A Genetic Screen for Dominant Modifiers of a Small-Wing Phenotype in Drosophila melanogaster Identifies Proteins Involved in Splicing and Translation

Abstract

Studies in the fly, Drosophila melanogaster, have revealed that several signaling pathways are important for the regulation of growth. Among these, the insulin receptor/phosphoinositide 3-kinase (PI3K) pathway is remarkable in that it affects growth and final size without disturbing pattern formation. We have used a small-wing phenotype, generated by misexpression of kinase-dead PI3K, to screen for novel mutations that specifically disrupt organ growth in vivo. We identified several complementation groups that dominantly enhance this small-wing phenotype. Meiotic recombination in conjunction with visible markers and single-nucleotide polymorphisms (SNPs) was used to map five enhancers to single genes. Two of these, nucampholin and prp8, encode pre-mRNA splicing factors. The three other enhancers encode factors required for mRNA translation: pixie encodes the Drosophila ortholog of yeast RLI1, and RpL5 and RpL38 encode proteins of the large ribosomal subunit. Interestingly, mutations in several other ribosomal protein-encoding genes also enhance the small-wing phenotype used in the original screen. Our work has therefore identified mutations in five previously uncharacterized Drosophila genes and provides in vivo evidence that normal organ growth requires optimal regulation of both pre-mRNA splicing and mRNA translation.

NORMAL biological development and homeostasis require tight control of growth at the level of individual cells, organs, and the whole organism. For example, unregulated cellular proliferation may result in too few or too many cells, leading to inappropriately sized, nonfunctional organs, which in turn can result in a variety of pathological conditions. Significantly, individuals of the same species raised in a given environmental niche grow to similar final sizes. This means that growth is controlled, at least to some extent, genetically. Such “growth genes” might encode key components or effectors of discrete growth-regulatory signaling pathways.

In recent years, the fruit fly, Drosophila melanogaster, has been used as a model organism to investigate the genetic basis of growth control (reviewed in Edgar and Nijhout 2004). Most of these studies have focused on the growth of the larval imaginal discs. Imaginal discs are epithelial structures that undergo massive growth during the ∼4 days of larval life. The size of the disc at the end of the larval period largely determines the size of the adult appendage (eye, wing, etc.) into which it ultimately develops (reviewed in Johnston and Gallant 2002).

Key regulators of Drosophila imaginal disc growth have been discovered through three main approaches. First, classical genetic research has produced many mutant strains that exhibit growth phenotypes. For example, Minute mutants, which correspond to at least 50 different genetic loci, have a slower growth rate and sometimes an altered adult size (Lambertsson 1998). These phenotypes are thought to be the result of a reduced capacity for protein synthesis. Indeed, several Minute mutations have been demonstrated to disrupt genes that encode ribosomal proteins (RPs) (Lambertsson 1998).

A second way in which Drosophila growth regulators have been discovered is through the study of proteins or signaling pathways whose mammalian orthologs have been implicated in cell proliferation and/or growth. An example of this approach is the characterization of the insulin receptor/phosphoinositide 3-kinase (InR/PI3K) pathway (reviewed in Leevers and Hafen 2004). For example, clonal loss of Dp110, which encodes the catalytic subunit of the Drosophila class 1A PI3K, reduces cell size and clonal growth in imaginal discs (Weinkove et al. 1999). Conversely, overexpression of Dp110 increases cell size, promotes cell cycle progression, and causes tissue overgrowth (Leevers et al. 1996). Similar results have been obtained through modulating the activity of other components of the pathway (Leevers and Hafen 2004). Significantly, final organ/body size is altered without affecting pattern or shape in all these experiments, thereby demonstrating that the InR/PI3K pathway regulates growth and determines final size per se rather than as a secondary result of effects on patterning. The exact mechanism through which InR/PI3K signaling results in growth is poorly understood, but is likely to involve targets of Akt and the Foxo transcription factor and the upregulation of protein synthesis (Leevers and Hafen 2004).

Drosophila Ras (Prober and Edgar 2000), Myc (Johnston et al. 1999; de la Cova et al. 2004), Tor (Zhang et al. 2000), and the Cyclin D–Cyclin-dependent kinase 4 complex (CycD–Cdk4) (Datar et al. 2000; Meyer et al. 2000) have also been demonstrated to regulate fruit fly growth subsequent to their initial characterization in other species. In response to high levels of nutrients and/or growth factors, Tor stimulates S6 kinase and the translation initiation factor eIF-4E and is thought to promote growth, at least in part, through upregulation of protein synthesis (reviewed in Neufeld 2004). CycD–Cdk4 is remarkable in that it promotes cellular growth and cell cycle progression in a coordinated manner and can therefore increase overall organ size (Datar et al. 2000; Meyer et al. 2000). This is in stark contrast to other components of the Drosophila cell cycle machinery, which alter cell division rates without directly affecting cell growth (Neufeld et al. 1998).

The third major way in which growth regulators have been discovered in Drosophila is through various ingenious genetic screens. Screening has proved to be a powerful and relatively unbiased approach to identify both characterized genes that were not previously known to have a growth-regulatory role and completely novel genes. For example, some of the earliest investigations in this field uncovered zygotic mutations that cause overgrowth of the larval imaginal discs (reviewed in Bryant et al. 1993 and Watson et al. 1994). A number of “tumor suppressor” genes, including fat and discs large 1, were discovered in this way. Taking the opposite approach, Galloni and Edgar screened homozygous mutant larvae for abnormally small sizes and developmental delay and so identified mutations in the genes encoding translation initiation factor eIF-4a and the mitochondrial RP, mRpS15 (Galloni and Edgar 1999).

Several laboratories have conducted clonal growth screens to circumvent the early lethality that often results from zygotic mutation of growth genes. In most of these screens, clones that are homozygous for random mutations were induced in the early eye imaginal disc and the adult eye was subsequently examined for over- or undergrowth. Genes identified in this manner include Pten (Gao et al. 2000), Tsc1 (Gao and Pan 2001; Potter et al. 2001; Tapon et al. 2001), Tsc2 (Ito and Rubin 1999), Rheb (Stocker et al. 2003), Tor (Oldham et al. 2000), warts (Justice et al. 1995; Xu et al. 1995), salvador (Kango-Singh et al. 2002; Tapon et al. 2002), and hippo (Harvey et al. 2003; Jia et al. 2003; Udan et al. 2003; Wu et al. 2003). Tor, Rheb, and the Tsc1–Tsc2 complex are all core components of the Tor signaling pathway (reviewed in Leevers and Hafen 2004). Interestingly, Tsc1–Tsc2 can be inhibited by InR/PI3K signaling, thus forming a link between the InR/PI3K and Tor pathways (reviewed in Marygold and Leevers 2002). Warts, Hippo, and Salvador form a complex that restricts growth by promoting both cell cycle exit and apoptosis (reviewed in Ryoo and Steller 2003).

As an alternative approach, other groups have screened for growth defects induced by overexpressing random genes in organs such as the developing eye or wing. Genes identified in this way include Rheb (Patel et al. 2003; Saucedo et al. 2003; Stocker et al. 2003), bantam (Hipfner et al. 2002; Raisin et al. 2003), and slimfast (Colombani et al. 2003). Growth regulation through the bantam and slimfast gene products occurs through very different mechanisms. bantam encodes a micro-RNA that promotes growth by simultaneously stimulating proliferation and preventing apoptosis, probably through suppressing the translation of specific mRNA targets (Brennecke et al. 2003). Slimfast is an amino acid transporter, the analysis of which revealed that the larval fat body acts as a nutrient sensor that can regulate organismal growth through a systemic mechanism (reviewed in Bradley and Leevers 2003).

Yet another screening method has been to search for genetic modifiers of growth-sensitized phenotypes. This more subtle approach can uncover important growth genes that might otherwise be missed. For example, a screen for modifiers of the big eye resulting from CycD–Cdk4 overexpression identified mutations in the genes encoding Hif-1 prolyl hydroxylase (Hph) (Frei and Edgar 2004) and the mitochondrial RP mRpL12 (Frei et al. 2005). In a different screen, overexpression of either scylla (scy) or charybdis (char) was found to suppress the big-eye phenotype achieved by coexpression of Akt1 and Pdk1 (Reiling and Hafen 2004). Interestingly, Hph, mRpL12, Scy, and Char are all involved in linking oxygen sensing to growth control.

It is clear that a wide variety of Drosophila growth genes have been identified through loss-of-function, overexpression, and modifier screens conducted in whole larvae, specific adult organs, or discrete clones of cells. We wished to identify additional factors that affect organ growth without altering pattern. To do this, we have screened for randomly induced mutations that dominantly enhance or suppress a small-wing phenotype obtained by expression of kinase-dead PI3K. Such a screen may be expected to identify factors that are limiting for organ growth in vivo.

MATERIALS AND METHODS

Fly stocks:

Fly strains are described at FlyBase (http://flybase.bio.indiana.edu/) and are available from the Bloomington Drosophila Stock Center when a reference is not given.

