1. Introduction
A major goal of genomic analysis is to identify novel loci and characterize their functions during normal and aberrant cellular processes. However, even for the case of well-established genetic models such as Drosophila melanogaster and others, our understanding of the role of most loci is still limited. For this reason, concerted genetic analyses are required to dissect the function and interrelationships of the identified sequences. Conventional genetic studies utilize chemical and radiation mutagenesis to disrupt a process (Greenspan, Reference Greenspan1997; Ashburner et al., Reference Ashburner, Golic and Hawley2005). These mutagens are advantageous as they sample the entire genome and are capable of producing a diverse array of mutations, including both loss-of-function and gain-of-function alleles. Conversely, they are complicated by the cumbersome and time-consuming tasks of mapping and isolating the associated DNA sequences. The advent of P transposon-based mutagenesis allowed tagging the mutated sequence that facilitated cloning of the relevant loci (Searles et al., Reference Searles, Jokerst, Bingham, Voelker and Greenleaf1982; Cooley et al., Reference Cooley, Kelley and Spradling1988; Yedvobnick et al., Reference Yedvobnick, Smoller, Young and Mills1988). However, the original schemes for transposon mutagenesis largely limited the class of mutations to insertional inactivation/loss of function. This limitation excluded production of rare gain-of-function mutations. A transposon-based system that allows both loss-of-function and gain-of-function genetic screening provides significant advantages, since a significant number of loci do not show obvious loss-of-function phenotypes (Miklos & Rubin, Reference Miklos and Rubin1996). This is possible using specialized transposon vectors, such as those exploiting the GAL4-Upstream Activating Sequence (UAS) cassette.
The development of the GAL4-UAS system in Drosophila allowed targeted expression of genes in a wide array of tissues (Brand & Perrimon, Reference Brand and Perrimon1993). The yeast GAL4 protein is a transcriptional activator that functions in Drosophila through interaction with its UAS target (Fischer et al., Reference Fischer, Giniger, Maniatis and Ptashne1988). A large number of Drosophila strains have been constructed that express GAL4 in a unique pattern based on the activity of distinct genomic enhancers (Manseau, Reference Manseau, Baradaran, Brower, Budhu, Elefant, Phan, Philp, Yang, Glover, Kaiser, Palter and Selleck1997; Duffy, Reference Duffy2002). When a GAL4 ‘driver’ line is crossed to a second transgenic strain containing a coding sequence under UAS-regulation, the downstream sequence is expressed in the pattern dictated by GAL4 expression. This technique has allowed directed expression of mutant versions of proteins in very limited domains, thereby facilitating genetic approaches that would otherwise be precluded by lethality. Brand & Perrimon (Reference Brand and Perrimon1993) validated this idea by expressing an activated form of Dras protein in eye and wing tissues and producing highly penetrant phenotypes that could be subjected to genetic modifier screening.
Rørth (Reference Rørth1996) combined GAL4-UAS and transposon mobilization to create a misexpression or overexpression genetic screening system. This method required construction of a specialized P element transposon named EP (Enhancer Promoter) derived from Casper 4. EP contains multiple copies of the UAS target site for GAL4 upstream of the hsp70 promoter; it also contains the mini-white marker to follow its movements. When EP is mobilized to new genomic sites it can position the GAL4-regulated hsp70 promoter upstream of coding sequences. After construction of a library of random EP insert lines, each line can be crossed to a specific GAL4 driver and the F1 progeny scored for phenotypes associated with gene overexpression or misexpression. As changes in the level or temporal expression of key regulatory proteins can elicit profound tissue patterning effects (Fortini et al., Reference Fortini, Rebay, Caron and Artavanis-Tsakonas1993; Neumann & Cohen, Reference Neumann and Cohen1996; Nagaraj et al., Reference Nagaraj, Pickup, Howes, Moses, Freeman and Banerjee1999; Hurlbut et al., Reference Hurlbut, Kankel, Lake and Artavanis-Tsakonas2007, Reference Hurlbut, Kankel and Artavanis-Tsakonas2009), this type of screen allows rapid identification of candidate loci necessary for a chosen developmental process. However, as described below, characterization of loss-of-function effects of such candidates is an essential corroboration for overexpression/misexpression analyses. Rørth (Reference Rørth1996) first utilized EP insertions to study genes involved in eye development through crosses with the sevenless-GAL4 driver strain that expresses in photoreceptor and cone cells. The GAL4 line was crossed with a collection of 352 EP lines each carrying a single insert. Six sets of progeny exhibited dominant eye phenotypes. Sequencing of the genomic DNA adjacent to the EP elements showed that five inserts were close to the transcription start sites of the driven loci. One of these loci was the Gap1 gene that encodes a Ras GTPase-activating protein, known to function during eye development. This study showed that an overexpression screen is a practical method for finding developmentally relevant genes in a directed, tissue specific manner. After the production of EP, additional vectors capable of mediating gene overexpression were designed, including P(Mae-UAS.611) (Merriam, Reference Merriam1997), EP yellow (EY) and EPg (Bellen et al., Reference Bellen, Levis, Liao, He, Carlson, Tsang, Evans-Holm, Hiesinger, Schulze, Rubin, Hoskins and Spradling2004), piggyBac WH (Thibault et al., Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004) and the bidirectional UAS vectors Gene Search (GS) (Toba et al., Reference Toba, Ohsako, Miyata, Ohtsuka, Seong and Aigaki1999) and XP (Thibault et al., Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004) that are capable of directing GAL4-regulated expression from loci on either side of the insertion site (Table 1). The bidirectional vectors offer the obvious advantage of a more rapid scan of loci for potential effects. However, when these insertions land between two loci, more extensive analysis of candidates is required to determine which one is responsible for the effects.
Table 1. Summary of potential misexpression/overexpression insertion strains
a The overlap of genes between all of the collections has not been determined.
For estimates of insert no. versus gene no. see Bellen et al. (Reference Bellen, Levis, Liao, He, Carlson, Tsang, Evans-Holm, Hiesinger, Schulze, Rubin, Hoskins and Spradling2004) and Thibault et al. (Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004). Insert no. for XP and WH derives from the original Exelixis collection of Thibault et al. (Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004). Insert no. for EY/EPg represents a selected subset within the Gene Disruption Project strains at Bloomington. Sense orientation was estimated from fraction of inserts 5′ to start site of loci (Bellen et al., Reference Bellen, Levis, Liao, He, Carlson, Tsang, Evans-Holm, Hiesinger, Schulze, Rubin, Hoskins and Spradling2004; Thibault et al., Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004), and the fraction of those inserts in the correct orientation to express the locus. This is a conservative estimate since insertions within certain exons and introns of loci can also drive a wild-type gene product. Antisense orientation was estimated from the fraction of inserts 3′ to start site of loci, and the fraction of those inserts in the orientation predicted to produce antisense RNA.
Locations of collections: Bloomington: Gene Disruption Project strains and Exelixis strains (XP, WH, EY/EPg, EP and pMae); Harvard: Exelixis strains (XP, WH); Kyoto Drosophila Genetic Resource Center strains (GS, pMae); Szeged strains (EP). Note: the Szeged Stock Center is closing and the remaining EP strains will be relocated.
Updated information on the availability and number of the above strains can be found on the Bloomington Stock Center webpage, the Harvard Medical School Exelixis Collection webpage and the Kyoto Drosophila Genetic Resource Center webpage.
