blistered: a gene required for vein/intervein formation in wings of Drosophila

The wings of Drosophila are derived from relatively simple epithelial bilayers that have proven useful for studying many aspects of development (Blair, 1994; Couso et al., 1994; Diaz-Benjumea and Hafen, 1994; Garcia-Bellido and De Celis, 1992; Schubiger and Palka, 1987; Willliams et al., 1993). Nonneural wing tissue is organized into a patterned distribution of two cell types, intervein (90%) and vein (10% of the wing surface). Each non-neural cell forms an apical cuticular hair so hair density is an indicator of cell surface area (Dobzhansky, 1929). Intervein regions consist of large cells (low hair density) whose major function is to provide an aerodynamic surface. As the fly ecloses intervein cells die leaving a double layer of transparent, hair-studded cuticle (Johnson and Milner, 1987). A reproducible pattern of five longitudinal veins (designated L1 to L5 from anterior to posterior) and two cross veins traverses the cuticular sheet. Veins consist of narrow channels lined by small living cells that provide structural support for the wing and passage for neurons and trachea.

During wing development intervein cells are responsible for connecting and holding the two surfaces of the wing together (Fristrom and Fristrom, 1993; Waddington, 1941). To that end, intervein cells differentiate a highly specialized system of cytoskeletal supports, the transalar array (Tucker et al., 1986), anchored in integrin-mediated basal adhesions (Fristrom et al., 1993). Vein cells, in contrast, are relatively unspecialized; they do not form connections with the opposite surface, do not express integrin and do not differentiate transalar arrays. We show here that vein cell differentiation involves an apical constriction leading to the characteristic small cell size and the acquisition of a laminin-containing basal matrix.

Numerous mutations that affect the pattern of wing veins have been isolated (Lindsley and Zimm, 1992). Many of these mutants have been classified according to their loss-of-function phenotypes (Diaz-Benjumea and Garcia-Bellido, 1990; Garcia-Bellido and De Celis, 1992) such as loss of veins (e.g. veinlet and vein), excess veins (e.g. plexus and net) and thick veins (e.g. Notch and Delta). Diaz-Benjumea and GarciaBellido deduced from extensive genetic and clonal analyses of such mutants that wing veins are specified in the third larval instar, the cell proliferation phase of wing development, and that genes involved in regulating cell proliferation are also involved in cell fate decisions. They also showed that loss of vein mutants are epistatic to excess vein mutants and proposed that veins are formed at the crests of waves of morphogens in the wing anlage with different sets of genes controlling the amplitude of the waves and the distance between crests (Diaz-Benjumea and Garcia-Bellido, 1990).

Some of the genetic predictions have recently been born out by molecular studies. The mutant veinlet (ve), a loss of vein mutant, is a wing-specific allele of rhomboid (rho) renamed rhove (Bier et al., 1990; Sturtevant et al., 1993). rho encodes a localized transmembrane protein that is component of the Drosophila EGF receptor (EGF-R) signaling pathway. Like its homologs in other organisms, Drosophila EGF-R is a transmembrane tyrosine kinase associated with cell division and determination of cell fate (Baker and Rubin, 1992; Clifford and Schüpbach, 1989; Zak and Shilo, 1992). Because of its localized expression rho is believed to restrict spatially the activity of EGF-R (Ruohola-Baker et al., 1993; Sturtevant et al., 1993). Three aspects of rho expression in wing discs implicate rho as a key element in determining the distal elements of the longitudinal veins. (a) In third instar wings, rho is expressed in narrow bands corresponding to the positions of the longitudinal veins with loss of expression in rhove leading to loss of vein. (b) Ectopic rho expression is associated with ectopic vein formation (Sturtevant et al., 1993). (c) Excess of vein mutants (e.g. px and net) show a corresponding pattern of excess rho expression in the third instar (Sturtevant and Bier, unpublished data). Thus, rho is important in the specification of wing veins.

