A Gain-of-Function Germline Mutation in Drosophila ras1 Affects Apoptosis and Cell Fate during Development

The RAS/MAPK signal transduction pathway is an intracellular signaling cascade that transmits environmental signals from activated receptor tyrosine kinases (RTKs) on the cell surface and other endomembranes to transcription factors in the nucleus, thereby linking extracellular stimuli to changes in gene expression. Largely as a consequence of its role in oncogenesis, RAS signaling has been the subject of intense research efforts for many years. More recently, it has been shown that milder perturbations in Ras signaling during embryogenesis also contribute to the etiology of a group of human diseases. Here we report the identification and characterization of the first gain-of-function germline mutation in Drosophila ras1 (ras85D), the Drosophila homolog of human K-ras, N-ras and H-ras. A single amino acid substitution (R68Q) in the highly conserved switch II region of Ras causes a defective protein with reduced intrinsic GTPase activity, but with normal sensitivity to GAP stimulation. The ras1R68Q mutant is homozygous viable but causes various developmental defects associated with elevated Ras signaling, including cell fate changes and ectopic survival of cells in the nervous system. These biochemical and functional properties are reminiscent of germline Ras mutants found in patients afflicted with Noonan, Costello or cardio-facio-cutaneous syndromes. Finally, we used ras1R68Q to identify novel genes that interact with Ras and suppress cell death.


Introduction
Multicellular organisms must extensively coordinate the activities of many diverse and highly specialized cells, requiring effective and flexible signaling mechanisms for both development and tissue homeostasis. Many inter-cellular signals are transmitted by receptor tyrosine kinases (RTKs), which control key aspects of cellular growth, differentiation, metabolism and cell death [1,2]. On the other hand, mutations that lead to abnormal activation of RTKs can generate oncogenes that promote tumorigenesis [3,4]. Ras proteins are guanine nucleotide binding proteins that act as molecular switches to transduce RTK-signals from the outside to the interior of the cell [5]. Remarkably, ,20% of all tumors contain an activating point mutation in Ras [6,7]. Consequently, this pathway has been extensively studied both in the context of normal signal transduction, and oncogenic growth. Because the RTK/RAS signaling network is highly conserved among animals, genetic model systems have made major contributions for elucidating this pathway [8,9,10,11].
One major physiologic function of the RTK/RAS pathway during development is the transmission of anti-apoptotic signals that suppress the activation of an intrinsic cell death program [12,13]. In Drosophila, as in mammals, cells are over-produced during development and compete for limiting amounts of extracellular survival factors in order to suppress the induction of apoptosis [14,15]. This strategy permits appropriate matching of different cell types in a tissue and allows for the elimination of any superfluous and potentially dangerous cells. A conserved mechanism for survival signaling involves activation of receptor tyrosine kinases (RTKs) at the cell surface, which in turn stimulates the antiapoptotic activity of Ras [13]. Active Ras promotes its antiapoptotic effect via several effector pathways, including the mitogen-activated protein kinase p42/44 (MAPK) of the ERKtype (extracellular signal-related kinase) via Raf [16,17,18], and the Akt kinase via Phosphoinositide 3-kinase [19]. In Drosophila, one major target for the anti-apoptotic activity of Ras is the proapoptotic Hid protein, which is inactivated via phosphorylation by MAPK [20,21,22]. Active Hid induces apoptosis by binding to and inhibiting Drosophila Inhibitor of Apoptosis Protein-1 (Diap1), an essential inhibitor of caspases in Drosophila [23,24,25,26,27]. In living cells, Diap1 inhibits both initiator and effector caspases, and its function is required for the survival of virtually all somatic cells [23,25,26,28,29,30,31]. In response to apoptotic stimuli, Diap1 is inactivated by natural IAP-antagonists, including Reaper, Hid and Grim (RHG proteins). The active forms of RHG proteins are generated in doomed cells by a combination of transcriptional induction and post-transcriptional regulation [32,33,34,35]. Once active, RHG proteins form complexes that both disrupt binding of Diap1 to caspases and also stimulate auto-ubiquitination and degradation of Diap1, thereby removing the ''brakes on death'' [27,36]. One important role of Hid is to recruit Reaper to the outer mitochondrial membrane, which is important for efficient inactivation Diap1 and apoptosis induction [27]. Survival signals, such as Spitz in the case of midline glia, inhibit the pro-apoptotic activity of Hid via activation of EGFR, Ras and MAPK, leading to direct phosphorylation of Hid by MAPK and inhibition of Hid pro-apoptotic activity [21,33].
We previously conducted large-scale dominant modifier screens in Drosophila to identify genetic modifiers of Hid-induced apoptosis [20]. These screens identified several loss-of-function alleles in sprouty and gap1, both negative regulators of the RAS/MAPK signaling pathway and helped define the mechanism by which MAPK signaling inactivates a critical component of the apoptotic machinery [22,37]. Here we report the identification and characterization of another Hid-modifier mutation, which maps to the switch II region of ras1 (also known as ras85D), the Drosophila homologue of mammalian N-ras, K-ras and H-ras. Although many loss-of-function alleles have been described for Drosophila ras1, this mutation is the first endogenous gain-of-function allele reported for this gene. We demonstrate biochemically that this viable hypermorph, ras1 R68Q , produces a defective Ras protein with reduced intrinsic GTPase activity, but normal sensitivity to GAP stimulation. These biochemical features are reminiscent of those recently described for mutant human H-ras and K-ras proteins known to underlie a group of related developmental disorders that includes Noonan syndrome, Costello syndrome and cardio-faciocutaneous syndrome [38,39,40]. Flies mutant for ras1 R68 exhibit a number of developmental defects that are characteristic of abnormally elevated RTK/RAS/MAPK signaling, including enhanced resistance to apoptosis, supernumerary R7 cells in the eye and ectopic wing vein formation, demonstrating that the mutant Ras protein has enhanced signaling capacity in vivo. This allele should be a useful tool to study the physiological consequences of modest activation of Ras signaling in vivo. Finally, we used this mutant to identify novel interactors of Ras that suppressors cell death.

