University, Ohio. Since nj42 and l(1 )k27e are (in this paper) shown to be allelic to mys and if, respectively, they are renamed mysnj⁴² and if^27emys⁴²if double mutant stocks were obtained as recombinants from a cross of y (1. – 0.0) cv (13.7) v (33.0) if3 (55) females and mysnj⁴² (1.-21.0) f (56.7) males. Of 18 y⁺cv⁺v f⁺recombinants, 5 (28%, expected 60%) were found to contain both mys^nj42 and if3. The presence of the mysnj⁴² allele was determined by non-complementation with the antimorphic mys^nj04 allele (mysnj42/mys^xR()4 is lethal), the presence of if by non-complementation with if^27e. Each stock had a similar penetrance and level of expressivity of the wing phenotype. 10 y⁺cv⁺v⁺f+recombinants each contained both mys^nj42 and if3 (expected, 95%). Adult wings were mounted in DEPEX (Gurr) and photographed on Kodak Technical Pan film using a Standard RA Zeiss microscope with bright-field illumination and a condenser from which the top lens had been removed.

Antibody labelling of imaginai discs was carried out essentially as described in Wilcox (1986). Wing imaginai discs from mid to late third instar larvae were dissected out in balanced salts solution (BSS, Wilcox, 1986), and allowed to attach to microscope slides which had been coated with poly-L-lysine (I mg ml-¹). Discs were first incubated with either PSlα-specific MAb, DK1A4 (Wilcox et al. 1981), PS2α-specific MAb, CF2C7 (Brower et al. 1984) or PSβ-specific MAb, CF6G11 (Brower et al. 1984). For staining with anti-PS2αand -PSβMAbs, discs were first fixed in 4% paraformaldehyde/ phosphate-buffered saline (PBS) for 10 min at room temperature, then washed in BBT (PBS containing 0.1 % BSA, 0.1 % Triton X-100 and 1 rrtM-NaN₃). Subsequent procedures were all carried out in the same buffer. For the anti-PSlα MAb, whose epitope is sensitive to fixation, the first incubation was carried out on living discs in BSS, and, after washing, the discs fixed before transfer to second antibody. The bound first antibody was visualized with a biotinylated second antibody and horseradish peroxidase coupled to avidin (Vectorlabs). Stained discs were dehydrated through a graded alcohol series, mounted in Araldite, photographed on Kodak Ektachrome EPT160 film, using Nomarski optics, and black and white prints made directly from colour transparencies. This technique produces high contrast images in which stained tissue shows as white.

Preparation of SDS lysates from small numbers of wildtype and mutant embryos, SDS-gel electrophoresis and Western blotting procedures were performed as described by Leptin et al. (1989). PS proteins were detected by PS2 α-specific MAb, PS2hc/l (Bogaert et al. 1987), β-specific MAb, bl2 or anti-PS1αrabbit antiserum (Leptin et al. 1989). Immunoprecipitation of PS complexes from non-ionic detergent lysates of wildtype and mutant embryos, using MAbs DK1A4, CF2C7 and CF6G11, was performed as described by Brower et al. (1984). Immunoprecipitated PS proteins were separated by SDS-gel electrophoresis and detected using anti-PS antibodies.

The mutants used in this study are listed in Table 1. The l(l)myospheroid (mys) gene encodes the PS integrin β subunit (MacKrell et al. 1988; Leptin et al. 1989). We have shown that a mutant in which the mys gene is deleted, Df(l)sn^CI28, and two ethane methane sulphon-ate (EMS)-induced lethal mys mutants each produce no PSβprotein and cause incorrect localization of the α subunits (Leptin et al. 1989).