Basic strains:

w¹¹¹⁸-iso, Oregon-R, or y w was used as a control strain. MS1096-GAL4 (Capdevila and Guerrero 1994), UAS-Dp110^KD (Leevers et al. 1996), UAS-Argos.M30-102-1 (A. Michelson, unpublished data), UAS-Argos.III (MacDougall et al. 2004), P[neoFRT]82B ry⁵⁰⁶, and P[hsFLP]1; P[neoFRT]82B P[Ubi-GFP^S65Tnls]3R were the basic strains used. The MS1096-GAL4 UAS-Dp110^KD, MS1096-GAL4 UAS-Argos.M30-102-1, and MS1096-GAL4; UAS-Argos.III stocks were made in our laboratory. Mapping strains are described in the mapping section below.

InR/PI3K pathway mutants:

InR/PI3K pathway mutants were InR³¹ (Fernandez et al. 1995), InR³³⁹ (Brogiolo et al. 2001), chico¹ and chico² (Bohni et al. 1999), Dp110^A P[gH] and Dp110^B P[gH] and p60^A; P[gR10] and p60^B; P[gR10] (Weinkove et al. 1999), Pten^dj189 (Gao et al. 2000), Pten³ (Goberdhan et al. 1999), Akt1^6M4 (H. Stocker and E. Hafen, unpublished data), Akt1¹ (Staveley et al. 1998), foxo²¹ and foxo²⁵ (Junger et al. 2003), Tsc1^Q87X and Tsc1^R453X (Tapon et al. 2001), Rheb^2D1 and Rheb^7A1 (Stocker et al. 2003), and Tor^ΔP and Tor^2L19 (Oldham et al. 2000).

Other mutants:

Other mutants were ncm^SH0931 (Oh et al. 2003), eIF-4a¹⁰⁰⁶ and eIF-4a¹⁰⁶⁹ (Galloni and Edgar 1999), eIF-3p40^k09003, eIF-4E⁰⁷²³⁸, dp^ov1, dp^lv1, and dp^olvR. RP mutants are listed in Table 2.

TABLE 2.

RP mutations tested for dominant modification of MS1096>Dp110^KD wing size

RP	Allele tested	Strength of Minute phenotypea	Enhancement of MS1096>Dp110KD small wingb
RpS3	RpS31	Strong	++++
RpS3A	RpS3A57g	Strong	++++
RpS5a	RpS5a1	Strong	++++
RpS13	RpS131	Strong	+
RpS17	RpS176	Strong	++++
RpS21	oho23B03575	Weak	++
RpS26	RpS2604553	Medium	+
RpS26KG00230	Medium	++
RpLP1	RpLP1beo	Medium	++
RpL5c	RpL52d1	Medium	++
RpL52d2	Medium	+++
RpL38c	RpL382b1	Medium	+
	RpL382b2	Medium	+

^aScored as described in materials and methods.

^bScored as described in Table 1, footnote a. P-values calculated using Student's t-test were <0.01 in every case.

^cData on RpL5 and RpL38 from Table 1 are reproduced here for completeness.

Deficiencies:

The X chromosome deficiency kit (c. 1998) from the Bloomington Drosophila Stock Center was used in the pilot screen. The interacting deficiencies shown in Figure 1E are: S-1, Df(1)BA1; E-1, Df(1)N-8 and Df(1)dm75e19; E-2, Df(1)G4e^LH24i^R; E-3, Df(1)KA14; E-4, Df(1)KA7 and Df(1)HA85; E-5, Df(1)N105 and Df(1)JA26; E-6, Df(1)C246; E-7, Df(1)sd72b; E-8, Df(1)N19; E-9, Df(1)JA27; and E-10, Df(1)HF396 and Df(1)A209. (A full list of the deficiencies used in the pilot screen is available upon request.) pix is deleted/disrupted in Df(3L)Scf-R6 and Df(3L)Scf-R11. prp8 is deleted/disrupted in Df(2R)BSC40, but is not deleted/disrupted in Df(2R)CB21. ncm is deleted/disrupted in Df(2L)TW137, Df(2L)M36F-S5, Df(2L)M36F-S6, and Df(2L)M36F2 [previously called M(2)36F²] but not in Df(2L)T317, Df(2L)H20, Df(2L)TW50, Df(2L)OD15, Df(2L)TW3, or Df(2L)VA16. [Additional data on deficiencies in the vicinity of ncm have been deposited at FlyBase (http://flybase.bio.indiana.edu/).] Deficiencies used in the mapping of RpL38 and RpL5 are described in Marygold et al. (2005).

Figure 1.

The small wing induced by overexpression of Dp110^KD can be dominantly enhanced or suppressed. (A–D, F, and G) Female wings are shown. (A) Wild type. (B) MS1096>Dp110^KD/+. (C and D) Tor mutations dominantly enhance the small-wing phenotype. (E) Several X chromosome deletions dominantly modify the MS1096>Dp110^KD small-wing phenotype. Schematic of the X chromosome shows the individual deficiencies tested. Rectangles indicate the regions deleted in each deficiency strain with lines indicating uncertain break points. Vertical dashed lines group together (conservatively) chromosomal regions that, when deleted, either enhance (solid-bar deficiencies; E) or suppress (shaded-bar deficiency; S) the phenotype. The deficiencies corresponding to these regions are listed in materials and methods. (F) An example of a deficiency that dominantly enhances the small-wing phenotype. (G) Df(1)BA1 dominantly suppresses the small-wing phenotype.

New dumpy alleles:

E-2f¹, E-2f², and E-2f³ are lethal in trans with each other and also in trans with a CyO or SM6a balancer chromosome. Both CyO and SM6a contain a lethal mutation in dumpy (dp) (Lindsley and Zimm 1992). E-2f¹, E-2f², and E-2f³ were subsequently found to be lethal in trans with independently derived lethal (l) dp mutations (dp^lv1 and dp^olvR) and show the characteristic oblique (o) wing and/or thoracic vortex (v) phenotypes in trans with the viable dp^ov1 mutation. We have named these three new dp alleles as dp^2f1-olv, dp^2f2-lv, and dp^2f3-olv, according to the nomenclature of Carlson (1959).

Mutagenesis:

FM7/Y males were fed with 25 mm ethyl methanesulfonate (EMS; Fluka, Buchs, Switzerland) in 10% sucrose solution as described (Newsome et al. 2000) and mated en masse with MS1096>Dp110^KD virgin females.

Mutant phenotype analyses:

Our standard fly culture conditions were to cross ∼10 virgin females to ∼5 males in fresh vials containing standard food; females were allowed to lay eggs for 2 days before transferring the adults to a new tube. Crosses were maintained at 25°. Where necessary, the desired F₁ progeny were identified by selecting against dominant markers on balancer chromosomes.

Crosses to assess the degree of dominant modification of the MS1096>Dp110^KD and MS1096>Aos wing phenotypes by E-Dp110^KD mutations were performed at least three times and two times, respectively. E-Dp110^KD mutations were outcrossed to a wild-type strain under our standard culture conditions to assess dominant effects on wing size; the area of E-Dp110^KD/+ female wings was then compared to that of +/+ female wings. At least two different alleles of InR/PI3K pathway mutants were used in crosses to MS1096>Dp110^KD. Appropriate controls were performed for each test cross under identical culture conditions. Approximately 20 wings were measured for control genotypes and 10 wings were measured for test genotypes. Male and female wings were measured separately. Adult wings were processed and their areas measured as described (Marygold et al. 2005). Statistical analyses were performed using Microsoft Excel: P-values were calculated using a two-tailed Student's t-test assuming equal variances.

Dominant Minute bristle phenotypes of E-Dp110^KD mutants were originally assessed in the stocks; where evident, this was confirmed by outcrossing E-Dp110^KD mutants to a wild-type strain under uncrowded culture conditions and examining F₁ progeny. The RP mutants listed in Table 2 were simultaneously outcrossed to the w¹¹¹⁸-iso strain under uncrowded culture conditions; F₁ progeny were ranked according to their dominant developmental delay and bristle phenotypes were compared to those of w¹¹¹⁸-iso flies raised under identical conditions.

For clonal analyses, lep^3c5 and lep^3c6 were recombined onto the P[neoFRT]82B chromosome arm and crossed to P[hsFLP]1; P[neoFRT]82B P[Ubi-GFP^S65Tnls]3R. Clones were induced at mid-third instar by a 34° heat shock for 10 min. Adult eyes were processed as described (Tomlinson and Ready 1987).

Mapping of mutations:

Mapping was performed for the third and second chromosome enhancer mutations and the second chromosome suppressor mutations using methods described separately below. In each case, the mapping chromosome itself does not dominantly affect wing size or morphology (data not shown).

Third chromosome E-Dp110^KD mutations:

To roughly map lep^3c5 and lep^3c6, recombination was allowed between the mutant chromosomes and a st¹ Sb^sbd-1 e^s ro¹ ca¹ mapping chromosome in the presence of MS1096>Dp110^KD. Approximately 250 lep^3c5 F₁ recombinants and ∼80 lep^3c6 F₁ recombinants were analyzed for the presence or absence of the enhancer mutation. To roughly map the other five E-3c mutations, recombination was allowed between the mutant chromosome and a ru¹ h¹ st¹ ry⁵⁰⁶ e¹ mapping chromosome. Initially, ∼50 individual male recombinants were crossed to MS1096>Dp110^KD female flies to test for the presence or absence of the enhancer mutation to map the mutation between two of the visible markers. In a second round of recombination, ∼100 males in which recombination had occurred between the two markers of interest were tested for the presence or absence of the enhancer mutation. The percentage of frequency of recombination between the E-3c mutation and the individual markers was used to calculate recombination distance in each case.