Genetic screens
Dominant modifier screens were performed as described in Fig.  S1. Approximately 170,000 F1 progeny from ENU and EMS mutagenized GMR-rpr 81 flies were screened for modification of a GMR-rpr 81 induced rough eye phenotype, yielding 25 enhancers and 5 suppressors (Table S1). Similarly, 300.000 F1 progeny from ENU, EMS and x-ray mutagenized flies were screened for suppression of a GMR-hid 10 induced rough eye phenotype, resulting in the recovery of 128 additional suppressors (Table  S2). In sum total, 158 dominant modifiers of GMR-rpr or GMR-hid were isolated in these screens.
Complementation analyses using phenotype and map information placed 133 of these modifiers into 13 complementation groups, while the remaining mutants represent single hits or have no recessive phenotype. To further enrich for mutants that are cell death specific, we eliminated general modulators of GMR promoter expression or eye development by testing modifiers against GMR-phyl and GMR-rho induced rough eye phenotypes, which are unrelated to cell death [43,44]. In addition, reasoning that mutants involving apoptosis genes should be able to modify cell death phenotypes in contexts other than the eye, suppressors from the GMR-hid screen were tested for their ability to suppress hs-hid induced embryonic lethality and vg-GAL4, UAS-hid induced wing ablation. On the basis of these secondary screens, we eliminated several complementation groups including glass, which encodes the transcription factor that drives GMR expression, Su(GMR)2A and su(GMR)3A, which are known to indirectly and non-specifically affect GMR promoter expression, and Su(GMR-hid)3A and Su(GMR-hid)3B, complementation groups that have not been assigned to a previously characterized gene [51,52]. We also eliminated 4 alleles linked to the parental GMR-rpr transgene. Our cell death enriched subset of modifiers therefore consists of 58 mutants in total, 40 that fall into 6 complementation groups and 18 single alleles.
All crosses and suppression experiments were carried out at 25uC except crosses with vg-GAL4 and UAS-hid, which were performed at both 18uC and 25uC. Suppression experiments with hs-hid were done by heat shocking 1 st instar larvae at 37uC for 15 minutes.

Biochemistry
A cDNA clone encoding Drosophila ras1 was obtained from the Drosophila Genomics Resource Center (clone ID: RE53955) and the entire ras1 ORF was subcloned into pBluescript (Stratagene). Mutant ras1 R68Q was generated using the QuikChange II XL Site-Directed Mutagenesis Kit (Stratagene). The Ras ORFs were then subcloned into pET-28a (Novagen) in frame for an N-terminal His tag. Catalytic human p120-Gap (GAP-285, amino acids 714-998, IMAGE Clone: 4829173, Open Biosystems) was subcloned into the pET41a vector (Novagen) to generate an N-terminal GST tag. Fusion proteins were expressed in BL21(DE3) E. coli (Invitrogen) and affinity purified on an AKTA Purifier (Pharmacia) using a HisTrap FF column (GE Healthcare) for Ras proteins and a GSTrap FF column (GE Healthcare) for GAP-285. Ras purification was done according to the procedure described for human H-Ras [53]. GAP-285 was expressed by inducing cells for 16 hours at 30uC with 0.2 mM IPTG.

Phenotypic Analyses
To visualize larval hemocytes, wandering 3 rd instar larva expressing UAS-EGFP driven by Hml-GAL4 were collected and immobilized on ice prior to imaging [56]. MG cells in stage 17 embryos were visualized using P[slit-1.0-lacZ] and ß-gal immunohistochemistry as previously described [57]. The number of MG was averaged for segments T2 to A5. Tangential sections (1 mm) of adult eyes were prepared according to standard protocols for analysis of ommatidia [58].

