Discovery of genes with highly restricted expression patterns in theDrosophila wing disc using DNA oligonucleotide microarrays

The imaginal discs of Drosophila melanogaster, which form the adult exoskeleton, offer an accessible model system for examining epithelial spatial patterning. In the larva, the discs show a prepattern of gene expression presaging the development of the body region and its specific structures. Knowing the number and identity of all genes involved is critical for a coherent understanding of genetic mechanisms operating to pattern a given structure.

Identifying genes has largely relied on the isolation of mutants with pattern defects, but strategies that detect genes by their expression patterns, enhancer and protein trapping using P elements with reporter genes(Bellen et al., 1989;Bier et al., 1989), or a GFP tag (Morin et al., 2001), have also identified genes involved in tissue-specific differentiation. However,P-elements show specificity in their insertion sites(Liao et al., 2000), so that screening the entire genome in this way may remain incomplete.

The annotated genome sequence (Adams et al., 2000) offers a systematic way to investigate spatial patterns of expression of all genes in the imaginal discs. Because the predicted gene number is approximately 14,000, this seems a large undertaking, but is being done for embryos that lend themselves more readily to high throughput in situ hybridization(http://www.fruitfly.org/cgi-bin/ex/insitu.pl). To identify a subset of genes that may have important roles in wing disc development, we used hybridization to high-density DNA-oligonucleotide arrays to define genes that show enriched spatial expression patterns. Cluster analysis of gene expression throughout the Drosophila life cycle has led to the identification of muscle-specific transcripts(Arbeitman et al., 2002) and genes expressed in specific imaginal discs, including the wing disc, have been discovered by profiling individual discs(Klebes et al., 2002). Here we compared RNA profiles from two complementary wing disc fragments, the presumptive wing/hinge and presumptive body wall regions(Fig. 1). These regions are separated by a lineage restriction that occurs in the first larval instar and genes preferentially expressed in one region may be important for that fate.

Our analysis identified many genes with uncharacterized roles in development that have striking spatial expression domains. We discuss the results in the light of the sequence information available for these genes and the role of known genes expressed in similar patterns. Some expression patterns suggest regulation by key morphogens. We have shown that three genes,with robust expression in the wing pouch, are sensitive to wingless(wg) signaling. This work makes a significant contribution to the goal of finding all the genes involved in wing disc development, by identifying a collection of genes that have not been implicated previously but which have expression patterns suggestive of potential region-specific roles.

Approximately 200 wild-type (Berlin strain) wandering third-instar larvae were dissected and the wing discs were collected in a drop of PBS on Sylgard(Dow Corning). Discs were cut between the presumptive hinge and the body wall regions using a 30-gauge syringe needle, and fragments were lysed in separate groups in RLT buffer (Qiagen). Total RNA was extracted from the tissue lysate using an RNeasy kit (Qiagen). Approximately 8 μg of total RNA from each of six independent samples (three wing/hinge and three body wall) was processed to produce biotinylated cRNA targets, which were hybridized toDrosophila Genechip 1 arrays following standard Affymetrix procedures(http://www.affymetrix.com).

Six arrays were used and each demonstrated control parameters within recommended limits (Raw Q<=30, background <=100, GAPDH 3′/5′ratios below four). To allow comparison between chips, each was analyzed using global scaling with a target intensity of 300. The scaling factors used to normalize to the target value were within four-fold of each other in all comparisons (range 0.36 to 1.34). Affymetrix Microarray Suite version 5.0 software was used to make each pairwise comparison between the three wing/hinge and the three body wall arrays. The data were then exported to Excel, in which the `Signal Log Ratios' were converted to fold changes. Only transcripts called as present in at least two samples, and showing a two-fold or greater difference (P≥0.95) between the wing/hinge and body wall samples in at least six out of the nine comparisons are included. The data were sorted by average fold-change to produce the lists of genes shown in Tables 1 and2. This filtering of the data means that some genes with spatially restricted patterns may be excluded; this is especially likely to be the case for genes with low expression levels. Therefore, the Excel spreadsheet with the full comparison set is given atSupplementary Information,and the array data have been deposited into the Gene Expression Omnibus (GEO)database(http://www.ncbi.nlm.nih.gov/geo/)(series accession number GSE93 and sample accession numbers GSM2583, GSM2584,GSM2585, GSM2586, GSM2587 and GSM2588).