The PS2αgene was localized, by in situ hybridization to polytene chromosomes, to chromosome band 15A (Bogaert et al. 1987), near the rudimentary gene (r, 1. –54.5; chromosome bands 15A1–2; Rawls et al. 1986)). Comparison of PS2αgene restriction mapping data (Brown et al. 1989) with a restriction map of the r region (kindly provided by Dr R. Miassod), placed the gene just proximal to r. Falk et al. (1984) produced a number of EMS-induced lethal mutants in this region, one of which, l(l)k27e, maps to a position very close to chromosome band 15A5, which is just proximal to r and hence close to the position of the PS2αgene. I(l)k27e embryos die late in embryogenesis. We tested whether this mutation was in the PS2α gene. Western blot analysis using a PS2α-specific monoclonal antibody (MAb), PS2αand an antiserum against the PSI αprotein N-terminal sequence peptide, revealed that lysates of l(l)k27e mutant embryos contained normal levels of PSI α protein but no detectable PS2αprotein (Fig. 1), indicating that the mutation was probably in the PS2α gene. Other data supporting this conclusion are discussed below. The PS2αheavy chain is a 140K (K=10³Ai_r) protein produced by proteolytic cleavage of a 160 K heavy/light chain precursor (Bogaert et al. 1987). Since the PS2hc/l MAb is able to recognize a truncated protein of ca 105K (Bogaert et al. 1987), it seems likely that any PS2aprotein produced in l(l)k27e embryos is less than 105K, and also lacks the light chain. Mutant embryos show few obvious defects other than greatly impaired muscle function; they fail to hatch.

From complementation studies, it has been suggested that the recessive mutation non-jumper (nj42) is probably a mutation in the mys gene (De la Pompa et al. 1989). Western blot analysis of the nj42 PSß chain shows that it has, compared to the wildtype protein, a slightly slower mobility on SDS gels (Fig. 2A –C). While these blots show comparable amounts of each of the PS components in nj42 and wildtype embryos (Fig. 2B –C and other data), we find that, from non-ionic detergent lysates of embryos, nj42 PS complexes are only poorly immunoprecipitated (Fig. 2D –K). The PSla-specific MAb precipitates much less of the PSl α βcomplex from a nj42 lysate (lanes F and J) than it does from a wildtype lysate (lanes D and H). From a wildtype lysate, the PS/J-specific antibody precipitates both PSI and PS2 complexes (the PSI α protein is shown in lane E) together with a large amount of residual PS β protein which is apparently not associated with either a chain (lane I). From nj42 lysates, the MAb precipitates only a very small amount of the PSI complex (lane G; the associated β chain is so low as to be undetectable, lane K) and is unable to precipitate the residual PS αprotein (compare lane K to lane I). These data confirm that nj42 is a mutation in the mys gene and suggest that the mutant PS βchain is altered in a way that either affects complex stability or else alters the complex conformation so that individual antibody epitopes of both α and β subunits are affected. In view of the earlier identification of the mvs locus (Wright, 1960), we rename this allele mys”¹.

mys’’¹⁴² causes severe wasting of the mesothoracic and metathoracic tergotrochanteral (TDT) muscles, which run from the thorax down into the leg of the fly and, as a result, flies show little or no escape jump response (Costello and Thomas, 1981; De la Pompa et al. 1989). The homozygous mutant flies show no abnormal wing morphology (Fig. 3A) although, presumably as a result of the thoracic muscle defects, their wings are sometimes held, abnormally, away from the body. The severity of this defect is dramatically increased in mys”‘⁴²/Df(l)mys (or mys’⁴²/ mys*^G4i) flies. Here, both wings usually are held out at 90° from the body midline, and often droop. An enhanced mutant muscle phenotype is seen in these heterozygous flies (De la Pompa et al. 1989). In addition, however, the flies show frequent wing defects; up to 50 % of wings have blisters caused by the incomplete apposition of the two surfaces. These blisters are often small and localized, usually to the central wing region (Fig. 3B), but sometimes a gross blistering of the whole wing is seen. Most interestingly, 10 –15% of the wings have extra or distorted crossveins, frequently accompanied by other vein distortions (Fig. 3C; compare to 3A). These results implicate the PS integrins in wing morphogenesis.

