An unusual recessive allele of the Drosophila groucho gene, which encodes a transducinlike protein, affects the fates of specific cells in the eye disc groucho is one of several transcription units in the Enhancer of split complex. Most groucho mutations are zygotic lethal due to the proliferation of embryonic neural cells at the expense of epidermal cells. In contrast, flies homozygous for the mutant allele described here, groBFP2, are viable but have abnormal eyes. The Drosophila compound eye is composed of several hundred identical facets, or ommatidia, each of which contains eight photoreceptor cells, R1-R8. In groBFP2 mutant retinas, most of the facets contain eight normally determined photoreceptor cells and one or two additional R-cells of the R3/4 subtype. The extra photoreceptors appear to arise from the mystery cells, which are part of the precluster that initiates the ommatidium, but do not normally become neurons. groBFP2 behaves as a partial loss-of-function mutant. Analysis of ommatidia mosaic for wild-type and groBFP2 mutant cells suggests that the focus of action of the groBFP2 mutation is outside of the photoreceptor cells. These results imply that one function of groucho is in a pathway whereby neuralization of the mystery cells is inhibited by other non-neural cells in the eye disc. In addition, determination of R3/4 photoreceptors usually requires contact with R2 and R5. Specification of the mystery cells as ectopic R3/4 subtype photoreceptors in groBFP2 mutant eye discs implies that induction by R2 or R5 is not absolutely necessary for R3/4 cell determination.

The Drosophila compound eye is composed of about eight hundred identical facets, or ommatidia, arranged in a precise hexagonal lattice. Ommatidia assemble stepwise within a monolayer of unpattemed epithelial cells in the eye imaginai disc in the wake of a visible depression called the morphogenetic furrow that moves across the disc from the posterior to the anterior (Ready et al., 1976; Tomlinson, 1985; Tomlinson and Ready, 1987a; Cagan and Ready, 1989a). The eight photoreceptor cells are recruited in the sequence R8, R2/5, R3/4, Rl/6 and R7, followed by four cone cells. The pigment cells and bristles are assembled later in the pupal disc. As the morphogenetic furrow advances at the rate of one row of ommatidia every two hours (Campos-Ortega and Hofbauer, 1977), facets at progressive stages of assembly are present behind the furrow in a single eye disc. The cells in a facet are not related by lineage (Ready et al., 1976; Lawrence and Green, 1979; Wolff and Ready, 1991a). Rather, ommatidial assembly is guided by a series of specific cell inductions (reviewed in Tomlinson, 1988; Ready, 1989; Zipursky, 1989; Banerjee and Zipursky, 1990; Moses, 1991; Rubin, 1991). The initial events in the assembly process are less well understood. As the initial stages of eye development involve choosing neurons from a pool of epithelial cells, many genes that mediate the decision between neural and ectodermal cell fates elsewhere in the fly also appear to function during eye development (Dietrich and Campos-Ortega, 1984; Cagan and Ready, 1989b; Baker et al., 1990; Mlodzik et al., 1990a).

The Enhancer of split (E(spff) gene complex of Drosophila is one of six loci referred to as “neurogenic” genes (recently reviewed in Campos-Ortega, 1991), which were first identified by their role in embryonic neurogenesis (Poulson, 1937; Lehmann et al., 1981, 1983). The Drosophila central nervous system arises from neuroectoderm cells which must choose between an ectodermal or neural fate. In embryos mutant for any one of the neurogenic genes, most or all of the neuroectoderm cells become neural and the embryo dies. Many experiments led to the conclusion that committed neuroblasts inhibit surrounding cells from also acquiring a neural fate by a cell-contact-mediated process (Taghert et al., 1984; Doe and Goodman, 1985; Technau and Campos-Ortega, 1986, 1987; Technau et al., 1988). The structures of the proteins encoded by the neurogenic genes, particularly Notch and Delta, are consistent with a role in cell commumcation (Wharton et al., 1985; Kidd et al., 1986; Vassin et al., 1987; Kopczynski et al., 1988). Most of the neurogenic genes also play a role in cell-contact-mediated epidermal/ neural commitment decisions in the peripheral nervous system, including the eye (Dietrich and CamposOrtega, 1984; Cagan and Ready, 1989b, Heitzler and Simpson, 1991; for reviews see Ghysen and DamblyChaudiere, 1989; Simpson, 1990). In addition, in the eye, Notch also mediates cell interactions involved in other types of cell commitment choices (Cagan and Ready, 1989b).