As further demonstrated by Rørth et al. (Reference Rørth, Szabo, Bailey, Laverty, Rehm, Rubin, Weigmann, Milán, Benes, Ansorge and Cohen1998) for the case of the slow border cells (slbo) mutation, overexpression or misexpression screening can be combined with a mutant phenotype to allow searches for genetic modifiers in a mutant/sensitized genetic background. Rørth et al. (Reference Rørth, Szabo, Bailey, Laverty, Rehm, Rubin, Weigmann, Milán, Benes, Ansorge and Cohen1998) showed that slbo sterility could be suppressed via EP directed gene expression mediated by slbo-GAL4. Likewise, loss of function phenotypes within the Notch pathway were shown to be enhanced or suppressed by GAL4-mediated overexpression of negative or positive regulators respectively, of Notch signalling (Hall et al., Reference Hall, Alexander, Chang, Woodling and Yedvobnick2004), shown in Fig. 1. Importantly, overexpression screens can in principle identify both complex pleiotropic loci, and redundant loci that are refractory to most loss of function screening strategies. It has been estimated that two-thirds to three-quarters of Drosophila genes are phenotypically silent upon loss-of-function, and some of these effects are due to redundancy (Miklos & Rubin, Reference Miklos and Rubin1996). Consequently, overexpression screens can complement those based on loss-of-function. These methods now have extraordinary potential for identifying new components of pathways, based on the availability of large collections of transposon inserts at the Bloomington Stock Center, Harvard University and the Drosophila Genetic Resource Center at Kyoto. Within these collections insertions in the correct transcriptional orientation and 5′ to the start site are the most likely to produce misexpression/overexpression of a wild-type product. Based on this assumption, it is estimated that from over 29 000 inserts, approximately 5900 protein-coding loci are positioned for GAL4-induction within these collections (Table 1). New transposon insertions can also be readily generated using standard mobilization schemes (Alexander et al., Reference Alexander, Woodling and Yedvobnick2006). However, hotspots for insertion decrease the efficiency of such new screens whereas redundant insertions have been identified in the available collections (Bellen et al., Reference Bellen, Levis, Liao, He, Carlson, Tsang, Evans-Holm, Hiesinger, Schulze, Rubin, Hoskins and Spradling2004; Thibault et al., Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004). Here, we review studies utilizing targeted misexpression/overexpression screens in Drosophila. The studies are summarized in Table 2.
Fig. 1. Combined overexpression-genetic modifier screen using GAL4-UAS and EP. (a) Diagram of full length and truncated forms of Mastermind (Mam) protein. The truncated MamH polypeptide is a potent inhibitor of Notch pathway function when expressed under GAL4-UAS regulation in various tissues (Helms et al., Reference Helms, Lee, Ammerman, Parks, Muskavitch and Yedvobnick1999). (b) A genotype constitutively expressing UAS-MamH across the dorsoventral margin of the wing imaginal disc is created by recombination with the C96-GAL4 driver transgene. This recombinant chromosome produces a 100% penetrant, dominant wing nicking phenotype, shown in panel (f). In a typical screen individual EP strains are crossed to the C96-GAL4+UAS-MamH recombinant strain and the F1 progeny are scored for changes in the wing phenotype (Hall et al., Reference Hall, Alexander, Chang, Woodling and Yedvobnick2004; Alexander et al., Reference Alexander, Woodling and Yedvobnick2006; Kankel et al., Reference Kankel, Hurlbut, Upadhyay, Yajnik, Yedvobnick and Artavanis-Tsakonas2007). The sensitized screening system can be pre-tested with known pathway components to determine if wing modifications follow predictions. For example, the loss of function MamH phenotype is suppressed by coincident expression of activated Notch (c) or wild-type Delta proteins (d), both of which elevate Notch signalling. Conversely, co-expression of the negative Notch pathway regulator hairless enhances the wing phenotype (e).
Table 2. Summary of screens
2. General considerations for screens
There are several issues to consider when designing and interpreting a misexpression/overexpression screen. Most importantly, identification of a GAL4- induced locus that affects development of a particular tissue does not establish a role for the locus within that tissue. Misexpression or overexpression can produce phenotypes irrelevant to wild-type function (Tseng & Hariharan, Reference Tseng and Hariharan2002; Molnar et al., Reference Molnar, López-Varea, Hernández and de Celis2006; Mindorff et al., Reference Mindorff, O'Keefe, Labbé, Yang, Ou, Yoshikawa and van Meyel2007; Gregory et al., Reference Gregory, Shandala, O'Keefe, Jones, Murray and Saint2007; Stofanko et al., Reference Stofanko, Kwon and Badenhorst2008), and it is likely that any large-scale screen will produce examples of this effect. Consequently, a normal function for the gene in the identified tissue needs to be corroborated through loss-of-function analysis. Genetic tests for loss-of-function can utilize canonical alleles, imprecise excision-induced deletions of the insertion, or strains derived from several other recent methods (see section 5). Additionally, based on the phenotypes associated with gene induction, a locus may appear to impact a well-described process or developmental pathway. In those cases, effects of expression on mutant phenotypes associated with known pathway components can be used to integrate the locus within the pathway (Abdelilah-Seyfried et al., Reference Abdelilah-Seyfried, Chan, Zeng, Justice, Younger-Shepherd, Sharp, Barbel, Meadows, Jan and Jan2000; Kankel et al., Reference Kankel, Hurlbut, Upadhyay, Yajnik, Yedvobnick and Artavanis-Tsakonas2007; Franciscovich et al., Reference Franciscovich, Mortimer, Freeman, Gu and Sanyal2008). If normal expression of the tagged locus has been described in the tissue under study that can provide additional support for a role (Maybeck & Roper, Reference Maybeck and Röper2009). Further, it is possible that background mutations in an insertion strain contribute to the phenotype. Evidence that induced expression of the tagged locus is solely responsible for the phenotype can be obtained by demonstrating the same phenotype after expression of a defined UAS-cDNA or gene construct (Stofanko et al., Reference Stofanko, Kwon and Badenhorst2008).
Insertion elements can also induce loss-of-function mutations via the insertion process. Such insertions should produce knockout effects independent of GAL4, and will be enhanced by other alleles or deletions for the locus. However, another critical consideration in these screens is the potential effects of RNA interference elicited by GAL4-induced antisense transcripts. A GAL4-regulated transposon that inserts within a transcription unit in the antisense orientation may be capable of producing such an effect, and examples have been discussed (Gregory et al., Reference Gregory, Shandala, O'Keefe, Jones, Murray and Saint2007; Maybeck & Roper, Reference Maybeck and Röper2009). We surveyed the major collections for inserts with such position and orientation, including EP (Rørth, Reference Rørth1996), WH and XP (Thibault et al., Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004), EY and EYg (Bellen et al., Reference Bellen, Levis, Liao, He, Carlson, Tsang, Evans-Holm, Hiesinger, Schulze, Rubin, Hoskins and Spradling2004) and GS and P(Mae-UAS.611) (Kyoto collection) and estimate that a significant fraction may elicit an RNAi effect (Table 1). Up to one-third of the GAL4-targeted loci in these collections may have the potential for RNA-mediated loss-of-function effects rather than misexpression or overexpression. Consequently, it is essential to perform genetic analyses to determine the basis for the GAL4-induced phenotypes. An RNA interference phenotype should be GAL4-dependent, in contrast to an insertion that produces a phenotype by damaging the locus.