For the fine mapping of pix, recombinants were generated between the pix^3c3 chromosome and a chromosome containing two closely spaced P[w⁺] elements [l(3)s2383^s2383-l(3)j2B9^j2B9, a recombinant chromosome made in our lab] that flanked the region of interest (ROI) defined by the initial meiotic recombination and deficiency mapping. Whereas the P[w⁺]-P[w⁺] parental flies have a relatively dark eye color because of having two [w⁺] transgenes, recombination events between the P[w⁺] elements generate recombinant flies with a single P[w⁺] element and a visibly paler eye color. Thus, informative recombinants with DNA breaks within the ROI were specifically selected. These recombinants were crossed to the pix^3c2 allele and to MS1096>Dp110^KD to test for the presence of the pix^3c3 mutation and were analyzed by PCR to identify which P[w⁺] element was present. Single-nucleotide polymorphisms (SNPs) in the ROI were identified (supplementary Table S2 at http://www.genetics.org/supplemental/) and used to genotype ∼200 recombinant chromosomes.

Second chromosome E-Dp110^KD mutations:

E-Dp110^KD chromosomes were allowed to recombine with an isogenized wg^Sp-1 Bl¹ L^rm Bc¹ Pu² mapping chromosome. Only wg^Sp-1, Bl¹, and L^rm were used for mapping and we refer to this chromosome as “Wg-Bl-L.” In the first stage of mapping using visible markers, ∼40 recombinant males that contained a chromosomal break between the two terminal markers (Wg and L) were isolated; additional recombinant males were stored at 18°. The initial 40 recombinants were tested for the presence or absence of the E-Dp110^KD mutation by testing for enhancement of MS1096>Dp110^KD and/or noncomplementation of the other mutant allele and then mapping the mutation either between two of the markers or distal to one of the terminal markers. A further ∼100 recombinants (from those stored at 18°) that contained the appropriate chromosomal break were then selected, and the presence or absence of the E-Dp110^KD mutation was again determined. The percentage of frequency of recombination between the E-Dp110^KD mutation and each adjacent marker in these 100+ recombinants was used to determine genetic linkage and so further narrow the ROI.

In the second stage of finer-scale mapping, SNPs between the E-Dp110^KD chromosome and the Wg-Bl-L chromosome were sought within the defined ROI. Initially, SNPs spaced widely in the ROI were found and used to genotype DNA prepared from all 100+ recombinants. More closely spaced SNPs were then used to genotype the progressively fewer informative recombinants remaining until either recombinants or SNPs were exhausted. Further details of this mapping strategy and a crossing scheme can be found in Martin et al. (2001). Second chromosome SNPs are described in supplementary Table S1 (http://www.genetics.org/supplemental/).

S-Dp110^KD:

To map S-2a, recombination was allowed between the S-2a² chromosome and the Wg-Bl-L mapping chromosome in the presence of MS1096>Dp110^KD. Approximately 250 S-2a² F₁ recombinants were analyzed for the presence or absence of the suppressor mutation: S-2a² was thus mapped to 54 cM, which is close to the apterous gene at 55 cM. The following observations strongly suggest that S-2a corresponds to apterous. First, Df(2R)nap1 (41D2–E1; 42B1–3) deletes apterous (41F8), is semilethal in trans with S-2a¹, and weakly dominantly suppresses the MS1096>Dp110^KD small wing. Second, S-2a¹/Df(2R)nap1 escapers and S-2a trans-heterozygous escapers lack wings, similar to certain apterous mutants. Finally, S-2a¹ fails to complement two previously described apterous mutations (ap⁴ and ap^rK568) with escapers again showing a no-wing phenotype.

Molecular biology techniques:

Isolation of genomic DNA from flies, primer design, PCRs, DNA sequencing, and sequence analysis were performed as described (Marygold et al. 2005). Crude DNA preparations from recombinant flies were made by crushing one to two prefrozen flies in 50 μl buffer [10 mm Tris-HCl (pH 8.2), 1 mm EDTA, 25 mm NaCl, 0.01 mg Proteinase K] in 0.2-ml PCR tubes or 96 well plates and incubating at 37° for 25 min and then at 95° for 5 min (to inactivate the Proteinase K); 2 μl of these crude preparations was used in a 25-μl PCR. Primers and specific PCR conditions used to amplify SNP regions, ncm, prp8, pix, and other candidate genes are available upon request; those used to amplify RpL38 and RpL5 are described in Marygold et al. (2005).

For single-strand conformation polymorphism (SSCP) detection, either Phast gels with the Phast electrophoresis system (Pharmacia Biotech, Piscataway, NJ) or GeneGel Excel kits (Amersham Biosciences, Little Chalfont, England) with the GenePhor electrophoresis system (Pharmacia Biotech) were used; PCR products were prepared according to the manufacturer's instructions and electrophoresis was performed at 10°. For restriction fragment length polymorphism (RFLP) detection, 10 μl of PCR product was digested in a total volume of 20 μl and incubated according to the instructions provided by the supplier [New England Biolabs (Beverly, MA) or Roche (Indianapolis)]; digested DNA was then subjected to electrophoresis on 1% agarose gels.

SNP identification and detection:

In a few cases, previously reported SNPs (Hoskins et al. 2001) were tested for dimorphism between the mapping and E-Dp110^KD chromosomes and used for genotyping recombinant chromosomes. New SNPs were detected in PCR-amplified genomic DNA by sequencing ∼1-kb regions or by performing SSCP analysis of 200- to 300-bp regions and then identifying bases/regions dimorphic between the mapping and E-Dp110^KD chromosomes. See supplementary Tables S1 and S2 (http://www.genetics.org/supplemental/) for a full listing of all SNPs identified. SNPs between the mutagenized second chromosome and the Wg-Bl-L chromosome were detected by sequencing DNA amplified from E-2-iso/CyO-iso and Wg-Bl-L-iso/CyO-iso heterozygous flies (where iso indicates an isogenized chromosome): as the CyO-iso chromosome is identical in these stocks, any SNPs detected result from differences between the E-2-iso and Wg-Bl-L-iso chromosomes. SNPs between the pix^3c3 chromosome and the l(3)s2383^s2383-l(3)j2B9^j2B9 chromosome were detected by sequencing DNA or by SSCP analysis of DNA amplified from pix^3c3-iso/+iso and l(3)s2383^s2383-l(3)j2B9^j2B9-iso/+iso flies (where the +iso third chromosome was derived from an isogenized y w stock). Recombinant chromosomes, in trans with either CyO-iso or +iso third chromosome, were genotyped by RFLPs, SSCPs, or direct DNA sequencing.

Bioinformatics:

The design of primer pairs and the determination of SNP locations on the physical map were based on the D. melanogaster Genome Release 3 and relied upon the BDGP GadFly annotation database. Orthologs of the Ncm and Prp8 proteins were identified by using the D. melanogaster sequences to perform BLASTp searches of the nonredundant peptide databases (http://www.ncbi.nlm.nih.gov/BLAST/). The highest-scoring hits for a variety of species are presented in Figure 7. CLUSTALW (http://pbil.ibcp.fr) was used for sequence alignment and calculation of percentage of identity. Pfam 16.0 (http://www.sanger.ac.uk/Software/Pfam) and SMART (http://smart.embl-heidelberg.de/) were used to identify protein motifs.

Figure 7.

E-2e and E-2a correspond to the splicing factors Prp8 and Ncm, respectively. (A) Schematic of the D. melanogaster Prp8 protein. The prp8^2e1 and prp8^2e2 mutations alter highly conserved amino acid residues within the 3′ splice site fidelity region 3.2 (3.2). Also shown is a putative nuclear localization sequence (NLS), a putative RNA recognition motif (RRM), and an MPN domain—see text for details. Amino acid numbering is given for each Prp8 ortholog. Note that the A. thaliana genome contains two different genes that encode slightly different Prp8 proteins. (B) Schematic of the D. melanogaster Ncm protein and orthologs from different species. Domains shown are: MIF4G domain (solid boxes), MA3 domain (shaded boxes), RS domains (diagonal hatching), and poly-serine tract (stippled). The ncm^2a2 and ncm^2a1 mutations each create premature STOP codons at amino acid positions 301 and 768, respectively. The ncm^SH0931 P-element insertion (triangle) is located within the predicted C-terminal RS domain. Ncm orthologs from different species are shown with their percentage identity over a ∼500-amino-acid stretch between conserved YIPP and IGLG tetrapeptides in the central region of the D. melanogaster protein; amino acid sequences outside of this central sequence are far less conserved. Note that the Anopheles gambiae Ncm protein is predicted from a partial cDNA clone that is missing amino- and/or carboxy-terminal sequences. Abbreviations and accession numbers (Prp8/Ncm) are: D.m., D. melanogaster (NP_610735.1/NP_609877.2); A.g., A. gambiae (XP_308873.1/XP_317618.1); H.s., Homo sapiens (NP_006436.2/XP_034594.2); C.e., C. elegans (NP_498785.1/NP_496363.1); A.t., Arabidopsis thaliana (NP_178124.1 and NP_195589.2/NP_178208.1); S.p., S. pombe (NP_593861.1/Q9P6R9); and S.c., S. cerevisiae (NP_012035.1/NP_011794.1).