Another category of wing mutant includes those with defects in the execution of a specified cell fate. For example, Drosophila integrins, transmembrane heterodimers, are components of the basal junctions that hold the two wing surfaces together (Fristrom et al., 1993). Mutations in integrin genes (e.g. inflated (if), an α chain mutation and myospheroid (mys), a β chain mutation) are defective in dorsal-ventral adhesion of intervein cells and result in wing blisters (Brower and Jaffe, 1989; Wilcox et al., 1989; Zusman et al., 1990). Thus, integrins are one of many effectors of intervein differentiation.

In this paper, we investigate the genetic and developmental properties of blistered (bs) (Lindsley and Zimm, 1992), a mutation that has received little attention in previous studies. In contrast to integrin mutants where the wings are blistered but the vein pattern remains relatively undisturbed, bs mutations lead to an excess of veins as well as blisters. We describe several new bs alleles and show that, with two notable exceptions, bs appears to function autonomously. The phenotypic analysis is facilitated by phalloidin and laminin staining of pupal wings where the extent of vein and intervein can be more readily analyzed than in adult wings. We have also examined wild-type and bs wings by conventional microscopy throughout metamorphosis to establish the onset and progression of bs defects. These phenotypic analyses suggest that bs is required for the initiation of intervein development at the onset of metamorphosis and for limiting the extent of veins.

Finally, we have investigated the genetic interactions of representative bs alleles with the rho, if and mys mutants described above. The observation that rho suppresses the vein defects of bs mutants, suggests a regulatory role for bs in vein formation. This is supported by the observation that the extent of rho expression in bs wings increases during metamorphosis. The domains of ectopic rho expression closely correspond to the regions of increased vein. We propose a model for wing development in which the positions of the major veins, determined in part by the distribution of rho, is established by the end of the third instar, but that the final extent of veins, determined in part by the bs product, is not established until the onset of vein/intervein differentiation during metamorphosis. At this time, cells that fail to differentiate as intervein can still enter the vein pathway. Thus, progressively more severe bs mutants have progressively wider veins until the entire wing is, in effect, a vein.

All stocks were maintained on standard cornmeal and molasses medium at 25°C. The origin of the various bs mutants is indicated in results.

Potential bs alleles were mapped with respect to the marker Irregular facets (If) which lies 0.3 map units distal to bs. A chromosome bearing a potential bs allele (bsⁿ) and If was placed in trans to a wild-type second chromosome. Females of this genotype were crossed to bs¹⁰If homozygotes. Progeny, heterozygous for If and displaying the bs phenotype (bsⁿ+/bs¹⁰If) or homozygous for If and bs⁺(bs⁺If/bs¹⁰If) were scored as recombinants. 800–1000 flies were scored for each new bs allele and in each case the new allele mapped to 0.3 map units of If (Gotwals, 1992).

To determine the lethal phase of the lethal bs alleles, bsⁿ (where n refers to any lethal bs allele) flies were crossed to wild-type flies. Heterozygotes were then backcrossed either to bsⁿ/Cyo or Df(2R)Px⁴/CyO animals. The only lethal progeny were bsⁿ homozygotes or bsⁿ/Df(2R)Px⁴. Typically, 100 embryos from this cross were picked and their development followed by counting first, second and third instar larvae, pupae, pharate adults and adults. The lethal phase was determined by observing when 25% of the animals in a given cross died.

Chromosomes carrying extreme bs alleles (bs¹², bs¹³ and bs¹⁴) were marked with pwn (a mutation that causes wing hairs to be small and barbed (Lindsley and Zimm, 1992) and balanced over CyO. Somatic recombination was induced by X-irradiation (1,000 rads over 10 minutes) at 36, 60 and 84 (+/— 12) hours after egg laying. Wings of eclosed flies were mounted in euparol. 45 pwn clones in 1200 wings were identified and examined.

Pupae were staged from pupariation (white prepupa) for 18 to 30 hours at 25°C. Staged pupae were immersed in 4% formaldehyde in PBS. A mid-dorsal incision was made with fine iridectomy scissors through both pupal case and body wall from the edge of the operculum to the posterior tip of the animal to admit fixative. Fixation was continued at room temperature on a shaker for at least one hour and the fixed animals stored for up to one week at 4°C. Dissection of the wings was accomplished by removing the pupal case, grasping the wing hinge with fine forceps and gently pulling off the wing.