For screens in a sensitized genetic background there are additional considerations. Most importantly, what appears to be a novel genetic interaction needs to be distinguished from a simple misexpression/overexpression effect. This can be checked by crossing the GAL4 driver strain with the candidate insertion modifier in an otherwise wild-type genetic background. Secondly, effects of insertions on the function of GAL4 activity itself need to be examined. For example, an insertion that generally decreases or increases transcription can impact the screen's baseline phenotype if that phenotype is produced via GAL4-UAS, which is a transcription-based system. Consequently, a suppression or enhancement effect may be elicited through changes in the ability of GAL4 to activate, rather than by an effect on a gene relevant to the phenotype. Modifiers of GAL4 can be identified by crossing candidate strains to a GAL4 driver strain that shows a phenotype by itself within an otherwise wild-type background, such as GMR-GAL4 (Kankel et al., Reference Kankel, Hurlbut, Upadhyay, Yajnik, Yedvobnick and Artavanis-Tsakonas2007). Alternatively, the candidate modifier can be tested against an independent GAL4-UAS phenotype that is unrelated to the screen phenotype. These are important controls since effects on GAL4 can be prevalent in screens. For example, in the large-scale screen of the Exelixis library of 15 500 insertions, 294/610 tested insertions (48%) altered the GMR-GAL4 eye phenotype (Kankel et al., Reference Kankel, Hurlbut, Upadhyay, Yajnik, Yedvobnick and Artavanis-Tsakonas2007). In contrast, a screen of a GS collection only found 7% of modifiers acting through a GAL4 effect (Gregory et al., Reference Gregory, Shandala, O'Keefe, Jones, Murray and Saint2007). Modifiers of GAL4 phenotypes can include loci encoding global transcription or translation factors, among other examples.
3. Misexpression/overexpression screening in wild-type genetic backgrounds
Transposon-based misexpression/overexpression screening has been used to identify candidates for loci involved in a wide variety of processes such as nervous system and muscle patterning, wing and eye development and immune system development among others. These studies assay the phenotypic effects of random gene overexpression or misexpression in an otherwise wild-type genetic background. Such screens are also referred to as gain-of-function, reflecting the fact that gene expression can be driven at anomalously high levels in a normal site of expression, or alternatively, the gene may be driven at a time or in a tissue that it is normally not expressed.
(i) Nervous system
The development of the central nervous system (CNS) requires an intricate and complicated system of regulation, specifying cell fate and creating proper axon guidance. In particular, axons are guided to their correct targets by many environmental and signalling cues. Since the nervous system depends on specific patterning of synaptic connections, appropriate cues for axon development and synaptic partner formation are crucial.
Mindorff et al. (Reference Mindorff, O'Keefe, Labbé, Yang, Ou, Yoshikawa and van Meyel2007) used the scrt11-6-GAL4 driver, a PGawB insertion in scratch, which expresses GAL4 throughout the CNS and peripheral nervous system (PNS) of the developing embryo. Over 1000 GS lines were tested and 142 insertions caused lethality. Lethal embryos were analysed by staining for Fasciclin 2, which is expressed strongly on all motorneurons and on a set of interneurons. Immunostaining of the nervous system showed that 56 (39%) of the lethal lines produced axonal defects, representing 51 unique genes. Thirty-eight of the genes had been previously studied, and 17 were implicated in the development of the nervous system. This set includes loci essential for neurogenesis, such as mam and H, and neural differentiation, including mirr and pnt. The identification of leak (roundabout 2), encoding the Slit receptor validated the scheme, which also found 13 novel genes, predicted to encode secreted and transmembrane proteins. The screen identified a large number of loci encoding transcription factors, including the NFκB product of dorsal. Additional analyses of dorsal misexpression in the visual system using GMR-GAL4 revealed mistargeting of photoreceptors cells R2–R5 to the medulla, rather than the lamina plexus layer of the brain optic lobe. However, a role for dorsal in axon targeting was not corroborated with loss-of-function mutations, indicating that high levels of dorsal expression may perturb the normal function of other essential loci.
Ou et al. (Reference Ou, Chwalla, Landgraf and van Meyel2008) used an overexpression screen to find candidate genes for dendrite morphogenesis in the CNS and PNS. Although many factors have been implicated in dendrite development, including transcription factors, GTPases, and cytoskeleton-associated proteins, current knowledge of their morphogenesis is incomplete (Grueber & Jan, Reference Grueber and Jan2004; Williams & Truman, Reference Williams and Truman2004; Parrish et al., Reference Parrish, Emoto, Kim and Jan2007). Additionally, dendrites of the CNS differ from those of the PNS, as the former are not specialized for receiving external stimuli (Hughes & Thomas, Reference Hughes and Thomas2007). Thus, it is important to ascertain if the development of these distinct dendrite types involve the same mechanisms. Two overexpression screens were conducted using 141 GS lines, one in the da neurons of the PNS and the other in RP2 motor neurons of the CNS. These 141 derived from a prescreen of 1127 lines for those producing lethality when expressed in the CNS (Mindorff et al., Reference Mindorff, O'Keefe, Labbé, Yang, Ou, Yoshikawa and van Meyel2007). Multiple GAL4 drivers were used, including ppk1.9-GAL4, which expresses in dendritic sensory neurons tiling the larval body wall. The gain-of-function screen in da neuron dendrites found 35 unique loci causing mutant phenotypes of five classes. These five classes resemble mutant phenotypes of known genes involved in dendrite growth. Inserts at tramtrack, encoding a zinc finger protein, resembled shrub mutants, which have reduced embryonic dendrite growth (Brenman et al., Reference Brenman, Gao, Jan and Jan2001; Sweeney et al., Reference Sweeney, Brenman, Jan and Gao2006). Inserts at six loci, including cbt, encoding a zinc finger transcription factor, are similar to mutations of Polycomb Group genes (Parrish et al., Reference Parrish, Emoto, Kim and Jan2007). Six additional targeted loci including abrupt and Hr38 elicited phenotypes related to cut mutants, and one GS line induced phenotypes similar to mutations of spineless with both increases and reductions in dendrite growth (Kim et al., Reference Kim, Zitnan, Galizia, Cho and Adams2006a). A last class included one line that resembled constitutive activation of the GTPase Cdc42, leading to reduced embryonic dendritic branching (Gao et al., Reference Gao, Brenman, Jan and Jan1999). In the RP2 CNS neuron screen, 51 genes elicited phenotypes, including a large fraction of loci encoding transcription factors such as mastermind and pointed (Bilder et al., Reference Bilder, Graba and Scott1998; Helms et al., Reference Helms, Lee, Ammerman, Parks, Muskavitch and Yedvobnick1999). None of these genes have been previously shown to affect dendrite growth, but further testing indicated that at least 43 of the 51 candidates are normally expressed in the CNS during periods of dendritic development. Overall, 39% of the genes identified were the same in the two screens, although the varying phenotypes suggest that mechanisms of development for dendrites of the CNS and PNS only partially overlap.
Kraut et al. (Reference Kraut, Menon and Zinn2001) designed a gain-of-function screen to alter synaptic connections between motor neurons and muscles. Two thousand two hundred and ninety-three EP insertions were driven in motor neurons via elav GAL4, which expresses in all post-mitotic neurons. To observe the neuromuscular junction (NMJ), the elav-GAL4 driver strain was combined with UAS-Green Fluorescent Protein (GFP). This allowed the NMJ to be visualized in live third instar larvae or after dissection and staining with anti-GFP. One hundred and fourteen EPs produced strong mutant phenotypes. Three classes were noted: errors in pathfinding, morphologically altered, reduced or missing synapses, and excess or ectopic synapses. Most insertions produced phenotypes in more than one class. The EPs landed adjacent to 41 known genes, of which more than three quarters are required for nervous system development based on loss-of-function analyses. Sixteen of these loci affect axon or synapse formation, including Fasciclin 2, neurexin, Laminin A and also roundabout 2. Thirty-five new genes, without a characterized loss of function phenotype were also identified, including sequences encoding kinases, phosphatases, ATPases, Rho family GTPases and others involved in protein modification or degradation.