In situ hybridization of DIG-labeled antisense RNA probes to dissected wandering third instar larvae followed a standard protocol(Sturtevant et al., 1993). Probes were generated using the appropriate RNA polymerase (SP6, T7 or T3)following the manufacturer's protocol (Roche). Synthesis was assessed by gel electrophoresis and the precipitated probe was dissolved in 50% formamide in TE. The DIG-labeled probes were diluted and tested over a broad range, as there was a great deal of variation in the effective probe concentration. Discs were mounted in Aquapolymount (Poly Sciences) and photographed using bright field or Nomarski optics.

When possible, DIG-labeled RNA probes were generated from cDNA clones obtained from the Drosophila Gene Collection (DGC,http://www.fruitfly.org/DGC/index.html)or from published sources. Most clones used were from the DGC1 release. DGC clones belonging to the unreleased DGC2 set were ordered from theDrosophila EST collections maintained by ResGen (Invitrogen). Clones used are listed in descending order according to the ranked lists (Tables1 and2), with the DGC clone identification number or literature citation as appropriate. Clones were linearized in the 5′ multiple cloning site (except where noted) and transcribed with the appropriate RNA polymerase.

pOT2a-tailup (GH12431), pOT2a-CG11100 (SD10763), pFlc-CG15064 (RE70039),pFlc-CG15353 (RH63135), pBS-CG6921 (LD14839), pFlc-BM40/SPARC (RH45818),pFlc-Obp56a/CG11797 (RE46170), pOT2a-zfh1 (SD06902), pFlc-viking (RE68619),pBS-Ef1alpha100E (RE68984), pFlc-Obp99a/CG18111 (RH70762), pOT2a-CG10126(GH22994), pOT2a-CG9338 (GH07967), pBS-Cg25C (GM04010) (this clone failed to grow and a PCR product was generated, see below), pBS-Act57B (LD04994)(linearized with AflIII to generate a 3′UTR probe specific to Actin57B that does not cross-hybridize with other Actins), pOT2a-CG5397 (GH04232),pFlc-Idgf4 (RE30918), pOT2a-CG4386 (LD47230), pFlc-CG3244 (RH18728),pFlc-CG2663 (RE73641), pOT2a-CG10275 (LD31354), pOT2a-CG8689/CG30359(GH18222), pOT2a-tsp/CG11326 (GH27479), pOT2a-Gs2 (GH14412), pFlc-Gld(RE20037).

pOT2a-CG17278 (SD04019), pBS-pdm2(Poole, 1995), pFlc-CG8780(RE33994), pOT2a-Nep1/5894 (GH03315), pBS-opa(Benedyk et al., 1994),pOT2a-Cyp310a1 (LD44491), pFlc-CG14534 (RE71854), pOT2a-CG9008/BG:DS00797.2(GH14910), pOT2a-zfh2 (GH11902,), pFlc-Doc2/CG5187 (RE40937) (linearized with AvaII to produce a specific 3′-end probe that does not cross-hybridize with Doc1), pOT2a-ana (GH07389), pOT2a-CG8381/CG30069 (LP06813), pOT2a-CG8483(LD39025), pCR-TOPOII-dorsocross (Lo and Frasch, 2001) (linearized with AflII to produce a specific 3′-end probe that does not cross-hybridize with Doc2/CG5187),pFlc-wengen/CG6531(RE29502).

For genes without cDNA clones available, gene-specific fragments were generated by PCR from genomic DNA using either Taq polymerase (Invitrogen) or Pfu Turbo polymerase (Stratagene). Genomic DNA was generated from adult Canton-S flies using the Qiagen DNeasy Kit. Primers were chosen that would specifically amplify exonic sequence from the gene. The primers were designed using the annotated gene information from GadFly, FlyBase and NCBI/GenBank. Exon numbers mentioned below refer to exon predictions or experimental information from the above sources. Exon fragments were cloned into pBS-KS and verified by sequencing. All clones have a 3′ T7 promoter. The fragments generated, the corresponding gene region and the primers used are listed below(written 5′ to 3′ with any 5′/3′ restriction sites introduced underlined) in the order that they appear on the wing/hinge enriched list (Table 2):

• dve exon 5, bp 520-769; primers 5Pdve CATGCACCACGATCACCATCATGG and Rdve CGTGCTGACTCAGATAGCCGTTCATGG