Dissected discs and wings were stained with FITC-phalloidin (Molecular Probes) or anti-laminin A (Henchcliffe et al., 1993) (kindly provided by Claire Hencliffe and Corey Goodman) as previously described (Fristrom et al., 1993). Stained wings were mounted in 50 to 75% glycerol in PBS containing 0.15% p-phenylenendiamine and examined by conventional immunofluorescence microscopy and by confocal microscopy (BIORAD 6000).

In situ hybribization of rhomboid probes used digoxigenin labelled RNA probes (Boehringer-Mannheim, 1093 657) as described previously (Sturtevant et al., 1993).

Larval and prepupal discs were dissected directly into fixative (1.5% glutaraldehyde, 0.5% formaldehyde in 0.1M sodium cacodylate). Pupal wings were dissected from animals fixed as described above then processed by standard procedures. 1μm sections for light microscopy were stained with toluidine blue and thin sections for electron microscopy were stained with uranyl acetate and lead citrate.

The blistered locus (bs; 2–107.3) is uncovered by Df(2R)Px2 and Df(2R)Px4 which define the cytological region 60C5-6; 60D1. bs has previously been represented by relatively weak viable alleles (Gotwals and Fristrom, 1991; Lindsley and Zimm, 1992). We have identified a new set of bs alleles including some with extreme wing phenotypes and others that are lethal as homozygotes. A set of EMS induced mutations uncovered by both Df(2R)Px2 and Df(2R)Px4 (hereafter Px4) were kindly provided by Elizabeth Raff and Robert Dettman. Five of these failed to complement viable bs alleles and mapped 0.3 map units centromere-proximal to If, coincident with the location of the blistered locus. We conclude that these five mutations are all bs alleles. Another allele pLFO6 was recovered in association with a nearby P-element insertion. Below, we rename and characterize each new bs allele in ascending order of severity (Table 1) – all are phenotypically more severe than existing alleles (Lindsley and Zimm, 1992). Two of the alleles are lethal but the basis for the lethality is not yet understood. bs14 behaves as a genetic null in that bs14 homozygotes and bs14/Df(2R)Px2 are both lethal as second to third instar larvae. Adult defects exhibited by the viable alleles appear to be restricted to the wings. Wing phenotypes range from ectopic venation and a moderate frequency of localized blisters (Fig. 1B) through very thick posterior veins and a high frequency of blisters (Fig. 1C) to a complete loss of adhesion between the two wing surfaces resulting in ballooned wings (Fig. 1D). Ballooned wings are typically composed entirely of small vein-like cells except for a strip in the center of the wing, corresponding to the intervein region between L3 and L4. This region is also resistant to ectopic vein formation in the weaker bs alleles and in other excess vein mutants (Diaz-Benjumea and Garcia-Bellido, 1990). In all the new bs mutants the wings are held out at an angle to the body axis and the animals are flightless. However, even in the most severely defective wings, neural tissues such as the bristles associated with the anterior margin and the campaniform sensillae appear to develop normally.

Mosaic analyses were undertaken to examine the wing phenotypes of the lethal alleles as well as to determine whether bs functions autonomously or non-autonamously. Homozygous mosaic patches of the lethal alleles bs13 and bs14 and the viable allele bs12 marked with the hair mutation pwn were generated. We observed both autonomous and non-autonomous aspects of the bs phenotype in mosaics.

Clones of the lethal allele bs14 were clearly autonomous with respect to the effects of bs on cell size. Homozygous bs14/bs14 clones of all sizes in all regions of the wing contained cells that are considerably smaller, as judged by hair density, than either the adjacent cells or the overlying cells (Fig. 2A,B). Clones were usually yellowish in color and the lateral spacing of clone hairs closer than the proximal-distal spacing. Where clones cross a vein, the boundary between vein and intervein is obscure because all the clone cells resemble vein cells. Nevertheless, the center of the vein could usually be identified by two rows of hairs directed towards each other. All homozygous bs14 patches, including those between L3 and L4 are characterized by small vein-sized cells. In contrast, in wings of the most severe viable alleles, a strip of large intervein-sized cells persists in the region between L3 and L4. Thus, bs14 has a more severe wing phenotype than any of the known viable alleles. This, combined with its genetic behavior (above), makes bs14 a likely candidate for a null allele.

