Cell, Vol.

67, 701-716,

November

15, 1991, Copyright

0 1991 by Cell Press

Rasl and a Putative Guanine Nucleotide Exchange Factor Perform Crucial Steps in Signaling by the Sevenless Protein Tyrosine Kinase
Michael A. Simon, David D. L. Bowtell, G. Steven Dodson, Todd R. Laverty, and Gerald M. Rubin Howard Hughes Medical Institute and Department of Molecular and Cell Biology University of California Berkeley, California 94720 generally phosphorylation of the PTK itself, but a number of other cellular proteins are also phosphorylated on tyrosine residues to varying extents. A number of proteins whose content of phosphotyrosine increases after a PTK is activated have been characterized. Among these molecules are several proteins that are known or thought to be elements of signaling pathways and are thus potential mediators of PTK signaling (reviewed by Cantley et al., 1991). These include phospholipase C-r, the ras GTPase activating protein (rasGAP), a subunit of a Ptdlns 3-kinase, and the c-rafl serinelthreonine kinase. A second consequence of PTK activation is the recruitment of several of these same cellular proteins into complexes with the activated PTK. The precise composition of these complexes is not clear, but in several cases the binding appears to be mediated by SH2 domains of the target proteins binding to the phosphotyrosine moieties of the activated PTK. Complex formation may act as a mechanism for localizing signal transducing molecules to their site of action. These studies have identified a number of proteins that may play important roles in signal transduction by PTKs. However, the functional significance of these potential components of PTK signaling pathways has been difficult to determine because there has been no easy way to ask whether these proteins are an essential part of the signaling pathway. Studies of mutations that affect development in flies and nematodes have provided in vivo evidence for the participation of two proteins in PTK signaling. The Drosophila I(l)polehole locus, which encodes a c-rafl homolog, is required for the action of the PTK encoded by the torso gene (Ambrosio et al., 1989; Sprenger et al., 1989; Casanova and Struhl, 1989). Studies of a Caenorhabditis elegans vulva1 development likewise indicate that a ras protein, the product of the let-60 gene, is essential for the action of the PTK encoded by the let-23 gene (Han and Sternberg, 1990; Beitel et al., 1990; Aroian et al., 1990). We have attempted to study the mechanism of PTK signaling by first systematically identifying genes whose products may be essential for normal strength signaling by a particular PTK, the product of the Drosophila melanogaster sevenless (sev) gene, and then characterizing the products of these candidate loci. Our hope is that these experiments will both provide in vivo evidence for the involvement of suspected elements of PTK pathways and perhaps identify new signaling proteins. The sev gene encodes a transmembrane PTK (Sevenless) that is most related to the vertebrate woos protein (Hafen et al., 1987; Basler and Hafen, 1988; Bowtell et al., 1988). The sole known function of Sevenless is for the development of the Drosophila compound eye. The eye is composed of approximately 800 repeating units called ommatidia. Each ommatidium consists of a precise array of eight photoreceptors and 12 accessory cells. In the absence of functional Sevenless, each ommatidium lacks a particular photoreceptor, the R7 cell. Studies of the development of flies mutant for sev show that the cell that would

Summary We have conducted a genetic screen for mutations that decrease the effectiveness of signaling by a protein tyrosine kinase, the product of the Drosophila melanogaster sevenless gene. These mutations define seven genes whose wild-type products may be required for signaling by sevenless. Four of the seven genes also appear to be essential for signaling by a second protein tyrosine kinase, the product of the Ellipse gene. The putative products of two of these seven genes have been identified. One encodes a ras protein. The other locus encodes a protein that is homologous to the S. cerevisiae CDC25 protein, an activator of guanine nucleotide exchange by ras proteins. These results suggest that the stimulation of ras protein activity is a key element in the signaling by sevenless and Elripse and that this stimulation may be achieved by activating the exchange of GTP for bound GDP by the ras protein. Introduction Protein tyrosine kinases (PTKs) are important regulatory proteins that control many aspects of cellular growth, differentiation, and metabolism (Cantley et al., 1991). Many polypeptide hormones are known to regulate the metabolism, differentiation, and growth of target cells by binding transmembrane receptors that possess intracellular PTK domains (Cross and Dexter, 1991). In addition, mutations that alter the normal regulation and activity of either receptor PTKs or the closely related cytoplasmic PTKs can generate oncogenes that are capable of altering cellular growth control (Bishop, 1991; Hunter, 1991). Understanding how the PTKs influence cellular activities could provide important insights into how these cellular processes are regulated and coordinated. Both the ligand-bound receptor PTKs and the oncogenic PTKs show elevated PTK activity relative to their unbound or unmutated counterparts. This increase in PTK activity is presumed to be a key element in initiating the signal transduction pathway, because mutations that destroy PTK activity almost invariably abolish function (for a partial exception, see Henkemeyer et al., 1990). The primary biochemical consequence of the increased PTK activity is

* Present address: Howard Florey Institute of Experimental Physiology and Medicine, University of Melbourne, Victoria, Australia,

Cell 702

A
Sevd2.
f

SeVB4 +

@ 22.7C

: e 0.0

R7 present

sevd2.

+.

sevB4

@ 24.3C

00 : :

R7 absent

Look for mutants such that:

@ 22.7C

.e l . l e

R7 absent

1. The Construction

Site-directed mutagenesis was used to generate five new sev alleles, which were then placed into seti2 flies by P element-mediated germline transformation. These mutations were based on the sequences of two temperature-sensitive alleles of v-src, LA24 and LA31, and two revertants, IA24 Rl and LA31 Rl (Fincham and Wyke, 1966). The central portion of the figure shows a comparison of the amino acid sequences of portions of the Sevenless protein and pp60*m. The phenotype of the V-SIC alleles and the changes in the ~~60~ amino acid sequence are shown below. The mutations that were induced in sev and their phenotypes are indicated above the comparison. The phenotypes of flies carrying the mutant sev alleles were examined by the reduced cornea1 pseudopupil method at 16OC and 29%.

0 p
Figure

sevd2;;

* + TM3 ry P[v, sev s4]

screen for absence of the R7 cell

2. The Scheme

normally become the R7 photoreceptor cell instead develops as a nonneuronal lens-secreting cone cell (Tomlinson and Ready, 1966, 1967). Analysis of ommatidia that are mosaic for the sev mutation has demonstrated that the requirement for Sevenless function is limited to the R7 cell itself. Thus, Sevenless is proposed to act as a receptor for a signal that determines whether the presumptive R7 cell becomes a photoreceptor or a nonneuronal cell. In this report we describe a genetic screen for mutations that decrease the effectiveness of signaling by Sevenless. This screen has allowed us to identify seven genes, Enhancers of sevenless, that when mutated can make the phenotype of a temperature-sensitive sev allele more severe. We have molecularly characterized two of the loci. One locus, Rasl (formerly Drasl; Neuman-Silberberg et al., 1964; Bock, 1967), encodes a ras protein. The other locus, Son of sevenless (SOS), has been previously identified on the basis of a dominant allele that enhances signaling by a mutant sev protein (Rogge et al., 1991). We show that SOS encodes a protein that is homologous to the CDC25 protein, an activator of guanine nucleotide exchange by RAS proteins in budding yeast (Broek et al., 1967; Robinson et al., 1967; Jones et al., 1991). These results imply that an increase of ras protein activity is a key element in signaling by the sev PTK and further suggest that this activation may be achieved by stimulating GDPGTP exchange of the Rasl protein. Results Enhancers of sevenless Define Seven Loci The simplest way to isolate mutations that might affect signaling by the sev PTK would be to isolate new recessive mutations that result in the absence of the R7 cell. This

(A) is a summary of the phenotypes of the sevB allele at different temperatures. The black circles are a depiction of the observed reduced cornea1 pseudopupil when the flies are reared at the indicated temperature. The reduced cornea1 pseudopupil represents an optical projection of the positions of the rhabdomeres, the light-sensing organelles of the photoreceptors, from a group of neighboring ommatidia. The R7 rhabdomere is represented by the central circle of the phenotypically wild-type image observed when seyB4 flies are reared at 22.7%. Each asterisk represents a mutagen-treated chromosome. (B) is a more detailed description of the screen for E(sev) mutations. Mutagenized seyd male flies were mated with se@*; CxD, DITMS, Sb, ry, P[ry, sep] females at 22.7%. The male and female Sb progeny (genotypes setiVY: I+; l ITM3, Sb, fy, P[ry, seva4] or sevd2/sevd2; I+; ITM3, Sb, ry, P[ry, seve4]) were scored for the absence of R7 photoreceptors. At this temperature, approximately 2% of the female Sb flies and less than 1% of the male Sb flies display a reduced cornea1 pseudopupil lacking an R7 image in the absence of mutagenesis. Flies with reduced cornea1 pseudopupils lacking the R7 image were backcrossed to sev? CxD, DITM3, Sb, ry, P[ry, sep] flies in order to determine which flies carried E(sev) mutations. The mutant progeny of this cross were then used to balance the E(sev) mutation. The mutagen-treated sevd flies were isogenic for both the second and third chromosomes.

approach has led to the discovery of two genes: bride of sevenless, which appears to encode the ligand for Sevenless (Reinke and Zipursky, 1966; Hart et al., 1990; Krlmer et al., 1991), and seven in absentia, which encodes a nuclear protein and is therefore unlikely to be directly involved in the early steps of Sevenless-mediated signaling (Carthew and Rubin, 1990). We chose a different strategy, because we suspected that much of the pathway downstream of the sev PTK would be shared with other PTKs. While Sevenless function is entirely dispensible for the viability of the fly, at least one other PTK is essential (Schejterand Shilo, 1969; Price et al., 1969). Furthermore, if other PTKs have crucial roles in the development of the eye, disruption of their signaling pathways might result in such disorganized eyes that the presence of the R7 cell would be impossible to score. Mutations that inactivate the common PTK signaling components would therefore be missed in a genetic screen for recessive mutations that cause the absence of the R7 cell. We therefore devised a

