Activation of the knirps locus links patterning to morphogenesis of the second wing vein in Drosophila

One of the important remaining questions in developmental biology is to understand how positional information is linked to the genesis of specific morphological structures in different locations of the organism. TheDrosophila wing is a system that is particularly well suited for studying the relationship between patterning and morphogenesis. For example,mechanisms for establishing positional information along the anterior-posterior (AP) axis have been analyzed in detail (reviewed byLawrence and Struhl, 1996;Strigini and Cohen, 1999;Klein, 2001) and morphological structures such as the five major longitudinal veins (L1-L5) form at invariant positions along the AP axis of the adult wing (reviewed byGarcía-Bellido and de Celis,1992; Bier, 2000). As described below, the five wing veins can be distinguished from one another by several criteria including: expression of specific combinations of genes during development; whether they are located primarily on the dorsal or ventral surfaces of the wing (odd versus even numbered veins respectively);and by the types or absence of sensory organs that form along them (e.g. sensory organs form on the L1 and L3 veins, but not on the L2, L4 or L5 veins). An interesting question then, is how vein-specific as well as more general vein developmental programs are initiated at various locations in response to different positional information.

Wing vein development in Drosophila can be subdivided into two temporally distinct stages: initiation and differentiation. Vein initiation begins during the mid-third larval instar(Waddington, 1940;García-Bellido, 1977;García-Bellido and de Celis,1992; Sturtevant et al.,1993; Sturtevant and Bier,1995; Bier, 2000)when the wing imaginal disc is a monolayer of cells. The wing blade proper derives from an oval region of the wing disc known as the wing pouch, while the remainder of the disc generates elements of the wing hinge and thoracic body wall. Vein differentiation, the second phase of vein development, occurs during metamorphosis as the wing disc buds out (or everts), folding along the future wing margin. Ultimately, this eversion leads to the apposition of the dorsal and ventral surfaces of the wing during pupal development, creating the bilayer of cells comprising the mature wing blade.

Genes involved in initiating wing vein development (vein genes) are expressed during the third larval instar in narrow stripes, corresponding to vein primordia, or in broader `provein' stripes, consisting of cells that are competent to become vein cells (Biehs et al., 1998). Some vein genes are expressed in all vein primordia,while others are expressed in subsets of veins or in single veins. For examplerhomboid (rho), which encodes an integral membrane protein(Bier et al., 1990;Sturtevant et al., 1996), is expressed in all vein primordia and promotes vein formation throughout wing development by locally activating the Egfr signaling pathway(Sturtevant et al., 1993;Noll et al., 1994;Bier, 1998a;Bier, 1998b;Guichard et al., 1999).caupolican (caup) and the neighboring gene araucan(ara) encode related homeobox genes that promote expression of other vein genes such as rho in odd numbered veins. Proneural genes such asachaete (ac) and scute (sc) promote neural development in the L1 and L3 primordium(Gomez-Skarmata et al., 1996).Delta (Dl) encodes a ligand for the Notch (N) receptor,which mediates lateral inhibitory interactions among cells in vein-competent domains during pupal development(Shellenbarger and Mohler,1978; Kooh et al.,1993; Parody and Muskavitch,1993). Delta is also expressed earlier during larval stages of wing development in all longitudinal veins except L2, and is likely to play a role in limiting vein thickness at this stage as well, since loss of Notch function during larval development leads to greatly broadened expression of the vein marker rho(Sturtevant and Bier, 1995).kni and knrl, which encode related zinc finger transcription factors in the steroid hormone superfamily(Nauber et al., 1988;Oro et al., 1988;Rothe et al., 1989), are expressed in a single stripe corresponding to the L2 primordium beginning in the mid-third larval instar, and are required to initiate L2 development(Lunde et al., 1998).kni also functions as a gap gene in early embryonic development(Nüsslein-Volhard and Wieschaus,1980; Nauber et al.,1988).

The L2 stripe of kni/knrl-expressing cells forms along the anterior border of a broad domain of cells expressing high levels of the related and functionally overlapping spalt-major (salm)(Kühnlein et al., 1994)and spalt-related (salr)(Barrio et al., 1996) zinc finger transcription factors. A variety of evidence indicates that central domain cells expressing the patterning genes salm and salr(together referred to as sal) induce their anterior neighbors, which express very low levels of sal, to become the L2 primordium(Sturtevant et al., 1997;Lunde et al., 1998). For example, in wings containing salm-mutant clones, ectopic branches of L2 are induced that track along and inside the salm^- clone borders, mimicking the normal situation in which an L2 vein forms just outside the domain of high-level sal-expressing cells(Sturtevant et al., 1997). This ability of sal-expressing cells to induce their anterior low-level sal-expressing neighbors to initiate L2 development requires the activity of the kni locus(Lunde et al., 1998). The induction of kni/knrl expression in the L2 primordium therefore provides an excellent system for studying the transition from spatial patterning to tissue morphogenesis.

Once activated along the anterior sal border, kni andknrl organize development of the L2 vein, in part by activating expression of the key vein-promoting gene rho and by suppressing expression of the intervein gene blistered (bs)(Montagne et al., 1996;Lunde et al., 1998). Thekni locus is highly selective in regulating downstream gene expression in the L2 primordium as revealed by the observation that several genes expressed in other veins, such as caup and ara, ac andsc, and Delta, are excluded from the L2 primordium. Thus,the kni locus links patterning to vein-specific morphogenesis by functioning downstream of sal and upstream of genes involved in vein versus intervein development.