In contrast to bs14, clones of the lethal bs13 showed a weak wing phenotype. Homozygous pwn patches only exhibited defects when they occurred in the vicinity of L5 and the anterior cross vein, a region that is particularly susceptible to ectopic vein formation. In this location, a bs13 clone typically contains an ectopic vein but not all the clone cells are necessarily vein-like. Thus, the wing phenotype of bs13 is comparable to that of the weak alleles bs2 or bs10.

(a) Widening of dorsal veins overlying ventral clones. Veins L3 and L5 are referred to as dorsal veins because they exhibit dorsal bulges (L2 and part of L4 bulge ventrally (Garcia-Bellido, 1977)). We found several examples in all three of the alleles tested where ventral clones incorporating L3 resulted in thickening of the dorsal aspect of the vein from 4 to about 8–10 cells in width (Fig. 2C-E). A similar phenomenon was seen with ventral clones underlying L5. This may represent true non-autonomy of bs, i.e. expression of bs in ventral cells is required to inhibit vein formation in the overlying dorsal cells. Alternatively, there may be a general requirement for dorsalventral adhesion between cells immediately adjacent to the vein in order to prevent them from becoming vein-like.

(b) Blisters extend beyond clone borders. Virtually all bs/bs clones were associated with blisters that typically extended well beyond the boundaries of the clone. This could be a non-specific effect resulting from hydrostatic pressure within the blister mechanically disrupting dorsal/ventral connections in adjacent heterozygous tissue. A similar phenomenon is seen in wing clones of integrin mutants (Brower and Jaffe, 1989; Zusman et al., 1990). As transmembrane adhesion molecules integrins presumably function autonomously. However, in the case of bs mosaics where the nature of the product is unknown and blisters can occupy up to six times the area of the marked clone, the possibility of a true non-autonomous effect on intervein adhesion cannot be ruled out.

We have examined the morphogenesis of wild-type and bs wings during pupal development to characterize further the bs phenotype and to describe the normal sequence of events in vein/intervein differentiation. Phenotypic analysis of adult wings is limited by the impermeable cuticle, the death of intervein cells and the tendency for blistered wings to crumple. We have therefore examined pupal wings at 30 hours apf (after puparium formation), a stage when the vein pattern is complete and the wing is free of cuticle. Confocal microscopy of phal-loidin-stained pupal wings was used to assess apical cell size and the extent of basal apposition. (Extensive actin filaments revealed by phalloidin-staining are located apically in association with the zonula adherens and in basal networks (Condic et al., 1990)). Laminin-staining was used to determine the extent of vein tissue (the acquisition of a laminin-containing basal lamina appears to be a specific property of vein cells (Fristrom et al., 1993; Murray et al., personal communication)). We also used these techniques to examine the course of vein/intervein differentiation between 18 and 30 hours apf in both bs and wild-type wings. These studies provide some new insights into the normal process of vein/intervein differentiation as well as some clues to the basis of the blistered phenotype.

Between 12 and 18 hours apf, the wild-type wing is a hollow sac composed of rapidly dividing unspecialized epithelial cells. The dorsal and ventral surfaces become reapposed as the basal surfaces of intervein cells extend processes to the opposite side (Fristrom et al., 1993; Waddington, 1941). Reapposition begins at 18 hours apf as longitudinal bands of cells, two to three cells in width, extend long processes basally to connect with their partners on the opposite surface (Fig. 3A). These ‘intervein bands (IBs)’ arise progressively from anterior to posterior in each of the five intervein regions so that by 21 hours apf, five IBs are present. Between 21 and 30 hours apf, the cells on either side of the IBs and adjacent to the wing margin reappose, progressively ‘zipping’ the intervein epithelia together until only the veins remain unapposed (Fig. 3C,E).

