ABSTRACT
The Drosophila rhomboid (rho) and Egf- and r genes are members of a small group of genes required for the differ- and entiation of various specific embryonic and adult structures. During larval and early pupal development expression of rho in longitudinal vein primordia mediates the localized formation of wing veins. In this paper we investigate the genetic hierarchy guiding vein development, by testing for genetic interactions between rho alleles and a wide variety of wing vein mutations and by examining the pattern of rho expression in mutant developing wing primordia. We identify a small group of wing vein mutants that interact strongly with rho. Examination of rho expression in these and other key vein mutants reveals when vein development first becomes abnormal. Based on these data and on previous genetic analyses of vein formation we present a sequential model for establishment and differentiation of wing veins.
INTRODUCTION
The Drosophila wing is emerging as an important model system for analyzing pattern formation in a fully cellularized and proliferating epithelial sheet. Because subtle wing defects can be readily identified, many mutants affecting the shape of wings or disrupting the normal pattern of veins have been recovered and are among the classic mutations used as genetic markers in Drosophila.
While there has been a significant amount of interest recently in early events contributing to anterior- and posterior and dorsal- and ventral pattering of imaginal discs, differentiation of adult structures such as wing veins has received less attention. Analysis of Drosophila wing vein morphogenesis in wild- and type (see Fig. 1) and mutant developing wings dates back to Waddington (1940) and more recently has been studied by García- and Bellido and colleagues (Díaz- and Benjumea and García- and Bellido, 1990a; García- and Bellido and de Celis, 1992). Compre- and hensive analysis of double mutant combinations of vein mutants (Díaz- and Benjumea and García- and Bellido, 1990a) and genetic mosaic analysis have lead to the formulation of a model of vein formation involving various forms of cell- and cell com- and munication (see García- and Bellido and de Celis, 1992 for a recent review).
Morphogenesis of the wing. (A) Diagram of wing development. Upper left: drawing of a third- and instar larval wing disc. The wing arises from the oval region of the disc known as the wing pouch. The remaining wing imaginal cells give rise to the thoracic body wall. Primordia for the longitudinal veins (L1- and L5) are stippled (resembling rho expression) with future dorsal (dark stipple) and ventral (light stipple) surfaces of veins confined to the wing pouch separated by a strip of cells that gives rise to the margin. Sensory organ precursors (open circles) form along the future anterior edge of the margin (M) and along the L3 vein at this stage. Upper right: drawing of an everting disc during the early prepupal stage. The pouch everts bringing the future dorsal and ventral surfaces into contact for the first time. Interactions between the dorsal and ventral surfaces of the wing ultimately lead to alignment of the dorsal and ventral components of the longitudinal veins. Bottom: Drawing of an adult wing. Veins bulge on either the dorsal surface (dark stipple) or the ventral surface (light stipple), defining a major and a minor surface for each vein. The pattern of major and minor surfaces known as corrugation tends to alternate for consecutive longitudinal veins. That this pattern of corrugation is highly conserved in diverse insect species provides some of the strongest evidence that wings evolved once during early insect evolution. The marginal vein (or costal vein) is designated by ‘M’ and the longitudinal veins are numbered beginning with L0 (corresponding to the subcostal vein in other nomenclatures) and ending with the partial vein L6. The anterior cross vein connects L2 to L3 proximally and the posterior cross vein connects L4 to L5 more distally. (B) A wild- and type adult wing corresponding to the bottom panel of part A.
Morphogenesis of the wing. (A) Diagram of wing development. Upper left: drawing of a third- and instar larval wing disc. The wing arises from the oval region of the disc known as the wing pouch. The remaining wing imaginal cells give rise to the thoracic body wall. Primordia for the longitudinal veins (L1- and L5) are stippled (resembling rho expression) with future dorsal (dark stipple) and ventral (light stipple) surfaces of veins confined to the wing pouch separated by a strip of cells that gives rise to the margin. Sensory organ precursors (open circles) form along the future anterior edge of the margin (M) and along the L3 vein at this stage. Upper right: drawing of an everting disc during the early prepupal stage. The pouch everts bringing the future dorsal and ventral surfaces into contact for the first time. Interactions between the dorsal and ventral surfaces of the wing ultimately lead to alignment of the dorsal and ventral components of the longitudinal veins. Bottom: Drawing of an adult wing. Veins bulge on either the dorsal surface (dark stipple) or the ventral surface (light stipple), defining a major and a minor surface for each vein. The pattern of major and minor surfaces known as corrugation tends to alternate for consecutive longitudinal veins. That this pattern of corrugation is highly conserved in diverse insect species provides some of the strongest evidence that wings evolved once during early insect evolution. The marginal vein (or costal vein) is designated by ‘M’ and the longitudinal veins are numbered beginning with L0 (corresponding to the subcostal vein in other nomenclatures) and ending with the partial vein L6. The anterior cross vein connects L2 to L3 proximally and the posterior cross vein connects L4 to L5 more distally. (B) A wild- and type adult wing corresponding to the bottom panel of part A.
The earliest known manifestation of differences between future vein and intervein cells is the localized expression of the rhomboid (rho) gene in rows of imaginal disc cells coin- and ciding with vein primordia (Sturtevant et al., 1993). rho is required for vein formation as the loss of function allele rhove results in truncated veins (Díaz- and Benjumea and García- and Bellido, 1990a; Sturtevant et al., 1993). Restricting rho to vein primordia is important for limiting vein formation to appro- and priate locations since ubiquitous expression of rho leads to the production of ectopic veins (Sturtevant et al., 1993; Noll et al., 1994).
rho is likely to contribute to signaling through the EGF- and Receptor (EGF- and R) as rho and Egf- and r belong to a small group of genes (ventrolateral or spi group genes) defined by similar complex embryonic mutant phenotypes (Mayer and Nüsslein- and Volhard, 1988; Bier et al., 1990; Rutledge et al., 1992; Kim and Crews, 1993; Raz and Shilo, 1993). The finding that the ventrolateral group gene spitz encodes an EGF- and like growth factor (Rutledge et al., 1992) is consistent with EGF- and R signaling serving as the focus of ventrolateral pathway. Further evidence for this hypothesis has been provided by strong genetic interactions between rho alleles and mutations in com- and ponents of the EGF- and R/RAS signaling pathway during embryo- and genesis (Noll et al., 1994; J.W. O’Neill and E. Bier, unpub- and lished data) and adult development (Sturtevant et al., 1993; Noll et al., 1994) and by interactions between Egf- and r alleles and ventrolateral mutants during embryogenesis (Raz and Shilo, 1993).
In this paper we identify a small group of mutants among the large collection of existing wing vein mutants that interact strongly with rho during wing vein development. We then examine the pattern of rho expression in these and other wing vein mutants throughout vein development. These experiments distinguish between mutants with similar final adult pheno- and types based on the stage at which rho expression first becomes abnormal. We propose a model for wing vein formation derived from these results and from previous double mutant and mosaic analyses (García- and Bellido, 1977; Díaz- and Benjumea et al., 1989; Díaz- and Benjumea and García- and Bellido, 1990a; García- and Bellido and de Celis, 1992).
MATERIALS AND METHODS
Fly stocks
All genetic markers and chromosome balancers used are described in Lindsley and Grell (1968) and Lindsley and Zimm (1992). Several wing vein mutants (tg, cg, vvl, and vn) were kindly provided by Antonio García- and Bellido. Other stocks were obtained from the Bloom- and ington, Indiana and Bowling Green, Ohio Drosophila Stock Centers.
Mounting fly wings
Wings from adult flies were dissected in isopropanol and mounted in Canada Balsam mounting medium (Gary’s magic mountant) following the protocol of Lawrence et al. (in Roberts, 1986). Mounted wings were photographed under Nomarski optics with a 4× lens on a compound microscope.
In situ hybridization to whole- and mount embryos or discs
In situ hybridization to whole- and mount discs and embryos was performed using digoxigenin (Boehringer- and Mannheim, 1093 657) labeled RNA probes (O’Neill and Bier, 1994) as described by Sturte- and vant et al., 1993.
