Functional Analysis of Genes Differentially Expressed in the Drosophila Wing Disc: Role of Transcripts Enriched in the Wing Region

Abstract

Differential gene expression is the major mechanism underlying the development of specific body regions. Here we assessed the role of genes differentially expressed in the Drosophila wing imaginal disc, which gives rise to two distinct adult structures: the body wall and the wing. Reverse genetics was used to test the function of uncharacterized genes first identified in a microarray screen as having high levels of expression in the presumptive wing. Such genes could participate in elaborating the specific morphological characteristics of the wing. The activity of the genes was modulated using misexpression and RNAi-mediated silencing. Misexpression of eight of nine genes tested caused phenotypes. Of 12 genes tested, 10 showed effective silencing with RNAi transgenes, but only 3 of these had resulting phenotypes. The wing phenotypes resulting from RNAi suggest that CG8780 is involved in patterning the veins in the proximal region of the wing blade and that CG17278 and CG30069 are required for adhesion of wing surfaces. Venation and apposition of the wing surfaces are processes specific to wing development providing a correlation between the expression and function of these genes. The results show that a combination of expression profiling and tissue-specific gene silencing has the potential to identify new genes involved in wing development and hence to contribute to our understanding of this process. However, there are both technical and biological limitations to this approach, including the efficacy of RNAi and the role that gene redundancy may play in masking phenotypes.

A fundamental question in development is how primordial cell groups are patterned to give rise to differentiated body structures with distinct morphologies. The Drosophila imaginal wing disc has been extensively studied as a model for tissue development. The single-layer columnar epithelium, which comprises the major cell layer of the disc, gives rise to two adult derivatives: the body wall (notum) and the wing/hinge (Bryant 1975). Many components of evolutionarily conserved signaling pathways are involved in division of the early wing disc into cell lineage compartments (reviewed in Cohen 1993; Blair 1995; Dahmann and Basler 1999; Strigini and Cohen 1999; Morata 2001; Tabata and Takei 2004). The subsequent specification and elaboration of the body-wall and wing/hinge regions is also controlled by the region-specific activity of signaling pathways and the expression of their target genes (e.g., Lawrence and Morata 1977; Halder et al. 1998; Diez Del Corral et al. 1999; Wang et al. 2000; Cavodeassi et al. 2002; Zecca and Struhl 2002a,b; Villa-Cuesta and Modolell 2005). Genes with potential roles in wing but not body-wall development are expected to be expressed primarily in the presumptive wing region. To identify such genes, a microarray-based approach was used by Butler et al. (2003) to assay differential expression between the presumptive body-wall and wing/hinge regions. Genes known to be required for development of a given region were identified (e.g., pannier, Bar, and stripe in the body wall and nubbin, Distalless, and vestigial in the wing). Additionally, many uncharacterized genes also displayed enriched expression in either the body wall or the wing pouch.

Here we used a reverse genetic approach to examine the developmental roles of several of these uncharacterized genes. We identified three genes required for wing development that had not previously been found using traditional forward genetic methods. Most genes, however, did not display loss-of-function phenotypes following RNA interference (RNAi). Each of the three genes with phenotypes affected a process specific to wing development that does not occur in the body-wall region. We conclude that transcript profiling, followed by a reverse genetic analysis, is an effective way to identify novel genes with functions in the developing wing. We also discuss limitations to this approach that are both technical and biological, including RNAi efficacy and the role that redundancy may play in masking phenotypes.

MATERIALS AND METHODS

Generation of misexpression constructs:

For genes with available full-length cDNA clones, cDNAs were excised from the parental plasmid and cloned into pUAST (Brand and Perrimon 1993). The Drosophila Gene Collection (DGC; http://www.fruitfly.org/EST/index.shtml) clones used were as follows: Cyp310a1/CG10391, LD44491; Doc2/CG5187, RE40937; Nep1/CG5894, GH03315; CG5758, GH27705; CG8483, LD39025; CG8780, RE33994; CG9008/BG:DS00797.2, GH14910; CG14534, RE71854; and CG17278, SD04019.

Generation of RNAi constructs:

To create hairpin double-strand RNA (dsRNA) constructs, a 650-bp or larger portion of a given gene, the RNAi “trigger,” was cloned into pBlueScript-KS, with a 360-bp heterologous spacer fragment as in Piccin et al. (2001). This sense-strand “trigger + spacer” fragment was then cloned into pUAST. The dsRNA construct was completed by adding the trigger fragment in reverse orientation into pUAST containing the “trigger + spacer” fragment. When necessary, appropriate restriction sites were added using PCR primers. For constructs containing smaller trigger sequences, an artificial intron derived from the third intron of the vn gene was used as a spacer between the inverted repeats to increase dsRNA activity (Schnepp et al. 1996; Reichhart et al. 2002). Sequences used were from cDNA clones or were cloned from genomic DNA using PCR.

