A screen for dominant mutations applied to components in the Drosophila EGF-R pathway

Abstract

The Drosophila epidermal growth factor receptor (EGF-R) controls many critical cell fate choices throughout development. Several proteins collaborate to promote localized EGF-R activation, such as Star and Rhomboid (Rho), which act sequentially to ensure the maturation and processing of inactive membrane-bound EGF ligands. To gain insights into the mechanisms underlying Rho and Star function, we developed a mutagenesis scheme to isolate novel overexpression activity (NOVA) alleles. In the case of rho, we isolated a dominant neomorphic allele, which interferes with Notch signaling, as well as a dominant-negative allele, which produces RNA interference-like flip-back transcripts that reduce endogenous rho expression. We also obtained dominant-negative and neomorphic Star mutations, which have phenotypes similar to those of rho NOVA alleles, as well as dominant-negative Egf-r alleles. The isolation of dominant alleles in several different genes suggests that NOVA mutagenesis should be widely applicable and emerge as an effective tool for generating dominant mutations in genes of unknown function.

The Drosophila epidermal growth factor (EGF)-receptor tyrosine kinase (EGF-R) controls a large array of cell-fate choices throughout the life cycle (1, 2) and can be activated by multiple ligands (3). Among them, Vein is directly secreted to signal to adjacent cells (4), whereas other EGF ligands such as Spitz (Spi) (5) and Gurken (Grk) (6) are similar to the human transforming growth factor α and are initially expressed as membrane-bound precursors. Numerous studies have provided corroborating evidence that these latter precursor ligands are initially inert and depend on two accessory membrane proteins, Rhomboid (Rho) and Star, to be processed into active diffusible forms (7–14). Rho is a predicted seven-pass transmembrane protein (15), and Star is predicted to be a type II single-pass transmembrane protein predominantly localized in the endoplasmic reticulum (ER) (9, 16), which acts as an obligate partner of Rho to activate EGF-R signaling in a cell nonautonomous fashion (10). Recent studies show that Star is necessary for Spi to translocate from the ER to the Golgi apparatus, where it is directly cleaved by Rho, a novel type of intramembrane serine protease (13, 14, 17).

Unlike the Egf-r, spitz (spi), and Star genes, which are expressed ubiquitously in most epidermal cells, rhomboid (rho) is expressed in a highly localized and dynamic pattern (15) that correlates with the in situ activation pattern of mitogen-activated protein kinase (MAPK), an essential downstream component of all tyrosine kinase receptors (18–20). This latter observation suggests that Rho provides the appropriate restricted spatial and temporal activation for membrane-bound EGF ligands. A good example of the localized activity of Rho is provided by the wing disc, in which the restricted expression of rho in longitudinal stripes controls the commitment of these cells to the vein fate through the activation of EGF-R/MAPK signaling. Thus, rho^ve mutants, which fail to express rho in vein primordia, lack sections of veins, whereas ubiquitous ectopic expression of rho converts the entire wing blade into a single solid vein (11, 21, 22).

In this report, we present an overexpression-based F₁ mutagenesis scheme applied to the rho, Star, and Egf-r genes. We efficiently recovered distinct dominant alleles of these structurally unrelated genes, indicating that novel overexpression activity (NOVA) mutagenesis is likely to become a widely applicable method for generating dominant alleles of various types of genes.

Materials and Methods

Fly Stocks and Crosses.

Upstream activation sequence (UAS)-rho and pUAS-Star stocks were described in ref. 10. The pUAS-Egf-r and pUAS-Egf-r^DN1 stocks were provided by Allan Michelson (Brigham and Women's Hospital, Boston). All crosses were performed at 25°C.

Molecular Analysis of Mutations in rho, Star, and Egf-r Transgenes.

For analysis of Δ2–3-induced mutants, sets of primers corresponding to sequences in the pUASt vector, the rho cDNA, or the Star cDNA were used to search for alterations in the various mutants by standard PCR or inverse PCR, with the Long Expand PCR system (Roche Molecular Biochemicals catalogue no. 1681842). Details are available on request.