The E(spl) locus, first identified by a dominant mutation, E(spl)D, consists of several closely linked transcription units with complex functional interactions. E(spl)D enhances the roughened eye phenotype of a unique recessive allele of the Notch gene called split (Welshons, 1956). In order to ascertain the function of the normal E(spl) gene, revertants of the dominant eye phenotype of the E(spl)D mutation were generated (Lehmann et al., 1983). The revertants, all of which are deficiencies that delete several transcripts, are embryonic lethal and have a typical neurogenic phenotype when homozygous. At least four of the transcripts within the deficiencies participate in neurogenesis and they have been divided into two functional units: the m5, m7, m8 group and m9/10 (Delidakis et al., 1991). The m5, tn7 and m8 transcripts encode proteins containing a helix-loop-helix (HLH) motif (Klambt et al., 1989) characteristic of some transcription factors (Murre et al., 1989). The E(spl)D mutation results in an altered form of the gene product of m8 (Klambt et al., 1989). The HLH proteins are at least partly functionally redundant as genetic screens for lethal mutations in trans to E(spl) deficiencies that delete the HLH protein transcripts and m9/10 have identified only mutations in the m9/10 transcription unit (Preiss et al., 1988). The m9/10 transcription unit was originally identified by a viable mutant called groucho which has specific head bristle duplications (Lindsley and Grell, 1968; Knust et al., 1987; Ziemer et al., 1988). For simplicity, the m9/10 transcription unit will be referred to as the groucho gene, groucho encodes a nuclear protein (Delidakis et al., 1991) with a repeated motif present in a G protein subunit, /J-transducin (Hartley et al., 1988), in the yeast cell cycle regulatory protein CDC4 (Yochem and Byers, 1987) and in PRP4, a spliceosome component (Dalrymple et al., 1989; Petersen-Bjprn et al., 1989). The functional relationship between groucho and the HLH proteins is not well understood.

Here we describe an unusual viable recessive allele of groucho, groBFP2, which has its principal effect on eye development. In homozygous groBFP2 adult eyes, most facets contain one or two extra photoreceptor cells. These cells are likely to originate from the mystery cells, which are part of an undifferentiated ommatidial precluster posterior to the morphogenetic furrow (Tomlinson and Ready, 1987a; Wolff and Ready, 19916). The mystery cells are located between the cells that will become R3 and R4 and are normally excluded from the precluster, but in groBFP2 discs the mystery cells become additional photoreceptors of the R3/4 subtype. Examination of the phenotypes of groBFP2 in trans to lethal groucho mutations and observation of clones of the lethal mutants in the eye suggests that groBFP2 is a unique partial loss-of-function mutant. Analysis of individual ommatidia mosaic for wild-type and groBFP2 R-cells suggests that the groBFP2 mutation acts outside of the R-cells. These results imply that cell communication, requiring groucho in non-neural cells outside of the developing facets, is necessary to exclude the mystery cells from the ommatidial precluster. In addition, determination of R3/4 subtype cells normally requires inductive signals from the neighboring R2 and R5 cells in the precluster (Tomlinson et al., 1988). Thus, the specification of the mystery cells as ectopic R3/4 subtype photoreceptors in groBFP2 eye discs implies that R3/4 cells can be recruited by an alternative route, not requiring contact with the R2/5 pair.