Genetic screens for dominant modifiers of apoptosis in Drosophila
Dominant modifier screens are designed to detect pathway components for which small perturbations in gene dosage can alter a sensitized phenotype, thus allowing for the recovery of both lossof-function and gain-of-function mutations. We used eye-specific expression of the Drosophila cell death genes hid or rpr under control of the GMR promoter to generate a dosage sensitive eye ablation phenotype and then screened for dominant modifiers of this phenotype to identify regulatory components of the intrinsic cell death program [20,24,41]. Although several genes identified in this way have been reported, details of these screens have not been previously published and are provided in the supplementary material (Fig. S1, Tables S1 and S2).
We identified a subset of 58 mutants that specifically affect rpr or hid-induced cell death. 40 of these correspond to 6 complementation groups, while 18 represent single alleles. Of the 6 complementation groups identified, 3 correspond to genes that regulate EGF receptor (EGFR) signaling. Five loss-of-function alleles each of gap1 and sprouty, both negative regulators of EGFR/ MAPK signaling, were recovered as strong, hid specific suppressors. These mutants have been characterized elsewhere [20]. We also isolated five loss-of-function Star alleles as enhancers of GMRrpr. Star is required for the correct processing of Spitz, a stimulatory ligand of EGFR [59]. Ten alleles of diap1, the major biochemical target for the pro-apoptotic activity of RHG proteins were isolated, including loss-of-function alleles that enhance rpr-, hidand grim-induced cell death and two distinct classes of gain-offunction alleles [23,28,31]. The fifth complementation group, consisting of 12 alleles, displayed a differential modulation of cell death phenotypes reminiscent of diap1 RING mutants and was found to encode dbruce, the Drosophila ortholog of mouse Bruce and human Apollon [60,61]. This very large (4852 amino acid) BIRcontaining protein is cytoprotective against caspases and required for spermatid survival [62,63,64]. Finally, 5 single alleles from the GMR-rpr screen likely represent weak hypomorphs of diap1 as they map close to the diap1 locus, and a sixth allele was identified as an allele of Delta.
The remaining complementation group, Su(GMR-hid)2A, and 12 additional single alleles were previously not characterized. We chose one allele, Su(21-3s) for further analysis based on its ability to potently suppress hid-induced phenotypes (eye/wing/organismal lethality) without non-specifically affecting GMR-phyl (Table  S2). Su(21-3s) mutants are homozygous viable and do not have overt abnormalities. This mutant was mapped by meiotic recombination to the right arm of chromosome 3, near the visible marker curled.

Characterization of Su(21-3s) suppressor phenotypes
We first examined more rigorously the suppression phenotypes of Su(21-3s) in the eye by testing the modifier effects of one or two copies of Su(21-3s) against various GMR expression constructs. This analysis confirmed that Su(21-3s) potently suppresses GMRhid induced cell death in a dosage dependent manner (Fig. 1A,B). We found suppression of GMR-rpr and GMR-grim phenotypes, however, to be extremely weak, even with two copies of Su(21-3s) (Fig. 1E,F). Given that hid is highly expressed in the developing eye, we believe the small effect exerted by Su(21-3s) on GMR-rpr and GMR-grim is due primarily to a suppression of endogenous Hid activity and not that of Rpr or Grim [24]. We conclude from these data that Su(21-3s) is a specific suppressor of hid induced cell death.
The activity of Hid is regulated by the EGFR/MAPK pathway in a manner that depends on intact MAPK phosphorylation sites in Hid. Our analysis here reveals that Su(21-3s) readily suppresses GMR-hid Ala3 , a hid allele lacking 3 of 5 predicted MAPK phosphorylation sites, but fails to suppress GMR-hid Ala5 , which is missing all 5 MAPK sites (Fig. 1C,D) [37]. This requirement for one or two of the predicted MAPK phosphorylation sites in Hid (Ser-121 and Thr-228), along with the observed specificity for GMR-hid suppression, strongly suggested that Su(21-3s) might be mediating its suppressive effects through the EGFR/MAPK pathway.
We further extended analysis of the Su(21-3s) suppression phenotypes to the developmental context of larval hemocytes, an important model system for the study of vertebrate haematopoiesis [65,66]. Drosophila hemocytes require trophic signaling from multiple pathways for their survival and in its absence undergo caspase dependent cell death [67,68]. Larval hemocytes also undergo caspase dependent cell death in response to ectopic hid expression [69]. Using a hemocyte specific promoter to drive expression of UAS-EGFP, we are readily able to visualize hemocytes in wandering 3 rd instar larvae (Fig. 1G) [56]. Ectopically expressing UAS-hid using the same driver results in near complete ablation of hemocytes by the 1 st instar larval stage (data not shown) and generates 3 rd instar larvae that are completely devoid of hemocytes (Fig. 1H). We found that the Su(21-3s) mutation is able to partially suppress this cell death such that EGFP expressing hemocytes are clearly visible anteriorly in the lymph glands of 3 rd instar larvae (Fig. 1I). Circulating hemocytes, however, appear to remain susceptible to hid-induced cell death and are missing, even in the presence of two copies of the Su(21-3s) allele. It may be that Su(21-3s) is a weak suppressor of cell death in hemocytes, sufficient to suppress Hid activity in young hemocytes localized to the supportive environment of a lymph gland, but insufficient in the context of a mature circulating hemocyte.