Trigger sequences used in upstream activating sequence (UAS)–RNAi constructs (sequence coordinates from DGC cDNA clones or FlyBase annotated exons) were Cyp310a1, LD44491 (nt 1–891); dNAB/CG15000, exon 2 from CG15000 (nt 22–840); Nep1/CG5894, GH03315 (nt 808–1573); Ugt86Di/CG6658, exon 3 (nt 57–707); LpR1/CG4861, exon 6 (316–1373); CG8483, LD39025 (nt 329–1120); CG8780, RE33994 (nt 740–1696); CG9008, GH14910 (nt 327–1442); CG14534, RE71854 (nt 513–1216); CG15489, single exon (nt 38–799); CG17278, SD04019 (nt 13–741); CG30069, LP06813 (3159–3995). Constructs made using an artificial intron as a spacer were dNAB/CG15001, CG15001 exon (1–321); Doc2, RE40937 (663–1104); CG5758, exon 8 (nt 23–417); CG15488, single exon (nt 1–432).

Generation of UAS lines:

Constructs were co-injected with the “wings clipped” helper plasmid (Karess and Rubin 1984) into yw embryos. Transformants were then mapped and homozygous or balanced stocks were generated and used for the GAL4 crosses. If fewer than three independent transformed lines were obtained, one of the existing inserts was mobilized to generate new insertions by crossing to the transposase stock w; Dr/TMS, P{ry[+t7.2]=Delta2-3}99B. The fly lines generated for each construct are detailed in supplemental data at http://www.genetics.org/supplemental/.

GAL4-UAS crosses:

Generally, we maximized GAL4 expression by carrying out crosses at 29° and, if a phenotype was seen, a given cross was repeated at 25° to see the effects resulting from lower activity of the temperature-dependent GAL4-UAS system (Duffy 2002). A minimum of three lines for each misexpression and RNAi construct were used for these analyses. The driver–responder crosses performed and the resulting phenotypes are detailed in supplemental data at http://www.genetics.org/supplemental/.

Fly stocks:

GAL4 driver lines used were Tub-GAL4/TM3 (Lee and Luo 1999), 71B-GAL4 (Brand and Perrimon 1993), 1348-GAL4 (Huppert et al. 1997), 69B-GAL4 (Brand and Perrimon 1993), da-GAL4-UH1 (Wodarz et al. 1995), en-GAL4 (Aza-Blanc et al. 1997), ey-GAL4 [P{GAL4-ey.H}3-8 from the Bloomington Stock Center], ap-GAL4/CyO (Calleja et al. 1996), and Act5C-GAL4 [P{Act5C-GAL4}17bFO1 from the Bloomington Stock Center].

PiggyBac insertion stocks from the Bloomington Stock Center used were CG5758[c01197], an insertion in CG5758 intronic sequence; CG5758[e01537], no sequence information; Ugt86Di[e00862], an insertion upstream of predicted coding sequence. Insertion sites were determined from the reported flanking sequences for the alleles. For CG5758[e01537], no flanking sequence data were available.

In situ analysis:

In situ analysis was carried out as described in Butler et al. (2003). Probes for examining RNA knockdown were generated using PCR to add a 3′ T7 site to a region of a given transcript not included in the corresponding RNAi construct. PCR fragments were gel purified and used as templates in the T7 transcription reactions. The T7 sequence added was 5′-GAATTTAATACGACTCACTATAGG-3′.

RT analysis of CG15488 and CG15489:

RNA was extracted from third instar wing discs using QIAGEN (Chatsworth, CA) RNeasy columns. Reverse transcription was carried out according to standard protocols from Ambion (Austin, TX). Sequence specific to the third exon of the known nubbin transcript [5′-CCACCATTGGCGTCATCGTAATCC-3′] was used as the 3′-primer for reverse transcription. 5′ primers containing sequence from CG15488 [5′-CTGCAACCAGCTCAGATTGCC-3′] or CG15489 [5′-GGCAACGGCAACAGCAACACC-3′] were used to amplify resulting cDNAs in a PCR reaction. Products were cloned into the pDrive vector from the QIAGEN PCR cloning kit and sequenced. The GenBank accession number for the reverse transcription product is DQ375488.

Sequence analysis:

Sequence analysis was performed using sequences, annotations, and tools from these databases: FlyBase (http://flybase.bio.indiana.edu), Ensembl (http://www.ensembl.org), EMBL-EBI (http://www.ebi.ac.uk/tools), SMART (http://smart.embl-heidelberg.de), NCBI (http://ncbi.nih.gov), BeetleBase (http://www.bioinformatics.ksu.edu/beetlebase), and SilkBase (http://papilio.ab.a.u-tokyo.ac.jp/silkbase). The MEME program for repeat analysis (Grundy et al. 1996) was obtained from http://meme.sdsc.edu/meme/website/intro.html.

RESULTS AND DISCUSSION

Functional analysis of uncharacterized genes using the GAL4-UAS system:

We used microarrays to identify 25 transcripts with enriched expression in a wing/hinge sample that did not have known wing expression patterns and confirmed their differential expression using in situ expression analysis (Butler et al. 2003; Figures 1 and 2). Here, 12 of these were selected for functional analysis. This set was chosen by excluding genes that have been characterized genetically (dve, pdm2, β-gal, ana, opa, Doc1, and wgn) and genes predicted to encode chitin-binding proteins (CG32039/CG6469 and CG14301). Transcripts, annotated as separate genes but which we or others in fact found to comprise segments of other genetically characterized genes, were also excluded: RT–PCR analysis showed that CG15489 and CG15488 comprise contiguous exons in an alternative transcript of the wing development gene nubbin (GenBank accession no. DQ375488). Depending on the pattern of RNAi transgene expression, reducing the levels of nubbin transcripts containing these two exons caused effects on wing morphogenesis and polyphasic lethality, indicating a requirement for this alternate transcript during development (supplemental data at http://www.genetics.org/supplemental/). CG15000 and CG15001 are reported to be exons in the dNAB gene (Clements et al. 2003). The remaining 12 genes represent loci for which no genetic data existed prior to the beginning of our analyses.