Immunoblot Analysis of Mutant Rho Proteins.

The anti-Rho serum (26) was used for immunoblotting at 1/1,000 dilution, in 0.25% Tween 20, 1% milk in PBS. Secondary antibodies (horseradish peroxidase-coupled anti-rabbit IgG, Jackson ImmunoResearch catalogue no. 111-035-003) were used at 1/5,000 dilution. Chemiluminescent detection was performed by using the Supersignal kit (Pierce catalogue no. 34080).

In Situ Hybridization and Histochemistry.

In situ hybridization, histochemistry (27), and detection of MAPK activation (10) were performed as described. Anti-Cut antibodies were obtained from the Developmental Studies Hybridoma Bank (University of Iowa).

Northern Analysis.

Northern blots were prepared by using standard methods, hybridized with a horseradish peroxidase-labeled rho RNA probe, and detected by using the Chemiluminescent system CDP-Star (Amersham Pharmacia catalogue no. RPN3690).

Results and Discussion

Isolation of rho Overexpression Alleles.

Rho has been recently defined as a novel type of intramembrane serine protease that cleaves the membrane-bound ligand Spi (mSpi) in the Golgi apparatus before its release into the extracellular space. Although a catalytic domain and key residues essential for proteolytic activity have been defined in Rho (14), it is not known which other parts of the protein fulfill regulatory functions and interact with functional partners such as Star or other components. Like many proteins recently identified in the Drosophila genome project, Rho does not contain any signature domains that could give clues about possible protein–protein interactions. In this context, we developed a new strategy to screen for potential NOVA alleles of rho, which might provide additional insights into the mode of Rho action.

The NOVA method makes use of the two-component GAL4/UAS expression system (23). The principle of this scheme is to expose a UAS transgene of interest to mutagenesis, express the mutated transgene in the F₁ progeny at high levels in a desired pattern by using a strong GAL4 driver, and then screen for novel visible phenotypes. In the present case, we exposed a UAS-rho transgene to the Δ2–3 transposase, which induces rearrangements such as small deletions, inversions, and duplications within or adjacent to P element insertions as a byproduct of gap repair after excision events (24, 25). We crossed individuals carrying the potentially mutagenized UAS-rho* transgene to flies carrying a strong ubiquitous wing-specific GAL4 driver (MS1096GAL4, referred to as wing-GAL4 hereafter) and then screened for novel dominant phenotypes in approximately 15,000 F₁ progeny of this second cross (Fig. 1A). Among the individuals of the relevant wing-GAL4>UAS-rho* genotype, most flies exhibited the all-vein phenotype resulting from strong misexpression of unaltered wild-type UAS-rho (Fig. 1C). A smaller fraction of the F₁ progeny had a wild-type phenotype, likely to reflect precise pUAS-rho excision events (Fig. 1B). In addition, we recovered two individuals exhibiting distinct dominant NOVA phenotypes. The first mutant has small blistered wings with thickened veins and margin defects (Fig. 1D Inset). In the most-affected individuals, wings are virtually absent. Because loss-of-margin structures and great reduction in wing size are not phenotypes observed in loss-of-function rho mutants or in flies misexpressing wild-type rho, we refer to this neomorphic rho mutant as rho^Neo. The second rho mutant had missing distal portions of wing veins (Fig. 1E) typical of rho loss-of-function situations (e.g., rho^ve, Fig. 1G). We considered this NOVA mutant to be a likely dominant-negative form of rho (rho^DN).

Figure 1.