Signal Transduction 703

by the Sevenless

PTK

Figure

3. The Phenotype

of E(sev)

Mutations

Each panel is a photomicrograph of a sectioned eye. The flies were reared at 22.7%. The anterior of each eye is to the right. (A) shows the phenotype of an eye from a female fly containing the set@ allele in the absence of E(sev) mutations. The genotype is set@*; +/TM3, Sb, ty, P[ry, se@]. The dark circular structures in each ommatidial cluster are the rhabdomeres, the light-sensing organelles, of the photoreceptors. The rhabdomere of the R7 cell is in the center of each cluster and is surrounded by the rhabdomeres of photoreceptors RI-R6. The rhabdomere of R6 lies proximally in the retina and is not visible in these sections. The R7 cell rhabdomere is present in all but one ommatidial cluster. The arrow points to the exception. (6) and(C) show the phenotype of female flies carrying the E(~ev)3C~ or E(sev)2A m alleles, respectively. The genotypes are(B) seti? E(sev)3C9/ TM3, Sb, ty, P[ry, sep] and (C) se@*; E(sev)2AWG/+; +/TM3, Sb, ry, P[ry. sep4]. In each fly, nearly all of the ommatidial clusters are lacking an R7 rhabdomere and thus an R7 photoreceptor. One exceptional ommatidium with an R7 cell is seen in each panel and is indicated with an arrow. In each panel, the most posterior row of ommatidia is approximately three rows from the posterior edge of the eye. We chose this region because this is the area of the eye that we assay by the reduced cornea1 pseudopupil method. There is an apparent difference in the quantity of Sevenless activity that is required in different regions of the eye. When Sevenless activity is limited, such as when flies carrying the se@ allele are reared at 22.7%, ommatidia in the anterior region of the eye generally lack an R7 cell, while more posterior clusters are still phenotypically wild-type.

more sensitive genetic screen that allows the identification of mutations that reduce rather than abolish the activity of downstream signaling proteins. Systematic studies of the effects of chromosomal deletions on fly development have indicated that there are very few loci that have observable haploid phenotypes (Lindsley et al., 1972). Other studies have shown that, in general, the level of expression of a locus is proportional to the gene copy number (Muller et al., 1931). Given these considerations, our strategy was to create conditions in which signaling through the Sevenless pathway is so barely adequate that even e-fold reductions in the gene activity of downstream elements of the pathway might disrupt signaling. Under these sensitized conditions, a recessive loss-of-function mutation in agene encoding a component of the signaling pathway might yield a dominant R7-minus phenotype even though it only inactivates one copy of the gene. We refer to such mutations as Enhancers of sevenless. An essential requirement for conducting such a genetic screen for Enhancersof sevenless is the ability to precisely control the strength of signaling by Sevenless. Since we could not predict the exact level of signaling that would be ideal for this genetic screen, we constructed two temperature-sensitive alleles of sev that allow fine control of signaling strength. The temperature-sensitive alleles were produced by site-directed mutagenesis of the sev gene based on the sequence of two temperature-sensitive v-src alleles that change amino acid residues conserved between

v-src and sev (Fincham and Wyke, 1986). Five sev mutations were constructed and placed in flies using P element-mediated transformation (Figure 1). Two alleles, se@ and seve4, are temperature-sensitive. The screen for Enhancers of sevenless was conducted as outlined in Figure 2 using the seve4 allele. The presence of the R7 cell was scored by the reduced cornea1 pseuof the sev dopupil method, which allows the rapid scoring phenotype in liveflies(Franceschini and Kirschfeld, 1971). At 22.7OC, the reduced cornea1 pseudopupils of flies carrying the set@ allele are wild type, while at 24.3OC, the reduced cornea1 pseudopupil is clearly lacking the R7 cell image. We screened for ethylmethylsulfonate-induced mutations that result in a reduced cornea1 pseudopupil lacking the R7 cell image at 22.7OC. Approximately 30,000 progeny of mutagenized flies were screened for the absence of R7 cells. Twenty Enhancer of sevenless (E(sev)) mutations were isolated and localized to individual chromosomes. None of these mutations as heterozygotes can prevent the development of the R7 cell of a fly that contains a sev+ gene (data not shown). Two examples of enhancement of the temperaturesensitive sev84 phenotype by E(sev) mutations are shown in Figure 3. Eyes from E(sev)l+ flies generally contain less than 10% wild-type ommatidia, while eyes from flies that do not carry an E(sev) mutation generally contain greater than 90% wild-type ommatidia. However, the extent of enhancement is variable and dependent on the exact temperature at which the flies are reared.

Cell 704

E(sev)lA
e3C e0P 1 X-l

Figure tions

4. Genetic

Mapping

of E(sev)

Muta-

2L

3L

The E(sev) mutations were mapped and tested for possible allelism as described in Experimental Procedures. The approximate recombination map positions (*5 centimorgans) of E(sev)2A each locus are shown below the arrows. The QH map positions of E(sev)3A and E(sev)3B are e2A only accurate to within 10 centimorgans due e4G to their distance from the nearest marker. The eOD E(sev)3A, E(sev)3B, and E(sev)3C mutations elA E(sev)lB l?6C define three essential genes. Animals carrying &SF eOA two mutations of the same groups rarely or a 1 never survive to adulthood. TheE(sev)2A mutav 2-48 Z-70 tions also define an essential gene; however, several of the alleles retain some activity. Animals carrying two strong alleles of the locus, E(sevpA such as E(.sev)2Ati and E(sev)2AUH, die before e3A E(seWB pupation, while animals that carry two weak elD e6B E(sev)SC alleles, such as E(sev)2Am and E(sev)2Ae1Afree4A e1C E(sev)lD quently eclose as small adults with extremely e1B e6A e4D e2F rough and disordered eyes. The placement of e6D e1E e0Q the two X-linked E(sev) mutations, f(sev)7Am a m and E(sev)lA@, into a single group is provi3-5 3-50 3-96 3R sional. The f(sev)lA& mutation is viable in hemizygous males and has a weak effect on the phenotype of the se+ allele, while the E(sev)7AaP mutation is tightly linked to a recessive lethal mutation and has a strong effect on the phenotype of seve allele. We have chosen to display these mutations as allelic because the two mutations map to the same region and are both capable of suppressing the phenotype of the E/p mutation. This interpretation of the data suggests that the f(sev)lAm allele retains some residual activity. However, the possibility that these two mutations affect two separate genes that each modify signaling by the sep and E/p PTKs cannot be excluded.
x

2R

LJ

The E(sev) mutations were then mapped by recombination to approximate chromosomal positions. To test whether mutations that map to the same region are allelic, we crossed flies carrying nearby E(sev) mutations to each other. Assuming that these genes are essential for viability, then an animal carrying two E(sev) mutations that inactivate the same gene should be inviable. The results of the mapping and complementation experiments are shown in Figure 4 (see also Table 1). The 20 E(sev) mutations define seven loci, which can be mutated to disrupt signaling by the seP protein. Multiple mutant alleles of E(sev)2A, E(sev)3A, E(sev)3B, and E(sev)SC, have been isolated and are recessive-lethal mutations, indicating that these loci encode essential functions. A fifth locus, E(sev)lA, may also have been identified by two mutations, but one allele

is viable (see Figure 4 for a discussion of this locus). Two loci, E(sev)28 and E(.sev)3D, have each been identified only once. Although the chromosomes carrying each of these unique mutations carry a recessive-lethal mutation, more alleles will be required to determine whether the recessive-lethal phenotypes are due to the E(sev) mutations. The distribution of the E(sev) mutations among only seven loci suggests that we have identified most of the loci that can be easily mutated to enhance the phenotype of the seP allele.

Four of the E(sev) Loci Appear to Be Required for Photoreceptor Development We next wished to examine the consequences of mutating both copies of the E(sev) loci for the development of the R7

Table

1. Summary

of Properties

of Enhancer

of sevenless Required for Photoreceptor Development Yes Yes Yes ?b ?b Yes No

Loci Suppressor of Ellipse Yes Yes Yes No No Yes No Acts in R7 Yes Ye.9 Potential Biochemical Activity

LOCUS E(sev)lA E(sev)2A/Sos E(sev)2B E(sev)3A E(sev)SE E(sev)3C/Flasl E(sev)SD

Number of Mutant Alleles 28 7 1 5 4 2 1

Putative guanine exchange factor

nucleotide

Yes

GTPase

B See Figure 4 for a discussion of this locus. b No clones were observed. These results are consistent with a requirement for cell growth c Occasional ommatidia with extra photoreceptors are observed. d Rogge et al. (1991) have shown that a dominant allele of this locus acts in the R7 cell.

or viability.