• CG15001 single exon, bp 1-321 (+5bp 5′UTR); primers 5P15001 GCGAATTCGCAATGGAGGCGAGCTCGAATCC; and 3P15001 GCTCTAGACTTGTTCCATTCGCATTCCTTCC

• CG15489 single exon, bp 38-727; primers 5PN15489 GCGAATTCAGCAACGGCAGCAGCGTGTTGC and 3PN15489 GCTCTAGACTGTTGGCGCCTTCAATGTTGTCG

• CG15488 single exon, bp 8-404; primers 5PN15488 GCGAATTCATGTCAGCGCAAATCACAGCAGCTCC and 3PN15488 GCTCTAGACACTGCTCTAAATCTCGCACATCC

• CG15000 exon 3, bp 22-840; primers 5PN15000 GCGAATTCAACCGGCATTGATCTTCAATCCGG and 3PN15000 GCTCTAGAGGAGAGTCGCTGTCGCTTGGAGG

• CG4861/CG31094 exon 3, bp 288-1374; primers 5P4861 CAACCAAGGAGGATGCAACGCAACC and 3P4861 GCTCTAGAACTTGTTCGCCTTGAAGACGG

• CG6469 exon 2, bp 24-346; primers 5P6469 GCGAATTCGATGTGAAGGACGACG and 3P6469 GCTCTAGACTGGTGGTACTGTGGCTGCTGC

• CG14301 exon 4, bp 2-227; primers 5P14301 GCGAATTCCTGGCATTACTGCGACATCGATGG and 3P14301 GCTCTAGATCAGGTGAAGATGTCCTCGACC

• Ugt86Di/CG6658 exon 8, bp 23-417; primers 5PUgt GCACTCACCCAACATGATTGAAGTGG and RUgt CCAGTCGGAAATGTAGACATTCTCCG

• CG5758 exon 8, bp 77-471; primers 5P5758N GCGAATTCTGCTGCTGAACCACTTCGTTAAGG and 3P5758N GCTCTAGATCCTCGAAGTCCTCGTCATCACCG

• Cg25C exon 4, bp 90-900; primers 5P25C GCGAATTCAGTTACGACATTGTAGACTCTGC and 3P25C GCTCTAGAGTCCAGTGTCGCCTTCAGGTCC

The following transgenic lines were used: 71B-GAL4(Brand and Perrimon, 1993),C96-GAL4 (Gustafson and Boulianne, 1996), UAS-wg and UAS-DNdTCF(van de Wetering et al.,1997). Larvae from crosses were raised at 29°C to enhance GAL4 activity and produce more extreme phenotypes.

To identify genes with expression patterns enriched in the presumptive wing/hinge or body wall regions, wing imaginal discs were cut into two fragments at the boundary between the body wall and the wing hinge(Fig. 1). Folds associated with the hinge provide morphological features to allow precise cutting(Fig. 1). RNA expression profiles of these samples were determined using oligonucleotide microarrays representing approximately 13,500 known and predicted genes in theDrosophila genome (Genechip Drosophila Genome Array 1,Affymetrix). The transcripts were ranked by average fold-change and those showing a two-fold or greater enrichment are shown in Tables1 and2. Information for all genes is available athttp://dev.biologists.org/supplemental/. Ninety-four transcripts show two-fold or greater enrichment in the body wall(Table 1) and 56 transcripts show two-fold or greater enrichment in the wing/hinge(Table 2). Several of these genes were also discovered by Klebes et al. as being more highly expressed in wing discs than leg discs or eye-antennal discs, suggesting they may also have appendage-specific roles (Klebes et al.,2002).

The rank order of transcripts correlates well with the spatial expression patterns of characterized genes. In the body wall, pannier(pan), twist (twi) and BarH1, which are enriched in the body-wall sample (Table 1), are known to be highly expressed in the presumptive body wall(Bate et al., 1991;Ramain et al., 1993;Sato et al., 1999). In the wing, knot (kn), nubbin (nub) andDistal-less (Dll) are expressed at levels greater than 10-fold above those in the body wall (Table 2). kn is expressed in the wing 3/4 intervein and hinge regions (Mohler et al., 2000;Vervoort et al., 1999),nub is strongly expressed in the entire wing pouch(Ng et al., 1995) andDll is expressed along the dorsal-ventral (DV) margin exclusively in the wing pouch (Campbell et al.,1993; Gorfinkiel et al.,1997).