Su(21-3s) is a viable gain-of-function allele of ras1 (ras85D)
In order to identify the gene responsible for the Su(21-3s) phenotype, we mapped it by a second, finer round of meiotic recombination to a 1 Mb interval between 85A and 85E, then further localized the mutation by P-element mediated male recombination to a 270 Kb interval between 85D11 and 85E1 ( Fig. 2A). Given that Su(21-3s) differentially suppresses hid, but not grim or rpr in a manner reminiscent of EGFR/MAPK pathway mutants, we suspected that Su(21-3s) might be a rare hypermorphic allele of ras1, or ras85D as it is otherwise known, because it is located within this interval. Indeed, when we sequenced ras1 in a candidate gene approach, a G to A transition in exon3 was identified. This mutation results in an amino acid substitution at position 68 of the Ras protein, replacing a positively charged arginine within the universally conserved switch II region of Ras with a neutral glutamine (Fig. 3).
The switch regions of Ras have been defined as regions that undergo a large conformational change when Ras transitions from the GTP-to the GDP-bound state [70]. Detailed biochemical analysis and crystal structures have revealed that residues in the switch II region of Ras contact and are stabilized by the GTPase Activating Protein (GAP), allowing them to participate up to a 1000 times more efficiently in the catalysis of GTP [71]. As a consequence, mutations in the switch II region of Ras interfere with its catalytic GTPase activity and prolong the time Ras remains bound to GTP. Mutations that reduce the GTPase activity of Ras are hypermorphic since it is the GTP-bound form of Ras that engages and activates downstream signaling effectors. The signaling activities of Ras are terminated when GTP bound by Ras is converted to GDP, explaining why the most frequently occurring oncogenic mutations in Ras, at amino acids 12,13 and 61, also render Ras biochemically inert as a GTPase (Fig. 3) [72,73]. It seemed feasible, therefore, that the R68Q mutation identified in Su(21-3s) flies could similarly result in a Ras protein with reduced GTPase activity, leading to a prolonged RAS/ MAPK signal that suppresses cell death induced by GMR-hid.
We reasoned that if the Su(21-3s) phenotype is due to a gain-offunction mutation in ras1, it should be revertible by introduction of a second, intragenic loss-of-function mutation. To test this, we conducted a reversion screen for loss of the Su(21-3s) suppression phenotype and successfully recovered two mutants containing intragenic loss-of-function ras1 mutations (Figs. 2 and S1). One revertant contains a 31 bp deletion in ras1 that results in a Ras protein truncated at amino acid 87. The second revertant contains an in frame 18 bp deletion of ras1 that removes amino acids 87-92 which are essential for Ras function [74]. These revertants have genetic properties of ras1 null alleles and fail to complement the known null alleles ras1 e1B and ras1 e2F and supporting the idea that the Su(21-3s) phenotype is due to a revertible gain-of-function mutation in ras1.
As an allele of ras1, Su(21-3s) should interact genetically with other members of the MAPK signaling pathway in a predictable manner. We crossed GMR-hid 10 flies in a Su(21-3s) background to mutants of MAPK signaling and observed the degree of cell death in the eye (Fig. S2). MAPK signaling mutants tested include argos, ras1, rolled and EGFR. In this analysis, we found that the Su(21-3s) mutant behaves in a manner consistent with that expected for a gain-of-function ras1 allele. For example, Su(21-3s) is not much affected by loss-of-function mutations in upstream components of MAPK signaling, such as argos or EGFR but is strongly ameliorated by loss of downstream components, such as rolled. Additionally, when a dominant negative form of Ras1 (sev-ras1 N17 ) is expressed in the eye, the suppressive effects of Su(21-3s) are completely abrogated. On the basis of our mapping data, the reversion screen, the sequence data and the genetic interaction data presented above, we conclude that Su(21-3s) is a hypermorphic allele of ras1, which we rename here, ras1 R68Q .

Biochemical analysis of recombinant Ras1 R68Q protein
To test the hypothesis that exchanging a positively charged arginine with a neutral glutamine at position 68 of Ras results in a protein with deficient GTPase activity, the intrinsic GTPase rates of wildtype and mutant Drosophila Ras1 proteins were measured. Full-length wildtype Ras1 and mutant Ras1 R68Q were bacterially expressed and purified as His-tagged fusion proteins, yielding large amounts of pure, catalytically active enzyme. Intrinsic GTPase activity rates were measured with a kinetic phosphate assay employing [camma-33 P]GTP as substrate. This sensitive assay revealed that Ras1 R68Q has an intrinsic GTPase activity that is approximately 1/3 that of wildtype Ras1, with enzymatic rates (k cat ) of 0.020 min 21 and 0.063 min 21 , respectively (Fig. 4C). Since many activating Ras mutations also result in an enzyme that is insensitive to GTPase activating proteins (GAPs), the ability of Ras1 R68Q to be stimulated by GAP was also assessed. Recombinant human GAP-285 protein was purified by affinity chromatography and its ability to stimulate wildtype and mutant Ras1 proteins was tested using a real-time fluorescent assay. These experimental data show that Ras1 R68Q remains amenable to GAP stimulation (Fig. 4D). This means that in contrast to constitutively active Ras mutants such as oncogenic Ras V12 , whose GTPase activity is completely refractory to stimulation by GAPs, Ras1 R68Q can be regulated and is able to cycle between on and off states [75]. Together, these biochemical data support the hypothesis that Ras1 R68Q has a reduced basal level of GTPase activity, remains in its active GTP-bound form for longer and thus has an enhanced  Su(21-3s) is a gain of function allele of ras1 (ras85D), the Drosophila homolog of human n-ras, h-ras and k-ras. (A) The cell death suppression phenotype of Su(21-3s) mutants was localized by meiotic recombination to the right arm of the 3rd chromosome as indicated by the large horizontal arrow. P-element induced male recombination mapping further localized this suppressor to the region depicted by the short arrow. An enlargement of this interval (5.162-5.452 Mb on the physical map) is shown below, illustrating the ORFs contained therein. The ras85D (ras1) locus, outlined with a red box, was sequenced in a candidate gene approach and a G to A transition in exon3 (G641A) was identified. (B,C) A screen for reversion of the dominant Su(21-3s) suppressor phenotype generated a number of genetic revertants, two of which, Su(21-3s) R11 and Su(21-3s) 41 , were molecularly determined to be intragenic loss of function ras1 alleles. (B) A schematic of the ras1 locus, with exons boxed and coding regions stippled, illustrates the Su(21-3s) point mutation in exon 3 (G641A) and the small intragenic deletions identified in Su(21-3s) R11 and Su(21-3s) R41 (labeled R11 and R41 respectively, with deleted sequences underlined in black.) The red arrows correspond to primers used in a PCR diagnostic (below) used to confirm that both revertants contain the original G641A mutation. (C) Sequence analysis of these lethal revertants using strand specific PCR revealed that Su(21-3s) R11 contains an 18bp deletion that removes essential amino acids 87-92 from the Ras1 protein. Su(21-3s) R41 was found to have a 31bp deletion, resulting in a frameshift that generates a truncated Ras1 protein. doi:10.1371/journal.pone.0023535.g002 signaling capacity, but is still amenable to regulation, making it compatible with nearly normal cellular function and organismal development. These biochemical features are highly reminiscent of those recently described for germline H-ras and K-ras mutants found in the developmental disorders Noonan syndrome, Costello syndrome and cardio-facio-cutaneous syndrome [38,40].