The role of the kni locus in L2 formation has been clarified by analysis of likely regulatory alleles of the kni locus, previously known as radius incompletus (ri)(Arajärvi and Hannah-Alava,1969; Lunde et al.,1998). Flies with this mutation lack large sections of the L2 vein. kni^ri mutants are homozygous viable, in contrast tokni null mutants, which die as embryos with a gap gene phenotype(Nüsslein-Volhard and Wieschaus,1980; Nauber et al.,1988). In support of kni^ri mutations being wing-specific regulatory alleles of the kni/knrl locus, expression ofkni and knrl in the L2 primordium is absent inkni^ri[1] mutants, and the L2 vein-loss phenotype can be partially rescued by ubiquitous expression of kni in the wing(Lunde et al., 1998).

The findings summarized above suggest that kni and knrlorganize the L2 vein developmental program by orchestrating expression of genes that execute distinct subsets of functions required for proper L2 development. Several important unanswered questions remain, including: what function(s) are disrupted in kni^ri[1] and other existingkni^ri mutants?; do these mutations eliminate the function of an L2-specific cis-regulatory element in the kni locus?;and finally, with respect to the definition of L2 versus other veins, is it important to exclude expression of genes expressed in veins other than the L2 primordium?

In this study we report the identification of an enhancer element upstream of the kni coding region that selectively directs gene expression in the L2 primordium in third instar larval wing discs. We show that three separate ri alleles have defects mapping within a minimal 1.4 kb L2 enhancer element. We demonstrate that two of these mutations eliminate activity of the L2 enhancer, kni^ri[1], which contains a 252 bp deletion, and kni^ri[53j], which harbors a single base-pair substitution. We find that truncation of the minimal L2 enhancer to a 0.69 kb fragment leads to ectopic reporter gene expression in the extreme anterior and posterior regions of the wing, indicating that repression contributes to restricting activation of the L2 enhancer. In addition, we show that the general wing promoting transcription factor Scalloped (Sd) binds with high affinity to several sites in the L2 enhancer and that sd is required for kni expression in the wing disc. We have also employed the L2 enhancer element as a tool to drive expression of various UAS transgenes in the L2 primordium. We find that the loss of the L2 vein inri mutants can be rescued by L2-specific expression of either thekni or knrl genes, or the downstream target generho. In addition, we find that misexpression of genes in the L2 primordium that are normally expressed in veins other than L2 results in abnormal L2 development. These results provide a framework for understanding how positional information is converted into morphogenesis of the L2 wing vein by `vein organizing genes' such as kni and knrl.

The weak kni^ri[M3] allele isolated by J. Díaz-Benjumea was kindly provided by Antonio Gárcia-Bellido(Universidad Autonoma de Madrid); Ruth Lehmann (Skirball Institute, New York)provided the strong viable Df(3L)kni^ri[XT2] allele and the Df(3L)kni^FC82 stock(Lehmann, 1985); Peter Portin(University of Turku) provided the kni^ri[92f] allele; Yuh Nung Jan provided the UAS-scute stock(Chien et al., 1996); Juan Modolell (Universidad Autonoma De Madrid) provided the UAS-ara and UAS-caup stocks; Amanda Simcox (Ohio State University) provided the UAS-vein stock (Simcox et al.,1996); Joachim Urban (Universitat Mainz) provided the UAS-eg stock (Dittrich et al.,1997); Ken Irvine (Waksman Institute) provided the w sd⁵⁸ P{ry⁺, hs-neo, FRT} 18A stock, and Seth Blair (University of Wisconsin-Madison) provided the hsflp3MKRS/TM6B stock. kni^ri[M*] andkni^ri[53j] were obtained from the Bowling Green Stock Center, where they have since been abandoned. kni^ri[53j]is a medium-strength homozygous viable allele isolated by Irwin Oster, a student of H. J. Müller, and kni^ri[M*] is a medium-strength homozygous ri allele from the collection of H. J. Müller (Müller,1941). Other mutant lines, including the medium-strengthkni^ri[1] allele, balancers and chromosomal markers were obtained from either the Bloomington Indiana Stock Center or the Tübingen Stock Center and are described by Lindsley and Grell(Lindsley and Grell, 1968) and Lindsley and Zimm (Lindsley and Zimm,1992).

Genomic fragments spanning the majority of the putative regulatory region of the kni locus (Fig. 1) delimited by Df(3L)kni^ri[XT2](Nauber et al., 1988) were subcloned into the pC4PLZ expression vector(Wharton and Crews, 1993).EcoRI fragments F, E, S, R and Q (isolated from phages provided by U. Nauber) were inserted into pBluescript II KS (+/-)(pBS, Stratagene) with the NotI and T7 promoter containing the side of the vector most proximal to the knirps coding region. The constructs were subcloned into the lacZ transformation vectorpC4PLZ. Subfragments of the 4.8 kb EcoRI fragment (fragment E in Fig. 1) that drive expression in the L2 primordium were then subcloned into pC4PLZ to define a minimal L2 enhancer element. To assay fragments of E, pBS-E was digested; the proximal most EcoRI-SalI (ES, 2.0 kb),EcoRI-XhoI (EX, 1.4 kb), and EcoRI-HincII(EC, 0.69 kb) fragments isolated and re-inserted into pBS. The 4.8 kb fragment E was also subcloned into the pC4PG4 expression vector, in which the lacZ gene in pC4PLZ has been replaced byGAL4 (Emery, 1996),and was used to drive expression of various UAS transgenes in the L2 primordium.