Figure 1.—

Wild-type expression patterns and visualization of RNAi-mediated knockdown by in situ hybridization. Wing imaginal discs are oriented with anterior to the left and dorsal at the top. For discs expressing RNAi transgenes, an inset diagram indicates the GAL4-driver expression pattern. en-GAL4 is expressed throughout the posterior compartment of the disc and ap-GAL4 is expressed throughout the dorsal compartment. All GAL4/UAS flies are heterozygous for the given driver and responder and grown at 29° to maximize GAL4 activity. (A) Cyp 310a1 wild-type expression. (B) Cyp 310a1 expression in en-GAL4; UAS-Cyp 310a1-RNAi disc. Posterior expression is strongly reduced. (C) Nep1 wild-type expression. (D) Nep1 expression in ap-GAL4; UAS-Nep1-RNAi disc. Dorsal compartment expression is absent (arrowhead). (E) Ugt86Di wild-type expression. (F) Ugt86Di expression in en-GAL4; UAS-Ugt86Di-RNAi disc. Posterior expression is strongly reduced. (G) CG5758 wild-type expression. (H) CG5758 expression in en-GAL4; UAS-CG5758-RNAi disc. Expression in the posterior compartment throughout the dorsal hinge is absent. (I) CG8483 wild-type expression. (J) CG8483 expression in ap-GAL4; UAS-CG8483-RNAi disc. The dorsal clusters of expression in the hinge (arrowhead) are strongly reduced. (K) CG9008 wild-type expression. (L) CG9008 expression in en-GAL4; UAS-CG9008-RNAi disc. Expression is significantly reduced in the posterior compartment of the wing pouch. (M) CG14534 wild-type expression. (N) CG14534 expression in en-GAL4; UAS-CG14534-RNAi disc. The intensity of expression along the posterior wing margin is significantly lower than in the wild-type disc. In some discs expression is completely abolished (data not shown). (O) Doc2 wild-type expression. (P) Doc2 expression in ap-GAL4; UAS-Doc2-RNAi disc. Dorsal compartment expression is unchanged. The disc in the inset shows that there is strong expression of the RNAi transgene in the dorsal compartment (hybridization with a probe that binds transcripts from both the transgene and the endogenous gene). (Q) LpR1 wild-type expression. The gene is expressed at a relatively low level and does not give a strong, clear pattern. (R) LpR1 expression in en-GAL4; UAS-LpR1-RNAi disc. Posterior compartment expression is unchanged. The disc in the inset shows that strong expression of the RNAi transgene in the posterior compartment (hybridization with a probe that binds transcripts from both the transgene and the endogenous gene).

Figure 2.—

Wing phenotypes resulting from RNAi gene silencing. All GAL4/UAS flies are heterozygous for the given driver and responder. (A) Wild-type wing. (B) 69B-GAL4; UAS-CG17278 (25°) wing showing typical blistering phenotype. Inset shows the wild-type CG17278 expression pattern. (C) 71B-GAL4; UAS-CG30069-RNAi (25°) wing showing typical blistering phenotype. Inset shows the wild-type CG30069 expression pattern. (D) Costal area of wild-type wing (29°). The different sections of the costa and the costal vein are labeled: mCo, medial costa; dCo, distal costa; L1, anterior wing margin. (E) Costal area from Tub-GAL4; UAS-CG8780-RNAi (29°) wing blade. Fusion is evident between the dCo and L1 eliminating the gap present in (D). Fusion is also observed between the mCo and dCo regions and there is narrowing and deposition of vein material between the costal and subcostal veins. Inset shows the wild-type CG8780 expression pattern. Arrowheads in D and E indicate the pattern changes.

RNAi experiments were carried out for all 12 genes and ectopic expression was performed for the 9 genes with complete cDNAs available. Each construct was tested with the same set of GAL4 drivers: Tub-GAL4, which directs ubiquitous expression (Lee and Luo 1999); 69B-GAL4, which is expressed in embryonic epidermis and in imaginal tissues (Brand and Perrimon 1993); 71B-GAL4, which is expressed in the wing pouch during the third larval instar and in intervein regions in the pupal wing (Brand and Perrimon 1993; Wessells et al. 1999); and 1348-GAL4, which is expressed in pupal interveins (Huppert et al. 1997). Some additional drivers were used for examining specific developmental effects of certain genes and for assaying the in vivo efficacy of RNAi.

Misexpression phenotypes were seen with eight of the nine genes tested while knockdown of three genes with RNAi caused phenotypes (Table 1; Figures 2 and 3; supplemental data at http://www.genetics.org/supplemental/). These three genes, CG8780, CG17278, and CG30069, are each novel single-copy genes with no previously described roles in development. Two of these, CG8780 and CG17278, showed phenotypes following both misexpression and RNAi. Overexpression of CG30069 was not tested due to the lack of a complete cDNA. The phenotypes observed for the various genes in the misexpression and RNAi experiments included effects on wing development, eye development, notal microchaete patterning, pupal morphogenesis, and eclosion and viability. A summary of these results is shown in Table 1.