Ras1 R68Q promotes survival of midline glia (MG)
The survival of Drosophila midline glia (MG) cells during embryonic development depends on survival signals mediated by the EGFR/Ras/MAPK pathway [21,76]. During formation of the Drosophila central nervous system, there are initially approximately ten MG cells per segment at stage 13. Most of these undergo apoptosis in a RHG-dependent manner such that by stage 17, only three MG per segment survive [45,77]. We tested the effect of ras1 R68Q in this system. MG cells were visualized in wildtype and ras1 R68Q embryos using the MG-specific pslit-lacZ reporter, and marked MG cells were carefully counted. This analysis revealed an increase in the number of MG cells in ras1 R68Q embryos as compared to wildtype embryos (Fig. 5E,F). Stage 17 wildtype embryos contained an average of 2.8 MG cells per segment (n = 448) whereas ras1 R68Q embryos contained an average of 3.3 MG cells per segment (n = 420). This difference is statistically significant by an unpaired t-test (p 95 #0.0001) and is consistent with an increase of Ras activity in ras1 R68Q flies (Fig. 5G).

Ras1 R68Q causes supernumerary R7 cells in the eye
The adult Drosophila eye comprises about 800 ommatidia, each with a precise, reproducible structure consisting of eight photoreceptors and 12 accessory cells [78,79]. Adoption of a neuronal cell fate by the precursor of the R7 photoreceptor requires an inductive signal from the neighboring R8 cell and is dependent on EGFR/MAPK signaling [80,81,82]. Furthermore, the cone cell precursors are capable of acquiring an R7 cell fate if MAPK signaling is ectopically activated in these cells, resulting in extra R7 cells that are easily visualized [83]. To determine if the ras1 R68Q mutation exerts effects in a paradigm other than apoptosis, semi-thin plastic sections of adult eyes were prepared and analyzed for defects in ommatidia formation. We observed two types of defect in ras1 R68Q flies typical for mutations that increase RAS/MAPK signaling during eye development. First, we detected ommatidia with supernumery R7 cells (Fig. 5B,C). Second, we noticed occasional ommatidia missing outer photoreceptor cells, also a phenotypic consequence of elevated MAPK signaling (Fig. 5D) [83]. The developmental defects in retinal cell differentiation observed here further supports our hypothesis that ras1 R68Q is a gain-offunction ras1 allele.

The wings of ras1 R68Q flies contain ectopic vein material
In addition to defects in the eye and midline glial cells, ras1 R68Q flies also show abnormalities in adult wing tissues. Homozygous ras1 R68Q flies have an extra longitudinal ''veinlet'' branching off the posterior crossvein (Fig. 5I). Additionally, an ectopic longitudinal vein was seen beneath the posterior crossvein and an ectopic crossvein appeared between the L4 and L5 wing veins near the hinge. These defects are remarkably similar to those observed in the wings of rl sem and DER Ellipse flies, which have elevated levels of MAPK signaling [84]. When UAS-ras1 R68Q was overexpressed in the wing using en-Gal4, extensive ectopic wing vein material and blisters developed (Fig. S3). Overexpression of wildtype Ras1 gave a similar but less severe phenotype. Finally, we attempted to express ras1 V12 in the wing using en-Gal4 but found this to cause organismal lethality. . Amino acid alignment of fly, worm and mammalian Ras1. These homologs have extensive primary sequence homology. Drosophila Ras1 (dmRas1), for example, is 87% identical to human K-ras (hsKras) at the amino acid level when C-terminal membrane-targeting sequences are excluded. Conserved regions are shaded in grey, with residues identical to the consensus sequence represented by a grey dot, while non-conserved residues remain unshaded. Five highly conserved signature motifs named ''G box'' sequences, labeled G1-G5 and boxed in red, are found in all families of small GTPases. Secondary structural elements are depicted as rectangles below the primary sequence alignment (alpha-helices, a1-a5, are dark grey and ß-sheets, ß1-ß6, light grey) and the phosphate-binding loop (P-loop), which binds the camma-phosphate of GTP, and the nucleotidesensitive switchI and II regions are indicated. The Switch regions are known to undergo large conformational changes upon exchange of bound GDP for GTP. The mutational spectrum of Ras is illustrated above the alignment, showing the distribution of amino acid substitutions encoded by germline mutations found for the developmental disorders Noonan, Costello and CFC syndromes and the most frequent cancer-associated somatic mutations (labeled in red). R68Q indicates the mutation characterized in this study, a non-conserved arginine to glutamine amino acid substitution within the switch II region of Drosophila Ras1 (dashed red box). hs, H. sapiens; dm, D. melanogaster; mm, M. musculus; ce, C. elegans. doi:10.1371/journal.pone.0023535.g003 Overexpression of ras1 R68Q in the eye induces both severe overgrowth and cell death It was previously shown that overexpression of wildtype Ras1 in the Drosophila eye, even at the high levels obtained by transgene expression, results in little or no observable phenotypic effect [83]. For this reason, studies of elevated Ras signaling in Drosophila regularly rely on a very strong, constitutively active Ras V12 mutant allele. We similarly observed that wild-type UAS-ras1 expressed by GMR-Gal4 had little effect on eye development in 11 independent transgenic lines (Fig. S4). In striking contrast, seven independent transgenic lines expressing UAS-ras1 R68Q resulted in highly distorted eyes that exhibited both hyperplastic tissue overgrowth and widespread cell death ablation phenotypes. For purposes of comparison, we attempted to express two different UAS-ras1 V12 alleles in the same manner, but again found this induced organismal lethality (likely due to the fact that GMR drives some expression in tissues other than the eye and Ras1 V12 induces non-cell autonomous cell death when overexpressed) [85]. We generated many more than seven UAS-ras1 R68Q transgenic lines but similarly found many of them to be lethal in combination with GMR-Gal4. This lethality was not observed in any of the 16 independent UAS-ras1 transgenic lines tested. These experiments further support the view that Ras1 R68Q is an activated form of Ras that nevertheless remains amenable to regulation and therefore is less biologically potent than the constitutively active Ras1 V12 protein.