RESULTS
rho interacts genetically with a small set of wing vein mutants
In a comprehensive genetic survey of wing vein mutants, Díaz- and Benjumea and García- and Bellido examined many double mutant combinations, leading these authors to propose various sub- and groupings of loss- and of- and vein and extra- and vein mutants. Assignment to various subgroups was based on superadditive interactions between members within subgroups and on consistent positive or negative interactions between members of different subgroups (Díaz- and Benjumea and García- and Bellido, 1990a). To extend these observations with respect to genes interacting with rho during vein development, we combined the loss- and of- and function rhove allele (Fig. 3A) or constitutive gain- and of- and function rhoHS alleles of differing strengths (Fig. 3B- and D; Sturtevant et al., 1993) with many of the currently available wing vein mutants (Fig. 2). In each case we scored for interactions with the test mutant as a heterozygote and in many cases also as a homozygote. The outcome of these crosses is summarized in Table 1 with mutants grouped according to general phenotypic class. Examples of strong genetic interactions are shown in Fig. 3E- and P.
Key wing- and vein mutant phenotypes. Wings shown are homozygous for the vein mutant shown unless specifically designated otherwise. (A) kn/kn (double arrow indicates that L3 and L4 are spaced closer together than in wild- and type), (B) ri/ri, (C) ab/ab, (D) vn1/vn1, (E) vn1/vnM1, (F) net/net (G) h1/h1, (H) Ser/+, (I) NAx/+, (J) Nts early (raised at 29°C during second through third larval instars), (K) Nts late (raised at 29°C from 0 hours AP through apolysis, e.g. 20 hours AP), (L) DpN Y/+, (M) Dl9P39/+, (N) tkv1/tkv1 raised at 18°C to enhance phenotype, (O) bs2/bs2, (P) Vno/+, (Q) Vno/Vno, (R) det/det (arrows point to where the posterior cross vein is detached from L4 and L5).
Key wing- and vein mutant phenotypes. Wings shown are homozygous for the vein mutant shown unless specifically designated otherwise. (A) kn/kn (double arrow indicates that L3 and L4 are spaced closer together than in wild- and type), (B) ri/ri, (C) ab/ab, (D) vn1/vn1, (E) vn1/vnM1, (F) net/net (G) h1/h1, (H) Ser/+, (I) NAx/+, (J) Nts early (raised at 29°C during second through third larval instars), (K) Nts late (raised at 29°C from 0 hours AP through apolysis, e.g. 20 hours AP), (L) DpN Y/+, (M) Dl9P39/+, (N) tkv1/tkv1 raised at 18°C to enhance phenotype, (O) bs2/bs2, (P) Vno/+, (Q) Vno/Vno, (R) det/det (arrows point to where the posterior cross vein is detached from L4 and L5).
Wing phenotypes resulting from interactions between vein mutants and rho. All crosses were performed at room temperature (22°C). (A) rhove/rhove (arrowhead points to the location of the missing L0 (subcostal) vein). (B) rhoHS- and Wk/+: weak constitutive rhoHS extra- and vein phenotype (arrowhead indicates subtle delta at the junction of L4 with the margin). (C) rhoHS- and Mod/+: moderate constitutive rhoHS extra- and vein phenotype (arrow indicates an ectopic vein spur between L3 and L4 typical of this line). (D) rhoHS- and Stg/+: strong constitutive rhoHS extra- and vein phenotype. (E) rhoHS- and Modkn/+ kn. (F) hh2 +/+ rhoHS- and Mod. (G) vnM1 +/+ rhove. (H) vn1rhove/vn1rhove. (I) net/net; rhove/rhove. (J) net/+; rhoHS- and Mod/+. (K) rhoHS- and Modh1/ + h1. (L) Ser +/+ rhoHS- and Stg. (M) DpN w+ Y/+; rhoHS- and Mod/+. (N) Vno +/+ rhoHS- and Wk. (O) rhoHS- and Wktkv1/+ tkv1. (P) bs2/+; rhoHS- and Mod/+.
Wing phenotypes resulting from interactions between vein mutants and rho. All crosses were performed at room temperature (22°C). (A) rhove/rhove (arrowhead points to the location of the missing L0 (subcostal) vein). (B) rhoHS- and Wk/+: weak constitutive rhoHS extra- and vein phenotype (arrowhead indicates subtle delta at the junction of L4 with the margin). (C) rhoHS- and Mod/+: moderate constitutive rhoHS extra- and vein phenotype (arrow indicates an ectopic vein spur between L3 and L4 typical of this line). (D) rhoHS- and Stg/+: strong constitutive rhoHS extra- and vein phenotype. (E) rhoHS- and Modkn/+ kn. (F) hh2 +/+ rhoHS- and Mod. (G) vnM1 +/+ rhove. (H) vn1rhove/vn1rhove. (I) net/net; rhove/rhove. (J) net/+; rhoHS- and Mod/+. (K) rhoHS- and Modh1/ + h1. (L) Ser +/+ rhoHS- and Stg. (M) DpN w+ Y/+; rhoHS- and Mod/+. (N) Vno +/+ rhoHS- and Wk. (O) rhoHS- and Wktkv1/+ tkv1. (P) bs2/+; rhoHS- and Mod/+.
Most wing vein mutants when heterozygous do not modify rhoHS ectopic vein phenotypes or exacerbate the rhove loss- and of- and vein phenotype (Table 1; Díaz- and Benjumea and García- and Bellido, 1990a). The nature of the relatively small number of dominant interactions we observed (i.e. suppression or enhancement) generally could be predicted from the phenotype of the test mutant alone (see legends to Table 1 and Fig. 3 for details). Thus, loss- and of- and vein mutants suppress rhoHS ectopic veins and enhance the rhove loss- and of- and vein phenotype, whereas extra- and vein mutants have opposite effects on rhoHS and rhove phenotypes. These data confirm previous interpretations of rhoHS extra- and vein phenotypes as gain of function rho alleles (Sturtevant et al., 1993). Additionally, we observed several superadditive inter- and actions between rhoHS phenotypes and homozygous mutants. We refer to wing vein mutants that interact strongly with rho in vein formation (hh, dpp, kn, vn, vvl, net, px, Ser, Dl, N, tkv, bs, Vno) and previously identified genes interacting with rho (e.g. ventrolateral group genes and components of RAS signaling cascade such as Star, Egf- and r, Star, ras1, gap1, and rl) as the rho interacting group. One obvious feature of this group is that it comprises examples of virtually every subclass of vein mutant listed in Table 1. As described below, these diverse mutants affect vein formation at different developmental stages, consistent with data indicating that rho functions throughout the course of vein formation.
Genes interacting with rho also interact with each other
To determine whether members of the rho interacting group are intimately involved in a common aspect of vein develop- and ment in addition to interacting with rho, we crossed these mutants to each other to generate a matrix of trans- and heterozy- and gous combinations. The results of these crosses are presented in Table 2. Examples of some of the most striking interactions are shown in Fig. 4 (see legends to Table 2 and Fig. 4 for details). The most prominent feature of Table 2 is the high frequency with which dominant trans- and heterozygous interac- and tions were observed within this pre- and selected group of mutants. In the extensive study of double mutant combinations described by Díaz- and Benjumea and García- and Bellido (1990a) very few dominant interactions between recessive wing vein mutants were observed. As virtually all interactions observed between mutants of the rho interacting group could be predicted based on how they interacted with rho in Table 1, it is likely that these dosage- and dependent genetic interactions reflect genes functioning in concert during vein formation.
Wing phenotypes resulting from interactions among vein mutants. (A) Egf- and rIK35/+; vn1/vn1. (B) Egf- and rElp/+; vn1/vn1 (arrow indicates rescued vein segment which is always missing in vn1/vn1 wings). (C) net Egf- and rElp/net + (arrow points to ectopic vein running parallel to L5). (D) net +/+ px. (E) net S /net + (arrows point to breaks in the ectopic veins) (F) net/net; ri/ri (arrow indicates the site at which L2 fuses with L3 – this is a highly penetrant phenotype). (G) net/net; det/det (arrows point to breaks in the ectopic veins and a floating vein segment). (H) Vno +/+ Dl9P39.