A NOVA screen uncovers novel activities of a UAS-rho transgene. (A) NOVA mutagenesis scheme. (B–G) Wings of the following genotypes: (B) wild type; (C) wing-GAL4>UAS-rho^wt; (D) wing-GAL4>UAS-rho^Neo. Insets here and in F show interrupted margin (bracket) in stronger examples. (E) wing-GAL4>UAS-rho^DN; (F) wing-GAL4>UAS-rho^Neo′; (G) rho^ve. (H–K) Structures of wild-type and mutant pUAS-rho constructs. Blue boxes indicate the transmembrane domains of the Rho protein. Triangles indicate the inverted terminal repeats of the P element. (H) Wild-type pUAS-rho^wt; (I) pUAS-rho^Neo; (J) pUAS-rho^DN; (K) immunoblot analysis of protein extracts from wild-type (lane 1), pHS-rho (lane 2), pHS-GAL4>UAS-rho^Neo (lane 3), and pHS-GAL4>UAS-rho^DN (lane 4) adult flies submitted to a 1-h heat shock at 38°C. (L) Northern blot of mRNA extracted from wild-type (lane 1), pHS-GAL4>UAS-rho^wt (lane 2), and pHS-GAL4>UAS-rho^DN (lane 3) adult flies following a 1-h heat shock.

Characterization of the Molecular Lesions in rho NOVA Mutants.

As a first step in analyzing the new rho NOVA alleles, we confirmed that the rho^Neo and rho^DN phenotypes were GAL4 dependent. We then used combinations of PCR primer sets to amplify rho sequences within the pUAS vector and/or surrounding genomic sequences to identify molecular lesions responsible for the dominant rho^Neo and rho^DN activities. This analysis revealed that the UAS-rho^Neo mutant carries a 5-kb deletion removing parts of both the rho cDNA and the adjacent white⁺ marker gene. This rho^Neo mutant construct is predicted to encode a truncated protein containing the first 140 amino acids of Rho (including the N terminus, TM1, half of the first loop) and 10 amino acids encoded out-of-frame by the 3′ end of the white⁺ gene, which are fused to the rho coding sequence (Fig. 1I; compare with wild-type structure in Fig. 1H).

Consistent with the predicted structure of Rho^Neo, immunoblotting of protein extracts from heat-induced HS-GAL4>rho^Neo flies by using an N-terminal specific anti-Rho antibody (26) revealed high levels of a shorter-than-normal Rho protein (27 kDa instead of 43 kDa for the wild-type species; Fig. 1K, lanes 3 and 2, respectively). Interestingly, this Rho^Neo protein comigrates with a smaller Rho protein species that is consistently observed on heat induction of full length Rho expression (Fig. 1K, lane 2).

Misexpression of components in the EGF-R pathway using the wing-GAL4 driver does not typically result in margin defects. It was therefore important to verify that the observed rho^Neo phenotype resulted from misexpression of the truncated rho mutant transgene rather than from some adjacent genomic sequence. Although the pUASt vector does not activate expression of endogenous genes efficiently (28), a general potential caveat to the NOVA method is that GAL4-dependent phenotypes could occasionally result from misexpression of an unrelated gene near the chromosomal site of a pUAS insertion. To address this concern, we cloned a PCR product containing the truncated cDNA of the rho^Neo gene back into the pUASt vector and retransformed this construct (named UAS-rho^Neo′) into flies. Misexpression of UAS-rho^Neo′ with the wing-GAL4 driver resulted in the same phenotype (Fig. 1F) as that obtained with the initial UAS-rho^Neo isolate (Fig. 1D). We conclude that the truncated rho^Neo allele is indeed responsible for the observed wing phenotypes.

Molecular analysis of the UAS-rho^DN mutation revealed an inverted duplication of the UAS and rho sequences with a spacer portion consisting of rho sequences (Fig. 1J). This UAS-rho^DN mutant is predicted to generate an RNA with a hairpin structure, which potentially could exert an RNA interference (RNAi) effect (see below). Consistent with this prediction, we observed a larger rho transcript (≈3.9 kb) in flies expressing rho^DN (Fig. 1L, lane 3) than in those expressing wild-type rho (2.3 kb; Fig. 1L, lane 2).

rho^Neo Interacts with the Notch Pathway.