TABLE 1.

Gene function assays and resulting phenotypes

Gene	Overexpression phenotype	RNAi phenotype	Wing-disc expression	Function/homology
Cyp310a1	None	None	Figure 1A	Cytochrome P450
Doc2	Extra veins, lethality	NDb	Figure 1O	T-box transcription factor
LpR1	NDa	NDb	Figure 1Q	LDL receptor
Nep1	Wing-expansion defects, lethality	None	Figure 1C	Neprilysin metalloprotease
Ugt86Di	NDa	None	Figure 1E	UDP-glucosyl transferase
CG5758	Wing adhesion/expansion defects, lethality	None	Figure 1G	β-Ig-H3/fasciclin domains
CG8483	Wing adhesion/expansion defects, lethality	None	Figure 1I	SCP domain
CG8780	Localized cell death, lethality	Costal vein patterning	Figure 2E	Unknown
CG9008	Bristle patterning defect, lethality	None	Figure 1K	Aldose epimerase-like
CG14534	Tubby phenotype, lethality	None	Figure 1M	DM5/DUF243 domain
CG17278	Abdominal swelling, melanization, lethality	Wing blistering, lethality	Figure 2B	Similarity to serpins
CG30069	NDa	Wing blistering and adhesion, lethality	Figure 2C	Unknown

ND, not determined.

^aNo full-length cDNA was available.

^bAssays were unable to determine if RNAi construct was active.

Figure 3.—

Nonwing phenotypes caused by overexpression. All GAL4/UAS flies are heterozygous for the given driver and responder. (A) Dorsocentral region of notum from Tub-GAL4 adult at 25°. (B) Dorsocentral region of notum from a Tub-GAL4; UAS-CG9008 adult at 25°. Several microchaete rows are disorganized with bristles displaying altered polarity. (C–E) Pupae raised at 25°. (C) TM6B, Tb/+ pupa showing Tubby phenotype. (D) Wild-type Canton S pupa. (E) da-GAL4; UAS-CG14534 pupa. The da-GAL4; UAS-CG14534 pupa shows a shortened, rounded phenotype similar to that seen in the Tubby mutant pupa. (F) Head from a ey-GAL4 (25°) fly. (G) Head from a ey-GAL4; UAS-CG8780 (25°) fly. Ommatidia are almost completely absent. (H) Abdominal region from a da-GAL4 (25°) male fly. (I) Abdominal region from a da-GAL4; UAS-CG17278 male fly showing typical abdominal swelling.

Ectopic expression of Doc2, Nep1, CG5758, CG8483, CG9008, and CG14534 results in developmental effects, while RNAi does not:

Dorsocross2 (Doc2/CG5187) encodes a T-box domain transcription factor, with similarity to the vertebrate Tbx6 subfamily of T-box genes, located in a complex with two closely related and redundant genes, Doc1 and Doc3 (Reim et al. 2003; Hamaguchi et al. 2004). Our examination of Doc2, which was initiated prior to these studies (Reim et al. 2003; Hamaguchi et al. 2004), complements the published findings. Overexpression of Doc2 protein caused complete lethality when misexpressed with all GAL4 drivers tested, with the exception of a single UAS-Doc2 line that gave a small number of survivors with 1348-GAL4 at 25° (supplemental data at http://www.genetics.org/supplemental/). Embryonic, larval, and pupal lethality, depending on the driver used, were observed. For drivers with broad early embryonic expression such as da-GAL4 and 69B-GAL4, no larval hatching was observed. Survivors from the cross to the 1348-GAL4 driver had ectopic vein material between L3 and L4 and in the region below L5 (data not shown), indicating that ectopic Doc2 can cause specific patterning defects during pupal wing development. These results indicate that Doc2 is toxic when ectopically expressed, consistent with its role as a transcription factor involved in the patterning of multiple tissues during development. RNAi resulted in some lethality when ubiquitously expressed with Tub-GAL4, but the results were inconsistent and no specific developmental effects were observed in any other GAL4-driver crosses (supplemental data at http://www.genetics.org/supplemental/). Experiments testing the UAS-Doc2-RNAi construct suggest that the construct may have been nonfunctional (Figure 1, O and P, and see In vivo analysis of RNAi efficacy).

Neprilysin1 (Nep1) together with three other genes in Drosophila, Nep2/3/4, encodes an M13-metallopeptidase with homology to the human Neprilysin gene (Turner et al. 2001). Mammalian neprilysins show a wide range of substrate specificities and play a role in regulating the action of several peptide signals and other proteins in the extracellular space through degradation of substrates (Turner et al. 2001). In third instar wing discs, Nep1 is expressed in limited areas of the mesopleura and wing hinge (Figure 1C; Butler et al. 2003). Silencing of Nep1 expression by RNAi had no effect on these or other tissues. Overexpression of Nep1 with Tub-GAL4 caused pupal and pharate lethality with no survivors. Crosses with 69B-GAL4, 1348-GAL4, and da-GAL4 resulted in many flies with bent or wrinkled wings, but most striking was that a significant percentage of flies had nonexpanded wings (supplemental data at http://www.genetics.org/supplemental/). Nep1 is predicted to have peptidase activity, so ectopic expression of the protein may cause degradation of some substrate(s) within the extracellular space required to mediate normal wing adhesion and expansion.