Identification of Novel Ras Interactors and Suppressors of Cell Death
In the reversion screen described above we were also able to recover additional suppressors of GMR-hid. We collected a number  . R68, located proximally to the catalytic site of Ras, also extends a positively charged guanidinium group towards the active site. The images were constructed using the Entrez software Cn3D with mmdbId:51925 (Chen et al., 2003). Guanosine diphosphate (GDP,brown); Mg 2+ (grey). (C) The intrinsic GTPase activities of affinity purified drosophila Ras1 wt (blue) and Ras1 R68Q (black) were determined using a kinetic phosphate assay employing [camma-33 P]GTP as a substrate. The conditions of the assay are such that the reaction proceeds with unimolecular kinetics and is insensitive to the amount of Ras protein employed (dashed vs. undashed lines). The mutant Ras1 R68Q has an intrinsic GTPase activity that is approximately 1/3 that of wildtype Ras1 (k cat = 0.020 min 21 and 0.063 min 21 respectively.) (D) Human GAP-285 protein was purified by affinity chromatography and its ability to stimulate wildtype and mutant Drosophila Ras1 proteins was tested using a real-time fluorescent assay. Both wildtype and mutant Ras1 proteins are sensitive to GAP stimulation (dashed vs. undashed lines). Data is the average of three independent experiments. Error bars are in red. doi:10.1371/journal.pone.0023535.g004 of strong dominant suppressors of the GMR-hid eye phenotype and mapped them using deletions on the 3 rd chromosome. As indicated in Table 1, we successfully recovered 14 suppressors that fall into 8 complementation groups. In most cases, we were able to identify a single gene that appears to be responsible for the suppressors phenotype (indicated in bold). In two cases, the mutations were narrowed to a small region, but we were unable to unequivocally identify a single candidate. We also recovered mutations in the glass gene, which affects expression from the GMR-driver [86]. We recovered mutations in Gap1 and Delta, both of which were identified in our original hid suppressor screen, indicating an overlap in the mutational spectrum between the two screens [20]. Based on previous reports, mutations in Gap1 are expected to suppress GMR-hid, and Delta/Notch signaling is known to intersect and cooperate with the Ras/Mapk pathway [87]. Interestingly, we also isolated a number of novel ras interactors, including four alleles of the predicted transcription elongation factor Su(Tpl) and an allele of notum, a component of the Wnt/Wingless signaling pathway. These results indicate that use of the ras R68Q allele in screens may indeed uncover novel regulatory interactions that have been missed with other strategies, including those that make use of the constitutively active, nonregulatable ras V12 hypermorph.