Ubiquitous expression of rho^Neo in the wing primordium disrupts formation of margin structures, suggesting that it interacts with a pathway involved in inducing margin cell fates or differentiation. To identify phenotypes specifically attributable to defects in wing margin cell fates, we expressed UAS-rho^Neo with the margin-specific vestigial-GAL4 driver and observed a more pronounced loss of margin structures (Fig. 2A) resembling that associated with reduction of wingless (wg) or Notch function (Fig. 2B). In addition, ubiquitous misexpression of UAS-rho^Neo results in thickened veins (Fig. 1D), another signature phenotype of Notch pathway mutants (Fig. 2B). Although ectopic or thickened veins can also result from misexpression of full-length rho (22, 29), it is unlikely that rho^Neo function is mediated by activation of EGF-R signaling or deregulated Rho protease activity, because this mutant lacks all sequences necessary for the proteolytic function of Rho (13, 14). It has also been reported that a dominant-negative form of EGF-R misexpressed in the margin causes notching (30). Because loss-of-function rho⁻ or Egf-r⁻ clones do not result in margin phenotypes (10, 31, 32), however, and because such loss-of-function mutations do not generate ectopic veins as observed in wing-GAL4>rho^Neo wings, it seems unlikely that rho^Neo functions by a dominant-negative mechanism specific to the EGF-R pathway.

Figure 2.

rho^Neo interferes with Notch signaling. (A–D) Wings of the following genotypes: (A) vg-GAL4>UAS-rho^Neo; (B) N^55e11/+; (C) N^Ax1/Y (male); (D) N^Ax1/+ wing-GAL4>UAS-rho^Neo (female); (E) wingless expression along the wing margin (M) of a wild-type third larval instar imaginal disk; (F) absence of wg expression in wing-GAL4>UAS-rho^Neo discs (arrow, margin primordium); (G) Cut expression along the wing margin (M) of a third instar imaginal disk; (H) reduced Cut expression in a wing-GAL4>UAS-rho^Neo wing disk (arrow, margin primordium).

We tested the hypothesis that rho^Neo interferes with Notch signaling by assaying the expression of two downstream targets of Notch, wg and cut. In wild-type discs, wg and Cut are expressed in a narrow row of margin cells at the dorso-ventral boundary (Fig. 2 E and G; refs. 33 and 34). In wing discs ubiquitously expressing the UAS-rho^Neo construct, expression of both wg and Cut was abolished or significantly reduced (Fig. 2 F and H), suggesting that rho^Neo reduces Notch signaling. Consistent with this proposal, all rho^Neo phenotypes are strongly suppressed by one copy of the activated N^Ax1 allele in heterozygous females (Fig. 2D, compare with Fig. 1D). This N^Ax1 allele on its own causes loss of veins when homozygous or hemizygous (Fig. 2C; ref. 21) but has little effect when heterozygous (data not shown).

One explanation for the rho^Neo mutant phenotype is that it interferes with other Rho-related proteins (35). Consistent with this possibility, a human Rhomboid-like protein has been reported to bind the Notch-activating protease Presenilin (36), suggesting a further connection between Rho-related proteins and Notch signaling. Additional analysis will be required to determine whether Rho^Neo interferes directly with the Notch pathway (e.g., by binding to a component of the Notch pathway and blocking its activity) or indirectly by impinging on other pathways interacting with Notch to promote margin development and vein formation.

An important question regarding the biological significance of Rho^Neo is whether a similar Rho fragment is generated in vivo and mediates a component of endogenous rho activity. It is relevant in this regard that an N-terminal Rho fragment of nearly identical size to rho^Neo is consistently produced in vivo when full length Rho is overexpressed (Fig. 1K) and is also present in wild-type embryo extracts (A.G., unpublished observations). One potential biological role for this N-terminal fragment of Rho could be the suppression of Notch activity in cells expressing high levels of Rho. Such negative feedback might aid in the creation of mutually exclusive domains of Notch and EGF-R activity in some developmental settings, as has been recently suggested for the partition of the eye disk into antenna and eye fields (37) and for bristle differentiation (38).

rho^DN Inhibits EGF-R Activity by an RNAi-Like Mechanism.