CG5758 is predicted to encode a secreted protein that contains two β-Ig-H3/FasI domains (SMART annotations), a domain known to be involved in cell adhesion (Elkins et al. 1990). In addition to other drosophilids, the mosquito and the beetle also contain likely homologs of CG5758 (data not shown). Expression of CG5758 is detected primarily in the dorsal hinge region of third instar wing discs (Figure 1G; Butler et al. 2003). RNAi silencing of CG5758 resulted in a variable effect on viability when expressed with the Tub-GAL4 driver (supplemental data at http://www.genetics.org/supplemental/). However, no defects were seen in derivatives of the dorsal hinge or elsewhere following RNAi with drivers expressed in the wing or more globally (supplemental data at http://www.genetics.org/supplemental/). Overexpression of CG5758 showed a range of developmental effects. Misexpression with Tub-GAL4 and da-GAL4 resulted in widespread pupal lethality at 29°. At 25°, seven of the eight independent UAS-CG5758 lines were completely lethal when crossed to Tub-GAL4. At 25°, misexpression with da-GAL4 resulted in pupal lethality but there were also some survivors. Most of these escapers displayed wing defects, including wrinkling, blistering, and nonexpansion (supplemental data at http://www.genetics.org/supplemental/). These results suggest that while reduction of CG5758 expression does not affect wing development, excess amounts of this putative adhesion protein can disrupt normal morphogenesis and adhesion in the wing.

CG8483 encodes a predicted secreted protein containing a domain variously referred to as an SCP/V5/TPX domain. This domain is found in proteins such as insect venom allergens, reptile toxins, plant pathogenesis proteins, and human GliPR/RTVP-1 (Henriksen et al. 2001). While the function of this domain is unclear, recent work suggests that the human GliPR/RTVP-1 may act as a tumor suppressor with proapoptotic activity (Ren et al. 2002). CG8483 is expressed along the wing margin and in clusters of cells in other regions of the third instar wing disc (Figure 1I; Butler et al. 2003). While RNAi silencing of CG8483 had no discernible effects, misexpression of the gene resulted in a mixture of low-penetrance adult phenotypes and lethality. Tub-GAL4-mediated overexpression at 25° and 29° resulted in nearly complete pupal or pharate lethality. Misexpression with other drivers resulted in variable levels of pharate adult lethality. Adults from some of these crosses showed low numbers of wing adhesion and expansion defects and the presence of blackened tissue behind the labial palps, depending on the GAL4-driver used (supplemental data at http://www.genetics.org/supplemental/). These results indicate that excess CG8483 is toxic during pupal development, although the cause of lethality is not clear. This is the first case in which a Drosophila SCP domain protein has been shown to have any developmental effects and may potentially be useful for examining general aspects of SCP domain function.

The predicted product(s) of the CG9008/BG:DS00797.2 gene displays limited similarity to the active site of known aldose epimerases (SMART and NCBI–CDS annotations). However, CG9008 shows no significant overall similarity to any vertebrate genes or to other Drosophila genes predicted to encode aldose epimerases. CG9008 homologs in mosquito, honeybee, beetle, and silkworm show a high degree of conservation with homologs present in yeast, bacteria, and Arabadopsis, as determined by BLAST analysis (data not shown), indicating that the proteins represent an evolutionarily conserved group. No function has yet been described for these “aldose epimerase-like” genes. Although CG9008 is expressed throughout the third instar wing pouch (Figure 1K; Butler et al. 2003) and appears to be a unique gene, RNAi silencing of expression had no effects. Overexpression of CG9008 using the ubiquitous Tub-GAL4 driver at 29° affected viability, with 2 of 10 UAS-CG9008 lines giving rise to no survivors and 3 other lines showing <5% survival when compared to the control class (supplemental data at http://www.genetics.org/supplemental/). At 25°, misexpression of all UAS-CG9008 lines with Tub-GAL4 resulted in survivors. Adults from these crosses showed patterning effects on the arrangement of notal microchaetes with the rows near the dorsal central region showing various degrees of disorganization and effects on bristle polarity (Figure 3B). Misexpression with other widely expressed drivers showed similar microchaete phenotypes (supplemental data at http://www.genetics.org/supplemental/).

CG14534 is predicted to be a secreted protein containing a DM5/DUF243 domain, an uncharacterized domain found in proteins from mosquito, honeybee, and silkworm as determined by BLAST analysis (data not shown). In third instar wing discs, CG14534 expression is limited to the posterior wing margin (Figure 1M; Butler et al. 2003). RNAi silencing of CG14534 had no effects on the posterior margin or any other structures. Overexpression of CG14534 with the Tub-GAL4 driver at both 29° and 25° resulted in pupae displaying a Tubby-like phenotype of shortened and rounded pupal cases (Figure 3E). At 29°, Tub-GAL4; UAS-CG14534 progeny died during the pharate stage. Pharate adults dissected from their pupal cases showed no obvious developmental defects. At 25°, a small number of escapers were seen, again with no apparent phenotypes other than a Tubby-like appearance. Overexpression using da-GAL4 resulted in different numbers of pupae with a Tubby-like phenotype depending on the UAS-CG14534 transgenic line used with one line tested showing 100% Tubby-like pupae (supplemental data at http://www.genetics.org/supplemental/). While CG14534 overexpression appears to phenocopy the classical dominant Tubby mutant, CG14534 and Tubby map to different regions of the genome, the 2L and 3R chromosome arms, respectively. The Tubby mutation has been localized by recombination mapping to a region of the third chromosome (3–90; Lindsley and Zimm 1992) that is near a cluster of several annotated genes containing DM5/DUF243 domains in a 22-kb region located in the 97C cytological region according to FlyBase cytological maps. However, whether Tubby corresponds to a gene in this cluster and how CG14534 elicits a Tubby-like phenotype remain to be determined.