Drosophila genetics

Fly lines

The E(spl) point mutants and deficiencies, and the ry+E8 transformant were gifts of A Preiss, C. Delidakis and S. Artavanis-Tsakonas, and are described in Preiss et al. (1988) and Delidakis et al. (1991). The boss deficiencies, gifts of A. Hart and S. L. Zipursky, are described in Hart et al. (1990). The enhancer trap lines that express β-gal actosidase in subsets of photoreceptor cell nuclei were gifts of members of the Rubin laboratory. BG9408 and BGP820 are marked with rosy+, and are on located in polytene chromosome bands 10A and 34A, respectively (M. Mlodzik and G.M.R., unpublished data). BGA2-6, N30, 032, AE127 and X81 are marked with white+. BGA2-6 is inserted into the scabrous locus (Mlodzik et al., 1990; Baker et al., 1990), AE127 is inserted into the seven-up gene (M. Mlodzik, J.S. Heilig and G.M.R., unpublished data) and X81 is in the rhomboid gene (M. Freeman, B.E Kimmel and G.M.R., unpublished data). N30 and 032 are inserted into polytene bands 34A and 65D, respectively (M. Freeman and G.M.R., unpublished data). Other mutant markers are described in Lindsley and Grell (1968). Flies were kept on standard food at 25°C.

Isolation of the groBFP2 allele

The groBFP2 allele was isolated during an extensive screen for viable recessive mutations on the third chromosome with abnormal eye morphology (J. A F.-V., R. W. Carthew and G M R, unpublished data). More than 20,000 lines of single mutagemzed st chromosomes were generated as follows. EMS-mutagenized (Lewis and Bächer, 1968) bw; st males, isogenic for st, were crossed to bw; TM3/TM6B virgins. Male progeny (bw, *st/TM3 or TM6B) were singly mated to bw, TM3/TM6B virgins. The progeny of this cross were intermated, and the resulting hnes were screened for white-eyed (bw; *st) flies, indicating viable mutagemzed third chromosomes. The white-eyed flies in over 8000 viable lines were examined with a dissecting microscope for eye roughness, and then for defects in the reduced corneal pseudopupil, a red trapezoidal image visible in white-eyed flies with incident polarized light (Francheschmi and Kirschfeld, 1971; see Banerjee et al, 1987) The groHFP2 mutant was recognized on the basis of an abnormal reduced corneal pseudopupil caused by retinal irregularities.

Meiotic mapping of groBFP2

The groBFP2 allele was first approximately mapped by crossing bw; st groBFP2/ th st eu sr e ca virgins with bw; th st eu sr e ca males. Many male progeny of all genotypic classes were individually mated with bw, st groBFP2/TM6B virgins to determine whether or not the recombinant chromosome contained groBFP2. By this process, groBFP2 was placed distal to ebony (e) and various marked groBFP2 chromosomes were obtained. The groBFP2 mutation was then positioned between e and Serrate (Ser) by crossing e groBFP2/ Ser virgins to e groBFP2 males and scoring all three markers in many progeny Finally, groBFP2 was mapped with respect to five P element transformant insertions marked with white+(P[w+]) in polytene bands 94D, 94E, 95F, 96C, and 96F (Rubin laboratory stock collection), by crossing w, e groBFP2/P[w+] virgins to w; e groBFP2 males and scoring many progeny for the P[w+], e and groBFP2 phenotypes The groBFP2 mutation mapped telomeric to all of the P[w+] insertions except for the one at 96F. No recombinants between groBFP2 and P[w+]96F were obtained as compared with 185 recombinants between e and groBFP2.

Other fly crosses

Combination of groBFP2 with deficiency chromosomes, other gro alleles, the ry+E8 transformant and enhancer trap lines were carried out using standard genetic crosses.