RESULTS

Characterization of the MS1096>Dp110^KD small-wing phenotype:

Overexpression of a kinase-dead version of the catalytic subunit of Drosophila PI3K (Dp110^KD) has a dominant-negative effect on growth (Leevers et al. 1996). Expression of Dp110^KD in Drosophila wing imaginal discs using the MS1096-GAL4 line results in small adult wings as a result of decreased cell size and cell number (Leevers et al. 1996). This reduction in wing area is dependent on transgene copy number: female flies heterozygous for a recombinant X chromosome containing MS1096-GAL4 and UAS-Dp110^KD (referred to as “MS1096>Dp110^KD”) have wings that are ∼25% smaller than those of wild type, whereas wings of MS1096>Dp110^KD homozygous females or hemizygous males are ∼50% smaller (compare Figure 1, A and B, to Figure 5, A and B).

Figure 5.

lep phenotypes. (A–D) Male wings are shown. (A) Wild type. (B) MS1096>Dp110^KD/Y. Note that MS1096>Dp110^KD/Y wings are more reduced in size compared to controls than MS1096>Dp110^KD/+ female wings (compare to Figure 1, A and B), and that the enhancement of the small-wing phenotype is correspondingly stronger in male compared to female wings. (C) pix^3c2 is a strong dominant enhancer of the small-wing phenotype. (D) Heterozygosity for lep^3c6 results in an extreme enhancement of the phenotype and wings are often reduced to stubs (not shown). (E) lep^3c5 clones in the adult eye do not comprise small cells. (Left) Cartoon showing the extent of a w^+/+ twin spot (dark shading) and its accompanying w^−/−, lep^3c5 clone (light shading). Most lep^3c5 clones are found at the dorsal-ventral boundary of the eye similar to the one shown. (Right) Cross section of the same eye with the mutant clone marked by absence of pigment granules (outlined). Note that the size and arrangement of the mutant ommatidial cells are similar to those in the surrounding wild-type tissue.

We investigated whether the MS1096>Dp110^KD small-wing phenotype is sensitive to heterozygous mutation of genes encoding known growth regulators. First, we tested mutations in genes encoding core transducers of InR/PI3K signaling. Surprisingly, null mutations in Dp110, p60, Pten, Akt1, and foxo each fail to dominantly modify the small-wing phenotype, although InR³³⁹ and chico² are weak enhancers (data not shown). We also tested mutations in genes encoding members of the Tor pathway, as this also promotes growth in Drosophila and there are several potential links between Tor and InR/PI3K signaling (Leevers and Hafen 2004). Although strong hypomorphic or null mutations in Tsc1 or Rheb do not dominantly modify MS1096>Dp110^KD wing size, null mutations in Tor clearly enhance the phenotype (Figure 1, C and D). The fact that the majority of InR/PI3K and Tor pathway mutants fail to dominantly modify the small-wing phenotype may be simply because these gene products are not limiting for wing growth, at least under conditions in which InR/PI3K signaling is already impaired. Tor itself may be an exception to this general observation because it promotes growth in response to multiple inputs and has several growth-regulatory targets (Neufeld 2004; see discussion). Together, these observations demonstrate that the small-wing phenotype is relatively insensitive to a reduction in the genetic dose of several well-characterized growth regulators.

We wished to use the MS1096>Dp110^KD small-wing phenotype in a large-scale modifier screen to identify additional molecules that are important for organ growth in vivo. However, in light of the findings described above, we first performed a small-scale pilot test using deficiencies that together span most of the X chromosome. Fourteen of the 41 deficiencies examined dominantly enhanced the small-wing phenotype whereas one deficiency suppressed it (Figure 1, E–G). Significantly, several of the enhancer deficiencies remove overlapping sections of the chromosome. This means that there are at least 10 regions of the X chromosome that, when deleted, enhance the small-wing phenotype (Figure 1E), suggesting that the MS1096>Dp110^KD phenotype is indeed sensitive to the dose of other genes. We therefore decided to carry out a larger-scale dominant genetic interaction screen.

Isolation of new growth genes through a dominant modifier screen:

Chemically mutagenized males were mated with MS1096>Dp110^KD homozygous females and the progeny were examined under the dissection microscope for enhancement or suppression of the test phenotype (Figure 2A and materials and methods). Compared to controls, MS1096>Dp110^KD/Y male wings are more reduced in size than MS1096>Dp110^KD/+ female wings (see above). We were therefore able to simultaneously screen for modification of our test phenotype at two different strengths (Figure 2A).

Figure 2.

Summary of the screen. (A) Crossing scheme for identifying dominant modifiers of the MS1096>Dp110^KD small-wing phenotype. EMS-mutagenized FM7/Y males were mated with homozygous MS1096>Dp110^KD virgin females. F₁ flies with mutations that either enhanced or suppressed the small-wing phenotype were individually backcrossed to the MS1096>Dp110^KD strain to check whether the modifying effect was transmitted to the next generation. The FM7 chromosome itself does not modify the small-wing phenotype (data not shown). Modifiers that bred true were retained, mapped to chromosomes, and balanced. “E/S” represents possible enhancer or suppressor point mutations, respectively, induced by the EMS mutagen. “Y” indicates the Y chromosome and thus identifies the male flies in the crossing scheme. (B) Flow diagram indicating the number and type of modifiers retained at each stage of the initial chromosomal mapping and complementation analyses. Numbers in parentheses in the bottom boxes indicate the number of mutant alleles in each complementation group. Note that the E-3c group was subsequently found to comprise two lethal complementation groups of two alleles and three single hits—see text for details.

The initial results of the screen are summarized in Figure 2B. We screened ∼150,000 flies and identified 38 enhancers, collectively termed E-Dp110^KD mutations, and 4 suppressors, termed S-Dp110^KD mutations. Most of the modifiers were initially identified in MS1096>Dp110^KD/Y male wings, suggesting that the stronger male phenotype provided the more effective background in which to detect modifying mutations. Where tested, the majority of these mutations also dominantly modified MS1096>Dp110^KD/+ female wings, although the effect was often weak.

Complementation and preliminary phenotypic analyses:

We conducted a lethal complementation analysis among the E-Dp110^KD and S-Dp110^KD mutations. As genes required for growth may not be essential for viability (e.g., Bohni et al. 1999; Montagne et al. 1999), we also grouped together mutations that, in trans, gave rise to viable adults with abnormally sized wings and/or bodies. Mutations that do not display lethality or any obvious size defects when in trans with other mutations were classed as “single hits” and were not analyzed further in this study. By these criteria, it is possible that mutant combinations that cause only mild effects on growth rate or final size may have been missed.

E-Dp110^KD mutations on the second chromosome:

Complementation analysis of the E-Dp110^KD mutants mapping to the second chromosome initially yielded six lethal complementation groups named Enhancer on chromosome 2-complementation groups a to f (E-2a to -f) (Figure 2B). Of these, E-2c and E-2f were not analyzed further. All five E-2c mutations cause ectopic wing vein tissue when combined with MS1096>Dp110^KD (data not shown), raising the possibility that their ability to reduce MS1096>Dp110^KD wing size may be an indirect consequence of their effect on patterning. Of the four mutations composing the E-2f group, E-2f¹, E-2f², and E-2f³ are lethal in trans with each other and were found to contain lethal mutations in the dumpy gene by complementation analyses (see materials and methods). However, independently sourced dumpy mutations (dp^ov1, dp^lv1, and dp^olvR) fail to enhance the MS1096>Dp110^KD phenotype (data not shown), suggesting that the E-2f¹, E-2f², and E-2f³ chromosomes contain second-site enhancer mutations. Moreover, it was possible to separate the dumpy mutation from the E-Dp110^KD mutation on the E-2f¹ chromosome through recombination (data not shown). Indeed, it has been noted previously that dumpy alleles are detected at high rates in EMS mutagenesis screens, probably owing to the enormous size (>100 kb) of the dumpy locus (Jenkins 1967; Wilkin et al. 2000). Significantly, the E-2f⁴ chromosome does not contain a dumpy mutation and is lethal in trans with E-2f¹ but not with E-2f² or E-2f³. We surmised that E-2f⁴ and the second-site enhancer mutation on the E-2f¹ chromosome are likely to disrupt the same gene and named this distinct enhancer locus “E-2g” (Figure 2B). The second-site enhancers on the E-2f² and E-2f³ chromosomes were reclassified as single hits and were not analyzed further.

Phenotypes exhibited by the second chromosome E-Dp110^KD mutants are consistent with the corresponding wild-type genes encoding proteins with vital or growth-promoting roles. E-2a trans-heterozygotes are embryonic lethal, and E-2b¹/E-2b² animals die either as embryos or soon after hatching (data not shown). In contrast, animals trans-heterozygous for E-2d or E-2e or E-2g mutations survive for several days as larvae, although they are abnormally small and never pupate (data not shown). This phenotype is reminiscent of mutations in known growth-promoting genes (e.g., Galloni and Edgar 1999; Weinkove et al. 1999). We also noted that adult flies heterozygous for E-2b or E-2d mutations have small, slender bristles and show a significant delay in their development to adulthood (Marygold et al. 2005). These dominant phenotypes are typical of the Minute class of mutations that are thought to disrupt RP genes (Lambertsson 1998) and are therefore consistent with a role for the E-2b and E-2d products in protein synthesis.