CG8780, CG17278, and CG30069 are unique genes required for wing development:

From the set of 12 genes tested, 3 showed developmental effects with RNAi-mediated gene silencing: CG8780, CG17278, and CG30069. All three are novel single-copy genes with no previously described roles in development.

CG8780 is required for patterning the costal region of the wing blade:

CG8780, a novel protein with limited similarity to any proteins outside of Drosophila species (data not shown), is expressed in regions of the third instar wing disc that give rise to the costal area of the wing (Figure 2E; Bryant 1975; Butler et al. 2003). Consistent with the expression pattern, RNAi silencing of CG8780 affects the development of the costal region in the adult wing. When induced with Tub-GAL4, da-GAL4, ap-GAL4, and 69B-GAL4 drivers, RNAi caused fusion of the costal vein with the L1 vein, resulting in a contiguous wing margin lacking any separation between the end of the costal vein and L1 (Figure 2E; supplemental data at http://www.genetics.org/supplemental/). Expression with Tub-GAL4 also caused fusion between the normally distinct medial and distal regions of the costal vein (Figure 2E). Thickening of the costal vein is also seen, sometimes to the extent of spanning the region normally found between the costal and subcostal veins of the wing (Figure 2E; supplemental data at http://www.genetics.org/supplemental/). Depending on the UAS-CG8780-RNAi line used, a variable reduction in the ability to fly was observed (data not shown), possibly resulting from a loss of flexibility at the fused junction between the costal vein and L1. None of the crosses demonstrated significant levels of lethality or overt signs that reduction of CG8780 activity affected any other tissue. These results suggest that CG8780 plays a role in patterning the specific region of the wing blade where it is expressed.

The cDNA clone available for CG8780 (RE33994) contains a single nucleotide deletion, which should result in a truncated protein 370 amino acids long vs. 717 amino acids long for the full-length protein predicted from the Drosophila melanogaster genomic sequence and an ortholog identified in D. ananassae. However, comparison of the phenotype from overexpressing this cDNA with similar experiments in D. ananassae suggests that the truncated protein retains function. Yoshida et al. (1994) showed that ectopic expression of a CG8780 ortholog, Om(2D), in the developing eye of D. ananassae results in extensive cell death and a reduced eye phenotype. We found ectopic expression of CG8780 resulted in complete lethality with all of the drivers tested except ey-GAL4 (supplemental data at http://www.genetics.org/supplemental/). Most ey-GAL4; UAS-CG8780 flies died before the pharate stage but those that reached this stage or eclosed showed a nearly complete loss of eye tissue (Figure 3G). The effects of overexpression are like those seen with Om(2D) while showing no resemblance to the RNAi phenotypes, likely ruling out the possibility that the truncated protein acts in a dominant-negative fashion.

The molecular role that CG8780 plays in wing patterning is difficult to predict due to the lack of any clear homologies to known functional domains or proteins in other non-drosophilid species. Its expression pattern and the RNAi results suggest that CG8780 may represent a gene that evolved to play only a very specific, although critical, role in patterning the costal region of Drosophilidae. The results of misexpression in the eye, as well as the lethal effects seen in other crosses, suggest that CG8780 has the potential to mediate cell death or suppress proliferation, processes central to differentiation and the morphogenesis of adult tissues.

CG17278 is required for viability and adhesion of wing surfaces:

CG17278 is predicted to encode a short secreted protein 80 amino acids long that shows similarity to Kazal-type serine protease inhibitors (SMART annotation). CG17278 also shows similarity to predicted Kazal domain proteins from the moth species Manduca sexta, Lonomia oblique, and Bombyx mori (data not shown). In third instar wing discs, CG17278 is highly expressed throughout the developing wing pouch except along the dorsal/ventral boundary where it is downregulated (Figure 2B; Butler et al. 2003). Both misexpression and RNAi-mediated gene silencing showed effects on wing development.

Overexpression using Tub-GAL4 and da-GAL4 at 29° resulted in nearly complete pupal and pharate adult lethality (supplemental data at http://www.genetics.org/supplemental/). At 25°, da-GAL4 crosses to the two lines tested resulted in many adults with wing morphogenesis defects such as bending or mild wrinkling of the wing blade. Most flies also displayed grossly swollen abdomens (Figure 3I). Additionally, many flies displayed dark spotting in the abdomen (supplemental data at http://www.genetics.org/supplemental/).

When CG17278 transcript levels were reduced using RNAi, clear effects on wing adhesion were seen (Figure 2B). Tub-GAL4-directed expression of the three strongest UAS-CG17278-RNAi lines caused pharate adult lethality at 29°. A weaker line resulted in many survivors with ∼70% of the flies showing blistering in at least one wing. At 25°, three of the four lines gave rise to survivors with wing blistering. Crossing the stronger transgenic lines to 69B-GAL4 and 71B-GAL4 also resulted in wing blistering and expansion defects with the wings partially inflated to different degrees (supplemental data at http://www.genetics.org/supplemental/).