Generation of mosaic eyes

groBFP2 clones

Clones of groBFP2/groBFP2 cells marked by the absence of the white gene product (no pigment granules associated with photoreceptor cells nor pigment cells) were generated by crossing w1118; st groBFP2/TM6B males with w11, P[w+]85D or w11, P[w+]90E virgins and X-irradiating (1000 rads) their progeny as first instar larvae. Thus, larvae of the genotype w”"; p[w+OgroBFP2 have clones of mutant cells of the genotype w28; groBFP2. P[w+]85D and P[w+]90E are P element transformant lines marked by expression of the white gene with insertions in polytene bands 85D and 90E, respectively (Rubin laboratory stock collection) w clones of cells in the eye were observed in ∼1 in 30 flies.

clones of lethal groucho alleles

Clones of the groucho alleles E(spl)E2S, E(spl)E48, E(spl)E73, E(spl)E7fE(spl)E77, E(spl)EI07, l(gro)X115 and the deficiency E(spl)BXi2 were generated as described above using p[w+]90E. All of the lethal groucho alleles were on e tx chromosomes, except l(gro)xi25 and E(spl)BX22 which were on ry506tx chromosomes E(spl)EI07 and E(spl)clones were obtained at frequencies similar to the groBFp2 clones. Clones of all of the other lethal alleles were usually small and were obtained at extremely low frequencies (∼1/200 flies)

Sections of eyes

The heads of various mutants and of eyes containing clones were fixed, embedded in plastic and sectioned as previously described (Tomlinson and Ready, 1987b)

Antibody staining eye discs

Third instar larval eye discs were dissected, fixed and stained with mAb22C10 (a gift of S. Benzer) as previously described (Tomlinson and Ready, 1987a). The enhancer trap lines expressing β-gal actosidase m cell nuclei in the eye disc were dissected, fixed and stained with a mouse monoclonal antibody raised against β-galactosidase as described previously (Tomlinson and Ready, 1987a; Heberlein et al., 1991).

The groBFP2 mutant phenotype

groBFP2 is a recessive mutation that was identified in a screen of the third chromosome for mutants with defects in eye morphology (J.A.F.-V., R.W. Carthew, and G.M.R., unpublished data - see Materials and methods). The external appearance of the eyes of homozygous groBFP2 flies is almost normal; they are slightly bulged and have some bristle spacing defects (data not shown). The reduced corneal pseudopupils (see Materials and methods) are blurred indicating some irregularity in the retina (data not shown). In addition to the eye defects, groBFp2 mutant males and females are only marginally fertile, and their wings are slightly broader and slightly held out.

The eyes of groBFP2 flies were examined in tangential sections. As shown in Fig. 1A and G, the retina is a hexagonal lattice of identical facets, or ommatidia. Each ommatidium in a wild-type eye has eight photoreceptor cells (R-cells) distinguished by their unique positions in a trapezoid. There are six outer photoreceptor cells, R1-R6, with large rhabdomeres (fightgathering devices), and two inner R-cells, R7 and R8, with small rhabdomeres. The trapezoids are all oriented in the same direction and are symmetrical with respect to a central equator. In the groBFP2 retina (Fig. IB and G), approximately two-thirds of the facets have one or two additional outer photoreceptor cells. The remaining facets have the normal number of photoreceptors. In addition, the orientations of the facets are irregular.

Fig. 1.