E-Dp110^KD mutations on the third chromosome:

The E-Dp110^KD mutations that mapped to the third chromosome did not fall into simple lethal complementation groups. Seven mutations, initially grouped together as E-3c^1-7, each exhibit either lethality or reduced body size when placed in trans with at least one other E-3c mutation (Figures 2B and 3A). However, the overall complementation pattern is rather complex. For example, E-3c⁵ is lethal in trans with E-3c² or E-3c⁴, but E-3c²/E-3c⁴ trans-heterozygotes are phenotypically wild type (Figure 3A). This may indicate either that E-3c is a genetically complex single locus or that the E-3c group actually comprises several distinct loci that can show particularly strong genetic interactions.

Figure 3.

Analysis and mapping of E-3c mutations. (A) E-3c complementation matrix. Crosses indicate lethality; “small” indicates that trans-heterozygous adults had abnormally small bodies and/or wings; WT, wild type. (B) The location of E-3c genes on a genetic map of the third chromosome is shown with respect to the ru h st ry and e recessive markers used in the mapping.

To further investigate the potential allelism of the seven E-3c mutations, their ability to enhance the MS1096>Dp110^KD small-wing phenotype was roughly mapped by meiotic recombination using visible markers (materials and methods; Figure 3B). E-3c² and E-3c³ each map 1 cM to the right of hairy and are lethal in trans, indicating that they are likely to disrupt the same gene. Similarly, E-3c⁵ and E-3c⁶ are probably allelic as each maps ∼2 cM to the left of ebony and they are lethal in trans. We have named the E-3c²/E-3c³ gene pixie (pix) and the E-3c⁵/E-3c⁶ gene leprechaun (lep), owing to the small body/wing phenotypes produced by these mutations when in trans with other E-3c mutations or when heterozygous in the MS1096>Dp110^KD background. Both E-3c¹ and E-3c⁷ map close to the centromere but are unlikely to disrupt the same gene as E-3c¹/E-3c⁷ flies are viable and phenotypically wild type (Figure 3A). Finally, E-3c⁴ maps 6 cM to the right of rosy and is a single hit. Thus the E-3c group appears to comprise mutations at five different loci: pix, lep, and the three single-hit loci E-3c¹, E-3c⁴, and E-3c⁷ (Figure 3B). We decided to include all the E-3c mutations in subsequent analyses, owing to the uniquely strong genetic interactions between them.

Of the lethal combinations among the E-3c group, several are larval lethal but survive as abnormally small larvae for an extended period, akin to E-2d, E-2e, and E-2g trans-heterozygotes. Furthermore, both pix mutants and E-3c⁷ mutants display the dominant Minute phenotypes of small bristles and developmental delay similar to E-2b and E-2d heterozygotes (data not shown). Together, these observations are consistent with the respective protein products having a positive role in growth.

Four additional E-Dp110^KD mutations on the third chromosome do not show obvious phenotypes in trans with the E-3c group of mutations. Among these four, two were provisionally grouped as E-3a because they have a trans-heterozygous phenotype of small wings and halteres (Figure 2B; data not shown), suggesting that they could affect thoracic disc growth by disrupting the same gene. However, one of these mutant stocks was lost and so E-3a could not be analyzed further. The remaining two third chromosome mutants show no phenotype in trans with any other third chromosome mutants and were not further analyzed (Figure 2B).

S-Dp110^KD mutations:

All four suppressors of the MS1096>Dp110^KD small-wing phenotype are on the second chromosome and form a single semilethal complementation group named S-2a (Figure 2B). Various mapping and phenotypic data suggest that S-2a corresponds to apterous (materials and methods). Apterous is a transcription factor that directs wing development from the earliest stages and, among other targets, induces expression of beadex (Milan et al. 1998). Significantly, the MS1096-GAL4 transgene used to drive Dp110^KD expression in the screen is inserted in the second intron of the beadex gene (Milan et al. 1998). We therefore considered it likely that that the S-2a/apterous mutations were isolated as suppressors of the MS1096>Dp110^KD phenotype because they dominantly reduce the expression levels of the MS1096-GAL4 driver. We therefore did not analyze this group further.

In summary, our screen initially yielded seven complementation groups on the second chromosome (E-2a, E-2b, E-2c, E-2d, E-2e, E-2g, and S-2a) and two groups of mutations on the third chromosome (E-3a and E-3c). Of these, we selected the E-2a, E-2b, E-2d, E-2e, E-2g, and E-3c groups for further analysis (Table 1, column 1). All these mutations were isolated as dominant enhancers of MS1096>Dp110^KD wing size (Table 1, column 3) and are therefore predicted to disrupt genes encoding proteins that normally promote/permit growth. Other phenotypes presented by these mutants are consistent with this idea (see above).

TABLE 1.

E-Dp110^KD complementation groups and summary of phenotypic analyses

					Enhancement of MS1096>Aosc
Complementation group	Alleles	Enhancement of MS1096>Dp110KD small winga	Minute bristle phenotype?b	Dominant reduction in wing size?	Small wing?	Wing-vein loss?
E-2a	E-2a1	+ + +	No	No	Yes	No
E-2a2	+ + +	No	No	Yes	No
E-2b	E-2b1	+	Yes	No	Yes	No
E-2b2	+	Yes	No	Yes	No
E-2d	E-2d1	+ +	Yes	No	Yes	No
E-2d2	+ + +	Yes	No	Yes	No
E-2e	E-2e1	+	No	No	Yes	No
E-2e2	+ +	No	No	Yes	No
E-2g	E-2g1	+ + +	No	No	Yes	Yes
E-2g2	+ +	No	No	Yes	Yes
E-3c	E-3c1	+ +	No	No	Yes	Yes
pix3c2	+ + +	Yes	Nod	Yes	No
pix3c3	+ +	Yes	No	Yes	No
E-3c4	+ +	No	No	Yes	No
lep3c5	+ + +	No	No	Yes	No
lep3c6	+ + + +	No	No	Yes	No
	E-3c7	+ +	Yes	No	Yes	No

^aThe relative degree of enhancement is a cumulative score that reflects results seen in both female and male wings collected from at least three separate crosses. Approximately, the number of “+” signs corresponds to the mean size of MS1096>Dp110^KD/+; E-Dp110^KD/+ female wings compared to MS1096>Dp110^KD/+ female wings as follows: +, >95%; ++, 91–95%; +++, 86–90%; ++++, <86%. P-values calculated using Student's t-test were <0.01 in every case.

^bNota were assessed for the Minute small and slender bristle phenotype.

^cMS1096>Aos/+; E-Dp110^KD/+ female wings were compared to MS1096>Aos/+ female wings. Wings were scored for significant size (mean wing area) and venation (presence and integrity of vein L3) differences.

^dpix^3c2 heterozygotes do show a mild reduction in wing size when reared under uncrowded, carefully synchronized culture conditions, but not under our standard culture conditions.

Dominant effects of the E-Dp110^KD mutations on wild-type wing size:

Before analyzing the E-Dp110^KD mutations further, we wished to test whether they cause a dominant reduction in wing size in a wild-type genetic background. If this were the case then their apparent ability to dominantly reduce MS1096>Dp110^KD wing size would be purely an additive effect rather than an informative genetic interaction. We outcrossed the E-2a, E-2b, E-2d, E-2e, E-2g, and E-3c mutations to a wild-type strain and measured wing size in the progeny. None of the E-Dp110^KD mutations were found to dominantly reduce wing size under standard culture conditions (Table 1, column 5), and they were therefore considered to be bona fide enhancers of the MS1096>Dp110^KD phenotype. However, we did observe that pix^3c2 heterozygotes are reduced in body and wing size when reared under uncrowded, carefully synchronized culture conditions (C. M. A. Coelho, unpublished results). Thus, it is possible that the dominant enhancement of the MS1096>Dp110^KD small-wing phenotype by the pix^3c2 allele, although not pix^3c3, is partly the result of an additive effect rather than a genuine genetic interaction. We also noted that several E-Dp110^KD mutants actually have larger wings than controls when crossed into a wild-type background (data not shown). Regardless of the explanation for this unexpected observation (see discussion), it makes the ability of these E-Dp110^KD mutations to dominantly reduce MS1096>Dp110^KD wing size more emphatic.