Restriction fragments isolated from a lambda phage walk(Nauber et al., 1988) covering over 50 kb of the kni upstream region, delimited by theDf(3L)kni^ri[XT2] deletion, were used as probes to determine the locations of potential chromosomal abnormalities such as deletions or rearrangements in ri mutants on Southern blots(Southern, 1975). Several different restriction enzymes were used to scan each region to distinguish single nucleotide polymorphisms from larger scale abnormalities such as those present in the kni^ri[1] and kni^ri[M3]mutants.

Primers AA and RK were used to amplify a region surrounding the proximalEcoRI-BglII (1.197 kb) fragment of E fromkni^ri[1], kni^ri[92f],kni^ri[M*], and kni^ri[53j]genomic DNA (AA=GACACAATGCTCCGAATTCC, the 5′ end is in F, the 3′end contains the EcoRI site; RK=CCCAATGGACCCCAATCTGGTTGGGG, the 5′ end is at 1.231 kb in E. Note that TGGGG was added to produce aKpnI site at the distal end of the resulting fragment). PCR fragments were purified for sequencing using the QIAquick PCR Purification kit. The blunt ended kni^ri[1] fragment was cloned intopCR-BluntII-TOPO (TOPO) and oriented so that the NotI site within TOPO and the BglII site within the fragment (N, B)are on opposite sides. The resulting TOPO-N,B construct andpBS-E were digested with NotI and BglII, fragments isolated and ligated to form pBS-EΔri[1]. TheEΔri[1] insert was sequenced to confirm that it differed from the wild-type sequence by only the 252 bp deletion. KpnI andNotI were used to remove EΔri[1], and it was cloned into the corresponding sites of C4PLZ to formEΔri[1]-lacZ. Similarly, primers ECO-ri and XBA-ri were used to PCR amplify the EcoRI-XhoI fragment of E (termed EX) from kni^ri[53j] genomic DNA(ECO-ri=GAATTCCAACGCGAAGCGTC; XBA-ri=TCTAGATGGGGCTGCTGCCA). The resulting 1.4 kb product was cloned into the TOPO vector. The EX(ri^53j) fragment was digested with EcoRI andXbaI and subcloned into the corresponding sites of pC4PLZ to form EX(ri^53j)-lacZ. A single subclone was sequenced to verify that only the single nucleotide change (C596A) had been incorporated.

Generation of reduction-of-function sd⁵⁸ clones and immunohistochemical analysis were performed as described previously(Guss et al., 2001).

Identification and analysis of Sd binding sites was performed as described by Guss et al. (Guss et al.,2001). Sequences of the upper strand of oligonucleotides used as probes for gel mobility shift assays and as PCR primers to introduce mutations into the Sd binding sites of the kni 0.69 kb element (fragment EC inFig. 1F) are listed below in the 5′ to 3′ orientation. Altered bases are shown in lowercase in the mutant version, and the corresponding bases are underlined in the native sequence as follows:

• kni 271 wild type:TTCCCCTCTTACATTTGTCGCATAGTTCCCATCTTGGCCA

• kni 271 mutant: TTCCCCTCTTtCATTaGgCtCATAaggCCCATCTTGGCCA

• kni 570 wild type:TTGCGGACACAGGACACGAAATGCGTTTTGTGCCTTAATT

• kni 570 mutant: TTGCGGACACAGGACACctAATGaGgTTTGTGCCTTAATT

• kni 640 wild type:TTGCTGGTGCACTGAAAGAAATAGTTGAAGGGATTATTG

• kni 640 mutant: TTGCTGGTGCACTGAAcctAATAGgTGAAGGGATTATTG

Reporter constructs with three mutated Sd binding sites were made by cloning PCR-generated fragments of fragment EC into the HsplacZ-CaSpeR plasmid (Nelson and Laughton, 1993). Four independentEC-3Sdmut-lacZ transformants were recovered and tested forlacZ expression.

Wings from adult flies were dissected in isopropanol and mounted in 100%Canada Balsam mounting medium (Aldrich #28,292-8) as described previously(Roberts, 1986).

In situ hybridization using digoxigenin-labeled antisense RNA probes(O'Neill and Bier, 1994) was performed alone or in combination with anti-β galactosidase (Promega)labeling as described previously(Sturtevant et al., 1993).

In a previous study, we presented genetic evidence that thekni^ri[1] mutation, as well as an approximately 50 kb deletion of kni upstream sequences(Df(3L)kni^ri[XT2]), disrupt the function of a wing-specific regulatory element in the kni locus that is responsible for driving expression of the kni and knrl genes in the L2 primordium (Lunde et al.,1998). As a consequence of the loss of kni/knrlexpression, kni^ri[1](Fig. 1B) andDf(3L)kni^ri[XT2] (Fig. 1C) flies have severely truncated L2 veins (compare with the wild-type vein pattern in Fig. 1A). As a first step in extending these studies, we gathered and characterized four other putative independently generated viable rialleles of the kni locus, which lack portions of the L2 vein,including kni^ri[53j](Fig. 1D),kni^ri[M3], kni^ri[92f], andkni^ri[M*] (not shown). As inkni^ri[1] wing discs, we found that kni expression is eliminated in the L2 primordium of each of these kni^rialleles (data not shown).