Signal Transduction 705

by the Sevenless

PTK

cell and other photoreceptors. Since the E(Sev) mutations are, with one exception, present on recessive-lethal chromosomes, we generated clones of cells that were homozygous for the E(sev) mutations. For the loci that are represented by multiple alleles, at least two alleles were tested. The analysis of these clones allowed us to divide the E(sev) loci into three classes (Figure 5 and Table 1). Mutations at four loci appear to greatly reduce the ability of a cell to develop as either an R7 cell or as any other photoreceptor. This class includes E(sev)lA, E(sev)2A, E(sev)2B, and &(sev)3C. Despite the fact that these mutations were identified owing to their ability to interfere with R7 cell development, our results suggest that they must also play a role in some process that is common to the development of all photoreceptors. Homozygous clones of two of the loci, E(sev)3A and E(sev)3B, were never or rarely seen. This result is consistent with the conclusion that these loci are essential either for cell growth or viability. One locus, E(sev)3D, appears to be dispensable for both photoreceptor and R7 cell development. The homozygous E(sev)3D ommatidia are generally wild type, though occasionally an ommatidium has an extra photoreceptor. However, this locus is represented by a single mutant allele, which may retain some activity. We examined embryos homozygous for mutations of either the E(sev)2A or the E(sev)3C loci to try to determine which steps of development require the function of these loci. In each case the mutant embyos hatch at frequencies equivalent to wild-type controls but die before the onset of pupal development. This may imply that the products of these loci do not have important roles during embryonic development. However, an equally likely possibility is that there are maternal contributions of these products that are sufficient for embryonic development. Analysis of embryos derived from homozygous mutant germline clones will be required to address this issue. E(sev) Loci Function in the Developing R7 Cell If the E(sev) loci do encode proteins that participate in the transduction of a signal from the activated sev PTK into the cell, then the proteins should function within the R7 cell itself rather than in the cell that transmits the signal. We have tested this prediction for two of the E(sev) loci, E(sev)lA and E(sev)3C, by generating marked clones of wild-type cells in a genetic background that contains the temperature-sensitive sep allele and is heterozygous for the E(sev) mutation. The ommatidia that are outside of the clone lack the R7 cell because of the enhancement of the set+ phenotype by the E(sev) mutation, while the ommatidia that are entirely within the clone are genetically wild-type for the E(sev) locus and thus are phenotypically wild-type. Since there are no absolute cell lineage relationships between cells within an ommatidium, the ommatidia that are at the boundary of the marked clone will contain cells that are heterozygous for the E(sev) mutation and cells that are wild-type for the E(sev) locus. If the E(sev) locus exerts its effect in the developing R7 cell, then we would expect R7 cells developing in a mosaic ommatidium to be wild-type for the E(sev) locus. The genotype of the other photoreceptors should be irrelevant.

The results of such an analysis of mosaic ommatidia are shown in Figure 6 (see also Table 1). For the E(sev)7ABoP mutation, 40 of 41 R7 cells from phenotypically wild-type ommatidia were wild-type for the E(sev)lA locus. Similar results (34 of 35) were obtained for the E(sev)3CF mutation. The two exceptions are most likely due to the presence of a small fraction of R7 cells even in the presence of the E(sev) mutations (see Figure 2). These results are entirely consistent with the conclusion that the E(sev)7A and E(sev)SC mutations act in the developing R7 cell to attenuate signaling by the sev PTK. The tendency of the Rl and R6 cells to be related by lineage to R7 and the incomplete penetrance of the E(sev) phenotypes makes it difficult to absolutely eliminate the possibility that these mutations also act in the Rl and R6 cells to block the formation of the R7 cell. However, the data clearly rule out the possibility that the two E(sev) mutations act by decreasing the strength of signaling by the R8 cell from which the bride of sevenless-mediated signal originates (Reinke and Zipursky, 1988; Kramer et al., 1991). Mutations of Four f(sev) Loci Attenuate Signaling by Another PTK An underlying assumption of our study was that different PTKs use at least some common signaling pathways. We asked whether the E(sev) loci might encode such elements of a common signaling pathway by testing whether the E(sev) mutations could attenuate signaling by a PTK other than Sevenless. The PTK that we chose to test was the Ellipse (E/p) allele of the faint little ball locus (also known as torpedo and DER). The faint little ballgene encodes a PTK receptor that is most homologous to the vertebrate EGF receptor and the neu proteins (Livneh et al., 1985; Schejter and Shilo, 1989; Price et al., 1989). Recessive loss-of-function mutations of faint little ball cause embryonic death. The mutation E/p is a dominant allele of the locus, which causes moderate roughness of the eye when heterozygous and severe eye roughness and reduced eye size when homozygous (Baker and Rubin, 1989). The basis for these phenotypes seems to be impaired formation of ommatidial clusters. The eyes of E/p/+ flies have normal ommatidia, but fewer of them than wild-type eyes. The E/p allele appears to encode an overly active protein that is so slightly activated that the activity provided by the normal wild-type copy of the gene is still a phenotypically significant fraction of the total activity; while E/p/+ flies have rough eyes, E/p/- flies have wild-type eyes. The sensitivity of the mutant eye to this slight change in gene activity suggested that if the E(sev) mutations did compromise signaling by the E/p PTK, the effect would be seen as a suppression of the E/p phenotype. Flies carrying each E(sev) mutation were crossed to flies carrying the E/p mutation in order to generate flies that were heterozygous for both mutations. These flies were examined for the roughness of their eyes. The results are illustrated in Figure 7 (see also Table 1). Mutations in the E(sev)3A, E(sev)3B, and E(sev)3D loci had no effect on the E/p phenotype. Mutations in the remaining four loci were capable of suppressing the E/p phenotype. Three loci,

Cell 706

Figure

5. The Requirement

for the E(sev)

Loci during

Photoreceptor

Development

Animals heterozygous for an Efsev) mutation were X-irradiated in order to generate clones of cells homozygous for the E(sev) mutation. The homozygous Efsev) cells are marked by the absence of a white gene. The white gene is required for the formation of pigment granules in both photoreceptors and pigment cells. The pigment granules of the photoreceptor cells are apparent as dark structures near the rhabdomere of each

;;?a1

Transduction

by the Sevenless

PTK

E(sev)lA, E(sevpA, and E(sev)PB, werevery effectivesuppressors and restored the eye to a nearly normal appearance. Mutations at a fourth locus, E(sev)X, were less efficient in their suppression, but were clearly distinguishable from E/p/+ flies. These results suggest that at least four of the E(sev) loci encode proteins that somehow participate in a signaling pathway that is shared by at least two PTK receptors. The same four loci are required for the formation of photoreceptor cells. These results suggest that the products encoded by these four loci may be functionally linked, perhaps as parts of a single pathway. We therefore chose to concentrate on the molecular cloning of two of these four loci. The E(sev)3C Locus Is the Rasl Gene The E(sev)3C locus is represented by two mutant alleles. Recombination mapping of the alleles placed the locus on the right arm of the third chromosome near curled. The position was further refined by complementation mapping using deletions for the region. Two deletions were particularly informative. The chromosome carrying M(3R)bylO is deleted for the 85D8-11 to 85ElO-13 region and is unable to complement the lethality of either E(sev)3C mutation. The chromosome carrying Df(3R)by62 is deleted for the region 85017-22 to 85Fl3-18 and rescues the lethality of either allele (Kemphues et al., 1980). These results place the E(sev)JC locus in the chromosomal region 85D8-22. One gene that is both believed to play a role in signal transduction and that is known to map to this region is Rasl, the Drosophila homolog of the c-H-rasl gene of vertebrates (Neuman-Silberberg et al., 1984). We sought in two ways to test directly whether the E(sevpC mutations mapped to Rash. Our initial approach was to place a 12 kb DNA fragment containing the Rasl gene into the germline by P element-mediated transformation and test whether this construct would rescue the lethality of animals carrying E(sev)SC mutations. The P element carrying Rasl, P[w,RaslA], was capable of rescuing the lethality of E(sev)3CYe/E(sev)3CF flies. This result indicates that the E(sev)3C gene lies within the 12 kb transformation fragment. The E(sev)3C mutations were then localized to the Rasl gene itself by sequencing the Rasl genes from the

E(sev)SC mutant chromosomes. The sequence of the entire coding sequence of each allele was determined and compared with a previously published sequence of Rasl. Both mutant allelescontainedchangesthat altertheamino acid sequence of the resulting Rasl protein (Figure 8). The Rasl allele from the E(sev)3PB chromosome contains a point mutation that changes glutamate-82 to a lysine. The E(sev)3@2F allele contains a point mutation that changes aspartate-38 to an asparagine. We conclude from these results that E(sev)3C is Ras7. The two mutant alleles have therefore been renamed RaslEmK and RaslD38N. The E(sev)2n Locus Encodes a Potential Guanine Nucleotide Exchange Factor E(sevp4 was the second gene that we sought to identify. This locus is represented by seven mutant alleles. The mutations differ significantly in their ability to enhance the sevfu phenotype and to suppress the E/p phenotype. When strong alleles, such as E(sev)2Ae2H and E(sevpA&, are crossed to each other, viable E(sevpA/E(sevpA progeny are not observed. However, combinations of weak alleles, such as E(sev)2Amo and E(sev)2Ae1A, are generally viable, but the mutant progeny have rough eyes (data not shown). Recombination mapping indicated that the E(sev)u\ locus is near the black gene (polytene position 34D4-8). Rogge et al. (1991) have recently reported the isolation of a dominant allele of the gene /(2)34Ea that maps adjacent to black (Woodruff and Ashburner, 1979; M. Ashburner and J. Roote, personal communication). This mutant allele was identified by its ability to suppress a particular mutant allele of sev. The locus was renamed Son of sevenless (SOS). These results suggested the possibility that the E(sev)2A alleles might also be allelic to /(2)34Ea mutations. Crosses between E(sev)2A alleles and two alleles of /(2)34Ea confirmed this hypothesis (data not shown). We will therefore refer to the E(sev)2A locus as SOS. The extensive genetic analysis available for the 340 region suggested that SOS is the first essential gene distal to black. This places SOS in the vicinity of polytene band 34D4. We isolated the DNA from this region by chromosome walking. The starting point for the walk was a hobo transposable element inserted in polytene band 34Dl (Blackman et al., 1989). We isolated overlapping cosmid