Wing phenotypes resulting from interactions among vein mutants. (A) Egf- and rIK35/+; vn1/vn1. (B) Egf- and rElp/+; vn1/vn1 (arrow indicates rescued vein segment which is always missing in vn1/vn1 wings). (C) net Egf- and rElp/net + (arrow points to ectopic vein running parallel to L5). (D) net +/+ px. (E) net S /net + (arrows point to breaks in the ectopic veins) (F) net/net; ri/ri (arrow indicates the site at which L2 fuses with L3 – this is a highly penetrant phenotype). (G) net/net; det/det (arrows point to breaks in the ectopic veins and a floating vein segment). (H) Vno +/+ Dl9P39.
rho expression in vein mutants reveals the developmental stage during which these genes function
Localized expression of rho in vein primordia (Fig. 5A,B) in combination with ubiquitous EGF- and R activity is required throughout the process of vein formation (Sturtevant et al., 1993; Noll et al., 1994; M. A. Sturtevant, K. Howard, and E. Bier, unpublished observations). The evidence that rho and Egf- and r are continuously required during vein development derives from combining a temperature sensitive Egf- and r allele (Egf- and rIF26) with the rhove mutation. At the non- and permissive tem- and perature there is a strong genetic interaction between rho and Egf- and r which leads to a nearly complete elimination of veins (M.A. Sturtevant, K. Howard, and E. Bier, unpublished data), similar to that observed using a null Egf- and r allele (Sturtevant et al., 1993). In a series of temperature upshift and downshift experiments we determined that the phenocritical period for this interaction spans the entire period of vein formation. Restricting rho expression to vein primordia is also important during all stages of vein formation as brief heat inductions of a rhoHS line supplied at any stage of vein development result in the production of ectopic veins (M.A. Sturtevant, K. Howard, and E. Bier, unpublished data). These data suggest that the pattern of rho expression is an ideal tool for visualiz- and ing developing veins. We therefore examined the pattern of rho expression throughout the course of wing vein development (see schematic in Fig. 1) in a variety of venation mutants to determine the stage when rho expression first deviates from the wild type pattern (Table 3). These experiments reveal a temporal order of gene activity during vein formation similar to that proposed originally by Waddington (Waddington, 1940) and more recently revised by García- and Bellido and coworkers (Díaz- and Benjumea and García- and Bellido, 1990a; García- and Bellido and de Celis, 1992). There are, however, several unanticipated results suggesting that genes with similar adult mutant pheno- and types may act at distinct developmental stages and that lateral inhibitory mechanisms may function much earlier than previ- and ously appreciated.
rho expression in mutants defective in different steps in the genetic hierarchy of vein formation. The pattern of rho expression in various mutants was examined during late third- and instar and prepupal stages, and in some case during early pupal stages by in situ hybridization with a digoxigenin- and labeled antisense RNA probe. (A) A wild- and type third- and instar imaginal wing disc. Vein primordia L1- and L5 are indicated. L1 is indicated, but is sometimes difficult to identify, and L0 and L6 are not resolved at this stage. (B) A wild- and type wing at approximately 30 hours AP. rho is expressed in a sharp pattern of longitudinal veins (1- and 3 cells wide) starting at approx. 18 hours AP when the wings first re- and establish contact following apolysis. Cross veins do not begin expressing rho until approx. 25 hours AP. (C) A kn/kn third- and instar disc. Double arrow indicates that the primordia for L3 and L4 are spaced closer together than in wild- and type discs. Vein primordia for L2 and L5 are indicated. (D) A ri/ri third- and instar disc. Arrow points to location of missing expression in L2. Vein primordia for L3- and L5 are indicated. (E) An ab/ab third- and instar disc. Arrow points to location of missing expression in L5. Vein primordia for L2- and L4 are indicated. (F) A vn1/vnM1 third- and instar disc. Reduced expression in L2 and L4 are indicated by bracketed numbers. Relatively unaffected vein primordia for L3 and L5 are indicated. (G) A net/net third- and instar disc. Note that ectopic rho expression is confined to discrete sectors bounded by vein primordia (L2- and L5 are indicated). (H) A net/net; rhove/rhove third- and instar disc. Location of L3 primordium is indicated. (I) A net/net; ri/ri third- and instar disc. Vein primordia L2- and L5 are indicated. (J) A NAx/+ third- and instar disc. The location of the L3 primordium is indicated. (K) An early Nts/Nts third- and instar disc (raised at 29°C during second through third larval instars). Vein primordia L2- and L5 are indicated (arrowheads point to missing sections of marginal staining). (L) A late Nts wing (approx. 20 hours AP) raised at 29°C from 0 hours AP through apolysis (e.g. 20 hours AP). Inset: a portion of L5 from a wild- and type wing (corresponding to the boxed region of Nts wing) at a comparable developmental stage. (M) A px/px wing (approx. 30 hours AP). Emerging ectopic vein segments are in various phases of development. The arrowhead points to isolated single ectopic rho- and expressing cells and the arrow points to a partially connected segment of ectopic vein. Dorsal- and ventral induction must be very rapid as ectopic vein rudiments are labeled on both the dorsal and ventral surfaces in all but a few rare cases. The dorsal and ventral components of these ectopic veins are strictly aligned. (N) A bs2/bs2 wing (approx. 30 hours AP). (O) A Vno/Vno wing (approx. 30 hours AP). Arrow points to wild- and type rho expression in the hinge region, indicating that the absence of expression elsewhere is not due to a poor staining reaction.
rho expression in mutants defective in different steps in the genetic hierarchy of vein formation. The pattern of rho expression in various mutants was examined during late third- and instar and prepupal stages, and in some case during early pupal stages by in situ hybridization with a digoxigenin- and labeled antisense RNA probe. (A) A wild- and type third- and instar imaginal wing disc. Vein primordia L1- and L5 are indicated. L1 is indicated, but is sometimes difficult to identify, and L0 and L6 are not resolved at this stage. (B) A wild- and type wing at approximately 30 hours AP. rho is expressed in a sharp pattern of longitudinal veins (1- and 3 cells wide) starting at approx. 18 hours AP when the wings first re- and establish contact following apolysis. Cross veins do not begin expressing rho until approx. 25 hours AP. (C) A kn/kn third- and instar disc. Double arrow indicates that the primordia for L3 and L4 are spaced closer together than in wild- and type discs. Vein primordia for L2 and L5 are indicated. (D) A ri/ri third- and instar disc. Arrow points to location of missing expression in L2. Vein primordia for L3- and L5 are indicated. (E) An ab/ab third- and instar disc. Arrow points to location of missing expression in L5. Vein primordia for L2- and L4 are indicated. (F) A vn1/vnM1 third- and instar disc. Reduced expression in L2 and L4 are indicated by bracketed numbers. Relatively unaffected vein primordia for L3 and L5 are indicated. (G) A net/net third- and instar disc. Note that ectopic rho expression is confined to discrete sectors bounded by vein primordia (L2- and L5 are indicated). (H) A net/net; rhove/rhove third- and instar disc. Location of L3 primordium is indicated. (I) A net/net; ri/ri third- and instar disc. Vein primordia L2- and L5 are indicated. (J) A NAx/+ third- and instar disc. The location of the L3 primordium is indicated. (K) An early Nts/Nts third- and instar disc (raised at 29°C during second through third larval instars). Vein primordia L2- and L5 are indicated (arrowheads point to missing sections of marginal staining). (L) A late Nts wing (approx. 20 hours AP) raised at 29°C from 0 hours AP through apolysis (e.g. 20 hours AP). Inset: a portion of L5 from a wild- and type wing (corresponding to the boxed region of Nts wing) at a comparable developmental stage. (M) A px/px wing (approx. 30 hours AP). Emerging ectopic vein segments are in various phases of development. The arrowhead points to isolated single ectopic rho- and expressing cells and the arrow points to a partially connected segment of ectopic vein. Dorsal- and ventral induction must be very rapid as ectopic vein rudiments are labeled on both the dorsal and ventral surfaces in all but a few rare cases. The dorsal and ventral components of these ectopic veins are strictly aligned. (N) A bs2/bs2 wing (approx. 30 hours AP). (O) A Vno/Vno wing (approx. 30 hours AP). Arrow points to wild- and type rho expression in the hinge region, indicating that the absence of expression elsewhere is not due to a poor staining reaction.
I. Establishment of positional values
Coordinate genes
The first group of genes to consider in vein patterning, which we refer to as the coordinate genes, function early during wing disc development to establish positional values. Coordinate genes include many of the segment polarity genes functioning during embryogenesis to establish positional values within each segment. These same genes then contribute to anterior- and posterior patterning during imaginal disc development. The earliest acting coordinate genes (e.g. engrailed) establish boundaries within imaginal discs during late embryogenesis, while others (e.g. wingless) function during the first or second larval instars (Struhl and Basler, 1993; Couso et al., 1993). Adult viable alleles of coordinate genes lead to an altered pattern or spacing of veins. Several coordinate mutants interact with rhoHS phenotypes (Table 1, Fig. 3E,F). As coordinate genes function early in disc development, initiation of rho expression during the third larval instar in a sharp pattern of stripes should reflect these alterations. This expectation was confirmed for each putative coordinate mutant examined. For example, veins L3 and L4 lie closer together in shifted (shf), fused (fu), and knot (kn; Fig. 2A) adult wings and the primordia for L3 and L4 in third- and instar wing discs of these mutants (visualized by rho expression) are shifted closer together (Fig. 5C) than in wild- and type discs (Fig. 5A). Also, the dominant segment polarity mutant ci57g and a viable recessive dppshv mutant have missing sections of L2 and L4, and rho expression is missing in L2 and L4 primordia in ci57g third- and instar discs and the L2 primordium is truncated in dppshv discs (data not shown).