CG17278, consistent with its high expression level in the wing pouch (Figure 2B), is required for adhesion of the wing surfaces and expansion. Although similarity to Kazal-type serine protease inhibitors suggests that CG17278 may be a serine protease inhibitor, whether it displays such activity and which molecules it may interact with during wing development remain to be determined. In addition to wing development, CG17278 is also required for viability, indicating a necessary role in other processes. One study has found that CG17278 expression is induced severalfold soon after bacterial infection (De Gregorio et al. 2001). These results suggest that CG17278 could play a role in activating an immune response that possibly underlies the melanization and apparent production of excess lymph seen in the abdomen when CG17278 was overexpressed.

CG30069, a protein containing a large tandem repeat region, is required for viability and adhesion of wing surfaces:

CG30069 is expressed throughout the third instar wing pouch except within the wing margin and presumptive vein regions (Figure 2C; Butler et al. 2003). RNAi-mediated silencing demonstrated a requirement for CG30069 in mediating adhesion between the two surfaces of the wing blade as well as an earlier requirement during embryonic development. Tub-GAL4-driven misexpression with all UAS-CG30069 lines resulted in embryonic lethality. RNAi with da-GAL4 resulted in embryonic and early larval lethality (supplemental data at http://www.genetics.org/supplemental/). RNAi with 71B-GAL4 and 1348-GAL4 showed some degree of lethality at 29° and 69B-GAL4-driven RNAi was completely lethal for all lines tested. Among the survivors of crosses at 29° and 25°, wing blistering was widespread especially in the relatively stronger RNAi lines (Figure 2C; supplemental data at http://www.genetics.org/supplemental/). Flies also showed other apposition defects such as curved/bent wings and wings that were partially inflated.

The predicted transcript for CG30069 encodes a large protein with >3600 residues. While BLAST analysis identifies potential homology to some proteins of known function (e.g., RNA pol II), these results appear to be due to the amino acid composition of a large tandem repeat array in CG30069 and not to any specific sequence homology. However, BLAST analysis of genomic sequences from several insects identified genes encoding orthologous proteins in mosquito, flesh fly, honeybee, silkworm, and beetle (data not shown). The fact that homologs of CG30069 have been found in all these insects suggests that it may play a critical role in wing development or in some other common developmental process for all insects. Analysis using the MEME program to detect motifs (Grundy et al. 1996) in CG30069 and its mosquito and flesh fly homologs (NCBI accession nos. EAA13867 and BAB16608, respectively) identified a 29-amino-acid unit repeated up to 80 times in a contiguous array in these proteins. The consensus sequences for the repeats from the three proteins are nearly identical (Figure 4), with some repeat-to-repeat variation within each protein. When the entire set of repeats among these three proteins are examined, the periodicity of the repeats, 29 residues, and the sequence Asp-Asn-Leu-X-X-Asp/Glu-Gly (D-N-L-X-X-D/E-G) are nearly invariant with only two repeats in the Drosophila and mosquito sequences showing variation in size and/or sequence (data not shown). Sequence conservation is also observed between different insects for some regions of the protein between the end of the repeat region and the C-terminal (data not shown).

Figure 4.—

CG30069 repeat motif alignment. Alignment of consensus motifs for the 29-amino-acid repeats found in CG30069 (Dm) and homologous proteins from flesh fly (Ff) and mosquito (Ag). Amino acids are indicated by an asterisk for fully conserved residues and by a colon for similar residues. Consensus motifs were determined using MEME analysis.

In the flesh fly (Sarcophaga peregrina), the homolog of CG30069 designated as the “210-kDa protein,” was identified as a substrate for the C1-peptidase cathepsin L and localized by antibody staining to the basal lamina of imaginal discs (Homma and Natori 1996; Fujii-Taira et al. 2000). While no secretion signals are evident in the predicted CG30069 protein, the flesh fly results suggest that the Drosophila protein may be a basal laminar protein and perhaps mediate adhesion between the apposing surfaces of the wing pouch. Adhesion of the wing surfaces during pupal development and metamorphosis is a multi-step process, with the dorsal and ventral surfaces adhering and disengaging before finally becoming stably apposed (Brown et al. 2000). Molecules such as integrins on the basal surfaces of the dorsal and ventral wing pouch play a necessary role in mediating this process and reduction of integrin function can lead to blistering in adult wings (Brown et al. 2000), which is similar to the effects seen with expression of UAS-CG30069-RNAi transgenes. Whether CG30069 interacts directly with integrins or other known molecules during adhesion is unknown. The early lethality observed indicates that CG30069 is also required during embryonic development.

In vivo analysis of RNAi efficacy:

Although RNAi assays were performed for all 12 genes, only the 3 genes described above had specific phenotypes. This lack of effect could be due to a number of reasons: the RNAi construct used was inactive, the protein encoded by a given transcript might play no significant role in development, or the gene is functionally redundant. To test the first possibility, in vivo analysis was carried out using in situ hybridization and genetic assays.