Other genes, known to have important roles in disc development, appear lower down the rank order (Table 2). vestigial (vg), a key gene for development of the wing and hinge regions (Williams et al., 1991), shows only two-fold enrichment but this is consistent with the expression pattern of vg in the wing disc that extends into the body wall region (Williams et al.,1991). Transcripts with expression patterns restricted to the posterior compartment, engrailed (en), invected(inv) and hedgehog (hh)(Coleman et al., 1987;Kornberg et al., 1985;Tabata et al., 1992), showed approximately two-fold enrichment in the wing/hinge sample(Table 2). The anterior-posterior compartment boundary splits the wing/hinge region into two equally sized compartments, but the position of the boundary in the body wall region produces a small posterior compartment representing approximately one-quarter of the total tissue (Fig. 1B). This is consistent with the approximately two-fold enrichment of posterior-specific transcripts found in the wing/hinge tissue sample. TheE(spl)-Complex genes are expressed in developing sensory organs found in both the body wall and wing margin regions. Hence, these genes are not enriched in any one sample (seehttp://dev.biologists.org/supplemental/). The m6 gene is an exception (enriched in the body wall sample)(Table 1) and is known to be expressed only in the body wall region(Wurmbach et al., 1999). In contrast, genes that show ubiquitous expression such as Ras or tubulin show no enrichment on the arrays (seehttp://dev.biologists.org/supplemental/).

Microarray analysis can therefore identify transcripts known to be differentially expressed in the wing/hinge and body wall regions of the disc. Further, the rank order of these by fold change reflects the level of enrichment so that transcripts with more restricted domains appear higher on the list. Few expression patterns of the genes on our list have been described, so to verify the validity of the approach, and to discover more genes with potential roles in the development of these specific regions, we made in situ hybridizations for some of these uncharacterized genes.

We analyzed 50 transcripts, shown in Tables1 and2, that had strong enrichment(mostly three-fold or greater). For the body wall-enriched transcripts, the larger set, we analyzed only transcripts for which clones are available in theDrosophila gene collections (DGC1 and DGC2, Berkeley Drosophila Genome Project). For the wing/hinge region, we examined transcripts with three-fold or greater enrichment, systematically in rank order from the top,and generated PCR probes when clones were not available. We found that all transcripts tested showed expression patterns that were consistent with the microarray data providing confirmation that the microarray analysis mirrors the spatial distribution of transcripts in vivo. Body wall-specific expression patterns are shown in Fig. 2and wing/hinge-specific patterns are shown inFig. 3.

The wing disc comprises three cell layers; the squamous epithelium of the peripodial membrane, the columnar epithelium that becomes the adult epidermis,and the adepithelial layer that includes myoblast cells that give rise to adult thoracic muscles and tracheal cells that form air passages(Fig. 1C). The adepithelial layer extends from the proximal disc dorsally into the hinge region(Fig. 1C). The body wall fragment includes cells of all three layers, so the arrays also identified transcripts specific to muscle and tracheal cells.

pan and BarH1, which encode transcription factors, are expressed in the body wall epidermis and are involved in bristle patterning(Ramain et al., 1993;Sato et al., 1999). Both transcripts were highly enriched on the arrays(Table 1). Also highly enriched was tailup (tup) (Thor and Thomas, 1997), which encodes a LIM domain homeobox protein,and is expressed in the epithelium in a large region of the posterior body wall encompassing the presumptive postnotum, scutellum and scutum(Fig. 2A). No role fortup in patterning the mesothorax has been described. Another transcript with broad expression was thrombospondin/CG11326 (tsp),which is expressed in a similar region of the body wall to tup(Fig. 2W). tsp is also expressed in the ventral hinge and hence shows lower enrichment on the arrays. The other genes found to be specific to the epithelium showed highly localized expression: Obp56a/CG11797 (Fig. 2G), CG10126 (Fig. 2L), CG3244 (Fig. 2S) and Glucose dehydrogenase(Fig. 2Y).Obp56a/CG11797 encodes an odorant-binding protein and interestingly three other odorant-binding proteins showed enrichment on the arrays(Table 1):Obp99a/CG18111 (Fig. 2K), CG9358 and Obp56d/CG1128. We foundIdgf4, encoding an imaginal disc growth factor(Kawamura et al., 1999), is expressed in the peripodial membrane, primarily in dorsal cells(Fig. 2Q). Presumably secretion of Idgf4 could influence development of the columnar epithelium.