II. Initiation of vein formation
Two mutually opposing sets of genes are likely to translate positional information generated by the coordinate genes into stripes of vein primordia, one group promoting the initiation of vein formation and the other group suppressing vein devel- and opment. Genes governing nervous system formation also play an early role in initiating vein formation and may contribute to the alignment of sensory structures with veins.
Vein- and promotion genes
In vein promotion mutants, veins fail to form at an early devel- and opmental stage (i.e. during the third larval instar). These loss of vein mutants may lack individual veins, as in radius incom- and pletus (ri), which lacks the majority of L2 (Fig. 2B), tilt (tt), which lacks a section of L3, and abrupt (ab) which lacks the distal portion of L5 (Fig. 2C), or may lack portions of several or all longitudinal veins such as vein (vn) (Fig. 2D,E), ventral veinless (vvl), and Hairless (H). The pattern of rho expression in third- and instar discs is consistent with the adult phenotypes of these mutants. For example, a single stripe of rho- and expressing cells corresponding to the L2 primordium is missing in ri (Fig. 5D), and L5 precursors fail to express rho in ab (Fig. 5E). tt also acts early as mutant discs exhibit a marked reduction of rho expression in the primordia for L2, L3, and L4. This reduction in rho expression is more general than the ultimate vein loss phenotype which is restricted to loss of a section of L3. Flies trans- and heterozygous for a strong viable combination of vn alleles (vn1/vnM1; Fig. 2E) or a combination of vvl alleles, lack sections of L2 and L4. rho expression in vn1/vnM1 discs is strongly reduced in vein primordia for L2 and L4 (Fig. 5F) and in vvl discs rho expression is specifically missing in cells giving rise to the ventral component of L4 (data not shown). Finally, a strong combination of H alleles that eliminates all longitudinal veins is associated with a virtual absence of rho expression in all longitudinal vein primordia except L3 in third- and instar discs (Table 3).
Vein- and suppression genes
Mutations in vein- and suppression genes such as net and plexus (px) produce a network of connected ectopic veins running between and parallel to longitudinal veins in intervein regions (see Fig. 2F). A notable feature of these ectopic anastamosing veins is that they are confined to particular intervein territories (e.g. extra veins do not form in the sector between veins L3 and L4). Ectopic rho expression is observed in net and px third- and instar wing discs, but is restricted to regions of the disc giving rise to ectopic veins (Fig. 5G). The domains of ectopic rho expression in these mutant discs alternate with regions devoid of rho expression. Interestingly, the boundaries between rho- and expressing and non- and expressing sectors coincide with the locations of normal longitudinal vein primordia. Ectopic rho expression in net or px mutants subsides during prepupal devel- and opment suggesting that other genes limit vein formation during this period.
Notch functions early to limit initiation of vein formation
Neurogenic genes such as Notch function at many stages of development to limit the number of various differentiating cell types. As Notch has been implicated in restricting the breadth of veins during later stages of wing development (see below) and is required for development of the margin (see Fig. 2J), we tested for a potential early role of Notch in initiating vein formation. The first indication that Notch does indeed play an early role in establishing the vein pattern is that expression of rho in discs isolated from a gain of function NAx mutant is dra- and matically reduced in all longitudinal vein primordia except L3 (Fig 5J), paralleling the adult NAx loss- and of- and vein phenotype (Fig. 2I).
We also examined the pattern of rho expression in wings derived from Nts individuals that were maintained at the per- and missive temperature (18°C) throughout embryogenesis and early larval development and then shifted to the non- and permis- and sive temperature (29°C) at different times during the second larval instar through early pupal stages. Examination of wings recovered from various temperature shift experiments (data not shown) confirmed the conclusions of previous studies, which distinguished two separate periods important for wing formation (Shellenbarger and Mohler, 1978). Early shifts to 29°C, starting in the second larval instar and lasting until the beginning of pupariation, lead to extreme notching of the margin and to the production of long paddle shaped wings. In extreme cases loss of anterior structures includes L2 and loss of posterior regions deletes L5. Remaining longitudinal veins, however, are of normal thickness (Fig. 2J). In contrast, late shifts to 29°C (0- and 50 hours AP) do not cause notching, but result in markedly thickened veins (Fig. 2K – see below). The pattern of rho expression in Nts third- and instar discs raised at 29°C beginning early in the second larval instar is shown in Fig. 5K. As expected from the final extreme wing margin defects resulting from this treatment (Fig. 2J), there are large gaps in rho expression along the presumptive margin (arrowheads in Fig. 5K). A striking and unexpected feature of rho expression at this stage, however, is that the longitudinal domains of rho- and expressing cells are greatly broadened. As this loss of function phenotype is opposite to that observed for the gain of function NAx phenotype described above, Notch may serve a lateral inhibitory role during this early period to restrict the number of cells initiating vein development.
Epistatis of vein- and promotion genes over vein- and suppression genes
Several key observations regarding the epistatic relationship of vein- and promotion genes over vein- and suppression genes have been made by García- and Bellido and co- and workers. With respect to com- and binations of rhove with net and px, two features of the double mutants are informative. First, net; rhove (or px; rhove) double homozygous mutant flies have nearly wild- and type wings in which the extra vein phenotype of net (or px) is completely sup- and pressed and the loss- and of- and vein phenotype of rhove is partially sup- and pressed (Fig. 3I; Díaz- and Benjumea and García- and Bellido, 1990a). The complete suppression of ectopic veins by rhove suggests that rho is required to mediate the effect of net and px, but the partial reverse suppression of rhove by these mutants makes this conclusion tenuous. We investigated this question by examining rho expression in double- and mutant discs. The pattern of rho expression in net; rhove (Fig. 5H) or px; rhove (data not shown) third- and instar discs is indistinguishable from that observed in the rhove single mutant (Sturtevant et al., 1993). This result demonstrates that rhove is completely epistatic over net and px mutants with respect to ectopic rho expression. Thus, it is likely that the partial suppression of the rhove loss- and of- and vein phenotype in net; rhove and px; rhove double- and mutants is due to the action of net and px on other gene(s) functioning in parallel to rho and not to a partial rescue of rho expression. The existence of genetic pathway(s) functioning in parallel to rho is also supported by the observation that elevated Egf- and r activity can suppress the rhove phenotype, while decreasing Egf- and r activity in combination with rhove leads to virtually complete vein loss (Sturtevant et al., 1993). A candidate parallel genetic element to rho is vn since there are dominant trans- and heterozygous interactions between rhove and vnM1 (Fig. 3G) that delete the same section of L4 missing in weak Egf- and r mutants. Further evidence that rho and vn act in concert is that rhovevn1 double mutants lack all veins in the wing blade (Díaz- and Benjumea and García- and Bellido, 1990a; see Fig. 3H in this man- and uscript) and the multiple combination of net dsr px; rhovevn1 mutants has the rhovevn1 complete loss- and of- and vein phenotype (Díaz- and Benjumea and García- and Bellido, 1990a). As vein- and promotion mutants are epistatic over vein- and suppression mutants, vein- and suppression genes most likely interfere with the ability of vein- and promotion genes to initiate vein formation in intervein regions.