The studies mentioned above suggested that there might be an L2-specific wing vein enhancer that controls expression of kni and knrlin the L2 primordium. We searched for such an enhancer element in two ways. First, we screened for the ability of genomic fragments upstream of thekni coding region to drive lacZ expression in the L2 primordium (summarized in Fig. 1E). Second, we used Southern blot analysis to identify deletions or rearrangement breakpoints in the genomic DNA of ri mutants in thekni upstream region uncovered by Df(3L)kni^ri[XT2](Fig. 1E, see below). As a result of the former effort, we identified a single 4.8 kb EcoRI fragment (fragment E in Fig. 1E-G) that drives lacZ reporter gene expression in a sharp stripe (Fig. 2B) similar to that of endogenous kni expression in the L2 primordium(Fig. 2A). Double label experiments confirm that lacZ expression driven by theE-lacZ construct coincides with that of endogenous kniexpression (Fig. 2C). Additionally, the E-lacZ expression pattern includes a weaker stripe in the posterior region of the wing (Fig. 2B, arrow). A faint endogenous kni stripe can also be observed in this location in overstained preparations of wildtype discs (data not shown). Consistent with previous studies(Lunde et al., 1998), theE-lacZ L2 stripe is located just anterior to the broad domain of strong salm expression (Fig. 2D). Notably, the weaker posterior E-lacZ stripe forms immediately adjacent to the posterior border of the salm domain(Fig. 2D).

In order to delimit a minimal L2 enhancer element, we tested the ability of 5′ truncated derivatives of the 4.8 kb fragment E(Fig. 1G) to drive reporter gene expression in larval wing imaginal discs. We observed that subfragments of 2.0 kb (EcoRI-SalI, data not shown) and 1.4 kb (theEcoRI-XhoI fragment EX,Fig. 1G) drive expression oflacZ (Fig. 2E) in patterns nearly indistinguishable from that of the 4.8 kb E-lacZconstruct (Fig. 2B). Further truncation generating a 0.69 kb EcoRI-HincII fragment (EC,Fig. 1G), however, results in high levels of ectopic lacZ expression in broad anterior and posterior domains (Fig. 2F). This observation indicates that regulatory element(s) required to repress gene expression in the extreme anterior and posterior domains of the wing primordium reside between the distal 0.69 kb and 1.4 kb of fragment EX(repression domain, Fig. 3B),and that sequences sufficient for activation are contained within the proximal 0.69 kb fragment EC (activation domain,Fig. 3A). As discussed below,we have identified sequences necessary for L2 activity in the activation domain as well as potential repressor binding sites in the proposed repression domain (Fig. 3A,B,Fig. 7).

As a complement to screening genomic fragments upstream of the knilocus for the ability to drive gene expression in the L2 primordium usinglacZ promoter fusion constructs, we scanned the approximately 50 kb region deleted in the Df(3L)kni^ri[XT2] allele for detectable aberrations in five putative independently isolated rialleles. Using a series of probes spanning that complete genomic region(Fig. 1E), we found that a single EcoRI fragment (the same fragment E that drives expression in the L2 primordium) contains aberrations in the kni^ri[1],kni^ri[92f], kni^ri[M*] andkni^ri[M3] mutants. For each of these alleles, mutations map within the minimal 1.4 kb L2 enhancer element (seeFig. 1F). Although no large-scale lesions were detected by Southern blot analysis within fragment E of the kni^ri[53j] mutant, subsequent sequence analysis identified a single base pair alteration within the minimal L2 enhancer element EX (see below). The kni^ri[53j] allele is also associated with a small deletion (approx. 50 bp) in fragment R.

Further molecular analysis of the various ri alleles revealed that the putative independently derived kni^ri[1],kni^ri[92f], and kni^ri[M*]alleles all contain the same 252 bp deletion. As this deletion is not flanked by duplicated sequences that could promote frequent identical recombination events, it may be that the kni^ri[92f] andkni^ri[M*] alleles are actually re-isolates of the original kni^ri[1] mutation. Consistent with these three alleles having a common origin, they also share a restriction enzyme polymorphism (i.e. loss of a HindIII site) in fragment Q(Fig. 1E) that is not shared by the other ri mutants or our white^- control stock. To test whether the 252 bp deletion within the minimal L2 enhancer element is responsible for the loss of kni and knrl gene expression inkni^ri[1] mutants, we made a kni^ri[1]-mutated E-lacZ construct. We amplified genomic DNA containing thekni^ri[1] mutation and substituted it for the corresponding part of the original E-lacZ construct to generate EΔri¹-lacZ(Fig. 1G) and transformed this construct into flies. We examined lacZ expression in third instar larval wing discs of EΔri¹-lacZ flies and found that they failed to express lacZ in the L2 primordium or anywhere else in the wing pouch (Fig. 2G), indicating that the 252 bp deletion inkni^ri[1] mutants eliminates the activity of the L2 enhancer element.

Since Southern blot analysis did not reveal any obvious lesions in fragment E of the kni^ri[53j] allele, we considered the possibility that this allele might have a subtler lesion in the L2 enhancer region. Accordingly, we PCR amplified and sequenced the 1.4 kb fragment (EX,Fig. 1G) corresponding to the minimal L2 enhancer from kni^ri[53j]. We found only a single nucleotide difference between the sequence of this mutant allele and the wild-type fragment EX (C596A, Fig. 3A), which lies within the same 252 bp region deleted inkni^ri[1]. Although it has been observed that single nucleotide mutations can eliminate endogenous enhancer function(Shimell et al., 1994), such cases are sufficiently rare that it was important, as in the case ofkni^ri[1], to demonstrate whether this point mutation was responsible for the loss of enhancer function. We used the EXkni^ri[53j] PCR fragment to make a lacZ construct,confirmed the sequence of the mutant construct, generated four independent transformants, and tested these flies for lacZ expression in third instar wing discs. We found thatEX(ri^53j)-lacZ mutant wing discs fail to express lacZ in the L2 primordium(Fig. 2H), strongly suggesting that this single base pair mutation is responsible for the observed L2 vein loss phenotype observed in kni^ri[53j] flies. One notable difference between the EX(ri^53j)-lacZand EΔri¹-lacZ mutant wing discs is that in the point mutant construct (ri^53j), lacZexpression is lost selectively in the L2 primordium and in the weaker posterior stripe (note the normal levels and pattern of expression in extreme anterior and posterior regions of the disc,Fig. 2H), while in the deletion mutant construct (EΔri¹), reporter gene expression is eliminated throughout the wing disc(Fig. 2G).