Adult eye phenotypes of groBFP2 mutant combmations. Shown in A-F are tangential sections of adult eyes (Materials and methods) The sections are apical so that R7 is apparent, rather than R8 (see (G)). The bar in F is 30 μm and applies to all panels. See the text and Table 1 for descriptions of the genotypes and phenotypes. (A) A wild-type eye in the region of the equator. (B) A groBFP2 homozygous eye, in which there are three type of facets, about two-thirds of the facets have 1, 2 or 3 extra outer photoreceptors, approximately one-sixth of the facets are wild-type and one-sixth have the normal number of photoreceptor cells, but R3 and R4 are symmetrical instead of forming a point of the trapezoid. The asymmetry of R3 and R4 is normally initiated in the most mature facets just before pupation (Tomlinson, 1985). In addition, the orientations of the facets are irregular. This particular eye is so that the pigment cells contain no dark staining pigment granules nor are there pigment granules near each rhabdomere. (C) A ry+E8, groBFP2 eye The arrow indicates one mutant in a field of wild-type facets. In D, E(spl)EI07/groBFP2, the arrows also indicate mutant facets. The genotypes in E and F are E(spl)B77/groBFP2 and E(spl)BX22/groBFP2, respectively (G) Wild-type and groBFP2ommatidia are depicted schematically, the circles denoting rhabdomeres. Note that R8 is located beneath R7. The groDFP2 mutant facet is shown with one ectopic R-cell (shaded) for simplicity. Because the identities of the R-cells are recognized only on the basis of the R-cell positions m a wild-type facet, it is impossible only by observing the groBFp2 mutant eyes shown m B to assign R-cell identities as shown in G. The R-cell labels shown are based on the antibody staining experiments in Figs 4, 5 and 6 Those experiments demonstrate that the ectopic cells are R3/4 subtype cells located adjacent to the normal R3 and R4, and that the other R-cells are normally determined in groBFP2 eye discs. Because groBFP2 facets have three or four cells in the R3/4 position, the labeling of individual R-cells of this group must be somewhat arbitrary. The cells labeled R3 and R4 m the groBFP2 facet in G were so assigned based on their positions next to R2 and R5, and the middle cell (shaded) labeled the ectopic one, primarily because that is where the mystery cells reside m the preclusters (see Fig. 4F).

Fig. 1.

Adult eye phenotypes of groBFP2 mutant combmations. Shown in A-F are tangential sections of adult eyes (Materials and methods) The sections are apical so that R7 is apparent, rather than R8 (see (G)). The bar in F is 30 μm and applies to all panels. See the text and Table 1 for descriptions of the genotypes and phenotypes. (A) A wild-type eye in the region of the equator. (B) A groBFP2 homozygous eye, in which there are three type of facets, about two-thirds of the facets have 1, 2 or 3 extra outer photoreceptors, approximately one-sixth of the facets are wild-type and one-sixth have the normal number of photoreceptor cells, but R3 and R4 are symmetrical instead of forming a point of the trapezoid. The asymmetry of R3 and R4 is normally initiated in the most mature facets just before pupation (Tomlinson, 1985). In addition, the orientations of the facets are irregular. This particular eye is so that the pigment cells contain no dark staining pigment granules nor are there pigment granules near each rhabdomere. (C) A ry+E8, groBFP2 eye The arrow indicates one mutant in a field of wild-type facets. In D, E(spl)EI07/groBFP2, the arrows also indicate mutant facets. The genotypes in E and F are E(spl)B77/groBFP2 and E(spl)BX22/groBFP2, respectively (G) Wild-type and groBFP2ommatidia are depicted schematically, the circles denoting rhabdomeres. Note that R8 is located beneath R7. The groDFP2 mutant facet is shown with one ectopic R-cell (shaded) for simplicity. Because the identities of the R-cells are recognized only on the basis of the R-cell positions m a wild-type facet, it is impossible only by observing the groBFp2 mutant eyes shown m B to assign R-cell identities as shown in G. The R-cell labels shown are based on the antibody staining experiments in Figs 4, 5 and 6 Those experiments demonstrate that the ectopic cells are R3/4 subtype cells located adjacent to the normal R3 and R4, and that the other R-cells are normally determined in groBFP2 eye discs. Because groBFP2 facets have three or four cells in the R3/4 position, the labeling of individual R-cells of this group must be somewhat arbitrary. The cells labeled R3 and R4 m the groBFP2 facet in G were so assigned based on their positions next to R2 and R5, and the middle cell (shaded) labeled the ectopic one, primarily because that is where the mystery cells reside m the preclusters (see Fig. 4F).

groBFP2 is an allele of groucho

By a series of meiotic mapping experiments with several mutant markers, groBFP2 was positioned very close to a white+ transposon inserted in polytene chromosome band 96F (see Materials and methods). Several chromosomes with deficiencies in the 96F region were crossed to groBFP2 (Fig. 2). Six of these deficiency chromosomes uncover the groBFP2 eye phenotype, and two do not. Thus, groBFP2 is located within a chromosomal region including the m8 and m9/10 (gro) transcription units of the E(spl) complex.