To confirm that rho^DN acts via a dominant-negative mechanism by inhibiting endogenous rho expression and hence EGF-R activity, we examined in situ activation of MAPK in wing discs expressing UAS-rho^DN. In the wing disk, MAPK activation revealed by an antibody specific for the diphosphorylated MAPK (18) is restricted to vein and margin primordia and depends on localized rho expression and subsequent activation of EGF-R signaling (Fig. 3A; refs. 10, 18, and 39). In line with rho expression defining the domain of EGF-R/MAPK signaling, ubiquitous activation of MAPK is observed in response to misexpression of UAS-rho (Fig. 3B; ref. 10). In contrast, overexpression of UAS-rho^DN abolishes MAPK activation in both vein and margin primordia (Fig. 3C), consistent with rho^DN inhibiting endogenous rho expression and subsequent EGF-R activation.

Figure 3.

rho^DN functions by an RNAi-like mechanism. (A) MAPK activation in a wild-type third instar imaginal disk in longitudinal vein primordia (L2–L5) and wing margin (M). (B) MAPK activation in a wing-GAL4>UAS-rho^wt disk. (C) Lack of MAPK activation in a wing-GAL4>UAS-rho^DN disk. (D) rho RNA expression in a wing-GAL4>UAS-rho^wt disk. (E) Undetectable rho RNA expression in a wing-GAL4>UAS-rho^DN disk. Endogenous rho expression (Inset) is also lost. (F) rho expression in a wing-GAL4>UAS-rho^wt/UAS-rho^DN disk. (G) Rho protein expression in a GAL4>UAS-rho wing disk. (H) Strong reduction of Rho expression in a wing-GAL4>UAS-rho/UAS-rho^DN disk. (I) wing-GAL4>UAS-rho wing. (J) wing-GAL4>UAS-rho^wt/UAS-rho^DN wing.

In support of rho^DN functioning by an RNAi-like mechanism (40), Rho protein could not be detected in extracts from the heat-induced UAS-rho^DN flies (Fig. 1K, lane 4). In addition, UAS-rho^DN transgene-derived RNA expression was nearly undetectable by in situ hybridization in wing-GAL4>UAS-rho^DN wing imaginal discs (Fig. 3E), consistent with the formation of double-stranded RNA hairpin structures that are then degraded to 21- to 23-nt fragments, which would be inaccessible to hybridization. Critically, endogenous rho expression in vein and margin primordia (Fig. 3E Inset) was also absent in these discs, indicating that rho^DN interferes with the expression or stability of the endogenous rho mRNA. We also coexpressed a wild-type UAS-rho transgene with UAS-rho^DN and observed a significant overall reduction in the level of the wild-type rho mRNA (Fig. 3F, compare with D) and only a faint trace of Rho protein staining (Fig. 3H) relative to that produced by the wild-type UAS-rho transgene alone (Fig. 3G). The fact that Rho protein levels were more severely reduced than rho RNA levels in discs coexpressing rho and rho^DN suggests that rho^DN compromises translation of the wild-type rho RNA as well as reducing its stability. In line with these various observations, the all-vein phenotype of wings misexpressing wild-type UAS-rho (Fig. 3I) appeared almost completely suppressed by coexpression with UAS-rho^DN (Fig. 3J). Cumulatively, these data indicate that rho^DN acts by inhibiting the activity of rho, most likely by promoting the degradation of its RNA and blocking its translation. This example illustrates the potential utility of NOVA mutagenesis for creating an RNAi version of a gene of interest by purely genetic means, as an alternative to the time consuming and often problematic construction of RNAi inserts by in vitro engineering.

Isolation of NOVA Alleles of Star.