Zhao et al. (Reference Zhao, Gu, Rice, McAdams, Roark, Lawson, Gauthier, Reagan and Hewes2008) used misexpression in a screen for candidate genes involved in metamorphic remodelling of the neuroendocrine system. The processes of molting and metamorphosis require the action and precise timing of the juvenile hormones and ecdysteroids family of hormones (Nijhout, Reference Nijhout1994). Proper levels of ecdysis-triggering hormone (ETH) and crustacean cardioactive peptide (CCAP) are required for normal ecdysis (Kim et al., Reference Kim, Jan and Jan2006b). Following ecdysis, neurons secreting CCAP and other neuropeptides such as bursicon, which functions during wing expansion, undergo major changes during the pupal stage (Luan et al., Reference Luan, Lemon, Peabody, Pohl, Zelensky, Wang, Nitabach, Holmes and White2006). Though it is understood that a myriad of extensive hormone and neuropeptide changes occur throughout metamorphic remodelling, the mechanisms are not fully described. The four GAL4 drivers: EH-GAL4, CCAP-GAL4, c929-GAL4 and 386Y-GAL4 directed transposon expression to various neuronal and peptidergic cells. This includes neurons in the brain and ventral nerve cord, endocrine cells in the corpora cardiaca, endocrine Inka cells, midgut cells and PNS neurons. These were tested with 6000 insertion lines, including EP, EY, WH and XP to identify potential regulators of ecdysteroid-dependent metamorphosis of neuropeptidergic cells. The screen revealed over 50 loci whose misexpression caused defects either in ecdysis or wing expansion. Additionally, 14 loci were found to disrupt the CCAP/bursicon neuron cell projections during metamorphosis. Of these loci, genes known to be involved in neuronal development were found, including cabut, an ecdysteroid-response gene, and fat facets, encoding an ubiquitin-specific protease (Hegde, Reference Hegde2004; Munoz-Descalzo et al., Reference Munoz-Descalzo, Terol and Paricio2005). Several of the loci had not been previously implicated in neuronal remodelling. This includes three novel inserts driving expression of micro-RNAs, small non-coding RNAs that play diverse regulatory roles (Kloosterman & Plasterk, Reference Kloosterman and Plasterk2006). Additionally, two loci mip120 and stonewall code for proto-oncoprotein-like transcription factors responsible for signalling by Myb-like protein complexes.
A PNS screen was conducted to find genes involved in adult mechanosensory bristle development (Abdelilah-Seyfried et al., Reference Abdelilah-Seyfried, Chan, Zeng, Justice, Younger-Shepherd, Sharp, Barbel, Meadows, Jan and Jan2000). External sensory (es) organs are comprised of a neuron and support cells. The es organ develops from a single sensory organ precursor (SOP) cell that is chosen from an equipotent proneural cluster during the process of lateral inhibition (Ghysen & Dambly-Chaudiere, Reference Ghysen and Dambly-Chaudiere1989). The SOP cell divides and asymmetrically differentiates into distinct daughter cells, with the help of Notch signalling (Hartenstein & Posakony, Reference Hartenstein and Posakony1989) and other genetic regulators. Over 2000 EP insertions were misexpressed in the SOP and surrounding cells using scabrous-GAL4 (sca-GAL4). Abdelilah-Seyfried et al. (Reference Abdelilah-Seyfried, Chan, Zeng, Justice, Younger-Shepherd, Sharp, Barbel, Meadows, Jan and Jan2000) observed that 105 (4·6%) of the EPs produced mutant phenotypes, exhibiting either a loss of external support cells, increase in es organs or support cells, or increase/decrease in a cell type after cell transformation. The 105 EPs tagged 78 unique loci, associated with lateral inhibition, cell differentiation, cell cycle regulation and es organ formation. Thirty-seven loci were previously characterized, including those implicated in es organ development. These include emc and big brain, known to have roles in lateral inhibition (Skeath & Carroll, Reference Skeath and Carroll1991; Rao et al., Reference Rao, Bodmer, Jan and Jan1992). Big brain, a member of the aquaporin family, influences endosome maturation and Notch activity (Kanwar & Fortini, Reference Kanwar and Fortini2008). Other candidates not previously associated with es organ development were also found, such as hedgehog, originally defined for its roles in segment polarity and eye morphogenesis (Mohler & Vani, Reference Mohler and Vani1992; Heberlein et al., Reference Heberlein, Wolff and Rubin1993) and yan, which encodes a transcription factor identified for its role in photoreceptor cell development (Lai & Rubin, Reference Lai and Rubin1992).
(ii) Muscle
EPs have been used to drive loci in mesodermal cells with the goal of finding regulators of muscle and heart cell fate specification (Bidet et al., Reference Bidet, Jagla, Da Ponte, Dastugue and Jagla2003). Normal myogenic development involves an array of cell–cell signalling and mesoderm-specific transcription factors, which create distinct competence domains. These domains eventually transform into clusters of equipotent cells via receptor tyrosine kinases (RTKs), epidermal growth factor receptor (EGFR) and fibroblast growth factor receptor (FGFR), which impinge on the mitogen-activated protein kinases (MAPKs) pathway (Carmena et al., Reference Carmena, Gisselbrecht, Harrison, Jimenez and Michelson1998). In these domains, one cluster will express the homeodomain transcription factor Even-skipped (Eve). In the same competence domain are cells that express the homeobox gene ladybird (lb). In normal development, lb and eve are never co-expressed, as they specify cell lineages. Bidet et al. (Reference Bidet, Jagla, Da Ponte, Dastugue and Jagla2003) compared the expression pattern of these homeobox genes in EP overexpression genotypes relative to wild-type. 24B-GAL4 was used to drive expression of 2293 EP insertions ubiquitously in mesodermal cells, including myogenic cell precursors. After scoring protein accumulation, 84 (3·7%) of the EPs were found to alter expression profiles in lb and/or eve. The insertions were proximal to 31 known genes and 18 predicted genes, and 80% of the identified candidates have vertebrate orthologues, suggesting essential functions. Among them is rhomboid, which is required for maturation and release of EGF signals, yan, which encodes an ETS transcription factor/negative RTK effector and rac2. Rhomboid and Yan are hypothesized to interact with Rac2, a Rho GTPase in the RTK signalling pathway during diversification of cardiac and muscle cell lineages. The majority of the other candidate genes encode DNA/RNA binding factors or proteins involved in signalling. For example, the extra macrochaetae (emc) gene, encodes a helix-loop-helix (HLH) transcription factor that does not contain a basic DNA-binding domain (Garrell & Modolell, Reference Garrell and Modolell1990; Cubas et al., Reference Cubas, Modollel and Ruiz Gomez1994). The absence of the DNA binding domain allows Emc to dimerize with other bHLH proteins and down-regulate their activity. This is consistent with an Emc role in binding and sequestering general myogenic bHLH factors in the mesoderm.