Global versus specific promotion of vein formation
Loss- and of- and vein mutants lacking individual veins raise a funda- and mental issue regarding the nature of vein promotion. Do these mutants compromise distinct vein promoting activities restricted to local regions of the wing disc, or do they disrupt a global promotion of all veins with more critical requirements for individual ‘sensitive’ veins? Double mutant combinations of net with the putative vein- and specific mutants ri and ab shed some light on this issue. The extra- and vein phenotype of net is most obviously suppressed in the region surrounding L5 in net ab double mutants and is most clearly reduced in the neigh- and borhood of L2 in net; ri double mutants (Fig. 4F; Díaz- and Benjumea and García- and Bellido, 1990a). However, we also observed a more general suppression of the net extra- and vein phenotype in both of these double mutant combinations. This widespread suppression of the net phenotype is manifest in reduction of early ectopic rho expression in net; ri double mutant imaginal discs (compare Fig. 5I and 5G). Similarly, tt mutant flies only lack a section of L3, but rho expression is reduced more generally throughout the primordia of L2- and L4. The selective loss of portions of L2 and L4 in vn1/vnM1 wings (Fig. 2E) and corresponding expression of rho in third- and instar discs (Fig. 5F) is another example of deceptive specificity. A more global role for vn is revealed in rhovevn1 double mutants, which lack all veins in the wing blade including all of L3 (Fig 3H). This is a striking example of parallel function since L3 is left largely intact in either single mutant. A ubiquitous require- and ment for vn is consistent with the poor viability of clones of lethal vn alleles in all locations of the wing blade (García- and Bellido and de Celis, 1992). Similar observations of illusory vein specificity have also been made in the case of rho and Egf- and r mutants (Sturtevant et al., 1993). rhove flies lack only distal portions of veins although rho expression is virtually eliminated in rhove wing imaginal discs (Sturtevant et al., 1993) and Egf- and rtop/DfEgf- and r lack only a segment of L4. However, Egf- and rtop/DfEgf- and r; rhove double mutants lack nearly all longitudinal veins (Sturtevant et al., 1993). Thus, in each of these cases (ri, ab, tt, vn, rho and Egf- and r) the apparent speci- and ficity of these mutants for particular individual veins or com- and binations of veins seems to obscure more global activities of these genes in promoting vein formation. The fact that these different mutants have distinct threshold requirements in par- and ticular regions of the wing does argue, however, that there are regional differences modulating the effects of these various genes.
Genes required for the integrity of the wing margin
Mutants leading to scalloping or notching of the wing margin do not have an obvious role in the formation of longitudinal veins within the wing blade. None- and the- and less, the observation that several genes of this class (e.g. Ser, Fig. 2H) strongly suppress rhoHS extra- and vein phenotypes (Table 1, Fig. 3L) suggested that these genes might play a general role in promoting vein formation. The pattern of rho expression in three mutants of this category, Ser/+, Ly/+, and sd is essentially normal during larval and prepupal development except for missing sections along the margin (data not shown). Thus, while these genes are likely to participate in wing morpho- and genesis by controlling processes such as cell proliferation (Speicher et al., 1994), they do not appear to play essential roles in initiating vein formation per se. The fact that Ser interacts strongly with rho in genetic tests but is not required for regulating the normal pattern of rho expression during vein development serves as a reminder that not all genetic interac- and tions are necessarily mediated at the level of transcription.
III. Vein differentiation
Differentiation of vein cells includes at least four independent processes: a lateral inhibitory mechanism limiting the number of vein- and competent cells assuming vein fates, a signal promoting vein differentiation along the axis of vein elongation, an inductive signal produced by dorsal vein cells required for maintenance of ventral vein differentiation, and suppression of intervein characteristics such as inter- and surface adhesion. We briefly describe each of these developmental events and provide examples of genes likely to contribute to these processes.
Genes restricting the thickness of veins
A prominent class of late acting vein mutants is the thickened vein group (see Table 1) which includes Notch, Delta (Dl), and thick veins (tkv). These mutants are members of the rho inter- and acting group suggesting that they may mediate an important function of rho. The basis for the thick vein phenotype has been attributed to the failure of a lateral inhibitory mechanism that normally restricts vein formation to a subset of cells having the potential to form veins (Díaz- and Benjumea and García- and Bellido, 1990a; García- and Bellido and de Celis, 1992). As mentioned above, temperature- and shift experiments performed with a Nts allele reveal a late requirement for Notch during pupal stages in limiting vein thickness (Shellenbarger and Mohler, 1978; M.A. Sturtevant, unpublished data).
When Nts individuals were raised at 18°C until pupariation and then shifted to 29°C during pupal development, the pattern of rho expression was broadened from rows 2- and 3 cells across (Fig. 5B) to strips 7- and 8 cells wide (Fig. 5L). This expansion of rho expression is consistent with the view that Notch con- and tributes to a lateral inhibitory mechanism restricting rho expression to the centers of broad ‘provein’ strips of cells competent to form veins. Expansion of the rho expression domains in Nts is first evident during prepupal stages, achieves its full extent by 25 hours AP (Fig. 5L), and then partially narrows later (30 hours AP) to reflect the final modest vein thickening phenotype (4- and 5 cells across). Notch functions together with Delta in restricting vein thickness as rho expression in trans- and heterozygous Dlvi/Dlts mutants is also broadened (data not shown). Consistent with previous obser- and vations that N /+; Dl /+ double mutants have more wild- and type wing patterns (Alton et al., 1989), N and Dl have opposite interactions with rho (Table 1) and other wing vein mutations (Table 2), supporting models in which the balance between these two genes rather than the absolute level of gene activity is critical for normal vein development.
To test whether the early role of Notch during the third larval instar (see above) would influence the response to late up- and shifts, we shifted second instar Nts larvae to 29°C and kept them continuously at the non- and permissive temperature through- and out early pupal stages (such individuals survive through early pupal stages but die before eclosing). The resulting phenotype assayed by rho expression at 30- and 35 hours AP or by examina- and tion of wings dissected out from pharate adults is essentially the superposition of the early (scalloped margin) and late (thick vein) Nts phenotypes (M.A.S. unpublished results). This simply additive phenotype supports the view that the early defects and the late thick vein phenotype result from two independent roles of Notch at distinct developmental stages (see discussion).
Vein extension
Mosaic analysis of extra- and vein mutants such as px suggests that mutant ectopic vein cells can recruit surrounding wild- and type cells to differentiate as veins to connect the mutant (vein) cells to the nearest longitudinal vein (García- and Bellido, 1977). This ability of differentiating vein cells to induce neighboring cells along the axis of vein elongation to assume vein fates can be observed during pupation in net and px wings. Ectopic rho expression in net or px mutants evident during the third larval instar (Fig. 5G), fades during prepupal stages and reappears in the pupa (approximately 25 hours AP) as isolated dots of rho- and expressing cells found near the middle of intervein territories (arrowhead in Fig. 5M). These dots then extend as narrow arcs of cells (arrow in Fig. 5M), which meander until they fuse with pre- and existing longitudinal veins to prefigure the final plexate vein phenotype (see Fig. 2F). detached (det, Fig. 2R) is likely to play a role in this vein extension process since net/net; det/det wings frequently have disconnected islands of ectopic veins running for short distances between and parallel to lon- and gitudinal veins (Fig. 4G). Vein extension is likely to require higher levels of vein promoting activity than those necessary to initiate ectopic rho expression as double mutant combina- and tions of net or px with vein loss mutants such as Star (e.g. net S/net +; Fig. 4E) or ab (e.g. net ab/net ab – data not shown) also have floating ectopic veins. Similar conclusions can be drawn from breeding experiments in which vein- and suppressing genetic backgrounds were selected (Thompson, 1974). Thus, while inhibitory interactions restrict the lateral dimension of veins, another signal(s) acting perpendicular to lateral inhibi- and tion promotes vein differentiation along the axis of vein elongation.
Dorsal- and ventral induction
Elegant use of mosaic analysis has revealed that the formation of ventral components of veins requires a signal(s) from the dorsal surface (García- and Bellido, 1977). Genes involved in the signaling between wing surfaces must, by necessity, act rela- and tively late in vein formation as the two surfaces only become apposed during prepupal stages (see diagram in Fig. 1). A highly conserved feature of venation across insect phylogeny is that alternating veins run predominantly along either the dorsal or ventral surfaces of the wing (see Fig. 1). Vno, which deletes sections of the L2 and L4 ventral veins (Fig. 2P), is an example of a mutation interfering with a late phase of vein differentiation. The pattern of rho expression is normal in third- and instar discs and prepupal wings of Vno/Vno homozy- and gotes, which lack all veins (Fig. 2Q), but vanishes abruptly during pupal stages (Fig. 5O). Thus, in Vno mutants vein formation appears to be initiated correctly but is disrupted at a later stage. The fact that Vno/+ heterozygotes specifically lack ventral veins (i.e. L2 and L4) and that vein segments at the edges of deleted veins often lack only the ventral component of the vein suggests that the Vno mutation may disrupt dorsal- and ventral induction since ventral veins would be expected to be most dependent on the dorsal- and to- and ventral signal.