Because novel dominant rho alleles were readily generated by NOVA mutagenesis, we wondered whether this scheme could be successfully applied to other UAS transgenes of interest. Because Star collaborates with rho and has no signature domain indicative of its function, it too is a good candidate for NOVA mutagenesis. Overexpression of wild-type Star typically results in no phenotype in the wing other than faint ectopic vein material near longitudinal veins (Fig. 4A). We submitted a UAS-Star transgene to the Δ2–3 transposase and used the same wing-GAL4 driver to misexpress mutagenized Star insertions at high levels. In this screen of ≈15,000 progeny, we recovered two types of GAL4-dependent dominant mutant alleles, referred to as Star^DN and Star^Neo, which resulted in phenotypes nearly identical to those produced by misexpression of rho^DN and rho^Neo. Thus, ubiquitous misexpression of Star^DN results in vein truncation (Fig. 4 B and C) similar to that observed in wings lacking Star activity (10), whereas misexpression of Star^Neo alleles causes reduction in wing size, thickened veins, blisters, and strong disruption of margin structures (Fig. 4 G and H). Consistent with Star^DN acting in a dominant-negative fashion, coexpression of this mutant with wild-type Star suppresses the vein-loss phenotype of Star^DN (Fig. 4D, compare with C). The nearly identical phenotypes produced by NOVA alleles of the rho and Star transgenes provide further evidence for the intimate functional relationship between rho and Star (10).

Figure 4.

Generation of dominant-negative and neomorphic Star mutants. (A–H) Wings of the following male genotypes: (A) wing-GAL4>UAS-Star; (B) wing-GAL4>UAS-Star^DN/UAS-Star^DN; (C) wing-GAL4>UAS-Star^DN/+; (D) wing-GAL4>UAS-Star^DN/UAS-Star^wt; (E) wing-GAL4>UAS-Star^DN/UAS-rho^wt; (F) wing-GAL4>UAS-Star^DN/UAS-m-grk(2X); (G) wing-GAL4>UAS-Star^Neo1; (H) wing-GAL4(2X)>UAS-Star^Neo2/UAS-Star^Neo2. (I–K) Structure of wild-type and mutant pUAS-Star^wt constructs. (I) pUAS-Star^wt, the blue box indicates the single transmembrane domain of Star. (J) pUAS-Star^DN, a 1.9-kb inversion with breakpoints in both the Star and 3′ untranslated simian virus 40 sequences of pUASt results in a C-terminal truncation of the Star protein. (K) pUAS-Star^Neo1and pUAS-Star^Neo2.

Molecular analysis of one Star^DN mutation revealed that it is associated with an inversion, which has breakpoints mapping within the Star coding sequence and its 3′ untranslated regions. As a result, the Star^DN transgene is predicted to encode a shortened protein with a large C-terminal truncation preceding the single transmembrane domain of Star at amino acid 124 (Fig. 4J, compare with wild-type structure in I). Star is a predicted type II membrane protein with a cytoplasmic N terminus, suggesting that the Star^DN peptide would be free in the cytoplasm. The C terminus of Star is required for binding to mSpi (39) and promoting its transport from the endoplasmic reticulum to the Golgi (13). Because Star^DN lacks the sequences necessary for interaction with mSpi, it is unlikely that it acts by titrating out Spitz-like ligands in an unproductive interaction. Consistent with this expectation, coexpression of Star^DN with mSpi or mGrk did not restore a wild-type phenotype but rather resulted in increased vein-loss (Fig. 4F). It seems more likely, therefore, that the N-terminal region of Star normally interacts with another partner required for Spi maturation (e.g., Rho), and that the Star^DN fragment binds and sequesters this partner in an inactive complex. In support of this view, coexpression of Star^DN with rho efficiently suppresses rho-induced ectopic veins (Fig. 4E). Star may thus function in a multimeric complex that includes mSpi and Rho and/or some other component. According to this model, coexpression of Star^DN with mSpi or mGrk would result in a greater titration of this endogenous factor and in an enhanced phenotype, as observed in Fig. 4F.