Dominant effects of the E-Dp110^KD mutations on the small-wing phenotype induced by Argos expression:

We next examined whether the selected E-Dp110^KD mutations specifically impair signaling through the InR/PI3K pathway. The Drosophila epidermal growth factor receptor (EGFR) pathway has been shown to promote wing growth (Diaz-Benjumea and Hafen 1994) and MS1096-GAL4-mediated expression of the EGFR inhibitor, Argos (Aos), reduces wing size in addition to affecting patterning of the wing (Howes et al. 1998; Figure 4B). We therefore tested whether the selected E-Dp110^KD mutations dominantly modify MS1096>Aos wing size. Significantly, all E-Dp110^KD mutations are dominant enhancers of the MS1096>Aos small-wing phenotype (Figure 4 and Table 1, column 6). This strongly suggests that the E-Dp110^KD mutations isolated in our original screen do not specifically affect either InR/PI3K- or EGFR-mediated wing growth. Instead, the E-Dp110^KD genes are likely to encode factors that either are common to both pathways or perhaps act in parallel growth-promoting pathways. One possible exception to this conclusion is lep: lep^3c5 and lep^3c6 are very strong enhancers of MS1096>Dp110^KD wing size but are among the weaker enhancers of MS1096>Aos wing size (Figure 5, A–D, and Table 1, column 3; data not shown). This suggests that the lep gene product could specifically affect InR/PI3K signaling.

Figure 4.

Dominant modification of MS1096>Aos wing phenotypes by E-Dp110^KD mutations. (A–D) Female wings are shown. (A) Wild type. (B) MS1096>Aos/+. (C) Heterozygosity for most E-Dp110^KD mutations (including E-2d², shown) enhances the size phenotype but does not enhance the venation phenotype. In fact, a slight suppression of the venation phenotype is often observed. (D) Heterozygosity for E-2g¹ enhances both the size and venation phenotypes of the MS1096>Aos wing. Similar results were obtained for E-2g² and E-3c¹.

To test this idea, we induced lep^3c5 and lep^3c6 clones in the eye imaginal disc and examined adult eyes for cell size defects. Previous studies have found that adult eye clones that are mutant in Dp110 or other components of the InR/PI3K pathway show a characteristic reduction in both cell size and clonal growth (e.g., Bohni et al. 1999; Weinkove et al. 1999). lep^3c5 clones in adult eyes are recovered at a reduced frequency and size compared to their twin spots, consistent with a role for the wild-type lep gene product in promoting growth (Figure 5E). However, cell size is not detectably reduced in the surviving lep^3c5 clones compared to neighboring wild-type tissue (Figure 5E). lep^3c6 eye clones could not be examined as they do not survive to adulthood. We conclude that Lep is unlikely to act specifically in InR/PI3K signaling but is probably required more generally to promote growth.

Dominant effects of the E-Dp110^KD mutations on the loss-of-veins phenotype induced by Argos expression:

A potential pitfall of screening for modifiers of a phenotype induced by transgene expression is that the isolated mutations affect expression of the transgene rather than modify the test phenotype per se. Indeed, we consider that S-2a/apterous was isolated in our screen for this reason (see above). Similarly, the ability of all the selected E-Dp110^KD mutations to enhance both the MS1096>Dp110^KD and MS1096>Aos small-wing phenotypes may arise simply because they dominantly increase MS1096-GAL4 expression. If this were so, the E-Dp110^KD mutations should also enhance other MS1096-GAL4-induced phenotypes that are unrelated to growth. We therefore examined whether the distinct loss-of-veins phenotype evident in MS1096>Aos wings is enhanced by the selected E-Dp110^KD mutations (Figure 4; Table 1, column 7). Only the E-2g mutations and E-3c¹ significantly enhanced the loss-of-veins phenotype, indicating that these mutations may dominantly increase GAL4 driver expression. Alternatively, these three mutations may somehow affect EGFR-mediated patterning processes in addition to growth. As either explanation is inconsistent with our intention of identifying novel factors that affect growth without disturbing pattern, the E-3c¹ and E-2g genes were not analyzed further in this study.

On the basis of the secondary screens described above, we conclude that E-2a, E-2b, E-2d, E-2e, pix, lep, E-3c⁴, and E-3c⁷ correspond to bona fide growth genes that appear to be generally required to promote growth in vivo. We next sought to finely map and thus identify some of these genes, focusing on those represented by more than one mutant allele as this greatly facilitates the mapping process.

pix is the ortholog of RLI1:

The ability of both pix^3c2 and pix^3c3 to enhance the small-wing phenotype was mapped initially by meiotic recombination to a region just proximal to hairy on chromosome arm 3L (see above and Figure 3B). Complementation tests with deficiencies in the region confirmed this location and delimited the chromosomal region of interest (ROI) containing pix to cytological bands 66E3–66F3 (materials and methods). Higher-resolution mapping of pix was achieved using a similar approach to that described in Jennings et al. (2004). Recombinants were generated between a pix^3c3 chromosome and a chromosome containing two closely spaced P[w⁺] elements that flank the ROI (materials and methods). In this way, informative recombinants were specifically selected and then genotyped using SNPs (see materials and methods).

SNP-based mapping of the pix^3c3 mutation showed that the corresponding gene lies in a ∼10-kb interval centered at ∼3L:8.9 Mb that contains five predicted genes. DNA sequence analysis of these five genes revealed that CG5651 contains missense mutations in both pix mutant strains. Furthermore, additional mutant alleles of CG5651, generated in an independent mutagenesis screen (Dahanukar et al. 1999), are lethal in trans with pix^3c2 and pix^3c3 (C. M. A. Coelho, unpublished results). CG5651/pix encodes the Drosophila ortholog of yeast RLI1, a protein implicated in translation initiation and ribosome biogenesis (C. M. A. Coelho, C. Bunn, D. Andersen and S. J. Leevers, unpublished results; Dong et al. 2004; Kispal et al. 2005; Yarunin et al. 2005). A detailed analysis of the pix gene and its function will be described elsewhere.

Identification of E-2a, E-2b, E-2d, and E-2e:

To map the four E-Dp110^KD's on the second chromosome, we adapted the hybrid mapping strategy developed by St. Johnston and colleagues for mapping mutations on the third chromosome (Martin et al. 2001). This two-step mapping process initially uses visible markers and a few recombinants to map mutations at a low resolution to a ROI and then uses SNPs and a high density of recombinants within the ROI to map the mutations to high resolution (Figure 6A; materials and methods). In addition, we used small-scale deletion mapping to confirm and sometimes improve upon the SNP-based mapping results.

Figure 6.

Mapping E-Dp110^KD mutations on the second chromosome. (A) Summary of the mapping of E-2e. A cartoon of the entire second chromosome and successive zooms of particular regions containing E-2e are shown on the left. The numbers to the right indicate the size of the euchromatic (euch.) interval known to contain E-2e and the number of informative recombinants (rec.) remaining at each stage of the mapping. (i) Initial mapping using dominant visible markers showed E-2e to be between Bristle and Lobe, with greater linkage to the Lobe locus at ∼2R:10 Mb. SNPs (asterisks) were therefore sought and used for genotyping of recombinants in an ROI of ∼2R:4–9 Mb. (ii–iv) Successive rounds of higher-resolution SNP-based mapping. (iv) The ∼94 -kb region defined by SNP-based mapping contains 12 predicted genes (solid rectangles). Mutations were identified in the CG8877 coding region in the E-2e¹ and E-2e² lines. Genomic coordinates given in i–iv refer to chromosome arm 2R, are given in megabases, and are derived from Release 3.2 of the genome sequence. (B) The location of E-Dp110^KD genes on a physical map of the second chromosome is shown with respect to the wg^Sp1, Bl¹, and L^rm dominant markers used in the mapping.

E-2e encodes the splicing factor Prp8:

As illustrated in Figure 6, E-2e was mapped by SNP-based methods to a ∼94-kb interval centered at ∼2R:7.2 Mb (∼48E1–4 on the cytological map) that contains 12 predicted genes. This location of E-2e was confirmed by deficiency analysis (materials and methods). Two of the 12 candidate genes for E-2e were eliminated through lethal complementation tests using preexisting mutations. Of the remaining 10 candidates, only 5 encoded proteins with identifiable Pfam domains, and, of these, 3 were considered to be possible growth genes. Sequencing of these 3 candidate genes in the E-2e¹ and E-2e² mutant chromosomes revealed mutations in the CG8877 coding sequence (Figure 7A). These are the first reported mutations in this gene.

Conceptual translation of the CG8877 coding sequence predicts a protein with remarkable similarity to the Prp8 pre-mRNA splicing factor of other species (Mount and Salz 2000; Figure 7A). For example, human and Saccharomyces cerevisiae Prp8 are, respectively, 89 and 60% identical to the D. melanogaster ortholog over their entire ∼2400 amino acids. Indeed, Prp8 proteins are the most evolutionary conserved proteins in the eukaryotic spliceosome and, moreover, Prp8 is one of the most highly conserved nuclear proteins known (Hodges et al. 1995; Grainger and Beggs 2005). We have therefore renamed the D. melanogaster CG8877 gene “prp8.”