Myoblast cells of the adepithelial layer develop into the direct and indirect flight muscles of the thorax, and genes involved in the development of these muscles have been shown to be expressed in the myoblasts during wing disc development. Several of these transcripts are enriched on the arrays(Table 1): Mef2(Cripps et al., 1998),twist (twi) (Bate et al.,1991) and heartless (htl)(Cripps et al., 1998).Act57B is known to be regulated by Mef2 in the embryo(Kelly et al., 2002), and we show Act57B is expressed in the myoblasts(Fig. 2O), suggesting this relationship also exists in these adult muscle precursors. Mef2expression is activated by twi(Cripps et al., 1998) and may be inhibited by the transcriptional repressor, zinc finger homology 1(zfh1) (Postigo et al.,1999). zfh1 is expressed in the myoblasts(Fig. 2H). stumps is also enriched on the arrays (Table 1) and expressed in the myoblasts(Sato and Kornberg, 2002). Together with htl, stumps has a role in the development of the tracheal cells (Imam et al.,1999; Sato and Kornberg,2002) (see also below). Viking (Vkg) encodes a component of collagen type IV and is known to be coexpressed withCg25C, another collagen IV subunit in the embryo and in blood cells(Yasothornsrikul et al.,1997). Both transcripts are enriched on the arrays(Table 1) and show similar expression patterns in the adepithelial myoblasts and blood cells(Fig. 2I,N). Other genes showing specific expression in the myoblasts are BM-40/SPARC, a calcium-binding glycoprotein, which is expressed in the embryonic mesoderm(Furlong et al., 2001;Martinek et al., 2002)(Fig. 2F), Elongation factor 1 alpha 100E (Ef1 alpha)(Hovemann et al., 1988)(Fig. 2J), CG8689, an alpha-amylase (Fig. 2V), and two transcripts encoding predicted proteins with unknown functionCG11100 (Fig. 2B) andCG15064 (Fig. 2C).

In the wing disc, cells of the larval and developing adult tracheal systems require activity of genes in the FGF pathway(Sato and Kornberg, 2002). Some of the key genes are expressed in the myoblasts (for example,htl and stumps), others in the epithelium (for example,branchless, bnl), and others in the tracheal cells themselves (for example, breathless, btl) (Sato and Kornberg, 2002). htl and stumps showed enrichment on the arrays but bnl and btl were not detectable. For bnl this may be because expression is highly localized and apparently at very low levels(Sato and Kornberg, 2002). However, it is not clear why the arrays failed to detect btlexpression because we did identify six genes that are also expressed specifically in tracheal cells. These are CG5397, an O-acyltransferase (Fig. 2P),CG4386, a serine-type endopeptidase(Fig. 2R), CG2663, an alpha-tocopherol transfer-like protein(Fig. 2T), and CG15353(Fig. 2D), CG6921(Fig. 2E) and CG9338(Fig. 2M) that have no known homologies. In particular, CG4386 is interesting as it is only expressed in the dorsal branch (Fig. 2R), and CG6921 is distinguished because it is very strongly expressed in the most proximal cells(Fig. 2E).

The wing/hinge fragment of the wing disc primarily contains cells of the peripodial membrane and the columnar epithelium(Fig. 1C), with only a few myoblasts that extend into the hinge region. Thus the genes detected by the arrays as enriched in this disc fragment were expressed in cells of one of the two epithelial layers.

Transcription factors comprise the largest category of genes (18/56) with elevated expression in the wing/hinge region. These are expected to have regulatory roles in patterning the region. Transcription factors with known expression domains and roles in wing development are present: kn, pox-n,nub, Dll, bifid/optomotor blind, rotund, ventral veins lacking, en, vgand in (Awasaki and Kimura,2001; Cohen et al.,1989; Coleman et al.,1987; de Celis et al.,1995; Grimm and Pflugfelder,1996; Kornberg et al.,1985; Mohler et al.,2000; Ng et al.,1995; St Pierre et al.,2002; Tabata et al.,1992; Vervoort et al.,1999). pdm2, which is highly related to nub,also shows wing-enriched expression on the arrays and is expressed in a similar domain to nub (Fig. 3C) (Ng et al.,1995). pdm2 apparently has no significant function in the wing (Yeo et al., 1995). The roles of the remaining seven predicted transcription factors is unknown,although the expression pattern of zinc finger homology 2(zfh2) and Sox 15 have been described and both are expressed specifically in the hinge region (Cremazy et al., 2001; Klebes et al.,2002). defective proventriculous (dve), which encodes a homeodomain protein (Nakagoshi et al., 1998), and CG15000, which is similar to NGFI-A-binding protein 2, are broadly expressed in the wing pouch, althoughdve is downregulated at the DV compartment boundary(Fig. 3B,L). odd paired (opa), known for a role in embryonic segmentation(Benedyk et al., 1994), is discretely expressed in cells of the presumptive mesopleura and dorsal hinge(Fig. 3G). No role foropa in wing disc development has been reported. Dorsocross1(Doc1) (Lo and Frasch,2001) and Doc2/CG5187 are T-Box related factors that are expressed in what appears to be an identical domain in the wing disc(Fig. 3R,X). Both transcripts also accumulate in body wall cells and this probably lowers their position in the overall ranked list (Table 2).