In light of the wing defect phenotype of the PSβsubunit mutant, we wondered whether we might find mutant alleles of the PS2a gene with a similar phenotype. We searched the literature to see if any ‘blister’ or wing venation mutant existed in the vicinity of the PS2 α gene. The mutation inflated (if) has been mapped to a position (1. –55) just proximal to r (Lindsley and Grell, 1968). 10 –15% of the wings of if mutant flies are blistered. The blisters are often small and centred on the anterior crossvein (see Fig. 4B), although cases where the wing surfaces remain more or less unattached are frequent. The mutation is not complemented by the putative PS2 αlethal mutant l(1)k27e described above. if3 l(1)k27e heterozygous females exhibit a much higher frequency of wing blistering (50 –80%) than is found in if³ /if³ homozygotes. These data indicate that if and l(1)k27e are allelic (l(1)k27e will therefore be called if²’⁴).

We find also that PS2 αprotein distribution is altered in wing imaginai discs from if³ larvae. In mature wildtype discs, the PS2 integrin is, in the main, restricted to cells of the ventral half (compartment) of the disc (Fig. 5A). Staining with the PS2 α-specific MAb, CF2C7, reveals that, in the more central region of the compartment (the cells of which will give rise to the lower, ventral, surface of the wing), the complex is expressed in an alternating series of weakly and strongly staining bands (Fig. 5A). In the peripheral region of the disc, weaker, but quite reproducible staining is seen in a number of places. In if3discs, the level of PS2αprotein expression is much reduced in the central region of the disc (Fig. 5B and C) but may be less affected in the peripheral regions. The banding pattern becomes much more pronounced and often (as in Fig. 5C), staining of the central weaker band disappears as though it may be preferentially affected. The cut-off in PS2 αexpression that occurs at the line of cells that will differentiate into the adult wing margin (arrows in Fig. 5A –C) is also much less distinct, especially at the edges of the disc. In addition, there seems often to be a rather higher level of expression of PS2 αprotein in the dorsal region of the disc. Normally,PS2αexpression is limited in the late disc to small isolated patches in the dorsal wing epithelium (Fig. 5A); in if³ discs larger regions of staining are observed (Fig.5D), which resemble more the pattern of PS2α expression in younger discs (data not shown, but see Brower et al. 1985). Alterations of the pattern of PS2 αexpression in if³ discs is also seen by Brower and Jaffe (1989). In combination, these data – the altered PS2α distribution observed in the viable if allele and the absence of detectable PS2α protein in the lethal allele, together with the wing phenotype resembling that found in the PSβ gene mutant – provide strong evidence that both if³ and if^k27e are mutations in the PS2α gene. The altered pattern of PS2α expression observed in if³ discs suggests that if³ may be a regulatory mutant affecting PS2α expression patterns. Consistent with this, normal levels of a normally sized PS2α protein were found in if³ mutant embryos (data not shown).

While mys (β ) and if (PS2α) are independent genes, their products combine to produce one functional protein complex. Therefore, one might expect combination of the two viable mutations to produce a greater than additive effect on the wing phenotype. To examine this possibility, we made a double mys^nj42if³ mutant (for details, see Materials and methods). The mutant exhibits a greatly enhanced phenotype: 90-95% of wings have largely unattached surfaces and remain folded and almost totally unexpanded after emergence of the fly from the pupal case (Fig. 4C). Other experiments show that the exacerbation of phenotype is only observed in the double hemi- or homozygous mutant; that is, neither allele acts as a dominant enhancer of phenotype.

We have used viable mutant alleles of genes coding for integrin subunits to investigate the functions of integrins during the later stages of Drosophila development. Flies carrying lethal alleles of these genes die before these stages are reached (mosaic analysis provides a method for investigating function using such alleles; see Brower and Jaffe, 1989). In the case of the PSβ chain, the genetically demonstrated allelism between the viable mutation nj42 and the lethal mutation l(l)mys (De la Pompa et al. 1989) was corroborated by comparison of mutant and wildtype protein products: the allelic variant nj42 (now mys^nj42 ) PSβ chain has an altered electrophoretic mobility. No such change was observed for the PS2 α chain in if³ homozygous flies. We therefore depend on less direct evidence to support our contention that if³ and if^k27e are allelic, and that both are mutations in the PS2α gene. However, this evidence is very strong. One mutation (if^27e) leads to the loss of detectable PS2α chain, the other (if³ ) to changes in the PS2α expression pattern. The two mutations do not complement each other. These results are consistent with if^k27e being a mutation that disrupts the coding region of the PS2α gene, and if³ a mutation in regulatory elements of the PS2α promoter required for proper expression of the a subunit in later phases of development.