Studies of the yeast and human orthologs have shown that Prp8 has fundamental roles at several steps in spliceosomal assembly and function (reviewed in Grainger and Beggs 2005). It is part of the U5 small nuclear ribonucleoprotein particle (snRNP); the U5·U4/U6 tri-snRNP; and the presplicing, activated, and postsplicing spliceosomal complexes. Furthermore, Prp8 is unique among spliceosomal proteins in that it contacts the intronic branchpoint region and both the 5′ and 3′ splice sites of the pre-mRNA. Thus, Prp8 is thought to provide a vital scaffold and catalytic function in the spliceosome of all eukaryotes. However, despite its remarkable evolutionary conservation, the only obvious motif within Prp8 is a carboxy-terminal MPN/Mov34 domain (Aravind and Ponting 1998). The function of this domain is unclear, but it is also found in proteasome regulatory subunits, eIF-3 subunits, and regulators of transcription factors (Aravind and Ponting 1998). Additionally, Grainger and Beggs have recently defined a putative nuclear localization sequence, an RNA recognition motif (RRM), and a third domain they call the “3′ splice site fidelity region” on the basis of the clustering of S. cerevisiae PRP8 mutations that suppress defects in splicing pre-mRNA 3′ splice site mutations (Grainger and Beggs 2005; Figure 7A). The carboxy-terminal half of this latter region, called “3.2,” shows exceptionally high sequence conservation through evolution and has been postulated to act at the catalytic center of the spliceosome (Grainger and Beggs 2005). Remarkably, both the prp8^2e1 and prp8^2e2 mutations we identified in D. melanogaster prp8 result in amino acid substitutions at closely spaced residues within region 3.2 (Figure 7A). Both these mutations are therefore predicted to cripple Prp8 and spliceosomal function and thus cause severe impairment to the expression of all intron-containing pre-mRNAs.

E-2a also encodes a splicing factor:

SNP-based mapping of E-2a resolved the ROI to a 300-kb interval centered at ∼2L:18.4 Mb (∼36E–F) containing 17 predicted genes (Figure 6B; data not shown). Deficiency-based mapping further delimited the ROI to an interval of 150 kb containing 9 predicted genes (materials and methods). DNA sequence analysis of the E-2a¹ and E-2a² chromosomes revealed mutations in the CG12750 coding sequence (Figure 7B). Furthermore, l(2)SH0931, a previously reported homozygous lethal P-element insertion within CG12750 (Oh et al. 2003; Figure 7B), fails to complement both E-2a¹ and E-2a² alleles and dominantly reduces MS1096>Dp110^KD wing size (data not shown). l(2)SH0931 is a weaker mutation than either E-2a¹ or E-2a² because l(2)SH0931 hemizygotes hatch as slow, sluggish first instars, whereas E-2a¹ or E-2a² hemizygotes are embryonic lethal (data not shown).

The CG12750 protein comprises 1330 amino acids and contains a MIF4G domain followed by a MA3 domain in the center of the protein (Figure 7B). MIF4G and MA3 are α-helical domains that are also arranged in tandem in the translation initiation factor eIF-4G (Ponting 2000). In eIF-4G, the MIF4G domain has been shown to act as a multisubstrate adaptor, binding RNA, DNA, eIF-4A, and eIF-3 (Ponting 2000). Despite the superficial resemblance between eIF-4G and CG12750, BLAST searches show that CG12750 is most similar to a distinct class of proteins that is found across species and includes Caenorhabditis elegans Nucampholin/LET-858 (Kelly et al. 1997), Schizosaccharomyces pombe Cwf22p and S. cerevisiae Cwc22p (Ohi et al. 2002), and human KIAA1604 protein (Figure 7B). As the C. elegans protein was the first member of this protein family to be named, we have also called the D. melanogaster ortholog “Nucampholin” (Ncm). The ncm^2a1 and ncm^2a2 mutations result in premature truncations of the protein within the MA3 domain and before the MIF4G domain, respectively (Figure 7B).

Several facts indicate that, like Prp8, Ncm and its orthologs perform a key role in pre-mRNA splicing. First, the yeast Cwf22p and Cwc22p proteins were originally identified as components of a Cdc5p/Cef1p-containing splicing complex (Ohi et al. 2002) that has since been implicated in the structural rearrangement and activation of the spliceosome (Makarov et al. 2002; Chan et al. 2003). Second, the KIAA1604 protein was identified in two independent analyses of human spliceosomal components (Jurica et al. 2002; Zhou et al. 2002). Notably, Moore and colleagues specifically isolated proteins within the spliceosomal C complex, thus placing KIAA1604 in the catalytically competent spliceosome (Jurica et al. 2002). Third, D. melanogaster Ncm and several Ncm orthologs contain domains rich in alternating arginine and serine residues (RS domains) and/or a poly-serine tract, in addition to the RRM within the MIF4G domain (Pestova et al. 1996; Boucher et al. 2001; B. J. Blencowe, personal communication) (Figure 7B). Together, these features make Ncm a member of the SR-related protein family of pre-mRNA splicing factors that promote formation of the spliceosomal complex in constitutively spliced pre-mRNAs, as well as regulate splice site selection in alternatively spliced pre-mRNAs (Blencowe et al. 1999). Finally, the strong evolutionary conservation between the different Ncm proteins (Figure 7B) further suggests that they play a critical role in splicing. Consistent with this view, D. melanogaster ncm^2a1 and ncm^2a2 are embryonic lethal mutations and ncm^2a1 or ncm^2a2 mutant clones induced in the imaginal wing disc do not survive (data not shown).

E-2b and E-2d encode RPs:

Both E-2b and E-2d were mapped to the centric heterochromatin of chromosome 2 (Figure 6B). This chromosomal location limited the resolution of the SNP-based mapping approach as both recombination and SNPs are relatively infrequent near the centromere (Berger et al. 2001; Hoskins et al. 2001; Martin et al. 2001). For this reason, we relied on deficiency-based mapping to improve the mapping resolution of the E-2b and E-2d genes. Gene identification was further aided by the fact that E-2b and E-2d mutations cause the dominant Minute bristle phenotype (Table 1, column 4), suggesting that they might disrupt RP-encoding genes (Lambertsson 1998). Indeed, further analyses demonstrated that E-2b corresponds to RpL38 at 41C–E/h46 on 2R, and E-2d corresponds to RpL5 at 40A–B/h35 on 2L. Both these genes encode proteins of the large ribosomal subunit and are thus required for protein synthesis. A detailed analysis of these genes is described elsewhere (Marygold et al. 2005).

Mutations in several other RP genes also reduce MS1096>Dp110^KD wing size:

Of the five E-Dp110^KD genes identified, three encode factors with direct links to mRNA translation (RpL5, RpL38, and Pix), while Prp8 and Ncm are likely to be required in the pre-mRNA splicing process that is prerequisite to efficient translation of most mRNAs (see discussion). We therefore addressed whether other translation factors, and especially RPs, are also limiting for growth of the MS1096>Dp110^KD wing.

As a first approach, we specifically tested a number of existing mutations in genes encoding translation initiation factors and RPs for their ability to enhance the MS1096>Dp110^KD small-wing phenotype. Our original screen was not saturating, and so mutations in such genes could have been missed. Hypomorphic mutations in eIF-3p40, eIF-4a, and eIF-4E each fail to dominantly modify the MS1096>Dp110^KD small-wing phenotype (materials and methods and data not shown). However, this negative result may simply be because these particular mutations do not significantly reduce eIF expression levels (see discussion). In contrast, all 13 RP mutations tested dominantly reduced MS1096>Dp110^KD wing size (Table 2).

All of the RP mutants tested exhibit the classic Minute phenotypes of developmental delay and/or short, slender bristles, albeit to differing extents (Table 2). Notably, there is a good correlation between the strength of the Minute phenotype and the degree to which the RP mutants dominantly enhance the small-wing phenotype (Table 2). As the strength of Minute phenotype reflects the extent to which protein synthesis is impaired in these animals (e.g., Saeboe-Larssen et al. 1998), it follows that the fly's protein synthetic capacity is a major determinant of MS1096>Dp110^KD wing size. RpS13¹ and oho23B⁰³⁵⁷⁵ appear to be exceptions to this general trend (Table 2), perhaps indicating a degree of specificity in the sensitivity of the MS1096>Dp110^KD wing size to particular RPs. However, these two discrepancies may be explained by the fact that dominant phenotypes, in this case both the Minute phenotype and the degree of enhancement of the small wing, are often subject to the effects of secondary mutations in the genetic background.

As a second way of testing whether lowering the genetic dose of other RPs or translation factors enhances the MS1096>Dp110^KD small-wing phenotype, we examined whether some of the single hits isolated in the original screen exhibit Minute bristles. Of 10 single-hit mutants examined, 4 have a clear Minute bristle phenotype (data not shown). We conclude that protein synthesis in general and RP levels in particular are limiting for growth under conditions where PI3K-signaling has been compromised.

DISCUSSION

Screening for dominant modifiers of the MS1096>Dp110^KD small-wing phenotype:

We have conducted a screen for genes that are important for organ growth in vivo. We used the GAL4/UAS system to overexpress a kinase-dead version of the catalytic subunit of PI3K in the Drosophila wing. The resulting small-wing phenotype provided a sensitized background in which mutations that affect wing growth could be detected. A total of 38 enhancer mutations and 4 suppressor mutations were initially isolated. Various secondary screens were employed that served to focus our studies on those modifiers that were considered most likely to correspond to bona fide growth genes. For example, three groups of modifiers (S-2a, E-2g, and E-3c¹) were eliminated from further investigation because they were judged likely to affect expression of the MS1096-GAL4 driver rather than modifying the MS1096>Dp110^KD wing size phenotype per se. Importantly, none of the selected enhancers, with the possible exception of pix^3c2, dominantly reduced adult wing size in a wild-type background. This demonstrates that these E-Dp110^KD mutations each reduce MS1096>Dp110^KD wing size as a result of a genuine genetic interaction rather than as an additive effect, and that their wild-type gene products are limiting for wing growth under conditions in which InR/PI3K signaling is compromised.