Discussion
We have conducted genetic screens for dominant modifiers of cell death induced by the Drosophila IAP-antagonists, hid and rpr. From over 150 mutants initially isolated, secondary screens allowed us to identify 58 cell death specific modifiers. Of these, 40 alleles were placed into six complementation groups that define both known and unknown genes. These include Star, gap and sprouty involved in EGFR/MAPK signaling, the known cell death regulator diap1, the very large BIR and UBC containing dbruce, and an unknown gene, Su(GMRhid)2A that remains unidentified. Here we focused on a previously uncharacterized cell death suppressor originally termed Su(21-3s). Using a combination of meiotic and P-element induced male recombination, genetic reversion, biochemistry and in vivo analysis, we demonstrate that this mutant is a gain-of-function mutation in ras1 (ras85D), the Drosophila homolog of human K-ras, Nras and H-ras. We also show that this allele affects cell fate decisions and the pattern of normal, developmental apoptosis in paradigms known to depend on Ras-signaling.
One important role of Ras signaling during development is the transmission of an anti-apoptotic signal [13,20,22]. As previously reported, the pro-apoptotic protein Hid contains 5 potential MAPK phosphorylation sites that are essential for its sensitivity to Ras-mediated inhibition [20]. A Hid protein with either 3/5 or 5/ 5 mutant MAPK sites (Hid Ala3 and Hid Ala5 , respectively) was refractory to suppression by the gain-of-function MAPK allele rl Sem (a very mild suppression by rl Sem is due to phosphorylation of the endogenous wildype Hid protein). In contrast, there was still some suppression of Hid Ala3 and Hid Ala5 by RasV12. It was postulated that this might be due to the ability of Ras, unlike MAPK, to exert additional anti-apoptotic effects through activation of the PI3-K/ Akt-kinase effector branch. In the current study, we found that ras R68Q was able to partially suppress Hid Ala3 but not Hid Ala5 (Fig. 1C,D). Because Hid Ala3 retains two phosphorylation sites, it appears that partial phosphorylation of Hid is sufficient for a mild inhibitory effect, and that all five phospho-acceptor sites need to be eliminated in order for Hid to become refractory to inhibition by MAPK. Furthermore, it appears that Ras R68Q , unlike Ras V12 , is unable to exert an additional suppressive effect via PI3-K/Aktkinase. Perhaps the enhanced signaling activity of Ras R68Q is able to activate the MAPK effector branch, but does not reach a required threshold to engage the PI3-K/Akt-kinase pathway [59,88,89,90,91]. This may also help to explain the organismal viability of Ras R68Q as compared to Ras V12 . Along the same line, Ras R68Q was able to suppress Hid-induced cell death of lymphocytes within the protective environs of the lymph gland but not of those that were circulating (Fig. 1G-I). In sharp contrast, over-expression of Ras v12 in hemocytes not only leads to survival of circulating hemocytes but in fact results in a massive overproliferation of hemocytes (Rodriguez and Steller, unpubl. data) These results serve to highlight the exquisite sensitivity of biological systems to the degree of Ras signaling and suggest that between the extremes of wildtype Ras and constitutively active Ras V12 lies a large spectrum of biological responsiveness.
Ras is highly conserved among metazoans and a number of Ras structures have been published that make it possible predict how mutations in specific regions might affect function. In the case of Ras R68Q , we considered that this change may affect the transition state of Ras. According to the ''arginine-finger hypothesis'' proposed by Scheffzek and colleagues, GTPase-activating-proteins (GAPs) dramatically accelerate the GTPase reaction of Ras by supplying an arginine side chain (arginine-789 in the case of GAP-334) into the active site of Ras to neutralize developing charges in the transition state [71]. A detailed analysis of the interactions between Ras and GAP-334 showed no role for R68 of Ras, explaining why Ras R68Q can be stimulated by GAP [71,74,92,93]. However, a close inspection of the Ras catalytic site (Fig. 4B) shows that R68 extends its side chain towards the catalytic center [94]. Mutating R68 to glutamine removes a stabilizing positive charge from the transition state and, according to the arginine-finger hypothesis, would be expected to result in less efficient hydrolysis of GTP. We tested this prediction biochemically and indeed found that Ras R68Q hydrolyzes GTP intrinsically at a reduced rate, approximately 30% of that of wild type GTP (Fig. 4C.) Oncogenic mutations in Ras occur most frequently at codons 12,13 or 61 and result in an enzyme with deficient GTPase activity. This renders Ras inactive because Ras is 'on' when bound to GTP and switches 'off' by hydrolyzing bound GTP to GDP. Inhibition of Ras GTPase activity therefore stabilizes Ras in its active conformation, prolonging its recruitment and activation of downstream signaling components [5,7,10,70,95]. The reduced GTPase activity of Ras R68Q means that it would remain in its active GTP-bound conformation for longer periods of time allowing for enhanced signaling to downstream effector pathways. As noted above, however, Ras R68Q may not remain in an active state sufficiently long to engage the catalytic p110 subunit of PI3K. An interesting alternative possibility however may be that R68 is directly involved in an interaction with PI3K and a mutation in R68 negatively affects this interaction. This raises the intriguing possibility that some of the phenotypes described for Ras R68Q (Fig. 5) may actually be due to a loss, rather than a gain of PI3K activity.