As noted above, the kni^ri[53j] mutant is also associated with a deletion of ≈50 bp in fragment R, which lies approximately 15 kb 5′ to fragment E within predicted intron sequences of a putative transcription unit (see kni locus map,Fig. 1E). As the ubiquitous expression of this transcription unit is not affected inkni^ri[53j] mutants (K. L., unpublished observations) and fusion of fragment R with lacZ does not result in gene expression in the L2 primordium, we have no evidence to suggest that this aberration has any relevance to the L2 vein loss phenotype of kni^ri[53j]mutants.

Analysis of the kni^ri[M3] allele indicates that this mutation involves an insertion and possibly rearrangements within fragment E. One breakpoint of the kni^ri[M3] allele falls within the 1.4 kb EX minimal L2 enhancer element (Fig. 1F). Because this is a relatively weak ri allele and is associated with a complex lesion, we did not characterize the nature of this rearrangement further.

The 4.8 kb L2 enhancer fragment E and truncated derivatives (fragments EX and EC) drive reporter gene expression specifically in the wing pouch. Recent work has shown that gene expression in the wing primordium is controlled by Sd, which functions with Vestigial (Vg) to define wing identity(Kim et al., 1996;Kim et al., 1997;Halder et al., 1998;Simmonds et al., 1998;Guss et al., 2001;Halder and Carroll, 2001). We tested whether Sd might similarly be required to activate endogenouskni expression by generating clones of cells homozygous for a strong hypomorphic allele of sd using FLP-FRT mediated mitotic recombination(Golic, 1991). In sdclones, β-galactosidase expression driven by the E-lacZ element is reduced (Fig. 3D). In clones falling along the future wing margin, however, reporter gene expression is unaffected (not shown; see Discussion). These results indicate that Sd plays an in vivo role in activating kni expression within the wing blade.

Because Sd, a TEA-domain protein, functions as a direct transcriptional regulator of several key developmental genes expressed in the wing disc(Halder et al., 1998;Guss et al., 2001), we tested whether Sd binds to the kni L2 enhancer. We searched for specific Sd DNA binding sites in the 0.69 kb EC activation domain using DNAse I footprinting and gel shift analysis (Fig. 3G and data not shown) and identified five sites bound by the Sd TEA domain, which conform well to the known Sd consensus binding site and consist of a tandem doublet of Sd binding sites and four single binding sites(Fig. 3A). The four single Sd sites are removed by the 252 bp deletion in the kni^ri[1]allele, however, none of them covers the kni^ri[53j] point mutation. To determine whether Sd plays a direct role in activating the L2 enhancer, we mutated a subset of these Sd binding sites in the context of theEC-lacZ construct, transformed these mutated constructs into flies and stained for lacZ expression in third instar wing discs. We found that mutation of the doublet at 271 in combination with the two single sites at 570 and 640 resulted in a complete loss of lacZ expression in the wing disc (Fig. 3H). These results indicate that Sd is required for activation of kni expression in the wing disc.

Having shown that the minimal L2 enhancer is mutated in several independently derived ri alleles (kni^ri[1],kni^ri[M3], and kni^ri[53j]), we wished to know whether driving expression of either the kni or knrlgenes with this element could rescue the ri vein-loss phenotype. We addressed this question using the conditional GAL4/UAS expression system of Brand and Perrimon (Brand and Perrimon,1993). We created transgenic flies that carry fragment E driving GAL4 expression (L2-GAL4) and used this driver to activate expression of UAS-kni or UAS-knrl transgenes in the L2 primordium ofkni^ri[1] mutants. As expected, GAL4 RNA is expressed in the L2 primordium of L2-GAL4 wing discs(Fig. 4A). We placed this L2-GAL4 driver and the UAS-kni and UAS-knrl transgenes into the kni^ri[1] mutant background and generated flies of the genotype L2-GAL4>UAS-kni;kni^ri[1]/kni^ri[1](Fig. 4B) and L2-GAL4>UAS-knrl;kni^ri[1]/kni^ri[1](Fig. 4C). These flies have fully restored L2 veins that resemble wild-type L2 veins in that the bulk of the vein is formed on the ventral surface of the wing.

Since rho is known to play a critical role in activating Egfr signaling in all vein primordia, we also determined whether expression of this downstream effector of the kni locus could bypass the requirement forkni in kni^ri[1] mutants. We found that L2 formation is indeed rescued in L2-GAL4>UAS-rho;kni^ri[1]/kni^ri[1] flies(Fig. 4D). As in the case of rescue by kni or knrl, the rescued L2 vein forms primarily on the ventral surface of the wing.