photoreceptor, but pigment granules are not present in every section. (A) shows a clone of white cells in the absence of E(sev) mutations. The spacing and construction of the ommatidial cl&ters is normal. (B) shows a clone of cells homozygous for the E(sevJ3PW mutation. Most of the ommatidia are phenotypically wild-type, but occasionally ommatidia contain an extra photoreceptor (see arrows). The remaining panels show the phenotypesof clonesof cells that are homozygousfor different Efsev) mutations: (C), f(sev)lA? (D), E(sev)2A? (E), E(sev)2B? (F), E(sev)3t7e. Analysis of a number of mosaic ommatidia, as judged by the presence of at least one wild-type and one mutant pigment cell, indicates that cells homozygous for any of these mutations are greatly underrepresented as photoreceptors. Approximately half of the photoreceptors from mosaic ommatidia are expected to be homozygous for white in the absence of E(sev) mutations. The fraction of photoreceptors from mosaic ommatidia 41345; E(sev)2PA, 2f411; .E(sev)3C+ls, 15i798. that were homozygous for the .E(sev) mutations were as follows: E(sev)lA *)p, O/334; E(sevpW, Similar results were obtained for clones homozygous for the E(sev)3cZF mutation. The analysis of other alleles of the E(sev)2A locus was more for complex. Clones of cells homozygous for the E(sev)2A MQ allele were similar to those for the E(sev)2A an allele. However, clones homozygous the f@ev)2A~D allele were more normal. Many ommatidia lacked one or two photoreceptors, but homozygous E(sev)2Ad0 photoreceptors were termination codons and produce substantially truncated proteins, we seen. Since the Efsev)2RzH and E(sev)2A MQ alleles both contain premature conclude that this is the null phenotype for this locus (see Figure 10). The phenotype of clones of cells homozygous for the E(sev)U\* mutation (see main text). When tested for lethality in is similar to that observed by Rogge et al. (1991) for an allele of this locus, called Son of seven/essX trans to other E(sev)2A alleles, the E(sev)2Aa0 mutation has a weaker effect than either E(sev)2AUn or E(sev)2A. Furthermore, the E(sev)2Areported allele, Son of severt/essx, retain partial mutation is a missense mutation (see Figure 10). We conclude that E(sev)2AmD and the previously activity.

Cell 708

.?: .o:.o: .e: 000 0000 080 ago 0% 00: 00: .O: ego 080 000 080 Ikkk
Figure 6. The E(sev)7AP and E(sev)3PB Mutations Act in the R7 Cell X-irradiation was used to generate marked clones of E(sev) cells in flies that carried the seP allele and were heterozygous for the E(sev) mutation. The cells of the clothes were genetically marked by the absence of a functional white gene and thus lack pigment granules. If an E(sev) mutation exerts its effect in a particular cell, then that cell should tendto be white in mosaic ommatidia that are wild-type iri construction. (A) and (6) are photomicrographs of such clones for the (A) E(sev)lAdP and (B) E(sev)3CYB alleles. In each case, the presence of the Qsevj mutation in the cells outside of the clone blocks the ability of the seP alleleto initiate fly development. (C) and (D) illustrate the genetically mosaic and phenotypically wild-type classes of ommatidia that were observed for E(sev)lAaP and E(sev)3PB, respectively. When a given genotypic class was observed more than once, the number of occurrences is indicated in the upper left of the box. Open circles represent white, E(sev) photoreceptors, while closed circles represent f(sey)/+ photoreceptors. The data can be summarized as the fraction of white cells for each photoreceptor. For E(sev)7AmP, the data are (41 total): RI, 1U; R2, 28; R3, 27; R4, 24; R& 25; R6, 10; R7, 1; R8, 19. For E(.ev)3cBiB, the data are (35 total): RI, 4; R2, 23; R3, 21; R4, 22; R5, 22; R6, 6; R7, 1; R8, 16. In,each case, the data are most consistent with the conclusion that the Qsev) mutation acts in the R7 cell. Sincethe extent of enhan:ement by the E(sev) alleles varies with verjl slight changes of the incubator temperature, we discarded clones when more than one phenotypically w,l(d-type ommatidium that was not part of the clone was apparent in the sectioned region. Occasionally, severely defective ommatidia were se& cilpng the edge of the clones. These ommatidia presumably contained homozygous f(sev) cells that were generated by the mitotic recombination.

clones

representing the approximately 150 kb of DNA in the 34Dl-8 region. The approximate position of the SOS gene within the cloded DNA was determined by analyzing the breakpoints of several chromosomal rearrangements.

This analysis indicated that at least part, if not all, of the SOS gene is contained within an 8 kb interval corresponding to coordinates 110 to 118 (see Figure 9). This prediction was tested by placing a 10 kb Bglll-Xbal DNA

Signal Transduction 709

by the Sevenless

PTK

Figure

7. The Effect

of E(sev)

Mutations

on the Phenotype

of the E/p Mutation

All of the eyes shown except those in (A) and (D) are from female flies heterozygous for the set@ mutation. (A) Scanning electron micrograph (SEM) of an eye from a female wild-type (Cantons) fly. (B) SEM of an eye from a fly heterozygous for the E/p mutation. The E/p mutation causes reduction in eye size and roughening of the eye surface (Baker and Rubin, 1999). (C) SEM of an eye from a fly heterozygous for both E(sev)u\tin and E/p. The presence of the E(sev) mutation strongly suppresses the phenotype of the E/p mutation. (D) to (L) show higher magnification SEMs that illustrate the effect on the E/p phenotype by different E(sev) mutations. (D) is an SEM of an eye from a wild-type (Cantons) female fly. (E) is an SEM from a fly heterozygous for the E/p mutation. The flies in(F) to (L) were heterozygous for the E/p mutation and the indicated E(sev) mutation: (F), E(sev)2AeH; (G), f(sev)2/V; (H), E(sev)3PB; (I), E@ev)3C? (J), E(sev)7AaP; (K), E(sev)2B? (L), E(sev)3EF. Since E/p disrupts the normal hexagonal array of the ommatidial lenses and bristles, the suppression of E/p can be scored as the return of the normal hexagonal array. E(sev)7AdP, E(sev)2Aa, &(sev)2Am, and E(sev)2ed* largely reverse the effects of the E/p mutation. The mutations f(sev)XYs (H) and E(sev)3F (I), partially suppressed the E/p phenotype. Note that there are still some irregularities in the shapes of the lenses and in the placement of the bristles. Mutations at the remaining E(sev) loci had no effect (see [L] for an example).

Cdl 710

e2F: D-RhS1 H-RAS

elf?,.

K t ii:

D-RASI H-RAS Figure

N 6 6. The E(sev)SC Locus Is Rasl

RRGRKNYKPNRR NPPbESGPGCnS

L s

189 189

The sequence of the Drosophila Rasl protein (Brock, 1967) is compared with that of the Human p21 KU (Capon et al., 1963). Identical residues are boxed in black. Similar residues are boxed in gray. The following amino acids were considered similar: E, D; V, L, I, M; A, G; F, Y, W; S, T; Cl, N. The mutation present in the E(sev)3pF allele changes the GAC aspartate-36 codon to an AAC asparagine codon. The mutation present in the .E(se~)3C~ codon changes the GAG glutamate-62 codon to an AAG lysine codon. No other changes were observed.

fragment (coordinates 110 to 120) into the germline by P element-mediated transformation. This P element, P[w, SosA] is capable of rescuing the lethality of animals carrying either strong allele, f{sev)2Ae2H or E(sev)2AtiG, This indicates and a deletion of the SOS gene, Df(2L)bso. that all essential functions of SOS are contained within this 10 kb region. We isolated eye-antenna1 imaginal disc cDNA clones derived from this region. Two classes of nonoverlapping cDNA clones were categorized. One class originates from the distal portion of this interval and from the neighboring Xbal fragment. This class is not fully contained within the 10 kb rescue fragment and was not characterized extensively. The second class of cDNA clones represents an mRNA transcript that spans most of the region implicated

by the deletion mapping. The sequences of a 5.3 kb cDNA clone and the genomic region encoding this transcript were determined. The deduced structure of the transcript is shown in Figure 9. Conceptual translation of the long open reading frame of this transcript from the first in-frame methionine shows that it is capable of encoding a protein of 1596 amino acids (Figure 10). We determined that this open reading frame actually encodes the SOS protein by cloning this region from three of the chromosomes containing SOS mutations and sequencing the open reading frames (Figure 10). Each mutant allele contains a point mutation that changes the predicted protein sequence. For instance, the strong alleles SOS~*~and SO&~, introduce termination codons at aminoacid positions 579 and 421, respectively. The weak

, I

IOkb 1

, I I I I I II 8 probe for RNA blot: I I

Df(2L) b em . b 85*2 II I 1 I ::

Df(2L) (SOS +)

ScoRv7

(SOS - )