Staudt et al. (Reference Staudt, Molitor, Somogyi, Mata, Curado, Eulenberg, Meise, Siegmund, Häder, Hilfiker, Brönner, Ephrussi, Rørth, Cohen, Fellert, Chung, Piepenburg, Schäfer, Jäckle and Vorbrüggen2005) utilized EP-elements driven in the epidermis to find genes implicated in muscle patterning. Muscle cell development involves processes where founder cells fuse with undetermined muscle cells to form myotubes; these myotubes migrate towards and attach with a set of epidermal border cells termed apodemes (Schnorrer & Dickson, Reference Schnorrer and Dickson2004). These processes are guided by a variety of genes, including those involved in transcription such as stripe, and cell communication, including slit/robo and fibroblast growth factor (FGF). To find additional loci involved in this process, Staudt et al. constructed stripe-GAL4 drivers to express in apodeme precursors and a group of cells that serve as substrate for muscle cell migration. Combining the drivers with 4500 EPs (3700 unique EP insertions), they looked for inserts that caused lethality, as disruption of muscle patterning prevents hatching. Antibody (anti-Myosin) staining of the differentiated but unhatched embryos was used to analyse for muscle pattern defects. Sixty-six (1·5%) EPs caused unambiguous gain-of-function phenotypes. This included 13 genes that code for membrane associated and secreted factors, eight coding for factors involved in protein modification, six coding for transcription factors, three coding for cytoskeleton binding proteins, and seven coding for functions involved with cell cycle control and biosynthesis. The screen also identified known muscle pattern formation genes such as Toll, gut feeling, and sulphateless and genes involved in cell migration and/or embryonic cell targeting including esg and sdc.
(iii) Wing and eye
Wing and eye development are popular choices for genetic screens since alterations in both structures are readily scored in adults, and formation of the wing and eye requires numerous signalling pathways that are used reiteratively during development (Neumann & Cohen, Reference Neumann and Cohen1996; Nagaraj et al., Reference Nagaraj, Pickup, Howes, Moses, Freeman and Banerjee1999; Freeman, Reference Freeman2005; de Iongh et al., Reference de Iongh, Abud and Hime2006; Jemc & Rebay, Reference Jemc and Rebay2006). Thus, gain-of-function screens in these tissues can allow detection of adult phenotypes associated with key loci that are normally early lethals when mutated. Molnar et al. (Reference Molnar, López-Varea, Hernández and de Celis2006) screened for genes involved in wing vein development. They expressed 13 000 GS insertions using GAL4-shv 3 kpm, a GAL4 driver expressing in the developing pupal veins. This screen uncovered 493 (4·2%) insertions that elicited modified wing vein phenotypes. The elements mapped to 254 insertion sites with 149 being single hits. Remarkably, of the identified genes, ~60% belong to the Notch, EGFR and decapentaplegic (Dpp) signalling pathways, previously known to affect vein formation. The remaining classes of loci, including numerous uncharacterized sequences, were enriched for putative functions in transcription and cell signalling.
Tseng & Hariharan (Reference Tseng and Hariharan2002) screened for loci that restrict cell growth or cell-cycle progression in the eye by looking for a small eye phenotype in adults. They employed EP lines driven throughout eye development using eyeless-GAL4. Of the 2296 EP lines tested, 46 (2·0%) displayed reduced eye size of varying severity. The 46 EP lines identified 32 different loci and 13 were in previously characterized genes. Some of these loci, such as hedgehog, dpp and fringe were known to be involved in eye development (Cho & Choi, Reference Cho and Choi1998). Two putative transcriptional regulators, Kruppel-homolog 1 and elbowB, and two regulators of the cytoskeleton, Rac2 and pebble, (Lehner, Reference Lehner1992; Harden et al., Reference Harden, Loh, Chia and Lim1995) were also identified. Nineteen of the loci were only identified via sequence as novel open reading frames.
(iv) Other tissues
A study of cytoskeleton formation employed 1001 EP insertions to identify genes required for salivary gland tubulogenesis (Maybeck & Röper, Reference Maybeck and Röper2009). The EP strains were crossed to forkhead-GAL4, which targets expression in the embryonic salivary glands, and also combined with either a GFP marker of the microtubule cytoskeleton (GFP-EFGas2) or a cell shape marker (SrcGFP). Fifty-one (5%) of the EPs produced significant mutant phenotypes when misexpressed, classified as either invagination defects, gland shape and lumen defects, position defects, or gland sub-fate defects. Of the 51 lines, seven corresponded to genes previously implicated in salivary gland morphogenesis: chickadee, tec29, doughnut on 2, rhomnoid1 and spitz, tap and slit. Forty-four EP lines corresponded to genes that encode proteins not previously associated with tubulogenesis, such as Egalitarian, Traf-4 and RanGAP, having functions in microtubule-based transport, Toll-like receptor signalling and nuclear transport, respectively (Minakhina et al., Reference Minakhina, Myers, Druzhinina and Steward2005; Takeshita et al., Reference Takeshita, Ishii, Kobiyama, Kojima, Coban, Sasaki, Ishii, Klinman, Okuda, Akira and Suzuki2005). In addition, 14 uncharacterized sequences were found with orthologues in other species. The range of potential functions identified in the screen was broad, including cytoskeletal and transcription factors, proteins involved in signalling, protein synthesis and degradation, membrane proteins, trafficking proteins and micro RNAs of the mir-310 cluster, previously implicated in epithelial morphogenesis.
Stofanko et al. (Reference Stofanko, Kwon and Badenhorst2008) analysed the immune system by screening for genes regulating larval haemocyte migration and differentiation. Haemocytes mediate defence against foreign pathogens. There are three types: plasmatocytes, crystal cells and lamellocytes and their differentiation requires numerous transcription factors such as Lozenge, Collier and the GATA factor Serpent. In the open circulatory system of Drosophila larvae, the majority of haemocytes freely circulate within the haemocoel; the rest remain sessile in patches or ‘islets.’ Haemocyte migration is regulated by either a response to signals from growth factors or a response to chemicals released due to wounding. However, the regulators that control haemocyte differentiation and attachment to the sessile islets are not well described. Using a Peroxidasin-GAL4 driver combined with UAS GFP, over 3400 EP and EY lines were expressed in plasmatocytes and crystal cells. One hundred and eight insertions, representing 101 loci were found to affect haemocyte development. The insertions caused disruption of sessile haemocyte compartments, changes in haemocyte number, improper positioning of haemocytes and changes in lymph gland size. Fifty-five EPs perturbed the formation of dorsal sessile haemocyte compartments; affected loci included RhoGeF2, involved in cell shape and mesodermal invagination during gastrulation (Barrett et al., Reference Barrett, Leptin and Settleman1997; Hacker & Perrimon, Reference Hacker and Perrimon1998) and C3G, a Ras family nucleotide exchange factor (Ishimaru et al., Reference Ishimaru, Williams, Clark, Hanafusa and Gaul1999). Overexpression of 36 loci increased haemocyte numbers, whereas 53 loci increased lymph gland size. Classification of the 101 candidate genes by gene ontology (GO) showed that more than half of the genes had no previously described function. Of the genes that did have described functions, the most abundant encoded transcription factors/nucleic acid binding proteins, such as Kr, esg, chn and broad.