Intervein differentiation
In parallel with the various genetic programs directing vein differentiation in vein primordia there are active intervein programs guiding differentiation of intervein cells. blister (bs) is likely to promote intervein differentiation by suppressing vein formation, since many properties of vein differentiation are observed in intervein regions in bs mutants (Fristrom et al., 1994). A primary function of bs is to suppress the action of rho, as bs; rhove double mutants display only the rhove loss- and of- and vein phenotype (Fristrom et al., 1994). Although the extra- and vein phenotypes of weak to moderate bs alleles (Fig. 2O) strongly resemble those of net and px mutants, rho expression is normal in third- and instar discs of bs mutants (Fristrom et al., 1994). The extra- and vein phenotype in bs mutants only becomes apparent during pupariation when ectopic rho- and expressing cells can be observed in regions giving rise to extra veins (Fig. 5N). Stronger bs alleles, which impart vein character to much of the wing surface, lead to ectopic rho expression as early as prepupal stages (Fristrom et al., 1994) but not during the third- and instar. This indicates that bs differs from net and px as it does not act to restrict initiation of vein formation, but rather sup- and presses vein formation later in differentiating intervein regions.
DISCUSSION
A sequential genetic model of wing vein formation
Several independent experimental methods have contributed to the wing vein development model presented in Fig. 6. A series of experiments using temperature sensitive alleles of Egf- and r (Egf- and rIF26) alone or in conjunction with rhove (M.A. Sturtevant, K. Howard, E. Bier, unpublished data) or Notch (Nts) (Shel- and lenbarger et al., 1978; M.A. Sturtevant, unpublished data), as well as staged heat inductions of rhoHS lines (M.A. Sturtevant, K. Howard, E. Bier, unpublished data) have identified a 35- and hour time period, beginning in the third larval instar and extending into early pupal stages, during which the vein versus intervein cell fate choice is decided. Mosaic analysis has also provided temporal information for the requirement of genes during wing vein development (García Bellido, 1977).
Model of the genetic hierarchy of vein formation. I. Establishment of Positional Values:Coordinate genes functioning to partition the segment (e.g. segment polarity genes and dpp) establish positional values along the anterior posterior axis of the wing. Other genes (e.g. apterous, aristaless ) determine dorsal- and ventral and proximal distal identities (Blair, 1993; Díaz- and Benjumea and Cohen, 1993; Campbell et al., 1993). We propose that these genes subdivide the disc primordium into a series of discrete sectors, the boundaries of which define locations of vein formation. Mutations in these genes shift or delete veins, or alter wing symmetry.
II. Initiation of Vein Formation: Positional information provided by coordinate genes is interpreted by vein- and promotion genes (e.g. vn, ri, ab, and tt) and the antagonistic vein- and suppression genes (e.g. net and px) to initiate vein formation (as visualized by early rho expression) at the correct locations. Genes directing nervous system development (N, h, emc, H, da, AS- and C) also provide an analogous function in vein formation.
III. Vein Differentiation:rho in combination with a parallel genetic pathway contributes to the activation of Egf- and r signaling (see Sturtevant et al., 1993) orchestrating the various aspects of wing vein differentiation. Key differentiation events include: lateral inhibition (an inhibitory process, active in broad regions with the potential to form veins, limits the lateral extent of veins – genes such as N, Dl, and possibly tkv contribute to this process); a vein extension function (a process by which vein segments once initiated tend to extend continuously along the axis of vein formation – det may participate in this function); dorsal- ventral induction (a signal provided by dorsal vein cells, perhaps involving the Vno gene, maintains the tendency of ventral vein cells to differentiate as such; Egf- and r does not seem to be required for this aspect of vein differentiation); and suppression of intervein differentiation such as adhesion between the two wing surfaces (mediated in part by integrins). Ultimately, densely packed vein cells secrete a thick cuticle and survive after adult eclosion, providing rigid open channels for fluid circulation. In contrast, intervein cells form strong inter- and surface bonds, flatten dramatically, and then die upon eclosion leaving a thin light cuticle behind (Fristrom et al., 1993).
Model of the genetic hierarchy of vein formation. I. Establishment of Positional Values:Coordinate genes functioning to partition the segment (e.g. segment polarity genes and dpp) establish positional values along the anterior posterior axis of the wing. Other genes (e.g. apterous, aristaless ) determine dorsal- and ventral and proximal distal identities (Blair, 1993; Díaz- and Benjumea and Cohen, 1993; Campbell et al., 1993). We propose that these genes subdivide the disc primordium into a series of discrete sectors, the boundaries of which define locations of vein formation. Mutations in these genes shift or delete veins, or alter wing symmetry.
II. Initiation of Vein Formation: Positional information provided by coordinate genes is interpreted by vein- and promotion genes (e.g. vn, ri, ab, and tt) and the antagonistic vein- and suppression genes (e.g. net and px) to initiate vein formation (as visualized by early rho expression) at the correct locations. Genes directing nervous system development (N, h, emc, H, da, AS- and C) also provide an analogous function in vein formation.
III. Vein Differentiation:rho in combination with a parallel genetic pathway contributes to the activation of Egf- and r signaling (see Sturtevant et al., 1993) orchestrating the various aspects of wing vein differentiation. Key differentiation events include: lateral inhibition (an inhibitory process, active in broad regions with the potential to form veins, limits the lateral extent of veins – genes such as N, Dl, and possibly tkv contribute to this process); a vein extension function (a process by which vein segments once initiated tend to extend continuously along the axis of vein formation – det may participate in this function); dorsal- ventral induction (a signal provided by dorsal vein cells, perhaps involving the Vno gene, maintains the tendency of ventral vein cells to differentiate as such; Egf- and r does not seem to be required for this aspect of vein differentiation); and suppression of intervein differentiation such as adhesion between the two wing surfaces (mediated in part by integrins). Ultimately, densely packed vein cells secrete a thick cuticle and survive after adult eclosion, providing rigid open channels for fluid circulation. In contrast, intervein cells form strong inter- and surface bonds, flatten dramatically, and then die upon eclosion leaving a thin light cuticle behind (Fristrom et al., 1993).
In this paper we examined directly the pattern of rho expression in various mutants to determine the earliest stage at which defects become apparent in developing veins. A strength of using rho expression as a marker for vein formation is that not only is rho expression in vein primordia required through- and out vein development, but restricted expression of rho is also necessary for achieving the normal vein pattern (M.A. Sturte- and vant, K. Howard, E. Bier, unpublished data). Consistent with continued requirement for localized rho expression during vein development, rho interacts genetically with genes functioning at all developmental stages (e.g. dpp, kn; → vn, net; → tkv, bs, and Vno). Thus, defects in the pattern of rho expression should translate into final wing phenotypes. This analysis has identi- and fied genes acting during the third larval instar based on mutant defects in the initiation of rho expression as well as genes acting later to mediate vein differentiation during prepupal and pupal stages. Some of the late genes may be mis- and classified as it is possible that certain aspects of early vein initiation might be disrupted without affecting rho expression. To address this possibility we have examined the pattern of Dl expression in several putative late mutants. Dl is expressed in provein regions early during the third larval instar and then becomes sharply restricted to veins during pupal stages (M. A. S. and E. B., unpublished data). These experiments reveal a similar temporal requirement for Dl and rho expression in mutant developing wings. Another caveat to this type of analysis is that we have used viable hypomorphic alleles of many genes. It is possible in some instances that stronger alleles would disrupt the process in question more profoundly, leading to the onset of observable defects at earlier developmental stages. bs is an example of this, since moderate strength alleles only show disruption of rho expression during pupal stages, while stronger alleles manifest defects during prepupal stages. With these qualifications in mind, however, the temporal data obtained from the use of temperature sensitive alleles, from mosaic analysis, and from examination of rho expression in developing mutant wings are remarkably self consistent. There is also good reason to believe that the use of hypomorphic alleles does not generally lead to grossly erroneous conclu- and sions. For example, the initial pattern of rho expression is disrupted as expected in each of the putative coordinate mutants examined (e.g. kn, fu, shf, ci57g, dppshv) even though these viable alleles are much weaker than the strong embryonic lethal alleles that have been isolated for most of these genes. We have also examined rho expression in a series of progres- and sively stronger allelic combinations of vn and px mutations (see Table 3). While the degree of rho mis- and expression depends on the strength of the allele examined, the developmental onset of abnormal rho expression occurs at the same stage for weak and strong alleles alike. Even in the case of bs, it should be noted that the earliest prepupal stage when ectopic rho expression can be observed in an extreme bs mutant is still several hours after ectopic rho expression has reached full intensity in net or px mutants, while net and px mutants have final vein pheno- and types equivalent to only weak or moderate bs alleles. These data indicate that examining rho expression in various mutants provides a good estimate of the developmental stage at which different genes function during vein development.