III

I I ~ II s -

I 8

II

II

B =
2 I L lkb ,

5
= 75 .s x

A I I F s

I I
11 FE $2 I

I
F i3

Figure

9. Structure

of the SOS Gene

(A) A map of the region near SOS. The map of the SOS region is shown for restriction enzymes Xbal (marks below the line) and BamHl (marks above the line). Distal is to the left. The coordinates are given as distance from the insertion site of the hobo[ry] element that served as the entry point for the chromosomal walk. The approximate positions of chromosomal breakpoints were determined by genomic DNA blots and are indicated above These chromosomes are deficient for the map. The solid bar represents DNA that is deleted from the Df(2L)bwh and Df(2L)Sco kv7 chromosomes. polytene bands 34D3 to 35A4 and 34D5 to 35D5-7, respectively (M. Ashburner and J. Roote, personal communication). The broken line represents is not entirely clear, but there appears to the uncertainty in assigning the position of the distal breakpoint. The nature of the besb2 rearrangement be a breakpoint in the region indicated by the broken line. (B) The structure of the SOS transcript. The region (coordinates 110 to 120) indicated by the bold line in (A) is shown. The deduced structure of the SOS mRNA is shown above the map. The exon and intron structure was deduced from a comparison of the genomic sequence of the 6601 bp Hindlll to Sall region with the sequence of a 5.3 kb cDNA clone that was isolated from an eye imaginal disc cDNA library. The cDNA clone was incomplete at both ends. The indicated position (nucleotide 6451) of the polyadenylation site was determined from an SOS cDNA clone derived from an embryonic cDNA library that included a poly(A) tail. The Blend of the mRNA was determined by Sl mapping and primer extension studies and is heterogeneous. There appear to be several start sites; the two major sites map to the 440 to 470 region. In the diagram of the mRNA structure, protein coding regions are indicated by the black line and 5- and 3-untranslated segments by the gray line. The 10 kb Bglll-Xbal fragment is capable of rescuing the recessive lethality of SOS mutations when placed into the genome by P element-mediated transformation. (C)Size of the SOS mRNA. Polyadenylated RNAwas extracted from imaginal discs. Two micrograms of RNAwas fractionated, blotted, and hybridized to =P-labeled DNA from the region indicated in (6). A single transcript of approximately 6 kb was detected. Analysis of embryonic RNA revealed a transcript of the same size (data not shown).

Signal Transduction 711

by the Sevenless

PTK

SOS SOS SOS

:9

SOS SOS SOS SOS SOS SOS SOS SOS CDC25 f% STE6 SOS CDC25 SDC25 BUD5 STE6

137 205 273 341 409 477 545 613 641 1121 7:: 493 709 1189 853 173 542

MFSGPSGHAHTISYGGGIGLGTGGGGGSGGSGSGSQGGGGGIGIGGGGVAGLQDCDGYDFTKCENAAR WRGLFTPSLKKVLEQVHPRVTAKEDALLYVEKLCLRLLAMLCAKPLPHSVQDVEEKVNKSFPAPIDQW ALNEAKEVINSKKRKSVLPTEKVHTLLQKDVLQYKIDSSVSAFLVAVLEYISADILKMAGDYVIKIAH CEITKEDIEVVMNADRVLMDMLNQSEAHILPSPLSLPAQRASATYEETVKELIHDEKQYQRDLHMIIR VFREELVKIVSDPRELEPIFSNIMDIYEVTVTLLGSLEDVIEMSQEQSAPCVGSCFEELAEAEEFDVY KKYAYDVTSQASRDALNNLLSKPGASSLTTAGHGFRDAVKYYLPKLLLVPICHAFVYFDYIKHLKDLS STOP (e4G) A SSQDDIESFEQVQGLLHPLHCDLEKVMASLSKERQVPVSGRVRRQLAIERTRELQMKVEHWEDKDVGQ NCNEFIREDSLSKLGSGKRIWSERKVFLFDGLNVLCKANTKKQTPSAGATAYDYRLKEKYF~RRVDIN DRPDSDDLKNSFELAPRMQPPIVLTAKNAQHKHDWWADLLNVITKSMLDRHLDSILQDIERKHPLRMP SPEIYKFAVPDSGDNIVLEERESAGVPM r, STOP (e2H)

HKWQKPSPFDSANL _____----------

TDNALLQELLLSYP SSSDLFSILmHFR

SOS CDC25 BUD5 SDC25 STE6 SOS CDC25 BUD5 SDC25 STE6 SOS CDC25 SDC25 BUD5 STE6 SOS CDC25 BUD5 SDC25 STE6 SOS SOS SOS SOS SOS SOS SOS SOS SOS Figure

776 1255 E: 610 844 1320 E8 677 910 1387 1034 377 743 975 1455 22 810 1021 1089 1157 1225 1293 1361 1429 1497 1565 YQNQPYCLNEESTIRQFFEQLDPFNGLSDKQMSDYLYNESLRIEPRGCKTVPKFPRKWPHIPLKSPGI KPRRQNQTNSSSKLSNSTSSVAAAAAASSTATSIATASAPSLHASSIWDAPTAAAANAGSGTLAGEQS PQHNPHAFSVFAPVIIPERNTSSWSGTPQHTRTDQNNGEVSVPAPHLPKKPGAHVWANNNSTLASASA RDVVFSPALPEHLPPQSLPDSNPFASDTEAPPSPLPKLVVSPRHETGNRSPFHGRMQNSPTHSTASTV TLTGMSTSGGEEFCAGGFYFNSAHQGQPGAVPISPHVNVPMATNMEYRAVPPPLPPRRKERTESCADM AQKRQAPDAPTLPPRDGELSPPPIPPRLNHSTGISYLRQSHGKSKEFVGNSSLLLPNTSSIMIRRNSA IEKRAAATSQPNQAAAGPISTTLVTVSQAVATDEPLPLPISPAASSSTTTSPLTPAMSPMSPNIPSHP VESTSSSYAHQLRMRQQQQQQTHPAIYSQHHQHHATHLPHHPHQHHSNPTQSRSSPKEFFPIATSLEG TPKLPPKPSLSANFYNNPDKGTMFLYPSTNEE 1596 Amino Acid Sequence of the SOS Protein

10. The Predicted

A conceptual translation of the long open reading frame of the SOS RNA transcript is shown. The translation begins at the first in-frame methionine codon, which is approximately 370 bases from the 5 end of the transcript (position 841 in the sequenced genomic DNA). The coding regions of three mutant alleles were sequenced. The positions of mutations that change the coding capacity of the transcript are indicated: SOS&, CAG codon to TAG; SOS&, TGG codon to TGA; SOS tiD , TCC codon to TTC. The SosHo allele contains one additional change (C to T at position 1473 in the genomic sequence) that does not affect the protein sequence. The predicted SOS protein sequence is compared with the sequences of four related proteins: the S. cerevisiae CDC25 protein (Camonis et al., 1986; Broek et al., 1987); the S. cerevisiae SDC25 protein (Damak et al., 1991); the S. cerevisiae BUD5 protein (Chant et al., 1991; Powerset al., 1991); and the S. pombe STEGprotein (Hugheset al., 19gO). Residuesare boxed wherever they are identical or similar to the SOS protein sequence. Identical residues are boxed in black. Similar residues are boxed in gray. The following amino acids were considered similar: E, D; V, L, I, M; A, G; F, Y, W; S, T; Q, N. The alignments were constructed using the Genalign program (Intelligenetics, Inc.)

allele, SOP, has a point mutation that changes serine783 to a phenylalanine. We therefore conclude that this transcript encodes the SOS protein. Comparison of the predicted SOS protein sequence with previously reported sequences shows considerable hoSDC25, and BUD5 proteins of mology with the CDC25, Saccharomyces cerevisiae and with the STEG protein of Schizosaccharomyces pombe (Figure 10). The SOS pro-

tein is most similar to the CDC25 and SDC25 proteins. These proteinsshare a homologous domain that is approximately 380 amino acids in length. For the alignment shown, the CDC25 and SDC25 proteins are 28% and 22% identical, respectively, to the predicted SOS protein in this region. Genetic studies have suggested that the CDC25 protein functions by activating ras protein activity (Broek et al.,

1987; Robinson et al., 1987). Biochemical studies have demonstrated that the CDC25 and SDC25 proteins can act as guanine nucleotide exchange factors to stimulate the conversion of inactive GDP-bound RAS2 protein to the active GTP-bound form (Jones et al., 1991; Crechet et al., 1990). These studies have further shown that a fragment of the SDC25 protein that includes the homology to SOS is sufficient to catalyze nucleotide exchange by either the RASP protein or by the human c-l-bras protein. These results suggest that the region of homology defines a domain that catalyzes guanine nucleotide exchange by ras proteins and that the SOS protein may also act by catalyzing the activation of a ras protein. Discussion In this study, we sought to identify genes that are required for signaling by the sev PTK. Our strategy was to limit the strength of signaling initiated by Sevenless so that the signal was barely adequate to achieve its function. We reasoned that when the sev PTK activity was barely above the threshold necessary for R7 cell formation, small reductions in the abundance or activity of other elements of the pathway might lower the signal strength sufficiently to cause a mutant sev phenotype. Our hope was that this system would be so sensitive to the levels of components of the pathway that mutations which inactivate only one copy of a gene that encodes a downstream element of the pathway would result in the absence of the R7 cell. Since the other copy of the gene would remain functional, these loci, Enhancers of sevenless, could be identified even if their function is essential for producing a fertile fly. Analysis of the first 20 E(sev) mutations identified seven E(sev) loci. For the two best characterized loci, SOS and Rasl, the mechanism of enhancement is inactivation of one copy of an essential gene. The nature of the E(sev) mutations in the other five loci has not been determined. However, for at least two loci, E(sev)3A and E(sev)36, the frequency of E(sev) alleles is consistent with the proposal that these mutations are loss-of-function alleles. Furthermore, our assumption that many of the genes whose products are required for normal signaling would be essential for viability appears to be correct for at least four of the seven E(sev) loci. We do not have enough alleles of the other loci to determine whether they are essential genes. The distribution of alleles among the E@ev) loci suggests that we have identified most, but probably not all, of the loci that can be easily mutated to enhance the se@ phenotype. This screen was designed to identify genes that might encode proteins that participate directly in signaling by Sevenless. Mutations in several other classes of genes might also be capable of producing an E(sev) phenotype. For example, mutations in genes whose products regulate the expression or stability of the temperature-sensitive sev protein might reduce the Sevenless-mediated signal. Another possible class of f@ev) mutations could affect the abundance of the ligand that activates the sev protein. It is interesting to note that a mutation that inactivates one allele of the bride of sevenless gene, the