As intervein cells become apposed basally, vein cells become reduced in diameter apically. In the 18 hour wing, the final round of cytokinesis is taking place (Schubiger and Palka, 1987) with numerous large premitotic cells interspersed with smaller postmitotic cells (Fig. 3B). The average diameter of the postmitotic cells is 4.5 microns. By 21 hrs, broad bands of cells of smaller apical diameter (2.5 microns) appear (Fig. 3D). When apical and basal sections are superimposed it is evident that the small cells lie between the IBs, centered over broad basal channels (Fig. 3C,D). The bands of cells with small apical diameter subsequently become narrower and more sharply defined until 30 hours apf, when the veins have reached their final dimensions (of 2–4 rows of cells in width; Fig. 3E,F). Note that the decrease in apical diameter of vein cells is accompanied by an increase in intensity of phalloidin staining suggesting that the apical constriction is mediated by contraction of the circumapical actin filaments (Condic et al., 1990).

When reapposition of intervein cells is complete, the remaining unapposed channels form the vein lacunae. These channels become lined with laminin (Fristrom et al., 1993; Murray et al., personal communication), as detected by staining with an antibody to the Drosophila laminin A chain (Henchcliffe et al., 1993). Thus, the 30 hour apf lamininstained wing resembles a miniature adult wing (Fig. 4A) with five longitudinal and two cross veins outlined by laminin staining (Fig. 4B). One difference between the pupal and adult wing is that a marginal channel extends around the entire periphery of the wing in pupae (and prepupae) but in adults there is no vein associated with the posterior wing margin. In pupae, not only is the entire marginal channel laminin stained, but the vein markers rho (Sturtevant et al., 1993) and star (Heberlein et al., 1993) are also expressed in this location, indicating that the posterior channel can be considered a transient vein.

In summary, intervein differentiation in wild-type wings involves the extension of basal processes that connect with cells on the opposite surface. This process begins in narrow intervein bands and progresses laterally until only the vein channels remain unapposed. Vein cells constrict apically and deposit a laminin-containing matrix basally.

At 30–36 hours apf there is a progressive increase in the extent of laminin-lined channels in progressively more severe bs mutants (Fig. 4C-F). We also find ectopic laminin-lined channels in areas that form ectopic veins in the adult (Fig. 4C,D). In bs12 wings, channels associated with L4 and L5 are often fused into a single large posterior ‘vein’ (Fig. 4E). In bs11/Px4 the entire wing is a laminin-lined tube, except for a narrow band along the posterior border (Fig. 4F).

There was no evidence of intervein blisters in pupal wings; those regions that were not laminin-lined were basally apposed. Even wings of bs12, which are typically completely ballooned in the adult, are always apposed to some extent in the anterior half of the wing and in a narrow band along the posterior margin (Fig. 4E). We conclude that blisters arise later, perhaps during the expansion stage, 50–60 hours apf (Fristrom et al., 1993) or during eclosion.

In phalloidin-stained pupal wings of bs mutants, we found, as expected, that cells of small apical diameter were associated with basal channels. Surprisingly, the distribution of cells with small apical diameter extends well beyond the corresponding basal channels. This is demonstrated in the bs11 wing shown in Fig. 3G,H. In basal view, the distal vein lacunae are only slightly wider than normal whereas in apical view small cells occupy the entire anterior and posterior thirds of the wing (as they often do in the adult wing). Thus, large areas of cells have properties of both vein and intervein: an apical diameter characteristic of vein cells and basal behavior characteristic of intervein cells. Cells of small diameter could arise in a number of ways; for example, defects in expansion during the preceding prepupal period could lead to abnormally small intervein cells. We therefore compared cell diameters in bs and wild-type wings at 18 hours apf when all postmitotic cells have a diameter characteristic of intervein cells. At this stage, wings of bs11 and bs12 were similar to wild type in both cell size (4.5 μm? and wing area (not shown). We tentatively conclude that wing cells of small apical diameter in at least some bs mutants arise by apical contraction during vein differentiation.

To identify the earliest detectable defects in bs wings, we compared prepupal wings from the most extreme viable bs combination (bs11/Px4) with those of wild type. The prepupal period, 0–12 hours apf, is the first stage of metamorphosis when the imaginal discs undergo morphogenesis to form the approximate shapes of the adult appendages. The wing disc, for example, is converted from a concentrically folded disc to a bilayered epithelium. Clear phenotypic defects in bs wings during the first few hours of the prepupal period presage equivalent defects seen during pupal development described above.