During the initial mapping and characterization of ras1 R68Q , we conducted a reversion screen in order to provide genetic evidence for our hypothesis that we had identified a rare gain-of-function allele in ras85D (Fig. 2). While searching for revertants, we also recovered several mutants that were strong suppressors of GMRhid. Recognizing that these mutants might be synergizing with ras1 R68Q to produce such a strong suppression, we successfully recovered and mapped 14 of these suppressors. As indicated in Table 1, most were mapped to a single candidate gene. Since these mutants were essentially derived from a dominant modifier screen for suppression of GMR-hid induced cell death, but within a sensitized ras1 R68Q background, we expected the mutational spectrum to be overlapping, yet distinct from that of previous GMR-hid or UAS-Ras V12 based screens. Indeed several suppressors turned out to overlap with ones identified previous screens. However, we also isolated two novel interactors: one allele of notum and four alleles of Su(Tpl). This demonstrates the utility of ras1 R68Q to identify novel genetic interactions. While notum affects the Wnt/ Wingless signaling pathway, Su(Tpl) is thought to function in the regulation of transcription in response to stress [96,97,98]. Much of our understanding of Ras-mediated signaling is derived from a combination of biochemical experiments conducted in mammalian tissue culture, and genetic studies in model organisms [10]. For example, Ras-mediated signaling regulates the specification and differentiation of R7 photoreceptors in the Drosophila eye [80,81,99]. However, until now, studies on the physiological consequences of elevated Ras in Drosophila have relied on overexpression of the activated ras1 v12 allele [83,85,100]. The viable hypermorphic ras1 allele described here, ras1 R68Q , represents the first endogenous gain-of-function mutation in Drosophila Ras and hence offers a new tool for the analysis of Ras biology in situ. In particular, certain aspects of Ras biology have remained largely inaccessible to the use of constitutively active versions of this protein. This is because mutants, such as ras1 v12 , do not cycle normally between off and on states, are insensitive to regulatory circuits and are generally not compatible with organismal development. As a consequence, in certain paradigms and contexts, ras1 v12 actually behaves as a loss-of-function mutant rather than a hypermorph, occluding the biological interpretation of Ras function in vivo [101]. Therefore, the use of milder, viable hypermorphs of Ras, such as ras1 R68Q , offers the potential for a refined understanding of the normal physiological roles of this important protein. Significantly, the ras1 R68Q allele described here shares overall biochemical properties with recently discovered mutations in k-ras and h-ras that underlie human developmental disorders, such as Noonan, Costello and CFC syndromes. Figure S1 Genetic schemes for dominant modifier and reversion screens. (A) GMR-rpr screen. yw; GMR-rpr 81 homozygous males were fed a solution of sucrose and 0.25mg/ml ENU or 25 mM EMS and mated to females of the same strain. F1 progeny were screened for suppression or enhancement of the parental rough eye phenotype. Of the 170,000 F1 progeny screened, ,95% derived from ENU treated males, while 5% were from EMS treated males (B) GMR-hid screen. yw males were treated as above or with 4500 rad x-rays and then crossed to GMR-hid 10 homozygous females. F1 progeny were screened for suppression of the GMRhid 10 rough eye phenotype. Of the 300,000 F1 progeny screened, ,49% derived from EMS treated males, ,49% from x-ray treated males and 2% from ENU treated males. (C) Reversion screen. Homozygous Su(21-3s) males were treated with 4000 rad x-rays and crossed to GMR-hid 1M ; Sb/TM6B females. 80,000 F1 progeny were screened for loss of the Su(21-3s) suppression phenotype. (TIF) Figure S2 The Su(21-3s) mutant differentially interacts with components of the EGFR/MAPK pathway. Suppression of the GMR-hid10 induced eye ablation phenotype by Su(21-3s) (A vs D) is not much affected by loss of function mutations in upstream components of MAPK signaling such as egfr (E) or argos (F), but is strongly ameliorated by loss of downstream components, such as rolled (B). Additionally, when a dominant negative form of Ras1 (sev-ras1 N17 ) is expressed in the eye, the suppressive effects of Su(21-3s) are completely abrogated (C). Genotypes: (A) GMRhid 10 /+, (B) GMR-hid 10 /rl 10A ;Su(21-3s)/+, (C) GMR-hid 10 /+;Su(21-3s)/sev-ras1 N17 , (D) GMR-hid 10 /+;Su(21-3s)/+, (E) GMR-hid 10 / egfr 2 ;Su(21-3s)/+, (F) GMR-hid 10 /+;Su(21-3s)/arg lD7 . (TIF) Figure S3 Overexpression of ras1 in the wing induces ectopic vein material. Overexpression of either wildtype ras1 (C) or mutant ras1 R68Q (D) using the en-Gal4 driver results in the deposition of significant amounts of ectopic wing vein material. This phenotype is much more severe with ras1 R68Q however, which frequently also results in wing blisters. Panels (A) and (B) are included for comparison only and are the same images shown in Figure 5. (TIF) Figure S4 Overexpression of ras1 in the eye induces developmental defects. Both overgrowth and cell death phenotypes are observed when Ras is overexpressed in the fly eye. Flies overexpressing wildtype ras1 (B,G) exhibit relatively minor disruptions in eye patterning and in the case of sev-Gal4 driven expression, a small but significant amount of overgrowth occurs in the anterior part of the eye (G). In contrast, overexpression of ras1 R68Q with GMR-Gal4 (C-F) causes severe overgrowth and patterning disruptions. An example from each of four independent transgenic lines is shown to illustrate the range of phenotypes. Likewise, overexpression of ras1 R68Q with sev-Gal4 elicits a much more pronounced overgrowth phenotype in the anterior part of the eye (H) compared to that of wildtype ras1 (G). Genotypes: (A) GMR-Gal4/ +, (B) GMR-Gal4/+;UAS-ras1/+, (C-F) GMR-Gal4/+;UAS-ras1 R68Q / +, (G) sev-Gal4/+;UAS-ras1/+, (H) sev-Gal4/+;UAS-ras1 R68Q /+.

(TIF)
Table S1 GMR-rpr modifiers: Summary of genetic interactions. Complementation groups are named for the known gene to which they correspond. The group named ''other'' consists of mutants that could not be placed into complementation groups.
-th-stindicates that the mutation was roughly mapped by meiotic recombination around the markers th and st and may be located on either side, whereas sr-e indicates that the mutation maps between sr and e. Alleles with the same map position and similar phenotypes are grouped together for simplicity. Rep, reduced eye pigmentation; Sup, suppressor; Enh, enhancer; -, no effect; ND, not done. (TIF) Table S2 GMR-hid suppressors: Summary of genetic interactions. Legend is as described in Table S1. -th-st-, -cu-and -sr-indicate that the mutation was roughly mapped by meiotic recombination around the designated markers and may be located on either side, whereas st-cu, cu-sr and sr-e indicate that the mutation maps between the designated markers. The mutation characterized in this study, Su(21-3s), is highlighted in yellow. Rep, reduced eye pigmentation; Ro, rough eye; Wv, extra wing veins; Wk, weak; Sup, suppressor; Enh, enhancer; -, no effect; ND, not done. (TIF)