As mentioned above (see Introduction), several genes that are important for vein development are expressed in veins other than L2. For example, the vein-and sensory organ-promoting gene ara is expressed in the odd numbered veins (L1, L3, and L5), the proneural genes ac and sc are expressed in L1 and L3, and the Notch ligand Delta is expressed in L1 and L3-L5, but not in L2. We were interested to know whether it is necessary to exclude expression of these genes from the L2 primordium for normal L2 development. To address this question, we misexpressed UAS-transgene copies of these genes in the L2 primordium with the L2-GAL4 driver and examined the L2 vein in adult wings. In each case, we observed defects in L2 development. Misexpression of the lateral inhibitory signal Delta in L2-GAL4>UAS-Dl flies results in modest truncation of the L2 vein in females (Fig. 5A) and almost entirely eliminates the L2 vein in sibling males(Fig. 5B), which are the consistently more severely affected sex. The near elimination of L2 development in severely affected Dl-misexpressing males is accompanied by a loss of rho expression in the L2 primordium of third instar larval wing discs (Fig. 5D), compare with wild-type rho expression inFig. 5C). Misexpression of the proneural gene sc in the L2 primordium of L2-GAL4>UAS-scflies results in L2 veins covered with ectopic bristles(Fig. 5E). The great majority of these ectopic bristles are strictly confined to the L2 primordium,consistent with the double label experiments(Fig. 2C) indicating that the L2-enhancer element drives gene expression precisely in the L2 primordium. In addition, ectopic bristles form sporadically in a broad posterior domain of the wing, which derives from a region of the wing disc in which the L2 enhancer element is also expressed in a weak diffuse pattern(Fig. 2B,D). Finally,misexpression of the ara gene in L2-GAL4>UAS-ara flies causes a reproducible thinning of the L2 vein(Fig. 5F). Cumulatively, these results underscore the importance of excluding expression of non-L2 vein genes from the L2 primordium.

As discussed above, the L2-GAL4 driver activates UAS transgene expression precisely in the L2 primordium. The ability to drive highly localized expression of genes in a non-essential tissue with L2-GAL4 suggests that this driver could be used as an effective tool to dissect the function and range of action of various genes. For example, since the eagle (eg)gene encodes a steroid hormone receptor closely related to Kni and Knrl(Rothe et al., 1989) we wondered whether this gene could substitute for kni or knrlin rescuing kni^ri[1] mutants. eg is involved in neuroblast specification during embryogenesis(Higashijima et al., 1996;Dittrich et al., 1997;Lundell and Hirsh, 1998), but is not expressed in the wing pouch (data not shown) and is not known to play a role in development of the wing proper. We found that, as in L2-GAL4>UAS-kni;kni^ri[1]/kni^ri[1] individuals, L2 formation is rescued in L2-GAL4>UAS-eg;kni^ri[1]/kni^ri[XT2] flies, however,this restored `L2' vein is decorated with a line of ectopic sensory bristles(Fig. 6A). As the same ectopic bristles are observed in L2-GAL4>UAS-eg individuals (data not shown), this phenotype can be attributed to misexpressing eg in the L2 primordium. By employing the L2 expression system it should now be possible to map the domain in Eg that is responsible for inducing ectopic bristles using chimeric Eg/Kni receptors.

The L2 expression system should also be useful in distinguishing cell-autonomous from non cell-autonomous gene activity. As a test of this idea we compared the phenotypes resulting from driving expression of an activated form of the Egfr (λtop) in the L2 primordium versus expression of the secreted Egf-like ligands Vein (Schnepp et al., 1996) and sSpitz(Schweitzer et al., 1995;Schnepp et al., 1998). Consistent with λtop acting in a cell autonomous fashion to promote vein versus intervein development, expression of this gene in the L2 primordium of L2-GAL4>UAS-λtop wings(Fig. 6B) had no effect on the development of the L2 vein or neighboring intervein cells. In contrast, when secreted Egf ligands were expressed in a similar fashion in L2-GAL4>UAS-vn wings (Fig. 6C) or in L2-GAL4>UAS-sspi wings(Fig. 6D), we observed ectopic veins that formed 2-3 cell diameters away from the L2 vein. Similarly, when we co-expressed rho and Star in the L2 primordium (e.g. in L2-GAL4>UAS-rho; UAS-Star wings;Fig. 6E), which also generates a potent secreted Egfr promoting activity(Guichard et al., 1999), we observed ectopic veins that were displaced from the endogenous L2 vein by a few cells. The ability of secreted ligands, but not the λtopconstruct, to induce ectopic vein formation at a distance cannot be attributed simply to the latter being a weaker activator of Egfr signaling since UAS-λtop is considerably more effective at generating ectopic veins than UAS-vn when other wing-GAL4 drivers are used (e.g. MS1096-GAL4 or CY2-GAL4, data not shown).

We also observed a non cell-autonomous activity of the Wingless (Wg) ligand in L2-GAL4>UAS-wg wings (Fig. 6F), which had ectopic veins displaced from the endogenous L2 as well as ectopic marginal bristles near the intersection of L2 with the margin. These ectopic marginal bristles always formed on the appropriate surface of the wing (Fig. 6F, insert),suggesting that this phenotype may result from the inappropriate activation of the wing margin genetic program.

An important step in understanding the link between patterning and morphogenesis is to identify key cis-regulatory elements that lie at the interface of these sequential processes. Formation of the L2 vein provides a potential paradigm for connecting patterning initiated by a morphogen (e.g. a centrally positioned BMP activity gradient) to the activation of a vein differentiation program in a narrow stripe of cells in the wing imaginal disc(e.g. controlled by kni/knrl). In this study we have identified acis-acting regulatory element that drives expression of thekni locus in the L2 primordium. We found that three distinct mutations causing L2 vein truncations alter this regulatory element and that in at least two of these cases the mutations disrupt the function of the enhancer. Dissection of this enhancer element indicates that a locally provided activating signal acts in conjunction with broadly distributed activating (sd dependent) and repressive systems to limit activation of the L2 enhancer to a sharp stripe of cells in the wing imaginal disc. We also found that excluding expression of non L2 vein genes from the L2 primordium is essential for normal L2 development.