Formation of the wing bilayer is accomplished largely by local changes in cell shape and size so that the wing disc folds along the wing margin bringing dorsal and ventral surfaces into apposition (Fristrom and Fristrom, 1993). The basal lamina of the larval disc forms a potential barrier to the formation of cellcell contacts between the apposed epithelia (Fig. 5A). Two different mechanisms appear to be involved in eliminating the larval basal lamina. (i) Beginning at the end of the third instar, cells in the vicinity of the wing margin detach from the basal lamina so that, as the wing margin expands and folds, the basal lamina is left behind. This detachment may be facilitated by the secretion of a proteoglycan-containing network between the cells and the basal lamina (Brower et al., 1987). Cells on either side of the margin elongate until their basal cell surfaces make direct contact (Fig. 5A). (ii) In more central regions of the wing, the basal lamina becomes trapped between the apposed surfaces and is subsequently degraded so that, by 6 hours apf, it is no longer detectable as a discrete structure. The apposed basal surfaces make contact and differentiate the junctions of the prepupal transalar apparatus (Fristrom et al., 1993). Throughout the process of dorsoventral apposition the basal cell surfaces are highly convoluted and vesicular suggestive of secretory and/or endocytotic activity. Laminin-lined channels or preveins lie in the approximate positions of the future veins L3-L5 (Waddington, 1941; Murray, personal communication) but are broader and less clearly demarcated than pupal wing veins. No clear differences in apical cell diameter between prevein and intervein cells are evident. As in pupal wings, a narrow laminin-lined channel encompasses the entire wing margin.

Defects in wing discs of bs11/Px4 become apparent in the first 2 hours after puparium formation. The distal and anterior submarginal cells detach from the larval basal lamina but do not become closely apposed resulting in an empty space between basal surface and basal lamina (Fig. 5B). In the central region, after the basal lamina disappears, the apposed basal surfaces are often in close proximity but show no tendency to make contact and form junctions. Ultrastructurally, the basal surfaces remain smooth and unvesiculated throughout the prepupal period (not shown). From 6 hours apf until the pupal period, a single, large, hemocyte-filled, central channel stretches from the anterior margin and includes most of the wing except for a narrow band of apposed cell along the posterior margin comparable to the pupal wing shown in Fig. 4F. Prepupal discs from bs11 homozygotes, are intermediate between bs11/Px4 and wild type with abnormally wide, laminin-lined, prevein lacunae alternating with narrow stretches where dorsal and ventral surfaces have adhered. Thus, bs appears to have similar prepupal and pupal phenotypes; failure in apposition of intervein cells and expansion of vein territories.

To investigate bs function further, we examined the interaction of various bs alleles with mutations in genes whose specific role in wing development is understood.

rhomboid (rho) is expressed where longitudinal veins will differentiate and is a component of the signaling pathway that specifies vein cell fate (see Introduction). rhove is a wingspecific allele of rho that results in lack of rho expression in wings along with the loss of the distal sections of longitudinal veins (Fig. 6A). We have constructed stocks homozygous for both rhove and the bs alleles bs10, bs11 and bs12. In general, rhove suppresses the vein defects associated with bs mutations but L2 to L5 remain truncated distally (Fig. 6B-D). rho is thus epistatic to bs with respect to vein formation. In combination with bs10 (Fig. 6B), rhove completely suppresses the formation of ectopic veins and blisters. Except for their ‘held out’ orientation, these wings cannot easily be distinguished from homozygous rhove alone. In combinations of rhove with the more severe bs alleles, veins remain wider than normal in the proximal part of the wing (Fig. 6D), an area beyond the domain of rho function.

The expression of rho in bs mutants (bs2, bs10 and bs11) was examined. In late third instar discs, rho expression is restricted to narrow stripes corresponding to longitudinal veins in both bs and wild type (Fig. 7A,D). In very young prepupal wings (2 hours apf; Fig. 7B,E), the bands of rho expression associated with L2 and L5 in bs are wider and more intense than in wild type, a difference that becomes more pronounced by the tongue stage (4 hours apf; Fig. 7C,F). The expression of rho in pupal wings (30 hours apf) is also expanded (not shown). In bs11 there are broad bands of solid rho expression similar to the prepupal pattern, whereas in the weaker mutants bs10 and bs2 there is a reticulate pattern of ectopic staining (Sturtevant and Bier, unpublished data). The increased area of rho expression thus anticipates the final vein phenotype of bs mutants. These observations lend strong support to the morphological evidence that the excess ‘veins’ of bs mutants represent true veins rather than a failure of intervein to differentiate. They also suggest that bs has a role in restricting the expression of rho during metamorphosis.