Subdivision of the wing primordium into discrete sectors
A temporal outline of developmental events and gene action during wing vein development is presented in Fig. 6. Early sub- and division of the wing pouch into longitudinal sectors is likely to be the end product of the action of segment polarity genes (e.g. en, ci, wg, hh, ptc, fu, shf, and kn) and other coordinate genes such as dpp. Consistent with these genes acting prior to the onset of vein formation, rho expression is not initiated normally in mutants of this class we have tested (fu, shf, kn, ci57g, dppshv), although other mutants included in this category must be directly examined before generalizing this finding to the group as a whole. The intense interaction of rho with kn but not with fu or shf, which have very similar early and late phenotypes, is noteworthy and may indicate a more intimate role for kn in initiating rho expression in L3 and L4. The formation of ectopic veins in hh/+; rhoHS (or dppshv/+; rhoHS) flies in the anterior compartment (Fig. 3F), which is a signifi- and cant distance from hh- and expressing cells confined to the posterior compartment (Lee et al., 1992; Tabata et al., 1992), is consis- and tent with the proposed roles of the hh and dpp products as secreted factors involved directly or indirectly in long range patterning (Heberlein et al., 1993; Ma et al., 1993; Ingham, 1993; Tabata and Kornberg, 1994 see also Smith, 1994 for review of vertebrate hedgehog homologues).
Based on results described in this study and on additional data indicating the presence of sharp boundaries coinciding with vein primordia (González- and Gaitán et al., 1994; M. A. Sturtevant and E. Bier, unpublished data), we propose that the coordinate genes subdivide the wing blade primordium into a series of discrete sectors and that vein formation is initiated at these boundaries. Consistent with this view, rho expression is directly initiated in a sharp pattern of stripes without an inter- and mediate stage of less localized expression. The clearest evidence that the developing disc is subdivided into a series of alternating sectors bounded by veins is provided by the pattern of rho expression in net or px mutant third- and instar discs (e.g. Fig. 5G). Further evidence that veins define the edges of discrete imaginal territories in the third- and instar disc is that stripes of rho- and expressing cells coincide with the boundaries of various gene expression domains (M.A. Sturtevant and E. Bier, unpublished data). Veins also serve as late clonal restrictions (Díaz- and Benjumea et al., 1989; Díaz- and Benjumea and García- and Bellido, 1990a; González- and Gaitán et al., 1994), suggesting these putative boundaries may be defined by the apposition of cells with distinct adhesive properties. Recent analysis of wing margin morphogenesis has revealed that signals generated at the interface between dorsal and ventral compartment cells induce cells along the wing margin to differentiate (Williams et al., 1994; Díaz- and Benjumea and Cohen, 1993). Thus, wing margin formation is an example of induction at the boundary between two lineage compartments. The fact that veins wrap around the edges of ptc clones (Phillips et al., 1990) could be explained by a similar mechanism in which differences in cell properties such as adhesion induce cells at clonal boundaries to differentiate as longitudinal veins.
Initiation of the vein pattern
Early acting vein- and promotion genes fall into two basic cate- and gories: those required for the formation of individual veins (e.g. ri and ab) and those required for several or all longitudi- and nal veins (e.g. rho, vn, H). It may be misleading, however, to make qualitative distinctions between these two classes of vein loss mutants as all of these genes may function more globally than is apparent from the single mutant phenotypes. These genes are likely to convert positional information into a com- and mitment to initiate vein differentiation. Vein promotion genes are not likely to be required for establishing positional values per se since the pattern of remaining veins is normal in these mutants. Furthermore, sensory organs normally associated with vein L3 are in the correct location in compound loss- and of- and vein mutant combinations (e.g. ve vn) that eliminate all longi- and tudinal veins. Thus, these genes act upstream of rho and most likely function downstream of the coordinate genes.
Vein suppression genes (e.g. net, px, h, emc) presumably function to limit vein formation to sector boundaries by inter- and fering with vein- and promotion in intervein regions. The epistasis of vein- and promotion over vein- and suppression is consistent with this view. The pattern of rho expression in double mutants of net or px with rhove is the same as in rhove. The lesion in rhove appears to be a deletion of only 600- and 800 bp of the rho wing vein enhancer (M. Roark, unpublished data). Thus, net and px are likely to impinge on cis- and acting response elements of the rho promoter that are very close to or interspersed with sites for activator binding. In addition to suppressing vein formation in intervein regions, it is possible that vein- and suppression genes also actively promote intervein differentiation, as is thought to be the case for the later acting bs gene (Fristrom et al., 1994; see below).
Several genes involved in specifying neuronal precursor cells also play a parallel role in vein development. Loss- and of- and function mutants in genes required for promoting neuronal precursor formation may lack veins (e.g. H), whereas gain of function alleles of these genes (e.g. AS- and CHw) and loss- and of- and function alleles of genes that suppress neuronal precursor formation (e.g. h, emc) produce ectopic veins. Additionally, the appearance of ectopic bristles in h1 mutants depends on rho function (García- and Bellido, 1977). We have observed that H is required for early expression of rho. Mosaic analysis suggested that both h and emc are also required prior to pupation (García- and Bellido and Merriam, 1971), although these genes may actually function at a somewhat later stage since ectopic AS- and C expression and the appearance of ectopic sensory organs are not detectable until after pupariation (Skeath and Carroll, 1991; Blair et al., 1992). The involvement of genes regulating neuronal precursor specification in vein formation may con- and tribute to the ultimate alignment of sensory organs along the marginal vein and L3. The placement of sensory organs along veins is not surprising since veins provide the only channels of living cells in the mature wing. Neurogenesis and vein formation are not strictly coupled, however, as veins form normally in sc10- and 1 flies lacking all L3 sensilla, while recipro- and cally, sensilla often form normally in ve vn1 flies lacking all longitudinal veins. Thus, the formation of vein and sensory- and organ precursors are likely to be independently initiated based on shared primary positional information (e.g. sector bound- and aries), and subsequently cross regulatory interactions reinforce collinear alignment of these two cell types.
One unexpected result was that Notch is required for estab- and lishing the early sharp pattern of rho expression in third- and instar wing discs. Adult flies lacking Notch activity during the second and early third larval instars exhibit strong defects in formation of the wing margin and deletions of extreme anterior and posterior structures, but the pattern of remaining longitudinal veins is not significantly affected. Despite the relatively normal width of these veins, rho expression is dramatically expanded during the third- and instar. Consistent with Notch acting to restrict the extent of vein initiation, rho expression is dramatically reduced in gain of function NAx mutant discs. It is unclear how Notch regulates the pattern of early rho expression. One pos- and sibility is that Notch mediates a lateral inhibitory signal to restrict the breadth of vein formation in these regions. This simple model is most similar to the well established role of Notch in a wide variety of other developmental contexts including a lateral inhibitory function later in vein development (see below) and is consistent with the opposite loss- and of- and vein phenotype of dominant NAx mutants. The absence of a final thickened vein phenotype resulting from early loss of Notch, however, is difficult to explain in this model. Even when Notch activity is continuously eliminated between the second larval instar and early pupal development, the thickened- and vein phenotype is no stronger than that observed with the late loss of Notch alone. Additionally, the vein thickening phenotype is less extensive during prepupal stages (4- and 9 hours AP) than either early during the third larval instar or later in the pupa (25 hours AP), suggesting that the early and late effects of Notch might represent independent activities of this gene. An alternative explanation for the early ectopic rho phenotype is that Notch mediates some aspect of signaling required for the action of the coordinate genes and that the broad stripes of ectopic rho expression are a manifestation of the failure to subdivide the disc into the normal array of discrete sectors with sharply defined borders. A role for Notch in mediating some aspect of the wingless signal during embryonic segmentation (Couso and Martinez Arias, 1994) and larval wing margin formation has been proposed (Couso et al., 1994; Hing et al., 1994; Couso and Martinez Arias, 1994), but the mechanism underlying the interactions between Notch and wingless remains unresolved. It is also unclear whether the early role of Notch in restricting rho expression during the third instar is related to its role in development of the wing margin. Further analysis will be required to determine the basis and signifi- and cance of the early Notch vein phenotype.