best candidate to be that ligand (Kramer et al., 1991) does not show a strong enhancement of the sep phenotype in our assay (data not shown). This failure to enhance the sep4 phenotype might be expected if the ligand is presented in considerably greater quantity than is necessary to saturate the sev protein ligand-binding sites. Some of the E(sev) mutations might also be expected to limit signaling by pathways that are parallel to the Sevenlessmediated pathway and are also essential for the formation of the R7 cell. We do not know whether such signaling pathways exist. A Role for the Rasl Protein in Sevenless Signaling The observation that the E(sev)SC locus is the Rasl locus, a gene whose product is closely related to the vertebrate ~21~ proteins (Neuman-Silberberg et al., 1984; Brock, 1987), suggests that at least some of the signals initiated by the activation of either the sev or E/p PTKs are transduced via activation of Rasl protein. Evidence for the involvement of ras proteins in signaling by other PTKs comes from several sources. The earliest indication that Ras activity is necessary for the action of PTKs came from experiments showing that the injection of monoclonal antibodies that inhibit Ras activity blocks the ability of oncogenie PTKs to transform cells (Smith et al., 1988). Similarly, the expression of dominant interfering ras proteins can block the action of pp80(Feig and Cooper, 1988). These results are consistent with the idea that Ras might be part of the signaling mechanism initiated by PTKs. Additional evidence for the association of Ras activity with signaling by PTKs comes from recent studies of vulva1 development in the nematode Caenorhabditis elegans. A PTK, the product of the let-23 gene, is essential for the decision by vulva1 precursor cells to adopt a primary vulva1 fate (Aroian et al., 1990). Mutations that reduce the activity of a ras protein, the product of the let-60 gene, also prevent the vulva1 precursor cells from adopting a vulva1 fate (Han and Sternberg, 1990; Beitel et al., 1990). Furthermore, dominant mutations that activate the let-60 protein, or the overexpression of the normal let-60 protein, cause too many of the vulva1 precursor cells to commit to the vulva1 fate and eliminate the need for the let-23 PTK. These data suggest that the function of the let-23 PTK in vulva1 development is to activate the let-60 (ras) protein. However, there is no evidence that the let-23 and jet-60 gene products act in the same cell, a requirement for their proposed relationship. All of these results are consistent with the hypothesis that activation of ras proteins is Controlled by PTK activity and is crucial for proper signaling by many PTKs. Our results are, however, also consistent with the hypothesis that Rasl protein acts in a separate pathway that is merely required for the action of the Sevenless-mediated pathway. Biochemical analysis will be required to address whether Sevenless activation leads to Rasl activation. A Model for Rasl Protein Activation The activity of a ras protein is regulated by the guanine nucleotide that is bound to the protein (reviewed by Bourne et al., 1990, 1991). The GTP-bound state is active, while

Signal Transduction 713

by the Sevenless

PTK

SEVENLESS

Figure 11. A Model and Rasl Proteins

for the Action

of the SOS

See text for description.

the GDP-bound state is not. The ratio of GTP:ras to GDP: ras is determined by two reactions. An active GTP:ras molecule is inactivated by the intrinsic GTPase activity of the ras protein. This inactivation reaction is greatly stimulated by the presence of a GTPase activating protein, rasGAP (Trahey and McCormick, 1987; Gibbs et al., 1988). An inactive GDP:ras molecule is activated by the exchange of the bound GDP molecule for a GTP molecule. The activation reaction has only been extensively characterized in budding yeast, where the CDC25 protein acts as a guanine nucleotide exchange factor to catalyze the reaction (Jones et al., 1991). The in vivo catalyst for guanine nucleotide exchange by Ras in metazoan organisms is not certain, although exchange-promoting activities have been detected in relatively crude extracts of cells and, in one case, purified to near homogeneity. The discovery that E(sev)2A, that is, SOS, encodes a protein that is related to the CDC25 protein suggests a possible biochemical link between the Sevenless and Rasl activity (see Figure 11). If the similarity between the CDC25 protein and the SOS protein is indicative of similar activities, then our genetic results are consistent with a model in which the activation of the sev PTK by ligand binding stimulates the activity of the SOS protein. The activated SOS protein would then stimulate Rasl protein activity by promoting the conversion of GDP:Rasl to GTP: Rasl. This model requires that the Rasl and SOS proteins each perform their roles in signal transduction in the developing R7 cell. We have demonstrated this directly for Rasl, while Rogge et al. (1991) have shown that the dominant activated Soslc2 allele acts in the R7 cell. The mechanism for stimulation of SOS protein activity by Sevenless could be achieved directly through tyrosine phosphorylation of SOS protein by the sev PTK or indirectly. Alternately, activation by Sevenless could be indirect. If so, then the products of the remaining E(sev) loci are the prime candidates for the roles of intermediaries. Our experiments cannot exclude the possibility that the

SOS protein might be a constitutive activator of Rasl and that Sevenless could act solely by inhibiting the activity of a negative regulator such as rasGAP. Biochemical studies will be required to determine whether SOS protein activity is actually regulated by Sevenless activity. The available data do, however, permit the conclusion that the level of SOS protein activity can be a limiting step in the decision by the presumptive R7 cell to become an R7 cell. The observations that the inactivation of one copy of the SOS gene decreases the effectiveness of signaling by Sevenless, while a dominant allele, which presumably elevates SOS activity, increases signaling show that small changes in the level of SOS function can be critical. Only two of the seven E(sev) loci have been molecularly characterized. The finding that two genes encode proteins well-suited to participate in signaling suggests that this genetic screen is capable of identifying many components of the signaling pathway initiated by Sevenless. What are some of the other participants that we might expect to find? Since the targets forras protein action are undefined, the most interesting possibility is that one of the proteins might be an effector molecule that is regulated by the Rasl protein. Since ras proteins appear to participate in the signaling by many PTKs, the strongest candidates for a gene encoding a Rasl effector molecule are probably the remaining two E(sev) loci that also affect the Ellipse pathway. The products of the E(sev) loci that affect signaling by the sev PTK but not by the E/p PTK are candidates for pathway components that differentiate the cellular response to Sevenless activation from the response to the activation of other PTKs.
Experimental Procedures

Genetics Fly culture and crosses were performed according to standard procedures. The mutagenesis screen for Enhancers ofsevenless mutations was performed as follows. Male seP2 flies that were isogenic for the second and third chromosomes were fed 25 mM ethylmethanesulfo-

Cdl 714

nate as described (Lewis and Bather, 1968) and mated to se@; CxD, DITM3, Sb, ry, P[ry,se+] females. The Fl progeny were reared at 22.7OC, and Sb individualswere assayed by the reduced cornea1 pseudopupil method for the presence of R7 photoreceptors (Franceschini and Kirschfeld, 1971). Approximately 30,000 Sb Fl progeny were screened. Individuals that displayed abnormal reduced cornea1 pseudopupils were crossed to s&*; CxD, DITM3, Sb, ry, P[ry,set@] flies and their progeny were examined for R7-minus reduced cornea1 pseudopupils. These individuals were used to map the E(sev) mutation to a chromosome and then balance the mutation. The chromosomes used for both segregation analysis and for balancing were FM7c, CyO, and TM3. The X-linked mutations were mapped meiotically with the markers y, cv, f, and cer. Females heterozygous for the .E(sev), seti2 chromosome and a y, cv, seti*, f, car chromosome were crossed to set@; CxD, DITMS, Sb, ry, P[ry,seve4] males. Individual virgin females were collected and scored for the E(sev) phenotype. These females were then mated toy, cv, set@, f, car males, and their progeny were scored for the presence of the markers. Similar mapping crosses were performed for the second and third chromosome E(sev) mutations using sev? b, pr, c, px, sp and sevd2, t/r, St, cu, ST, e, ca flies, respectively. However, in these cases, the E(sev) phenotypes were more difficult to score, and we therefore only backcrossed females that were obviously E(sev). In all cases, approximately 50 to 100 potentially recombinant chromosomes were scored. We estimate an error rate of approximately 2% in scoring the E(sev) phenotpye. The E(sev)24 locus was further mapped by the failure of mutant alleles to complement chromosomes carrying either /(2)34,QDM7 or /(2)34EssF5 (Woodruff and Ashburner, 1979; M. Ashburner and J. Roote, personal communication). Histology Fixation and sectioning (2 mm) of adult Drosophila eye was performed as described (Tomlinson and Ready, 1987). Scanning electron microscopy was performed as described (Kimmel et al., 1990). Mosaic Analysis Clones of cells homozygous for E(sev) mutations were generated by X-ray-induced mitotic recombination between heterozygous E(sev)+ and E(sev) mutant chromosomes. Clones were visualized by marking the E(sev)+ chromosome with the cell-autonomous w/rife+ gene. Clones were identified as white patches in otherwise pigmented eyes. For E(sev) mutations on the autosomes, the white+ gene was provided on a transposable element. The following elements were used: 2L, P[(w, ry)A-R]4-2 in 39E-F; 2R, P[(w, ry)A-RIO42 in 49D; 3L, P[w]33 in 70C; 3R, P[(w, ry)D]3 in 90E (Hazelrigg et al., 1984; Levis et al., 1985; our unpublished data). For the X chromosome, the .E(sevj chromosome was marked with white and the E(sev)+ chromosome was marked by the normal chromosomal copy of white+. All flies carried one copy of the sev+ gene provided either at its normal chromosomal location or on an autosome as part of a P element. Flies were irradiated as first instar larvae (1 OOOR). The flies were reared at 25OC. The frequency of clones was approximately 1 in 100 flies. Clones of wild-type cells were generated in a similar manner, except that the E(sev) chromosome rather than the .E(sev)+ chromosome was marked by the white+ gene. The E(sev)+ cells are therefore marked as white. For the E(sev)lA locus, E(sev)7AdP, sevd2 females were crossed to w, se@*; CxD, DITM3, Sb, ry, P[ry, sev] males. For the Rasl locus, w~, se@; RaslW/TM3 females were crossed to w~, sev? P[ry, se+]A males. P[ry, se@]A is a second chromosome insert. The flies were irradiated as above and reared either at 22.7OC (for .E(sev)7A experiments) or at 23.5OC (for Rasl experiments). The difference in temperature is due to a slight difference in the level of expression from the two P[ry, se@] inserts. Progeny of these crosses that carried the E(sev) mutation and the P[ry, se+] insert were examined for clones and sectioned as described above. Generation of Temperature-Sensitive SW Alleles Site-directed mutagenesis was used to generate the amino acid substitutions shown in Figure 1 (Sambrook et al., 1989). The 17.5 kb DNA fragment containing the mutated sev allele was placed in a modified Carnegie 20 P-transformation vector and injected into set@, Iy embryos as previously described (Karess and Rubin, 1984). The 17.5 kb fragment consists of a 15 kb EcoRl fragment that has previously been