The effects of bs on rho expression do not account for all aspects of the bs phenotype and suggest that bs mutants have primary defects in intervein specification and/or differentiation. For example, rho ve only weakly suppresses blisters. With the more severe alleles, bs11 and bs12, a high frequency of blisters remains (about 50% with bs11 and 95% with bs12). Indeed, in combination with bs12, the wings often remain balloon-like. So even though, in the absence of rho, veins are suppressed, the remaining intervein regions often fail to adhere dorsoventrally. We note that rhove does effectively (100% of the progeny) suppress the posterior blisters associated with bs10. This might be an indirect result of eliminating ectopic veins in this region of the wing i.e. increased area of adhesion may compensate for a presumed decrease in strength of adhesion between the two wing surfaces. rho ve also fails effectively to suppress the non-vein-associated decrease in cell size (see Fig. 3H) and consequently the wings remain small (Fig. 6C,D).

A role for bs in intervein formation is further supported by genetic interactions between bs and integrin mutants. Integrins, transmembrane adhesion proteins, are expressed only in the intervein regions of pupal wings where they eventually become localized to intercellular junctions between the dorsal and ventral wing surface (Fristrom et al., 1993). Viable mutations of if (α2 integrin) and somatic clones of lethal mys (β integrin) mutations result in adult wing blisters similar to those seen in weak bs mutations (Brower and Jaffe, 1989; Zusman et al., 1990) but cause little or no disruption in the normal vein pattern. We looked for dominant interactions between integrin and bs mutations by crossing females carrying if (if3) or mys (mysnj) to bs males. There was an increase in frequency and severity of blisters in most of the combinations examined particularly with the more severe bs alleles. (see also Wessendorf et al., 1992). For example, if3/Y; bsdefF31/+ results in 100% of wings having blisters (compared with 24% in if3/Y; +/+). For the most part blistered/integrin combinations did not affect venation even when the blisters were very large.

The metamorphosis of the larval wing disc into an adult wing takes place in two stages: the prepupal period (0 –12 hours apf) and the pupal period (12 –96 hours apf). In the prepupal period, the dorsal and ventral wing epithelia become apposed, converting the concentrically folded wing disc to a bilayered wing. A coarse pattern of unapposed ‘prevein channels’ is established, presaging the events of the pupal period. At the onset of the pupal period, the wing layers separate, revert to an unspecialized state and undergo further cell division. Definitive wing differentiation begins around 18 hours apf with the progressive reapposition of dorsal and ventral epithelia (Fig. 3). Reapposition begins at the center of intervein regions and proceeds laterally until, by 30 hours apf, only the vein channels remain unapposed. After reapposition of intervein regions has commenced, future vein cells constrict in apical area. When intervein reapposition is complete, vein cells become coated basally by a laminin-containing ECM and laminin is cleared from the intervein regions.

We have identified a series of mutants that fail to complement each other and existing bs mutants and that map to the bs locus 0.3 map units proximal to If. These have therefore been designated as bs alleles and numbered bs10-bs14 in increasing order of severity. Of these new alleles, three are viable and two are lethal as homozygotes. Viable alleles and mosaic patches of lethal alleles are phenotypically characterized by defects in wing development.

Increasingly severe bs mutants differentiate vein over increasingly large areas of the wing as judged by the extent of laminin-lined channels and the expansion of rho expression in prepupal and pupal wings as well as by the adult phenotype. The suppression of extra vein in bs mutants by rhove (a mutant lacking rho expression in wings) further supports the view that the extra vein in bs wings represent true veins. We conclude that loss of bs+ function results in the conversion of intervein to vein. In addition to extra vein, bs mutants show intervein defects such as loss of adhesion between dorsal and ventral surfaces leading to the blisters from which the mutant is named. Rhove only weakly suppresses such defects. In contrast, mutations in integrin genes (effectors of intervein adhesion during pupal development) interact dominantly with bs mutants to increase the frequency of intervein blisters. Taken together, the results indicate that bs functions both in intervein formation and in determining extent of vein. We suggest that these two processes are closely linked during development and propose the following model.