Vein differentiation
The mechanism by which rho expression mediates vein formation is likely to involve hyperactivation of EGF- and R signaling (Sturtevant et al., 1993; M. A. Sturtevant, K. Howard, E. Bier, unpublished data). The best characterized aspect of vein differentiation is the lateral inhibitory mechanism restricting vein formation to a row 2- and 3 cells across from a 7- and 8 cell wide competent provein domain. This provein region includes cells most easily converted to the vein fate by ectopic rho expression (Sturtevant et al., 1993). Lateral inhi- and bition seems to be mediated in part by Notch and Delta. At 25 hours AP the pattern of rho expression in Nts wings raised con- and tinuously at 29°C (beginning in prepupal stages) includes the entire strip of provein cells having compact Nomarski mor- and phology (Sturtevant et al., 1993). At later developmental stages rho expression begins to recede from the full provein territory indicating that other lateral inhibitory mechanisms in addition to Notch restrict vein formation. Based on their thickened vein mutant phenotypes thick veins (tkv), thickened, and thick are additional candidates for lateral inhibitory genes.
A thickened vein phenotype might also result from defects in processes other than lateral inhibition. For example, a viable allele of tkv, which encodes a dpp receptor (Brummel et al., 1994; Nellen et al., 1994; Penton et al., 1994), interacts strongly with rhoHS alleles (Fig. 3O) and Notch (Díaz- and Benjumea and García- and Bellido, 1990a) consistent with this receptor mediating a parallel lateral inhibitory signal. However, the genetics of the tkv locus is not straight forward (Terracol and Lengyel; 1994). Some combinations of apparent loss of function tkv alleles generate loss of vein phenotypes, which are enhanced by reduction in the dosage of dpp. Other combinations of tkv alleles, however, yield thick veins and this latter phenotype is enhanced by increasing the level of dpp. Perhaps there are early and late functions of dpp that have opposite consequences on vein formation. Alternatively, some tkv alleles may be partial gain of function mutations. The fact that mosaic analysis shows that dpp mutant cells located on the dorsal surface of the wing can lead to loss of ventral veins (Posakony et al., 1991) may indicate a role in D- and V signaling (see below). The thick vein phenotype could then be an indirect result of hyperactive D- and V signaling.
In addition to lateral inhibitory interactions restricting vein thickness, there is a paradoxical tendency for developing veins to promote vein extension along the axis of vein elongation which is likely to assure vein continuity. The emergence of ectopic veins in pupal wings of net or px mutants illustrates the vein extension process. Ectopic rho expression is first observed as isolated islands which then extend branches to connect to a pre- and existing vein. The isolated intervein cells first expressing rho most likely recruit their neighbors since an isolated clone of px mutant cells can induce surrounding wild- and type cells to develop as veins, connecting the mutant patch to nearby lon- and gitudinal veins (García- and Bellido, 1977; Díaz- and Benjumea et al., 1989; García- and Bellido and de Celis, 1992). The det gene may participate in this vein extension function since det mutants have a detached posterior cross vein and net; det double mutants have disconnected ectopic vein segments (Fig. 4G).
Another requirement for vein formation is dorsal- and ventral induction (García- and Bellido, 1977; Díaz- and Benjumea et al., 1989; García- and Bellido and de Celis, 1992) which takes place following disc eversion when the dorsal and ventral surfaces first come into contact. Mosaic analysis of various vein- and promotion mutants has revealed the existence of a dorsally provided signal required for the differentiation of ventral com- and ponents of veins (García- and Bellido, 1977; García- and Bellido and de Celis, 1992). Reciprocal, albeit weaker, signals emanating from the ventral surface also reinforce the developmental com- and mitment of dorsal vein components (García- and Bellido, 1977; García- and Bellido and de Celis, 1992; J. Chacko and E. Bier, unpublished data). These inductive signals are likely to con- and tribute to refining the alignment of dorsal and ventral vein com- and ponents to ensure precise register of the independently specified dorsal and ventral vein primordia. Vno may disrupt dorsal- and ventral induction since rho expression in this mutant is normal until prepupal or pupal stages. Vno most strongly affects veins having the major component on the ventral surface (e.g. L2 and proximal L4), which might be expected to be most dependent on this trans- and surface induction. It is also possible that Vno affects some other process required for main- and tenance of the vein fate. Although the vvl mutation results in a vein phenotype similar to that of Vno, it appears to disrupt vein formation before disc eversion, which is prior to the onset of dorsal- and ventral signaling. Mosaic analysis has shown that rho is required for production of the dorsal signal (García- and Bellido, 1977), but that Egf- and r function is not (Díaz- and Benjumea and García- and Bellido, 1990b). Thus, although rho and Egf- and r activity are intimately linked in many developmental settings, rho may mediate some aspects of vein formation through additional pathways. Alternatively, the use of hypomorphic Egf- and r alleles in mosaic studies may conceal a role for Egf- and r in dorsal- and ventral induction.
Finally, genes governing intervein development (e.g. bs) must also be considered, as they define the alternative cell fate in the wing. For example, expression of genes required for inter- and surface adhesion is restricted to intervein cells (Fristrom et al., 1993), permitting the non- and adherent strips of vein cells to form open channels between the two surfaces. Adhesion between intervein surfaces depends on the activity of genes encoding integrins such as l(1)mys (β integrin) and if (α integrin). Various allelic combinations of integrin mutations lead to formation of blisters and ectopic veins (Wilcox et al., 1989; Brower and Jaffee, 1989; Zusman et al., 1990, 1993). The fact that these mutants, even when homozygous, do not enhance HS- and rho phenotypes (which also cause blistering) indicates that similarity in final phenotype does not necessar- and ily lead to synergistic genetic interactions. Another indication that adhesion is an important characteristic of intervein differ- and entiation is that integrin mutants interact strongly with bs, a key gene mediating intervein differentiation (Wilcox, 1990; Wessendorf, 1992; Fristrom et al., 1994). Adhesion molecule related products encoded by the ft (Mahoney et al., 1991) and ds (C.S. Goodman, personal communication) genes may play a role in maintaining the compact morphology of veins within each wing surface since these mutations suppress HS- and rho extra- and vein phenotypes when homozygous.
It is possible that some genes required for initiating the rho expression pattern also function during later developmental stages. rho and Egf- and r are themselves examples of genes func- and tioning throughout vein development. Genes such as net and px, which interact with late acting mutants (e.g. N, tkv, and bs), as well as with early acting genes (e.g. vn, ri, and ab), may also act at more than one stage of vein development.
How various aspects of vein versus intervein differentiation ultimately lead to the extremely different morphogenic fates of these two alternative cell types is currently unknown. It appears, however, that these two differentiation programs can be partially uncoupled since under certain circumstances (e.g. in small blistered regions) structures having properties inter- and mediate between veins and intervein can be observed.
CONCLUSIONS
Genes regulating vein development can be placed in a hierar- and chical model of pattern formation. Coordinate genes first subdivide the wing into a series of alternating sectors. We propose that wing veins are induced at boundaries between these sectors through the action of vein promoting genes and counteracting vein- and suppression genes. Epistasis analysis indicates that vein- and suppression genes act by blocking vein- and promotion in intervein regions. Localized expression of rho in developing vein primordia then mediates vein formation throughout development. Key events required for vein differ- and entiation include: lateral inhibition to restrict the breadth of veins within a wing surface; vein extension to promote vein elongation and assuring vein continuity; signaling between the dorsal and ventral surfaces to maintain and perhaps refine the register of veins on the two surfaces; and restriction of inter- and surface adhesion to intervein regions.
Important questions to be resolved in the future include: (1) what are the mechanisms by which the wing is subdivided into discrete sectors by the coordinate genes; (2) how do vein- and promotion and vein- and suppression genes transform this posi- and tional information into the pattern of early rho expression; and (3) how do rho and Egf- and r direct distinct steps in vein morpho- and genesis and differentiation?
ACKNOWLEDGEMENTS
We thank Jason O’Neill for synthesizing RNA probes used for whole- and mount in situ hybridization; Dan Lindsley, Antonio García- and Bellido, Sean Carroll, Michael Levine, Margaret Roark, Reviewer 1, and Kathryn S. Burton for helpful discussions and critical comments on the manuscript; Nickolina Cataulina and her colleagues for photo- and graphic reproductions; and Kathryn S. Burton for preparing the figures. This work was supported by NIH Grant no. RO1- and NS29870- and 01, NSF Grant no. IBN- and 9318242, Research Grant no. 5- and FY92- and 1175 from the March of Dimes Birth Defects Foundation, and an ACS Insti- and tutional Award. E. B. was supported by funds from the McKnight Neuroscience Foundation, Sloan Foundation, and an ACS Junior Faculty Award.