Eight transcripts encoding enzymes are enriched two-fold or greater in the wing/hinge region (Table 2). This group includes the most highly enriched transcript detected in the analysis, a kazal-type serpin gene CG17278 (68-fold,Table 2). CG17278shows a strong and specific expression pattern in the wing encompassing most of the wing pouch. One of the potentially most interesting wing-enriched enzymes is a cytochrome P450 gene, Cyp310al. This gene is strongly expressed in the dorsal and ventral parts of the wing pouch but excluded from the DV and AP boundaries (Fig. 3J). Variable expression in anterior body wall cells is also observed which is consistent with the array data that indicateCyp310al transcripts are also present in body wall RNA. The role of cytochrome P450 genes in development has recently been reviewed(Stoilov, 2001), and the list of these genes with roles in development is growing. Surprisingly, theβ-galactosidase gene (CG3132) was found to be enriched in the wing/hinge region (Table 2).β-gal expression in Drosophila has been analyzed and although it is expressed in some discs, wing expression was not reported(Schnetzer and Tyler, 1996). We did find weak expression in a cluster of cells in the hinge but the majority of expression is in blood cells, which adhere preferentially to the distal disc margin (Fig. 3E). Thus the β-gal transcript probably appears as wing/hinge enriched primarily because it is expressed in blood cells. We also determined the expression pattern of two other enzymes; the metalloendopeptidaseNep1/CG5894 (Fig. 3F)and UDP-glucosyl transferase (Ugt86Di)(Fig. 3S).

The α-integrin, inflated, which has a role in cell adhesion,is expressed in the ventral compartment(Brower et al., 1984) and is thus enriched on the wing/hinge arrays(Fig. 1,Table 2). A novel gene,CG5758, is potentially involved in cell adhesion as it encodes a predicted protein with β-Ig-H3/Fas domains and its expression is restricted to the dorsal hinge (Fig. 3U). CG8381 encodes a proline-rich protein with repeated`PEVK' motifs also found in titin. This gene is strongly expressed in the wing pouch but repressed in cells of the future veins and cells at the DV margin(Fig. 3V). Despite intense expression in the wing pouch, CG8381 shows only modest enrichment on the arrays (Table 2), probably reflecting the fact that the gene is also expressed in several groups of cells in the body wall region (Fig. 3V).

The expression of two receptors was determined. CG4861 encodes an ldl-receptor-like protein and is expressed at very low levels throughout the wing pouch (Fig. 3O).wengen/CG6531, which is a receptor of the TNFR family(Kanda et al., 2002), is expressed strongly in the wing pouch and weakly in the body wall(Fig. 3Y). On the arrays, its ligand, eiger (Kanda et al.,2002), was undetectable in the wing/hinge region sample but enriched in the body wall sample (Table 1).

Two structural proteins, CG6469, a larval cuticle protein, andCG14301, a chitin-binding protein, are the only genes we identified as being expressed in the ventral peripodial membrane. CG6469 is expressed broadly in the peripodial membrane but at a higher level in the ventral region (Fig. 3P).CG14301 is expressed in cells of both epithelial layers, in the columnar epithelium at the anterior disc margin and in four patches of cells in the wing pouch and the overlying peripodial membrane(Fig. 3Q).