Analysis of the mys^nj42 and if³ mutants shows that one of the late developmental roles of the PS integrins is to ensure the correct apposition and patterning of the wing epithelia. During the pupal stage, wing morphogenesis proceeds through a complicated series of events that are not easily accessible to experimentation (Waddington, 1940). However, it may be possible to deduce from the final mutant wing phenotypes during which of these phases PS integrins might be required. The imaginai disc epithelium carrying PS integrins on its basal surface evaginates at the onset of metamorphosis. Following the evagination, the dorsal and ventral wing blade epithelia become apposed, but the two epithelial sheets subsequently separate and re-attach twice more. During these separations the cells of the two surfaces remain attached, via long cellular processes, to basal desmosomes. Inside the cell, the desmosomes attach to large microtubule bundles (transalar arrays) which span the cell to the apical surface and serve to connect the two wing surface cuticular sheets (Mogensen and Tucker, 1988). While the longitudinal veins form (or, at least, are specified) early in imaginai disc development, the crossveins form only during the second apposition event. The mys^nj42/Df(l) mys phenotype includes rather localized vein and crossvein defects, suggesting that the integrins are required during this second apposition. The failure, in mys^nj42if³ double mutant flies, of wings to undergo their final expansion and unfolding indicates that there is also a later integrin function. This suggests that integrins may act through much of wing morphogenesis. The rather localized defects often seen in both β and PS2α mutants seem more likely to reflect ‘weak points’, in time or space, in the global morphogenesis process rather than a localized position of function.

Although we can test genetically only the function of PS2 (since we have no mutant as yet in the PSI α gene), it seems likely that both integrins are required for correct wing development. Together, the two integrins cover most of the mature wing disc epithelium, PSI and PS2 occupying complementary areas. Following the évagination of the wing disc the two integrins come to lie facing each other on the basal surfaces of the two apposed wing epithelia. The complementary distribution of the two occurs at other times and places during fly development; for example, PSI and PS2 complexes are found on adjacent tissues, the tendon cells and the attached muscle, respectively, in embryonic muscle attachments. This frequent juxtaposition suggests that the two may cooperate in some way in their function. In the developing wing, the cytoskeletal transalar array–basal desmosome structures which connect the dorsal and ventral epithelia resemble closely those found in myoepidermal attachments (Mogensen and Tucker, 1988) like those of the embryonic body wall where we have previously shown that the integrins have an important function (Leptin et al. 1989). In both locations the basal desmosomes may resemble adherens junctions rather than true desmosomes (Mogensen and Tucker, 1988), and thus are a likely place for the integrins to be localized and to operate as cytoskeleton–extracellular matrix connectors.

The integrins are part of a large multi-component system; the α and β subunits are themselves associated and, in other organisms, the integrin complexes have been shown to interact both with extracellular ligands and with intracellular cytoskeletal proteins (for review see Hynes, 1987). Further genetic analysis should enable us to find other molecules that interact with integrins both inside and outside the cell. We can, for example, screen existing mutations that cause wing blister or venation defects for their ability to exacerbate the integrin mutant phenotypes. Alternatively, it should be possible to screen for new mutants that act as dominant enhancers of phenotype. A finer analysis will also require knowledge of the contribution of other adhesion systems that we assume also operate in wing development. The putative adhesion molecules fasciclin III (Brower et al. 1980; Patel et al. 1987) and the 1(2) giant larva gene product (Lützelschwab et al. 1987), for example, are both expressed in the wing disc. If the products of these genes and integrins are involved in similar morphogenetic processes, then mutants in these, in combination with the integrin mutants, might also cause greatly enhanced wing phenotypes.

We are grateful to Richard Smith for excellent technical assistance, to Tom Barnes, Mark Bretscher, Jonathan Hodgkin, Daniel St Johnston and Marcel Wehrli for critical comments on the manuscript and to Dr Raymond Miassod for sending us his restriction map of a large region of the Drosophila genome around the rudimentary gene. A.D. was supported by a Herchel Smith Harvard Scholarship.