It is well established that the positions of the longitudinal veins are defined by the end of the third larval instar. We propose (1) that the lateral extent of veins remains labile until the pupal period and (2) that the progress of intervein specification and/or differentiation during metamorphosis limits the extent of vein formation. Characteristics of mutations in bs (described here) and Notch (N), a transmembrane receptor involved in cell fate determination (Kidd et al., 1989) support this view as follows.

(a) In severe bs mutants much of the wing is converted to vein but rho expression in such mutants is not affected until early in the prepupal period. Thus, the excess veins exhibited by bs mutants presumably arise after the onset of metamorphosis. This is consistent with the observation that rho functions in vein formation throughout most of metamorphosis (Sturtevant and Bier, unpublished data). We conclude that one function of bs is directly or indirectly to delimit the prepupal and pupal expression of rho and thereby prevent the acquisition of extra vein during metamorphosis.

(b) Recent work with N also demonstrates that vein width can be modified as late as 24 hours apf. One of many phenotypes associated with N, is a slight increase in vein thickness. This defect occurs during the pupal period as demonstrated by the temperature-sensitive period of Nts (Shellenbarger and Mohler, 1978) and for N deletion constructs missing all or part of the extracellular domain (Rebay et al., 1993).

(a) Three lines of evidence indicate that bs functions in intervein cells. First, morphological defects in severe bs mutants are evident by 2 hours apf suggesting that bs has a primary defect in intervein formation, which precedes its effect on rho expression. Second, intervein defects persist in severe bs mutants even when excess veins are suppressed by rhove. Third, bs appears autonomously to convert intervein to vein in all regions of the wing. It is therefore likely that bs is expressed throughout intervein regions at least by the onset of metamorphosis. (It is also possible that bs has an earlier function (in larval discs) that is not manifest until morphogenesis begins.)

(b) N is expressed predominantly in intervein regions (Fehon et al., 1991) and is presumably involved in signalling formation of intervein at the expense of vein (Heitzler and Simpson, 1991). The slight increase in vein width associated with loss of N function is strongly dominantly enhanced by bs (L. Wessendorf and D. Fristrom, unpublished data). Although it is premature to conclude that bs is part of a N signalling pathway, both these genes are evidently expressed in intervein cells and affect vein width.

The morphogenetic events involved in the pupal differentiation of intervein and vein (summarized in Fig. 8) are consistent with this model. Intervein differentiation begins in longitudinal stripes midway between the future veins and expands laterally until the veins are reached. At about the same time, stripes of smaller vein cells become evident. The filling in of the regions between the alternating stripes of vein and intervein may involve a complex two-way lateral signalling system that establishes cell identity as morphogenesis proceeds.

In the context of this model, we view the bs product as necessary for the triggering or initiation of intervein formation, first manifested by the apposition of dorsal and ventral epithelia. When bs+ function is limiting, intervein differentiation fails to expand laterally or in the most severe cases fails to begin; rho expression expands and vein forms in place of intervein. We cannot yet predict the nature of the bs product; possibilities range from a signalling molecule required for intervein identity to a transcription factor required for regulating downstream components of the intervein pathway or even an effector molecule (e.g. a basal lamina degrading protease) required early in intervein differentiation. The cloning and characterization of the bs gene will undoubtedly contribute to a better understanding of the processes involved in cell fate decisions during wing development.

We are indebted to Elizabeth Raff for providing mutants and Corey Goodman for providing anti-laminin antibodies. We also thank Marjorie Murray and Lisa Wessendorf for sharing unpublished data and for helpful discussions and comments on the manuscript. Part of this work was carried out by D. Fristrom while a visiting scientist at the C.S.I.R.O. Division of Animal Production, Prospect, N.S.W. Australia. The work was supported by grants from NIH and ACS.