We identified the genes corresponding to five of the bona fide E-Dp110^KD enhancer groups: two encode splicing factors and the other three encode factors required for protein translation (Table 3). The mutations generated in these five genes have allowed the first genetic and phenotypic characterization of these loci in Drosophila. In addition to these five genes, several other modifiers remain to be identified. These include at least 11 gene regions uncovered by our pilot screen using X chromosome deficiencies, the lep gene, and several single-hit loci. On the basis of our preliminary characterization of the MS1096>Dp110^KD small-wing phenotype, a mutation in Tor or possibly chico or InR might be found among the single-hit enhancers, although this has not been tested.

TABLE 3.

Molecularly identified E-Dp110^KD genes

Complementation group	Gene disrupted	Predicted molecular function
E-2a	ncm (CG12750)	Splicing
E-2b	RpL38 (CG18001)	Translation
E-2d	RpL5 (CG17489)	Translation
E-2e	prp8 (CG8877)	Splicing
E-3c2/E-3c3	pix (CG5651)	Translation

Although a large number of flies were screened, relatively few modifying mutations were identified, relatively small complementation groups were obtained, and several mutations that might have been predicted to have a dominant modifying effect did not do so. One trivial explanation for the former two of these observations is that we were screening for a quantitative phenotype (wing area) that lacked an internal control, meaning that subtle interactions could have been missed. It is also possible that only a restricted set of factors—perhaps those that integrate a number of growth regulatory signals—are limiting for growth of the MS1096>Dp110^KD wing. This idea is consistent with the fact that over half of the mutations isolated in the original screen were put into just 10 complementation groups. Furthermore, of mutations in several core members of the InR/PI3K or Tor pathways, only mutations in Tor itself, which promotes growth in response to multiple inputs and has several growth regulatory targets, clearly dominantly affected MS1096>Dp110^KD wing size.

A third possibility that could explain the output of our screen is that the MS1096>Dp110^KD small-wing phenotype or growth in general is relatively insensitive to small changes in gene dosage/protein levels. In this scenario, our screen would have been biased toward detecting rare antimorphic mutations or hypomorphic/null mutations of haplo-insufficient genes. Indeed, our analyses of the five E-Dp110^KD genes we identified support this hypothesis. For example, genomic deletions that remove the pix locus do not dominantly enhance the MS1096>Dp110^KD small-wing phenotype, and flies heterozygous for the deletion have a weaker growth phenotype than flies heterozygous for the pix^3c2 or pix^3c3 mutations (C. M. A. Coelho, unpublished data). Also, the various RP mutants identified as enhancers in this study are all dominant mutations as a result of the haplo-insufficiency of these loci. Finally, both ncm mutations result in premature STOP codons within the coding sequence of the gene that may produce truncated Ncm proteins that could act as dominant-negative proteins.

The role of splicing and translation in organ growth:

It is well established that cellular and organ growth depends upon the synthesis of new proteins (Jorgensen et al. 2004). Indeed, several growth-stimulatory pathways, including InR/PI3K signaling, are known to result in increased ribosome biogenesis and/or protein synthesis (Thomas 2000). From this perspective, it is perhaps not surprising that the MS1096>Dp110^KD small-wing phenotype is enhanced by mutations in genes encoding factors required for efficient protein synthesis, such as RPs or Pix. Alternatively, it is possible that suboptimal levels/activity of Pix or certain RPs result in lowered synthesis of a more specific set of growth regulatory targets, as has been proposed for eIF-4a (Galloni and Edgar 1999).

Although translation initiation is thought to be the rate-limiting step in cellular protein synthesis, it is at least feasible that disruption of the preceding steps, such as the processing of pre-mRNA to form mature mRNA transcripts, can also reduce the overall rate of protein production (Matthews et al. 2000). Thus, the identification of splicing factors in our growth screen may reflect that protein synthesis at the level of transcript maturation can be limiting for organ growth. This might especially be the case for prp8 mutant heterozygotes. Here, a 50% reduction in functional Prp8 protein would be expected to decrease the splicing efficiency of all intron-containing pre-mRNAs, but would particularly impair the expression of highly expressed genes required for growth and protein synthesis, such as RP transcripts. Indeed, it is known that expression of certain RPs is controlled at the level of splicing (Jorgensen et al. 2004). However, we note that neither ncm nor prp8 mutants exhibit a dominant Minute bristle phenotype (Table 1, column 4), suggesting that protein synthesis is not severely compromised in these heterozygous mutant flies.

An alternative explanation for the identification of Prp8 and Ncm in our screen would be that they specifically disrupt splicing of pre-mRNAs that act in growth-promoting pathways: many “growth genes” can be alternatively spliced to produce different protein isoforms and abnormal expression/phosphorylation of splicing factors has been observed in certain cancers (Kalnina et al. 2005). Although such a mechanism seems unlikely for a core spliceosomal protein such as Prp8, it is an attractive hypothesis in the case of the RS domain-containing Ncm protein. Interestingly, the activity of RS domain splicing factors can be regulated by phosphorylation (Stamm 2002) and the RS domains themselves are potentially substrates for Akt kinase as they often contain sequences that conform to the Akt optimal phosphorylation motif of RXRXXS/T (Obata et al. 2000). Indeed, the amino- and carboxy-terminal RS domains of D. melanogaster Ncm are predicted to contain 10 potential phosphorylation sites for Akt under high-stringency conditions (http://scansite.mit.edu/). As Akt is itself activated by InR/PI3K signaling, Ncm could feasibly regulate the production of specific splice variants that promote growth in an InR/PI3K-dependent manner. Although this idea remains to be tested, a precedent exists in the case of PI3K-dependent regulation of the human SR protein SRp40. Here, activation of InR/PI3K signaling results in phosphorylation of SRp40 by Akt; phospho-SRp40 then acts on an intronic element in the PKCβII pre-mRNA to promote incorporation of exon 2 into the mature PKCβII transcript (Patel et al. 2005).

Surprisingly, we found that several E-Dp110^KD mutations, including those in ncm, prp8, pix, RpL5, and RpL38, can dominantly increase wing size in the absence of the MS1096>Dp110^KD transgenes (data not shown; Marygold et al. 2005), suggesting that these E-Dp110^KD gene products limit growth in a wild-type background. Although we do not understand the mechanism involved, this unexpected observation may be the end result of suboptimal rates of splicing or translation throughout the development of these flies: this may disturb the normal tight regulation of growth, leading to an increase in final organ size. Indeed, there are a number of precedents for tissue overgrowth occurring despite an overall reduction in the rate/level of protein synthesis. For example, decreased levels of any of several different RPs can give rise to tumors in zebrafish (Amsterdam et al. 2004), and mutation of Drosophila RpS21 (Torok et al. 1999) or RpS6 (Watson et al. 1992; Stewart and Denell 1993) results in overgrowth of the imaginal discs and/or hematopoietic organs. Quite how certain E-Dp110^KD mutations dominantly increase wing size in a wild-type background yet dominantly reduce MS1096>Dp110^KD wing size is a mystery at the present time. Further insight into this apparent paradox will require study of the growth of E-Dp110^KD mutants, with and without the MS1096>Dp110^KD transgenes, at earlier time points in wing development.

Identifying novel growth genes in Drosophila:

Genetic screens have successfully identified a number of different growth promoters and growth inhibitors in Drosophila (see Introduction). It is evident from these studies that screens of a similar type have often yielded the same growth genes. For example, several independent clonal screens in the eye identified Tsc1 (Gao and Pan 2001; Potter et al. 2001; Tapon et al. 2001), salvador (Kango-Singh et al. 2002; Tapon et al. 2002), and hippo (Harvey et al. 2003; Jia et al. 2003; Udan et al. 2003; Wu et al. 2003). Similarly, multiple overexpression screens isolated bantam (Hipfner et al. 2002; Raisin et al. 2003) and Rheb (Patel et al. 2003; Saucedo et al. 2003; Stocker et al. 2003). It is equally evident that different types of growth regulator have been discovered through different types of growth screen. Indeed, the screen we describe in this article is different from those conducted previously, and the growth genes we have isolated have not been detected in other screens. One explanation for this is that diverse growth genes are susceptible to discovery only in certain types of screen. That is, an obvious growth phenotype may be obtained through gain-of-function but not loss-of-function experiments, through clonal but not whole-organ analyses, or through studies in the wing but not the eye. For example, it is unlikely that RP gene mutations could be identified in either clonal or overexpression screens because homozygous mutant clones are cell lethal whereas overexpression of a single RP is not expected to be able to promote growth (Lambertsson 1998). Indeed, to our knowledge, splicing factors have not previously been identified to be limiting for growth, while core translation factors have been isolated in only two other Drosophila growth screens: mutations in eIF-4a, bonsai/mRpS15, and possibly RpL30 were identified in a screen for growth-defective homozygous mutant larvae (Galloni and Edgar 1999), while a mutation in mRpL12 was identified as a suppressor of a CycD–Cdk4 overexpression phenotype (Frei et al. 2005). We envisage that more context- and screen-dependent growth genes are to be discovered.