The results described in this study demonstrate definitively thatri mutations are regulatory alleles of the kni locus disrupting the function of a cis-regulatory enhancer element that drives gene expression in the L2 primordium of wing imaginal discs. A crucial line of evidence supporting this conclusion is that mutant versions of the L2 enhancer incorporating either the 252 bp deletion present inkni^ri[1] or the single base pair substitution present inkni^ri[53j] eliminate the ability of this element to direct gene expression in the L2 primordium. In addition, it is possible to completely rescue the vein-loss phenotype of kni^ri[1] by expressing either the UAS-kni or UAS-knrl transgenes with an L2-GAL4 driver. Consistent with activation of rho being one of the key effectors of kni/knrl function, it is also possible to rescue the L2 vein-loss phenotype of kni^ri[1] by expressing a UAS-rho transgene in L2, although rescue is less complete and penetrant than that observed with UAS-kni or UAS-knrl.

The isolation of the L2 enhancer also addresses an unresolved question regarding the basis for the failure of Df(3L)kni^ri[XT2]and Df(3L)kni^FC82 to complement(Lunde et al., 1998). As these two deletions have endpoints that break within the same 1.7 kb EcoRI fragment (indicated by the dashed vertical lines inFig. 1E), one explanation for the failure of complementation could be that this 1.7 kb fragment contains the L2 enhancer. The alternative explanation for the ri phenotype ofDf(3L)kni^ri[XT2]/Df(3L)kni^FC82 is that transvection (Lewis, 1954;Geyer et al., 1990;Pirrotta, 1999), which normally occurs between regulatory and coding region alleles of thekni locus (Lunde et al.,1998), is interrupted by the large divergent deletions that have little if any overlap. Since the 4.8 EcoRI fragment containing the L2 enhancer maps nearly 10 kb upstream of the 1.7 kb EcoRI fragment, and a lacZ fusion construct (abd-lacZ,Fig. 1E) containing the 1.7 kbEcoRI fragment does not drive any gene expression in the wing disc(Lunde et al., 1998), the latter hypothesis that Df(3L)kni^ri[XT2] andDf(3L)kni^FC82 are unable to engage in effective transvection is the most likely explanation for the failure of these two mutations to complement.

The results of these and previous studies of L2 vein initiation lead to a model in which localized vein inductive signaling acts against a background of regional repression and global wing-specific activation(Fig. 7A). Previous genetic experiments have suggested a model in which sal-expressing cells produce a short-range L2 inducing signal (X) to which sal-expressing cells themselves cannot respond(Sturtevant et al., 1997;Lunde et al., 1998). According to this `for export only' signaling model, the signal X diffuses into adjacent anterior cells and activates expression of the kni/knrl genes in the L2 primordium, in much the same fashion as Hh, expressed from the refractory posterior compartment, activates expression of target genes such asdpp or ptc in the anterior compartment (reviewed byBier, 2000). Based on the analysis of the L2 enhancer element described in this study, the 252 bp region deleted in kni^ri[1] mutants may contain response element(s) to this putative factor X. It is worth noting, however, thatlacZ expression is lost in all cells of the wing disc (posterior and circumferential cells as well as the L2 primordium) when these sequences are deleted from the 4.8 kb fragment. This observation suggests that sequences mediating more general activation (e.g. Sd binding sites, see below) may also be contained within this 252 bp region. The sequence surrounding the single nucleotide alteration in kni^ri[53j] mutants (C596A),however, is a particularly intriguing candidate for mediating the L2-specific activation of kni since introducing this mutation in the context of the minimal 1.4 kb enhancer leads to selective loss of lacZexpression only in the stripe of cells adjacent to the saldomain.

The hypothetical transcription factor X′(Fig. 7B) that binds to the region surrounding the kni^ri[53j] point mutation and mediates the inductive signal X presumably collaborates with the more generally required wing selector Sd(Halder et al., 1998;Guss et al., 2001), since mutation of four of the Sd binding sites (the doublet and two single sites) in the L2 activation domain completely eliminates enhancer activity in the wing disc. Clonal analysis with a hypomorphic sd allele also indicates that sd is required for high-level expression of the full 4.8 kb L2 enhancer element in the wing disc. It is notable that the reduction inlacZ expression in these clones is not as dramatic as the complete loss of L2 activity observed when Sd binding sites in the activation domain are mutated. There are several possible explanations for this discrepancy. Firstly, the sd mutation used in these experiments is a hypomorphic allele and therefore has residual activity. Unfortunately, strongersd alleles produce even smaller viable clones in the wing disc and thus were not used. Secondly, as only small clones can be generated, they must typically have been produced with only two or three intervening cycles of cell division. Consequently, the sd^— cells may still contain functional levels of wild type Sd (protein perdurance). Another possibility is that other activators can partially substitute for Sd, at least in certain regions of the wing. Based on the absence of L2 activity when Sd binding sites are mutated and the reduction in L2 activity insd^— hypomorphic clones, we conclude that Sd plays an important role as an activator of the L2 enhancer. These results support the view that Sd functions as a general transcriptional activator of genes expressed in the wing field.