shown to be providing full sev function and an additional 2.5 kb 5 of the sev gene (Hafen et al., 1987). The transformants were assayed by sectioning and by the reduced cornea1 pseudopupil method. The TM3, Sb, ry, P[ry, sever] chromosome was generated by transposition of the P[ry, sep] element. lsolatlon and Transformation of the Rash and SOS Regions The Rasl region was cloned from a library of Drosophila genomic DNA cloned into the SFixll (Stratagene, Inc.) vector, using a 700 bp NcolEcoRl fragment as a probe (Neuman-Silberberg et al., 1984). A 12 kb Xhol-Notl fragment was cloned into the pW8 transformation vector (Klemenz et al., 1987) and injected into w~, sevd2embryos as previouslydescribed (Karess and Rubin, 1984). The Xhol site liesapproximately4.5 kb 5of the initiating methionine of the Rasl gene, The Notl site is derived from the vector polylinker sequences and corresponds to a position approximately 5 kb 3of the Rasl termination codon. The Rasl alleles from the .E(sev)3C mutant chromosomes were cloned as 5.7 kb EcoRl fragments into the Zapll (Stratagene, Inc.) vector. The mutant alleles were then cloned into pBluescript(+) or (-) for sequencing. The mutant alleles could be distinguished from the balancerderived alleles on the basis of a sequence polymorphism in a noncoding region. The SOS region was cloned by chromosome walking using a cosmid library prepared by J. Tamkun. The starting point was the Hobo[ry+] element inserted in polytene band 34Dl (Blackman et al., 1989). A library of genomic DNA was constructed in XFixll and screened with a probe from the rosy gene. The walk was oriented and monitored by hybridization of cosmid DNAs to wild-type and SOS deficiency chromosomes. Genomic blot analysis of deficiency chromosomes was performed by standard techniques using probes from the SOS region (Sambrook et al., 1989). The SOS alleles from mutant chromosomes were cloned into SFixll as an approximately 18 kb Bglll fragment that includes the entire gene. The alleles from the mutant chromosomes were distinguished from the balancer-derived alleles on the basis of polymorphic Hindlll sites. The mutant alleleswere subcloned into plasmid vectors for sequencing. The SOS transformation construct was made by inserting the IO kb Bglll-Xbal fragment into BarnHI-Xbalcut pW8. The resulting plasmid was injected into w~, seti2 embryos as previously described (Karess and Rubin, 1984). The 10.3 kb Xbal fragment that contains the entire SOS gene was used to screen an eye imaginal disc cDNA in I@10 constructed by Dr. A. Cowman. A single cDNA lsosT was then subcloned into pBlueScript(-) for sequencing. Transcript Analysis Polyadenylated RNA was prepared from imaginal discs and from embryos, electrophoresed through a 1% agarose-formaldehyde gel, and blotted onto nitrocellulose by standard methods (Sambrook et al., 1989). The blot was probed with the fragment of the sosT cDNA shown in Figure 9. Nuclease Sl analysis and primer extension were performed exactly as described (Sambrook et al., 1989). The probes used for Sl analysis were uniformly labeled single-stranded probes corresponding to the reverse complements of nucleotides 0 to 610 and 0 to 731. The primers used correspond to nucleotides 581 to 610 and 706 to 731. DNA Sequencing All DNA sequences were determined using the dideoxy chain termination method (Sanger et al., 1977) using Sequenase (U.S. Biochemical). Templates for the determination of the 6601 bp genomic SOS sequence were generated by sonication of plasmid DNA and insertion of the sonicated fragments into the vector MlSmplO. The entire sequence was determined on both strands. The sequences of SOS cDNAs were determined on one strand using templates generated from plasmids containing EcoRl fragments inserted into the pBlueScript(-) vector using an Erase-a-Base kit (Promega). Mutant alleles of SOS were sequenced on one strand using oligonucleotide primers spaced approximately every 200 bp. The exons of the mutant Rasl alleles were sequenced on one strand using oligonucleotide primers. Acknowledgments We wish to thank M. Cyert, W. Ferguson, J. Kaplan, B. Meyers, H.

Signal Transduction 715

by the Sevenless

PTK

Varmus, and members of our laboratory for their helpful suggestions regarding the manuscript. In addition, we wish to acknowledge the important contributions of J. Kaplan and I. Hariharan to the design of these experiments. We also thank M. Ashburner and J. Roote for informationaboutthe34Dregion and for Drosophilastrains, J. Tamkun for providing the Drosophila genomic cosmid library, D. Tyler for DNA sequencing, D. Hackett for microinjection of embryos, T. Cutforth for deficiency mapping of the E(sev)3C locus, Jorge Santiago-Blay for scanning electron microscopy, and I. Hariharan for an RNA blot. D. D. L. B. was a C. J. Martin Fellow. M. A. S. is a Lucille P. Markey Scholar in Biomedical Sciences. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 USC Section 1734 solely to indicate this fact. Received August 21, 1991; revised September 12, 1991.

required 561-577.

for specification

of R7 cell fate in the Drosophila

eye. Cell 63,

Casanova, J., and Struhl, G. (1989). a receptor tyrosine kinase, specifies ila. Genes Dev. 3, 2025-2038.

Localized surface activity of torso, terminal body pattern in Drosoph-

Chant, J., Corrado, K., Pringle, J. R., and Herskowitz. I. (1991). Yeast BUD5, encoding a putative GDP-GTP exchange factor, is necessary for bud site selection and interacts with bud formation gene SEW. Cell 65, 1213-1224. Cross, M., and Dexter, T. M. (1991). Growth factors in development, transformation, and tumorigenesis. Cell 64, 271-280. Crechet, J.-B., Poullet, P., Mistou, M.-Y., Parmeggiani, A., Camonis, J., Boy-Marcotte, E., Damak, F., and Jacquet, M. (1990). Enhancement of the GDP-GTP exchange of RAS proteins by the carboxy-terminal domain of SCD25. Science 248,866-868. Damak, F., Boy-Marcotte, E., Le-Roscouet, D., Guilbaud, R., and Jacquet, M. (1991). SDC25, a CDCZdlike gene which contains a RASactivating domain and is a dispensable gene of S. cerevisiae. Mol. Cell. Biol. 77, 202-212. Feig, L. A., and Cooper, G. M. (1988). Inhibition ation by a mutant ras protein with preferential Cell. Biol. 8, 3235-3243. Fincham, sensitive Sarcoma of NIH 3T3 cell proliferaffinity for GDP. Mol.

Ambrosio, L., Mahowald, A. P., and Perrimon, N. (1989). Requirement of the Drosophila raf homologue for torso function. Nature 342, 288291. Aroian, Ft. V., Koga, M., Mendel, J. E., Ohshima, Y., and Sternberg, P. W. (1990). The let-23 gene necessary for Caenorhabditis e/egans vulva1 induction encodes a tyrosine kinase of the EGF receptorsubfamily. Nature 348, 893-699. Baker, N. E., and Rubin, G. M. (1989). Effect on eye development of dominant mutations in Drosophila homologue of the EGF receptor. Nature 340, 150-l 53. Basler, K., and Hafen, the sevenless protein Cell 54, 299-311. E. (1988). Control of photoreceptor cell fate by requires a functional tyrosine kinase domain. Caenorhabditis of vulva1 inducCell 64,235-

V. J., and Wyke, J. A. (1988). Localization of temperaturetransformation mutations and back mutations in the Rous Virus src gene. J. Virol. 58, 894-699. N., and Kirschfeld, K. (1971). Ioeil compose de Drosophila. Les phenomenes de pseuKybernetik. 9, 159-162.