Huang & Haddad (Reference Huang and Haddad2007) used overexpression to screen for genes involved in anoxia sensitivity. Drosophila embryos can survive in severe hypoxic conditions for days and adult flies can survive without oxygen for hours without apparent injury (Haddad et al., Reference Haddad, Sun, Wyman and Xu1997; Krishnan et al., Reference Krishnan, Sun, Mohsenin, Wyman and Haddad1997). Though Drosophila is known to be tolerant to hypoxia, the mechanisms to sense and respond to low O2 are relatively unknown. Using daughterless-GAL4 (da-GAL4), which expresses ubiquitously, 1600 EPs were overexpressed throughout development. The screen measured changes in the time required for flies to recover after oxygen deprivation. Four EP lines showed significantly longer recovery times. Remarkably, each of these EPs drove the same downstream transcript, CG14709, encoding a protein in the multidrug resistance protein (MRP) subfamily, a member of the ATP-binding cassette (ABC) transporter superfamily (Dean et al., Reference Dean, Rzhetsky and Allikmets2001; Dean & Annilo, Reference Dean and Annilo2005). Other studies have suggested that CG14079 is an orthologue of MRP4/ABCC4, responsible for nucleotide transport (Dean et al., Reference Dean, Rzhetsky and Allikmets2001). However, a role for CG14079 (renamed Drosophila anoxia-sensitive dMRP4) in response to hypoxia had not been characterized. CG14079 was overexpressed only in neurons using the elav-GAL4 driver and a similar delayed recovery to oxygen deprivation was observed, demonstrating that overexpression in neurons alone causes the anoxia-sensitive phenotype. Interestingly, CG14079 had been identified in an earlier gene overexpression screen (Monnier et al., Reference Monnier, Girardot, Cheret, Andres and Tricoire2002). This screen of 2500 P(Mae-UAS-6.11) elements driven ubiquitously by da-GAL4 was designed to find loci that affected lifespan in the presence of excess reactive oxygen species via exposure to H2O2. Five lines were characterized, with overexpression of CG14079 leading to increased sensitivity to oxidative stress. These studies suggest a potential link between responses to both anoxia and oxidative stress.
4. Misexpression/overexpression screening in sensitized genetic backgrounds
GAL4-regulated overexpression/misexpression is also applicable to genetic interaction screening. To implement a sensitized screen, a GAL4 driver and UAS-regulated insertion collection are combined with a canonical mutant genotype, or a UAS-regulated construct that produces a phenotype readily scoreable for changes, such as a dominant-negative eye or wing phenotype. In designing a sensitized screen, it is crucial to select a phenotype that can be pretested for predicted responses to gain-of-function and loss-of-function modifiers. For example, expression of a dominant-negative construct should down regulate a pathway and produce a loss-of-function phenotype. This phenotype is predicted to be suppressed by overexpression of positive effectors of that pathway, or enhanced by overexpression of negative effectors (Fig. 1). Likewise the phenotype associated with the construct should respond appropriately to loss-of-function mutations in pathway components. Thus, sensitized screens that investigate processes that are genetically well described are more likely to succeed.
(i) Wing
Hall et al. (Reference Hall, Alexander, Chang, Woodling and Yedvobnick2004) and Alexander et al. (Reference Alexander, Woodling and Yedvobnick2006) screened for Notch pathway components by each testing approximately 2000 EP insertions for their ability to modify a wing nicking phenotype (Fig. 1). The wing phenotype derived from dorsal–ventral wing margin expression of a truncated version of the Mastermind (Mam) protein. Mam contains an amino terminal charge cluster of basic amino acids that mediates a physical association with the intracellular domain of Notch (Kitagawa et al., Reference Kitagawa, Oyama, Kawashima, Yedvobnick, Kumar, Matsuno and Harigaya2001). A shortened version of Mam (UAS-MamH) that terminates directly after the basic domain acts as a dominant negative for Notch pathway function when expressed under GAL4 regulation (Helms et al., Reference Helms, Lee, Ammerman, Parks, Muskavitch and Yedvobnick1999). The C96-GAL4 driver was used to express UAS-MamH across the developing wing margin and to simultaneously drive the library of single EP inserts. These screens identified canonical Notch pathway components such as Delta and Mam, as well as other functions known to impact Notch signalling, including the glycosyltransferase Fringe (Moloney et al., Reference Moloney, Panin, Johnston, Chen, Shao, Wilson, Wang, Stanley, Irvine, Haltiwanger and Vogt2000) and the negative regulator of EGF receptor function, Kek1 (Ghiglione et al., Reference Ghiglione, Carraway, Amundadottir, Boswell, Perrimon and Duffy1999). More novel loci that influence Notch signalling were also identified, such as domino, which encodes products of the SW12/SNF2 class of DNA-dependent ATPases (Ruhf et al., Reference Ruhf, Braun, Papoulas, Tamkun, Randsholt and Meister2001), poils aux pattes (pap) which encodes a protein related to the TRAP 240 component of the mediator transcription complex (Boube et al., Reference Boube, Faucher, Joulia, Cribbs and Bourbon2000) and a novel Minute locus encoding ribosomal protein L13a.
The GAL4-UAS-Mam truncation-based screen for Notch pathway modifiers was greatly expanded by Kankel et al. (Reference Kankel, Hurlbut, Upadhyay, Yajnik, Yedvobnick and Artavanis-Tsakonas2007), who utilized the Exelixis collection of transposon insertions as the source of modifiers (Thibault et al., Reference Thibault, Singer, Miyazaki, Milash, Dompe, Singh, Buchholz, Demsky, Fawcett, Francis-Lang, Ryner, Cheung, Chong, Erickson, Fisher, Greer, Hartouni, Howie, Jakkula, Joo, Killpack, Laufer, Mazzotta, Smith, Stevens, Stuber, Tan, Ventura, Woo, Zakrajsek, Zhao, Chen, Swimmer, Kopczynski, Duyk, Winberg and Margolis2004). This collection derives from several transposable elements with different insertion sequence site preferences. It contains 15 500 unique insertions, in or near 53% of Drosophila coding regions. As above, dominant negative Mam was driven via C96-GAL4 across the wing margin and tested against the insert library. Insertions targeting 408 specific loci identified 31 previously characterized Notch interactors. Additionally, this screen identified 160 novel candidates for loci that affect Notch signalling, including those with RNA recognition motifs, linking Notch to RNA processing. Associations between Notch, cell proliferation and other pathways such as EGF receptor were also uncovered. Significantly, a novel class of interactors was found that appeared specific to loss of mam function, rather than Notch signalling. One example of this involved the armadillo (arm) locus, encoding the Drosophila β-catenin homolog. A mam-specific interaction with arm was also seen cross-species using a cell culture assay.
(ii) Eye
Kazemi-Esfarjani & Benzer (Reference Kazemi-Esfarjani and Benzer2000) screened for suppressors of polyglutamine (polyQ) toxicity mediated by a UAS-driven run of 127 CAG codon repeats. The transgene encoding a polyQ product was driven in all cells of the retina by GMR-GAL4, which contains the rhodopsin1 gene enhancer. Flies co-expressing these constructs exhibit a severely disrupted eye morphology that is associated with aggregates of polyQ. To look for modifiers of this rough eye phenotype, 7000 EP transpositions were generated and each was tested as a transheterozygote for modification of the GMR-GAL4+UAS-polyQ phenotype. Fifty-nine modifiers were identified and two suppressor loci were characterized further: dHDJ1 that encodes a protein homologous to the human HSP40/HDJ1 and dTPR2 that encodes a product related to human Tetratricopeptide 2. These results were of interest since both proteins contain a J domain, found in chaperone proteins, and their activity could influence polyQ aggregate formation.