In a group of genes with miscellaneous functions(Table 2) we determined the expression of three genes. anachronism (ana), a secreted glycoprotein (Ebens et al.,1993), is expressed in five clusters of cells including one in the body wall region and in some individual neuroblasts(Fig. 3T). ana null mutants are viable and have no observable defects suggesting it is not required, or functions redundantly, in the wing(Park et al., 1997).CG14534, which has a domain that has been recognized in several proteins but has an unknown function (DUF243), is expressed only in cells that will give rise to the posterior wing margin(Fig. 3M). CG8483,which has homology to a venom allergen, is expressed in a complex pattern suggestive of expression in peripheral sense organ precursors(Fig. 3W).

We determined the expression pattern for five of eight genes for which the sequence reveals no homology to known protein domains. CG15489 andCG15488 (Fig. 2I,K)are in a cluster of genes also including nub and pdm-2 that are expressed in similar domains and are adjacent in the genome.CG15001, consisting of only a single exon, is adjacent to another gene (CG15000), also discovered on the arrays, with a similar expression domain (Fig. 3H,L).BG:DS00797.2/CG9008 is expressed strongly in the wing pouch and also in the adepithelial cell layer (Fig. 3N). CG8780 is highly enriched on the arrays (31-fold,Table 2) and expressed specifically in the hinge and ventral pleura(Fig. 3D).

The genes, CG17278, Cyp310a1 and CG8381 all show very intense expression in the wing pouch but reduced expression at the DV margin(Fig. 4A,D,G). Wg is expressed at the DV margin forming a gradient that regulates the expression of target genes in a concentration-dependent manner(Strigini and Cohen, 2000). To determine whether Wg signaling represses the expression of CG17278,Cyp310a1 and CG8381, we ectopically expressed wg in the dorsal and ventral wing-pouch regions (71B-gal4; UAS-wg), or inhibited Wg function at the DV margin by expressing a dominant-negative form of TCF (van de Wetering et al.,1997), a transcription factor required for Wg-signal transduction(C96-GAL4; UAS-DN-dTCF). With higher levels of Wg activity in the wing pouch, expression of all three genes was inhibited(Fig. 4B,E,H). In contrast,inhibition of Wg signaling at the DV margin allowed ectopic expression ofCyp310a1 in all margin cells and increased the number of cells expressing CG17278 and CG8381(Fig. 4F,C,I). In the presumptive margin, cells continue to express wg in the absence of Wg activity, cell replication increases(Phillips and Whittle, 1993),and ectopic expression of dmyc appears in margin cells(Johnston et al., 1999). Therefore, ectopic expression of the genes studied here is caused by loss of Wg-dependent repression rather than loss of the non-expressing cells from the presumptive margin. This does not imply that Wg-dependent repression must be direct. Without functional data on these potential target genes, their relationship to wg and their role in wing patterning remain unknown.

We have shown that microarray analysis of RNA profiles, followed by in situ hybridization, can rapidly identify candidate genes that warrant investigation for their role in wing disc development. Most genes identified here are not represented by mutant alleles. This may reflect any of several situations. (1)Some genes may be refractive to mutagenesis. This is unlikely to be the case for chemical mutagens or ionizing radiation, but P elements show specificity in insertion site (Liao et al.,2000). As P elements are the mutagen of choice in most current screens, for example, the Berkeley Drosophila Genome Project (BDGP) screen(Spradling et al., 1999), some genes may not be susceptible. (2) Redundancy masks gross phenotypes. This may be a factor for highly related genes such as Doc1 andDoc2/CG5187, which are also expressed in very similar patterns(Fig. 3R,X), and for members of multi-gene families such as the cytochrome, Cyp310a1 and the serpin,CG17278. (3) The genes have no crucial function. Despite having a localized expression pattern, some genes may play a minor role or no role in the cells in which they are expressed.

The challenge will be to decide among these possibilities in a new round of genetic analysis that uses techniques such as RNA silencing or homologous recombination to reduce function of specific genes(Fortier and Belote, 2000;Kennerdell and Carthew, 1998;Kennerdell and Carthew, 2000;Martinek and Young, 2000;Piccin et al., 2001;Rong and Golic, 2000;Rong et al., 2002), and if necessary simultaneously in several genes in a family, to determine if phenotypic change occurs. Whole genome profiling is a powerful method to identify genes that then become a high priority for such analysis.

We thank S. DiNardo, S. Poole and M. Frasch for clones. This work was supported by a Case Studentship and travel grant from the BBSRC to M. B.; a Fogarty senior fellowship to A. S.; grants awarded to A. S. from the NSF and to R. P. from Cancer Research UK and to J. R. S. W. from BBSRC.