Franceschini, dopille dans

Gibbs, J. B., Schaber, M. D., Allard, W. J., Sigal, I. S., and Scolnick, E. M. (1988). Purification of ras GTPase activating protein from bovine brain. Proc. Natl. Acad. Sci. USA 85, 5026-5030. Hafen, E., Basler, K., Edstroem, J. E., and Rubin, G. M. (1987). Sevenless, acell-specific homeoticgeneof Drosophila, encodesaputative transmembrane receptor with a tyrosine kinase domain. Science 236, 55-63. Han, M., and Sternberg, fates during C. elegans 63, 921-931. P. W. (1990). let-60, a gene that specifies vulva1 induction, encodes a ras protein. cell Cell

Beitel, G. J., Clark, S. G., and Horvitz, H. R. (1990). elegans fas gene let-60 acts as a switch in the pathway tion. Nature 348, 503-509. Bishop, 248. J. M. (1991). Molecular themes in oncogenesis.

Blackman, R. K., Koehler, M. M. D., Grimaila, R., and Gelbart, W. M. (1989). Identification of a fully-functional hobo transposable element and its use for germ-line transformation of Drosophila. EMBO J. 8, 211-217. Bourne, H. R.. Sanders, superfamily: a conserved 125-132. Bourne, H. R., Sanders, superfamily: conserved 349, 117-127. D.A., and McCormick, F. (1990). switch for diverse cell functions. TheGTPase Nature 348,

Hart, A. C., Kramer, H., Van Vactor, D. L., Jr., Paidhungat, M., and Zipursky, S. L. (1990). Induction of cell fate in the Drosophila retina: the bride of sevenless protein is predicted to contain a large extracellular domain and seven transmembrane segments. Genes Dev. 4, 18351847. Hazelrigg, T., Levis, R., and Rubin, G. M. (1984). Transformation of white locus DNA in drosophila: dosage compensation, zesre interaction, and position effects. Cell 36, 489-481. Henkemeyer, M., West, S. R., Gertler, F. B., and Hoffmann, F. M. (1990). A novel tyrosine kinase-independent function of Drosophila abl correlates with proper subcellular localization. Cell 63, 949-960. Hughes, activators Hunter, D. A., Fukui, Y., and Yamamoto, M. (1990). Homologous of ras in fission and budding yeast. Nature 344, 355-357. T. (1991). Cooperation between oncogenes. Cell 84,249-270. to

D.A., and McCormick, F. (1991). TheGTPase structure and molecular mechanism. Nature

Bowtell, D. D., Simon, M. A., and Rubin, G. M. (1988). Nucleotide sequence and structure of the sevenless gene of Drosophila melano. gaster. Genes. Dev. 2, 620-834. Brock, H. W. (1987). Sequence and genomic structure of ras homologues Dmras85D and Dmras64B of Drosophila melanogaster. Gene 51, 129-137. Broek, D., Toda, T., Michaeli, T., Levin, L., Birchmeier, C., Zoller, M., Powers, S.. and Wigler, M. (1987). The S. cerevisiae CDC25 gene product regulates the RASladenylate pathway. Cell 48, 789-799. Camonis, J. H., Kalekine, M., Gondre, B., Garreau, H., Boy-Marcotts, E., and Jacquet, M. (1966). Characterization, cloning and sequence analysis of the CDC25 gene which controls the cyclic AMP level of S. cerevisiae. EMBO J. 5, 375-380. Cantley, L. C., Auger, K. R., Carpenter, C., Duckworth. B., Graziani, A., Kapeller, R., and Soltoff, S. (1991)Oncogenesandsignaltransduc lion. Cell 64, 281-302. Capon, D. J., Chen, E. Y., Levinson, A. D., Seeburg, P. H., and Goeddel, D. V. (1983). Complete nucleotide sequences of the T24 human bladder carcinoma oncogene and its normal homologue. Nature 392, 33-37. Carthew. R. W., and Rubin. G. M. (1990). seven 1 absentia, a gene

Jones, S., Vignais, M.-L., and Broach, J. R. (1991). TheCDC25protein of S. cerevisiae promotes exchange of guanine nucleotides bound ras. Mol. Cell. Biol. 17, 2641-2846. Karess, element R. E., and Rubin, G. M. (1984). Analysis of P transposable functions in Drosophila. Cell 38, 135-146.

Kemphues, K. J.. Raff, E. C., Raff, R. A., and Kaufman, T. C. (1980). Mutation in a testis-specific f3-tubulin in Drosophila: analysis of its effects on meiosis and map location of the gene. Cell 21, 445-451. Kimmel, 8. E., Heberlein, U., and Rubin, G. M. (1990). The homeo domain protein rough is expressed in a subset of cells in the developing Drosophila eye where it can specify photoreceptor cell subtype. Genes Dev. 4, 712-727. Klemenz, R., Weber, U., and Gehring, W. J. (1987). The white gene as a marker in a new P-element vector for gene transfer in Drosophila. Nucl. Acids Res. 75, 3947-3959. Kramer, H., Cagan, R. L., andZipursky,S. L. (1991). lnteractionofbride

Cdl 716

ofsevenless membrane-bound ligandand receptor. Nature 352, 207-212.

thesevenless

tyrosin-kinase

GenBenk

Accession number

Number for the sequence reported in this article is

Levis, Ft., Hazelrigg, T., and Rubin, G. M. (1985). Separable c&acting control elements for expression of the white gene of Drosophila. EMBO J. 4,3489-3499. Lewis, E. B., and Bather, sulphonate to Drosophila F. (1968). Method of feeding ethyl methanemales. Dros. Inf. Serv. 43, 193-194.

The accession M77501.

Lindsley, D. L., Sandier, L., Baker, B. S., Carpenter, A. T. C., Denell, R. E., Hall, J. C., Jacobs, P. A., Miklos, G. L. G., Davis, 8. K., Gethmann, R. C., Hardy, R. H., Heasler, A., Miller, S. M., Nozawa, H., Parry, D. M., and Gould-Somero, M. (1972). Segmental aneuploidy and the gross genetic structure of the Drosophila genome. Genetics 77, 157184. Livneh, E., Glazer, L., Segal, D., Schlessinger, J., and Shilo, 8. Z. (1985). The Drosophila EGF receptor gene homolog: conservation of both hormone binding and kinase domains. Cell 40, 599-607. Muller, H. J., League, B. B., and Offermann, C. A. (1931). Effects of dosage changes of sex-linked genes, and the compensatory effects of other gene differences between male and female. Anat. Rec. 51 (SUPPI.), 1 IO. Neuman-Silberberg, F. S., Schejter, E., Hoffmann, F. M., and Shilo, B. 2. (1984). The Drosophila ras oncogenes: structure and nucleotide sequence. Cell 37, 1027-1033. Powers, S., Gonzales, E., Christensen, T.. Cubert, J., and Broek, D. (1991). Functional cloning of BUD!& a CDC2brelated gene from S. cerevisiae that can suppress a dominant-negative RAS2 mutant. Cell 65, 1225-1231. Price, J. V., Clifford, R. J., and Schiipbach, T. (1989). The maternal ventralizing locus torpedo is allelic to faint lilt/e ball, an embryonic lethal, and encodes the Drosophila EGF receptor homolog. Cell 56, 1085-l 092. Reinke, R., and Zipursky, S. L. (1988). Cell-cell interaction in the drosophila retina: the bride ofsevenless gene is required in photoreceptor cell R8 for R7 cell development. Cell 55, 321-330. Robinson, L. C., Gibbs, J. B., Marshall, M. S., Sigal, I. S., and Tatchell, K. (1987). CDC25: a component of the RAS-adenylate cyclase pathway in Saccharomyces cefevisiae. Science 235, 1218-1221. Rogge, R. D., Karlovich, C. A., and Banerjee, U. (1991). Genetic dissection of a neurodevelopmental pathway: Son of sevenless functions downstream of the sevenless and EGF receptor tyrosine kinases. Cell 64, 39-48. Sambrook, J., Fritsch, E. F., and Maniatis, T. (1989). Molecular Cloning: A Laboratory Manual, Second Edition (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press). Sanger, F., Nicklen, S., and Coulson, A. (1977). DNAsequencing with chain terminating inhibitors. Proc. Natl. Acad. Sci. USA 74, 54635467. Schejter, E. D., and Shilo, B.-Z. (1989). The Drosophila EGF receptor homolog (DER) gene is allelic to faint little ball, a locus essential for embryonic development. Cell 56, 1093-l 104. Smith, M. R., DeGudicibus, S. J., and Stacey, ment for c-ras proteins during viral oncogene 320,540~543. D. W. (1988). transformation. RequireNature

Sprenger, F., Stevens, L. M., and Niisslein-Volhard, Drosophila gene torso encodes a putative receptor Nature 338, 478-483.

C. (1989). The tyrosine kinase.

Tomlinson, A., and Ready, D. F. (1988). Sevenless: a cell specific homeotic mutation of the Drosophila eye. Science 237, 400-402. Tomlinson, A., and Ready, D. F. (1967). ommatidium. Dev. Biol. 123, 264-275. Trahey, M.. and McCormick, lates normal N-ras21 GTPase, Science 238,542-545. Cell fate in the Drosophila

F. (1967). A cytoplasmic protein stimubut does not affect oncogenic mutants.

Woodruff, R. C., and Ashburner, M. (1979). The genetics of a small autosomal region of Drosophila melanogaster containing the structural gene for alcohol dehydrogenase. II. Lethal mutations in the region. Genetics 92, 133-149.