We also analyzed the structures of two Star^Neo mutants and found lesions mapping virtually to the same location. In both cases, a deletion fused C-terminal sequences of Star with the white marker. As a consequence, Star^Neo1 and Star^Neo2 encode Star proteins lacking 15 and 17 C-terminal amino acids, respectively, followed by different residues encoded by frameshift fusions to distinct white sequences (Fig. 4K). As in the case of the rho^Neo mutant, overexpression of Star^Neo1 resulted in loss of wg and Cut expression in the margin primordia of third instar larval discs (data not shown). That Star^Neo2 is somewhat weaker than Star^Neo1 in inducing Notch-like phenotypes may result from the different C-terminal amino acids encoded by white sequences or may reflect a function of the two amino acids present in Star^Neo1 that are missing in Star^Neo2.

The strong similarity between rho^Neo and Star^Neo phenotypes suggests that they arise from interference with a common process. As a recent report shows that Star and Rho act sequentially rather than simultaneously (13), it is possible that the component(s) interacting with rho^Neo and Star^Neo act in the Golgi apparatus, where Star may hand over mSpi or other substrates to Rho for the final step of maturation.

Recovery of Distinct Classes of Dominant-Negative Egf-r Alleles in NOVA Screens.

The Egf-r gene is another good test case for NOVA analysis, because in vivo generated NOVA alleles could be compared with an existing dominant-negative Egf-r (Egf-r^DN1) construct created by in vitro genetic engineering (42). Strong ubiquitous expression of Egf-r^wt in the wing results in moderate ectopic vein formation (Fig. 5A), whereas overexpression of Egf-r^DN1 results in small wings and vein loss (Fig. 5B). EGF-R^DN1 is a truncated version of the EGF-R lacking the cytoplasmic kinase domain (Fig. 5H, compare with Egf-r^wt in G) and has been proposed to act primarily by forming nonfunctional heterodimers with endogenous wild-type EGF-R chains (43). We submitted the UAS-Egf-r^wt transgene (Fig. 5G) to mutagenesis using either Δ2–3 transposase or ethyl methanesulfonate (EMS), and crossed the mutagenized males to wing-GAL4 females in two screens of ≈4,000 individuals each. We recovered new putative dominant-negative Egf-r alleles in both screens, which we refer to as Egf-r^DN2 and Egf-r^DN3, respectively. Misexpression of either of the pUAS-Egf-r^DN alleles with the wing-GAL4 driver results in small curved wings and vein truncations (Fig. 5 C and D), which are similar to, although somewhat weaker than, the phenotypes generated by misexpression of the reference Egf-r^DN1 construct (Fig. 5B).

Figure 5.

Generation of dominant-negative forms of UAS-Egf-r. (A–F) Wings of the following male genotypes: (A) wing-GAL4>UAS-Egf-r^wt; (B) wing-GAL4> Egf-r^DN1; (C) wing-GAL4>Egf-r^DN2; (D) wing-GAL4>Egf-r^DN3; (E) wing-GAL4>Egf-r^DN2/Egf-r^wt; (F) wing-GAL4>Egf-r^DN3/Egf-r^wt. (G–J) structures of wild-type and mutant pUAS-Egf-r constructs: (G) pUAS-Egf-r^wt; (H) pUAS-Egf-r^DN1, which has a STOP codon 20 amino acid after the transmembrane domain; (I) pUAS-Egf-r^DN2, which has a breakpoint mapping 13 amino acid downstream of that of Egf-r^DN2; (J) pUAS-Egf-r^DN3, which has a point mutation (R1062K) in the kinase domain. Stars (*) indicate consensus residues conserved across a wide variety of tyrosine kinases.