Ferres-Marco et al. (Reference Ferres-Marco, Gutierrez-Garcia, Vallejo, Bolivar, Gutierrez-Aviño and Dominguez2006) employed Delta overexpression in the eye to screen for factors that contribute to cell growth and tumorigenesis. GAL4 under eyeless (ey) regulation was used to drive Delta anterior to the morphogenetic furrow and create a large eye phenotype. The ey-GAL4 construct simultaneously drove random inserts of the bidirectional GS vector. The screen led to identification of an insert that overexpressed both the longitudinals lacking (lola) and pipsqueak (psq) loci, which produced invasive tumours derived from eye tissue when co-expressed with Delta. The lola and psq loci each encode multiple forms of transcription factors that contain a BTB domain. The BTB domain has been previously associated with epigenetic silencing of loci through recruitment of Polycomb Group and HDAC complexes (Melnick et al., Reference Melnick, Carlile, Ahmad, Kiang, Corcoran, Bardwell, Prive and Licht2002). Ferres-Marco et al. (Reference Ferres-Marco, Gutierrez-Garcia, Vallejo, Bolivar, Gutierrez-Aviño and Dominguez2006) identified the Retinobastoma family protein (Rbf) as one key target for gene silencing in these eye tumours. They also found that Delta-Notch signalling contributes to Rbf silencing through hypermethylation of DNA around the promoter, and suggested that both gene-silencing events contribute to malignancy.
A sensitized misexpression screen was done to study genes involved in neuronal plasticity, the process in which neural circuits change in response to environmental alterations (Franciscovich et al., Reference Franciscovich, Mortimer, Freeman, Gu and Sanyal2008). Stable changes in neurons involve transcription or translation of specific products, and a crucial transcription factor AP-1, a heterodimer of Fos and Jun, plays diverse roles in their plasticity (Sanyal et al., Reference Sanyal, Sandstrom, Hoeffer and Ramaswami2002; Etter et al., Reference Etter, Narayanan, Navratilova, Patel, Bohmann, Jasper and Ramaswami2005). However, despite knowledge of the various roles of AP-1, little is known about the upstream or downstream factors that mediate its effects in neurons. This study used a UAS-activated dominant negative truncation of the Fos gene, which inhibits AP-1, to generate a small adult eye phenotype when combined eyeless-GAL4 (Eresh et al., Reference Eresh, Riese, Jackson, Bohmann and Bienz1997; Sanyal et al., Reference Sanyal, Sandstrom, Hoeffer and Ramaswami2002). Four thousand three hundred EPs were tested for rescue or enhancement of the eye phenotype. Genes known to function solely in eye development were not further investigated. Candidates from this screen were also driven with elav C155-GAL4 in post-mitotic neurons to examine effects in neural development. The screen found over 300 candidates, with 249 verified and 54 predicted genes. Several products previously implicated with AP-1 function were found, such as Ras85D, Bsk (dJNK), CycB and Men, confirming the validity of this screen. Ten suppressors and 15 enhancers were further tested in the NMJ. Thirteen of these resulted in mutant synaptic phenotypes. Two genes producing the strongest phenotypes, sprouty and shaggy, are inhibitors of signalling cascades. Shaggy, a GSK3-β kinase, was shown to inhibit AP-1-dependent synaptic growth via interactions with the Jun-N-terminal kinase pathway.
Zhu et al. (Reference Zhu, Wilson and Leptin2005) screened for genes involved in the FGF signalling pathway. The two FGF transmembrane receptors, Breathless (Btl) and Heartless (Htl), phosphorylate target kinases upon ligand binding. Signalling involves the cytoplasmic molecule downstream-of-FGF-receptor (Dof) and others such as Corkscrew (Battersby et al., Reference Battersby, Csiszar, Leptin and Wilson2003; Petit et al., Reference Petit, Nussbaumer, Dossenbach and Affolter2004; Wilson et al., Reference Wilson, Battersby, Csiszar, Vogelsang and Leptin2004). The signalling mechanism from the activation of the FGF receptor to its intracellular targets was investigated. Although FGF signalling is not required for eye development, Zhu et al. genetically targeted this tissue for their modifier screen. GMR-GAL4 (Freeman, Reference Freeman1996) was used to drive constitutively activated Btl plus Dof in the eye, eliciting a rough eye phenotype (Casci et al., Reference Casci, Vinos and Freeman1999). This genotype was then combined with 2100 EPs, and 26 enhancers' plus 24 suppressors of the rough eye phenotype were identified. The candidate genes encode proteins of diverse functions, including kinases, transcription factors, membrane proteins and mitochondrial proteins. Only a minority of these modifiers interacted with an eye phenotype produced by ectopic EGFR expression, suggesting that the loci scored did not overlap in function with this pathway. In contrast, more significant interactions were observed with an eye phenotype produced from an activated form of PVR, the receptor for both platelet-derived and vascular endothelial growth factors. In follow up analyses, two of the modifier genes, sar1 and robo2, encoding a GTPase and a protein involved in axon guidance respectively were further examined. Loss of function for these loci produced genetic interactions with a dof hypomorphic allele, affecting endogenous FGF signalling within the trachea of embryos, validating the genetic scheme.
Park & Song (Reference Park and Song2008) screened for genes involved in the checkpoint (Chk2) signalling pathway, which plays key roles in DNA damage response. Upon phosphorylation following DNA damage, the Chk2 receptor protein kinase relays signals to a variety of downstream proteins to bring about DNA repair, cell cycle arrest and apoptosis (Sancar et al., Reference Sancar, Lindsey-Boltz, Unsal-Kaçmaz and Linn2004). Due to the critical functions of Chk2 signalling, pathway genes are involved in tumorigenesis or have oncogenic properties. To search for undocumented members of the Chk2 pathway, Park & Song (Reference Park and Song2008) used the GMR-GAL4 driver to express UAS-Chk2, producing a rough eye phenotype, and to simultaneously drive EPs. The screen was validated using a mutation in the ATM-kinase, which is required for activation of Chk2. The mutation suppressed the rough eye phenotype. A counterscreen of EPs against a rough eye phenotype derived from UAS-Notch expression was also employed to eliminate modifiers common to Notch signalling. Using 2240 EPs, they identified 25 candidate genes not previously associated with Chk2 signalling; 22 EPs suppressed and three enhanced the original phenotype. Several of the inserts led to defects in G2 arrest after irradiation, implicating the loci in the response to DNA damage. One of the enhancers, ballchen (Aihara et al., Reference Aihara, Nakagawa, Yasui, Ohta, Hirose, Dhomae, Takio, Kaneko, Takeshima, Muramatsu and Ito2004), encodes a protein kinase that phosphorylates histone H2A, implicated in the DNA repair response (Tanaka et al., Reference Tanaka, Huang, Halicka, Zhao, Traganos, Albino, Dai and Darzynkiewicz2007). The suppressors included calmodulin, which may modulate cell cycle progression via spindle effects (Goshima et al., Reference Goshima, Wollman, Goodwin, Zhang, Scholey, Vale and Stuurman2007), kayak (Drosophila Fos), which functions at the G2-M transition (Hyun et al., Reference Hyun, Becam, Yanicostas and Bohmann2006) and melted, which regulates cell growth via effects on the insulin-signalling pathway (Teleman et al., Reference Teleman, Chen and Cohen2005).