In situ hybridization was used to determine if the endogenous transcript was downregulated by the RNAi transgene. The nine genes without RNAi phenotypes were examined (Figure 1). RNA probes that did not hybridize to the RNAi trigger sequences were used to monitor endogenous gene expression (materials and methods). Clear knockdown of transcript levels was seen for seven genes: Cyp310a1, Nep1, Ugt86Di, CG5758, CG8483, CG9008, and CG14534 (Figure 1, A–N). No obvious reductions were seen in transcript levels for Doc2 or LpR1 (Figure 1, O–R) even though the RNAi transgenes were expressed robustly (insets in Figure 1, P and R). This suggests that the RNAi constructs were either nonfunctional or unable to lower RNA levels sufficiently to be recognized by this assay.

A genetic approach was also used to test the activity of the RNAi transgenes. Six genes had lethal overexpression phenotypes but no effects from RNAi. Thus we tested the ability of the RNAi transgenes to suppress lethality associated with overexpression. The genes tested were Doc2, Nep1, CG5758, CG8483, CG9008, and CG14534. The UAS-Doc2-RNAi transgene was unable to suppress lethal and visible phenotypes caused by misexpression of Doc2 (Table 2). This finding and the inability to reduce transcript accumulation suggest that the Doc2 RNAi transgene is ineffective. In contrast, due to their ability to suppress overexpression effects (Table 2), we conclude that the RNAi transgenes for Nep1, CG5758, CG8483, CG9008, and CG14534 were functional at least to some degree. As a control, we also tested whether co-expression of UAS-GFP had any rescuing ability. This was to exclude the possibility that suppression was due to a titration of GAL4 activity caused by introducing an additional transgene. Overexpression phenotypes were unchanged in the presence of UAS-GFP (supplemental data at http://www.genetics.org/supplemental/).

TABLE 2.

RNAi transgenes suppress lethal overexpression phenotypes

Gene	Overexpression	RNAi	Overexpression/RNAi
Transgenes expressed with Tub-Gal4
Doc2a	Extra veins, few survivors	No effect	Same as overexpression
Nep1	Pupal lethal	No effect	Adult viable
CG5758	Pupal lethal	No effect	Adult viable
CG8483	Pupal lethal	No effect	Adult viable
CG9008	13% survival, bristle phenotype	No effect	30% survival, bristles normal
CG14534	Pupal lethal, Tb-like pupae	No effect	Adult viable, fewer Tb-like pupae

^aDoc2 was assayed using 1348-GAL4.

While RNAi is a powerful means to reduce gene expression, only a null mutation can provide a definitive indication of gene function. There are currently no alleles other than wild type for the genes Nep1, CG8483, CG9008, and CG14534. Currently, the Bloomington Stock Center has two PiggyBac insertion stocks for CG5758 and one for Ugt86Di (see materials and methods). We tested these over appropriate deficiencies and also made trans-heterozygotes for CG5758. All crosses yielded viable adults displaying no phenotypes (data not shown). While these findings are in keeping with our results from RNAi, the insertion mutants have not been characterized in detail and additional alleles will need to be examined before a definitive conclusion can be made that these genes do not have loss-of-function phenotypes.

Redundancy may mask phenotypes:

We effectively reduced the function of 10 genes using RNAi but only 3 showed phenotypes. Two of the 7 genes without RNAi phenotypes, CG9008 and CG5758, appear to be unique genes in Drosophila on the basis of BLAST analysis, although each contains domains with limited similarity to known gene families. Even though these genes are expressed specifically, they may have no physiological role in the wing or could be functionally redundant with proteins displaying no clear molecular similarity. The remaining 5 genes, Cyp 310a1 (cytochrome P450), Nep1 (Turner et al. 2001), Ugt86Di (Luque and O'Reilly 2002), CG8483 (SMART/Ensembl annotations), and CG14534 (SMART/Ensembl annotations), are members of multi-gene families. Therefore, it is possible that effects from loss of function are masked by the activity of another family member. In keeping with this idea, at least one other gene from each of these families is expressed at similar levels in the wing pouch according to our microarray data: 9 cytochrome P450 genes; the neprilysins Nep3 and Nep4; UDP-glucosyltransferase Ugt86Da; SCP domain proteins Ag5r2 and CG5106/scpr-C; DM5/DUF243 domain proteins CG14254, CG14643, CG8986, and CG5812/GCR(ich). These data are available from the Gene Expression Omnibus database (Butler et al. 2003). To test the potential effects of redundancy, it will be necessary to simultaneously reduce the function of multiple genes in a given family.

Concluding remarks:

Reverse genetics was used to assay the function of 12 uncharacterized genes with expression in the developing Drosophila wing. Ten genes showed effective silencing with RNAi but only 3 of these had resulting pattern defects. Most of the genes failing to show defects are members of gene families, suggesting that functional redundancy will mask phenotypes. The findings highlight the difficulty whereby the function of a large number of Drosophila genes, not already discovered by traditional forward genetics, may prove refractive to single-gene analysis. The 12 genes were selected for analysis by virtue of high specific expression in the wing region (Butler et al. 2003); however, no genes, even those for which RNAi produced a phenotype, had a dramatic effect on wing development. This may reflect the fact that these genes belong to a large category of Drosophila genes that have not been identified by mutation because they have subtle roles. We conclude that transcriptional profiling followed by RNAi knockdown is an efficient way to analyze genes expressed in a given tissue but the approach is limited to some extent by the efficacy of RNAi and gene redundancy. Furthermore, it is likely that in many cases resulting phenotypes will be subtle.