Molecular analysis of the Δ2–3 induced Egf-r^DN2 showed that this mutant is deleted for the entire kinase domain of the receptor and a small portion of the white marker, resulting in an “out-of-frame” fusion between the two genes, mapping 33 amino acids after the transmembrane domain of EGF-R (Fig. 5I). This structure is very similar to that of Egf-r^DN1, which has a STOP codon 20 amino acids after the transmembrane domain (Fig. 5H). The EMS-derived Egf-r^DN3 mutation results from a single amino acid substitution, R1062K, in the kinase domain of the receptor (Fig. 5J), consistent with the propensity of EMS to act as a point mutagen. This substitution is conservative; however, it alters a residue that is absolutely invariant among all tyrosine kinases and is immediately adjacent to the catalytic aspartate residue in the active site of the kinase domain (44). Although Egf-r^DN2 and Egf-r^DN3 generate similar phenotypes in the wing, we found that their activities differ when expressed with Egf-r^wt. Coexpression of Egf-r^DN2 (or Egf-r^DN1) with Egf-r^wt results in mutual suppression of the Egf-r^wt ectopic vein phenotype and of the Egf-r^DN2 vein-loss phenotype (Fig. 5E). In contrast, Egf-r^DN3 did not suppress the Egf-r^wt phenotype but instead interacted positively to produce smaller rounded wings with a strong all-vein phenotype (Fig. 5F). As the Egf-r^DN3 mutant retains a complete cytoplasmic domain that could interact with effector molecules, coexpression of this mutant with Egf-r^wt may lead to the formation of EGF-R^DN3/EGF-R^wt heterodimers with partial activity, as has been suggested for Egf-r alleles that exhibit interallelic complementation (45).

Perspectives on the General Utility of the NOVA Method.

An attractive feature of the NOVA mutagenesis scheme described above is that the transgene of interest is modified in vivo, without the need of in vitro genetic engineering followed by time-consuming transformation procedures. More critically, NOVA mutants are recovered in an unbiased fashion solely on the basis of the nature of the phenotypes they induce. This phenotype-based screening eliminates the need for making predictions regarding likely functions of specific protein domains. Given that a substantial fraction of proteins predicted by genome sequencing have yet-unknown functions, the ability to screen for mutants without any advance knowledge of functional domains should be of significant utility. Although transposase is a particularly efficient mutagen for NOVA screening because it targets P element sequences, EMS can also be used in this scheme to create random point mutations. The different behaviors of the truncated Egf-r^DN2 and substitution Egf-r^DN3 mutants when coexpressed with wild-type Egf-r illustrate the value of screening for mutants with more than one type of mutagen. Point mutagens may also prove useful in creating NOVA mutations in endogenous genes adjacent to EP insertion lines (28), which carry UAS sequences that can activate expression of neighboring genes in a GAL4-dependent fashion.

The ability of transposase to target rearrangements to sequences within P elements has been exploited previously to induce loss-of-function mutations in a hs-fused transgene (46). An important difference between this earlier screen and the NOVA mutagenesis described here is the use of a high-level expression system, which we believe is critical for the recovery of novel activities that would otherwise go undetected. For example, the two novel dominant-negative Egf-r alleles and the dominant-negative Star^DN are very dosage sensitive. The requirement for high-level expression of dominant-negative constructs presumably reflects the need to produce a significant excess of the mutant protein relative to the endogenous protein to titrate out a sufficiently large fraction of the proteins or factors interacting with the wild-type protein. However, these high expression levels may also occasionally result in artifactual phenotypes caused by low-affinity interactions between proteins that would not ordinarily occur to a significant extent. It is therefore important to conduct additional experiments (e.g., with loss-of-function mutants, deficiencies, or duplications) to confirm the biological relevance of each NOVA phenotype.

The results reported in this paper suggest that the NOVA method could be employed to generate dominant alleles of a wide range of genes using various mutagens and provide insight to the function and mechanism of action of novel genes. This approach should be of particular utility in investigating the function of human disease genes, which have no known functional motifs but have homologues in Drosophila (47). Furthermore, NOVA mutagenesis should be applicable to any organism in which it is possible to misexpress transgenic constructs at high levels in a conditional fashion.

Abbreviations

EGF-R

epidermal growth factor receptor

MAPK

mitogen-activated protein kinase

NOVA

novel overexpression activity

RNAi

RNA interference

UAS

upstream activation sequence

EMS

ethyl methanesulfonate