Shalaby et al. (Reference Shalaby, Parks, Morreale, Osswalt, Pfau, Pierce and Muskavitch2009) used the Exelixis collection in a screen for new Notch pathway modifiers. Their study utilized a Delta gain-of-function phenotype mediated via GMR-GAL4 driven Delta overexpression posterior to the morphogenetic furrow. This expression results in a dominant eye phenotype that derives from cell fate alterations associated with aberrant Notch signalling. After outcrossing to the insertion library they obtained 170 candidates, including loci known to function in Notch signalling, such as numb and kuzbanian as well as a number of novel candidates. Two hormone receptor loci, Hr38 and Hr39, and others involved in intracellular trafficking, including Vha68-2, which encodes a component the v-ATPase proton pump complex were found. A large number of eye modifiers were not identified in the Kankel et al. (Reference Kankel, Hurlbut, Upadhyay, Yajnik, Yedvobnick and Artavanis-Tsakonas2007) screen, possibly reflecting the different tissue contexts (eye versus wing). One modifier was studied in some detail: CG2446 (Amun). Amun is predicted to encode a DNA glycosylase, an activity associated with base excision repair processes, and also transcriptional regulation. Evidence was presented linking Amun to the regulation of Achaete, a transcription factor necessary for specification of sensory cells within equivalence groups associated with Notch activity.
(iii) Other tissues
Laviolette et al. (Reference Laviolette, Nunes, Peyre, Aigaki and Stewart2005) performed a gain-of-function-modifier screen for loci involved in the formation of the larval NMJ. N-ethylmaleimide sensitive factor (NSF) has roles in vesicular trafficking and can influence neurotransmitter receptors levels and synaptic strength. The starting phenotype for this study was overgrowth of the neuromuscular synapse, elicited by expression of a dominant negative form of NSF2. The construct was driven in neurons by elav-GAL4 and the NMJ visualized with a GFP fusion to the CD8 protein transmembrane domain and the C-terminus of the Shaker channel. Three thousand GS insertion lines were analysed and 99 suppressors were obtained, with 89 being near identifiable gene sequences. The 89 loci specify numerous products previously associated with nervous system function, for example, the lola, nejire and E2F loci encode transcription factors. Loci encoding cytoskeletal components, including actin5C, myosin binding protein and signalling proteins, such as methuseleh, which functions in G protein receptor coupled signalling, were also uncovered. The signalling class of modifier loci included frizzled 4, encoding a Wnt receptor, implicating Wg signalling in NMJ formation.
Gregory et al. (Reference Gregory, Shandala, O'Keefe, Jones, Murray and Saint2007) used a sensitized overexpression screen to find genes that modulate Rho signalling during cytokinesis. Rho is a GTPase with crucial roles during cytokinesis in animal cells (Lu et al., Reference Lu, Longo, Zhou, Massa and Chen2009). In Drosophila, Rho is activated at the cell equator by Pebble (Pbl), a Rho GTP exchange factor, initiating cytokinesis (Hime & Saint, Reference Hime and Saint1992; Lehner, Reference Lehner1992). Other proteins involved include RacGAP50c (Jantsch-Plunger et al., Reference Jantsch-Plunger, Gonczy, Romano, Schnabel, Hamill, Schnabel, Hyman and Glotzer2000) and Pavarotti (Pav-KLP) (Somers & Saint, Reference Somers and Saint2003). Expression of a UAS-driven dominant negative version of Pbl via GMR-GAL4 produces a rough eye phenotype through failure of Rho activation. Two thousand one hundred and ninety GS lines were tested for effects in this hypomorphic Rho signalling background and 112 modifiers were analysed. More than half of the loci encode cell cycle or signalling proteins, and others encode enzymes involved in metabolism, particularly regulators of phospholipids. The screen found known genes encoding proteins involved in cytokinesis, such as Diaphanous, a Rho effector involved in actin organization (Afshar et al., Reference Afshar, Stuart and Wasserman2000) and Four wheel drive, a phosphoinositide-4 kinase involved in cytokinesis during spermatogenesis (Brill et al., Reference Brill, Hime, Scharer-Schuksz and Fuller2000). Four genes not previously implicated in cytokinesis were scored: thread/diap1, cdc 14, Pitslre and phosphoinositide-dependent kinase 1 (PDK1). Thread/Diap1 is an inhibitor of apoptosis (Lisi et al., Reference Lisi, Mazzon and White2000) and Cdc 14 is a protein phosphatase regulator of late mitotic events in Caenorhabditis elegans (Stegmeier & Amon, Reference Stegmeier and Amon2004). Pitslre is a homolog of cyclin-dependent kinase 11 (CDK11) known to promote centrosome maturation and spindle formation during mitosis (Petretti et al., Reference Petretti, Savoian, Montembault, Glover, Prigent and Giet2006), whereas PDK1 is involved in cell growth regulation (Bimbo et al., Reference Bimbo, Liu and Balasubramanian2005).
5. Conclusions
GAL4-UAS-mediated overexpression/misexpression allows rapid screening of coding sequences in Drosophila for potential roles in specific tissues. This method has the capacity to uncover novel loci refractory to standard loss-of-function genetic approaches, such as redundant or pleiotropic genes. Additionally, the large extant libraries of strains harbouring sequenced transposon insertion sites often allow direct correlation of phenotypes to gene functions. Studies outlined in this review have validated this type of genetic screen by uncovering loci previously associated with the processes targeted. Further, in each case novel loci were also identified. It is important to stress that overexpression or misexpression alone cannot be used to implicate a gene in a developmental process. However, the large and growing list of genetic resources in Drosophila, including mutations in many loci, allow further tests for loss-of-function phenotypes associated with the candidate loci. For example, nested sets of chromosomal deficiencies are available that cover the majority of the Drosophila genome (Parks et al., Reference Parks, Cook, Belvin, Dompe, Fawcett, Huppert, Tan, Winter, Bogart, Deal, Deal-Herr, Grant, Marcinko, Miyazaki, Robertson, Shaw, Tabios, Vysotskaia, Zhao, Andrade, Edgar, Howie, Killpack, Milash, Norton, Thao, Whittaker, Winner, Friedman, Margolis, Singer, Kopczynski, Curtis, Kaufman, Plowman, Duyk and Francis-Lang2004; Tweedie et al., Reference Tweedie, Ashburner, Falls, Leyland, McQuilton, Marygold, Millburn, Osumi-Sutherland, Schroeder, Seal and Zhang2009), allowing rapid dosage effect assays for loci with no mutant alleles. Additionally, the generation of imprecise excisions (Xu et al., Reference Xu, Condell, Plesken, Ma, Ren and Sclame2006), targeted gene knockouts (Gong & Golic, Reference Gong and Golic2004), libraries of RNAi strains (Dietzl et al., Reference Dietzl, Chen, Schnorrer, Su, Barinova, Fellner, Gasser, Kinsey, Oppel, Scheiblauer, Couto, Marra, Keleman and Dickson2007; N. Perrimon, personal communication [http://www.flyrnai.org/TRiP-HOME.html]) and zinc finger nucleases (Beumer et al., Reference Beumer, Trautman, Bozas, Liu, Rutter, Gall and Carroll2008) provide methods to inactivate specific loci if canonical alleles are not available. The possibility of rapid candidate gene identification through overexpression or misexpression analysis, in combination with these techniques for gene inactivation, provides a powerful tool for analysis of gene function.
This work was supported by NSF grant IBN-0415453 and gifts from the Ammerman Foundation, and David and Karen Smoller. We thank Bob Levis and Steve Thibault for helpful advice.