Repression also plays a key role in restricting L2 enhancer expression to a narrow stripe of wing disc cells. It has been shown previously thatsalm and salr, which are expressed strongly in the central region of the wing, repress expression of kni and knrl(Lunde et al., 1998), although low levels of sal may also be required to activate kniexpression (de Celis and Barrio,2000). In the current study we also find evidence for repression of L2-enhancer activity in peripheral wing disc cells abutting those expressing high levels of sal. Truncation of the minimal 1.4 kb(fragment EX) L2 enhancer element results in reporter gene expression expanding to fill the anterior and posterior regions of the wing pouch in a pattern complementary to that of sal. The region deleted from the EX fragment contains several consensus binding sites for Brinker(Fig. 3B,Fig. 7B)(Rushlow et al., 2001;Zhang et al., 2001), which may mediate this repression since the pattern of brinker (brk)expression in the wing pouch (Campbell and Tomlinson, 1999; Jazwinska et al., 1999) is very similar to that of lacZ expression driven by the 0.69 kb fragment EC. One way to integrate the action of the localized inductive signal X→X′ pathway with that of abutting central and peripheral repressive factors is to propose that the signal-dependent activator X′ can overcome the repressive action of the peripheral inhibitor but not repression by Salm/Salr (see legend toFig. 7). One feature common to several prominent signaling pathways is that activation of the pathway converts a resting repressor into a transcriptional activator(Barolo and Posakony, 2002). In the kni L2 enhancer, activation of the hypothetical signal X pathway may relieve repression by a heterologous repressor since the putative activator (i.e. X′) and repressor (e.g. Brk?) sequences in the L2 enhancer are separable. One example of a signaling pathway that functions by relieving inhibition by a heterologous repressor is the recent report of Egfr signaling inactivating repression by Suppressor of Hairless (Su(H))(Tsuda et al., 2002). In addition to central and peripheral repression in the wing disc, there may also be a repressor in posterior compartment cells (e.g. En,Fig. 7A,B) to prevent activation of the L2 enhancer in cells posterior to the salexpression domain. Perhaps the 4.8 kb L2 enhancer element lacks some sites for this putative repressor since E-lacZ drives significantly higher levels of gene expression in the posterior stripe than that observed for endogenous kni or knrl expression.

All longitudinal veins share several morphological characteristics such as being composed of densely packed cells on both the dorsal and ventral surfaces of the wing that secrete a thickened cuticle. The primordia of all longitudinal veins also express the rhomboid gene, which is required for activating the Egfr pathway in vein but not in intervein cells. In addition to these shared properties, each vein can be distinguished by expression of other vein genes (Biehs et al., 1998). For example, vein L2-specific characteristics include:expression of kni and knrl, lack of Delta expression, lack of cauplara expression, and lack of ac/sc expression. Given that all veins are ultimately quite similar morphologically, it is relevant to ask whether the differences in gene expression patterns observed in different veins are important.

In this study, we examined the necessity of excluding expression of non-L2 vein genes in the L2 primordium by forcing expression of genes such as Dl,ara, ac and sc in L2 and asking whether this manipulation had any impact on L2 development. These experiments strongly suggest that exclusion of non-L2 vein genes from the L2 primordium is indeed important since misexpression of each of these genes resulted in abnormal L2 development. The phenotypes resulting from misexpressing non-L2 vein genes in the L2 primordium can largely be reconciled with the normal functions of these genes. For example, forced expression of Dl leads to loss of L2,consistent with suppression of vein formation by Notch signaling. The ectopic bristles that form strictly along the L2 vein in wings misexpressingac or sc are also consistent with the neural promoting function of proneural genes. It is less clear why misexpression of theara gene causes thinning of the L2 primordium, since this gene normally activates expression of the vein-promoting gene rho in odd numbered veins and ac and sc in the L3 primordium. Perhaps expression of a gene normally involved in development of the odd numbered dorsal veins is somehow incompatible with development of the ventral L2 vein.

The sharp stripe of gene expression driven by L2-GAL4 can also be used as a rapid assay to test the function or range of action of various genes in the wing. For example, in the case of the Egfr pathway, it is possible to distinguish components exerting cell-autonomous versus non cell-autonomous functions. In line with the fact that activation of the Egfr pathway in the wing leads to the formation of veins, L2-GAL4-driven expression of the secreted ligand sSpi generates ectopic veins in the neighborhood of L2 while expression of the equally potent constitutively active λtop form of the EGF receptor does not result in a non-autonomous phenotype. The L2 expression system may also be used to compare the functions of related genes such as the nuclear receptors Kni, Knrl, and Eg. Although these three transcription factors share nearly identical DNA binding domains and can all rescue the vein-loss phenotype of ri mutants, they differ in that misexpression of eg in the L2 primordium induces the formation of ectopic bristles along L2. This distinct activity of eg may relate to its normal embryonic function in directing cell fate choices in the CNS. Using the L2 expression system as an assay, it should be straightforward to map the domain in eg that is responsible for its neural inducing capacity by constructing chimeric molecules composed of different domains of Eg and Kni. Finally, the observation that L2-driven expression of Dl leads to a highly penetrant loss of L2 in males and a consistently weaker phenotype in females provides the basis for a modifier screen to identify mutations that either suppress the phenotype in males or enhance the phenotype in females. Some of the modifier loci identified in such a screen, which are not specifically involved in Notch signaling, may encode components of the hypothesized factor X→X′ pathway.

We thank Orna Cook and Margaret Roark for critical comments on the manuscript, and Dan Ang for help with preparing figures. This work was supported by NIH grant NIH R01 GM60585 and NSF grant NSF IBN-0094634. K. A. G. acknowledges the support of Sean B. Carroll, in whose lab the studies on the role of sd were performed. This work was funded by NIH 1F32HD08326 to K. A. G., and the Howard Hughes Medical Institute, of which Sean B. Carroll is an Investigator.