Fig. 2.

The groBFP2 mutant phenotype is uncovered by chromosomes deficient for E(spl) transcription units m8 and m9/10. At the top is shown an approximately 25 kilobase (kb) portion of the E(spl) gene complex (Preiss et al., 1988, Knust et al., 1987). The m5, m7 and m8 transcripts are described in the text. The m9[10 transcnpts, which correspond to the groucho gene, have different 3’ ends but encode the same protein (Hartley et al., 1988). The arrows indicate the direction of transcription. The striped bar indicates the genomic DNA fragment cloned into the P element within the ry+E8 transformant fine (Preiss et al., 1988). ry+E8 complements groucho point mutations, including groBFP2. Shaded bars indicate the regions deleted in the deficiency chromosomes indicated at right. Arrows at the ends of the bars indicate that the deletions extend beyond the 25 kb DNA segment depicted. The regions deficient within the chromosomes have been mapped cytologically. l(gro)xl and l(gro)2172 delete polytene bands 96F5/7-96F12/14 and 96F5/7-97B1, respectively (Preiss et al, 1988). boss2 and boss3 delete 96E6-97B4/5 and 96F8/11-97B, respectively (Flart et al., 1990). E(spl)BX22 is cytologically normal, but has been shown by molecular analysis to contain a deletion of 14 kb (as depicted) and also an inversion of the 14 kb just upstream (to the left) (Preiss et al., 1988; Shepard et al., 1989). boss709 contains a deletion of 96F10/11-97D1/2 (Delidakis et al., 1991). boss15 and boss16 contain deletions of 96F3/5-11/12 and 96F5/7-12/13, respectively (Hart et al., 1990).

Fig. 2.

The groBFP2 mutant phenotype is uncovered by chromosomes deficient for E(spl) transcription units m8 and m9/10. At the top is shown an approximately 25 kilobase (kb) portion of the E(spl) gene complex (Preiss et al., 1988, Knust et al., 1987). The m5, m7 and m8 transcripts are described in the text. The m9[10 transcnpts, which correspond to the groucho gene, have different 3’ ends but encode the same protein (Hartley et al., 1988). The arrows indicate the direction of transcription. The striped bar indicates the genomic DNA fragment cloned into the P element within the ry+E8 transformant fine (Preiss et al., 1988). ry+E8 complements groucho point mutations, including groBFP2. Shaded bars indicate the regions deleted in the deficiency chromosomes indicated at right. Arrows at the ends of the bars indicate that the deletions extend beyond the 25 kb DNA segment depicted. The regions deficient within the chromosomes have been mapped cytologically. l(gro)xl and l(gro)2172 delete polytene bands 96F5/7-96F12/14 and 96F5/7-97B1, respectively (Preiss et al, 1988). boss2 and boss3 delete 96E6-97B4/5 and 96F8/11-97B, respectively (Flart et al., 1990). E(spl)BX22 is cytologically normal, but has been shown by molecular analysis to contain a deletion of 14 kb (as depicted) and also an inversion of the 14 kb just upstream (to the left) (Preiss et al., 1988; Shepard et al., 1989). boss709 contains a deletion of 96F10/11-97D1/2 (Delidakis et al., 1991). boss15 and boss16 contain deletions of 96F3/5-11/12 and 96F5/7-12/13, respectively (Hart et al., 1990).

To test if groBFP2 is a mutation in the groucho gene, seven lethal gro point mutations (Preiss et al., 1988) and the original viable gro allele (Knust et al., 1987; Ziemer et al., 1988) were tested for complementation by groBFP2 (Table 1). Except for the viable gro allele, all of these mutations cause eye defects in trans to groBFP2 similar, but not identical to groBFP2 homozygotes (see below and Table 1). Moreover, one copy of the P transformant ry+E8, which contains the gro transcription unit and complements gro mutants completely (Preiss et al., 1988), also rescues the homozygous groBFP2 mutant phenotype (Fig. 1C). AU of these data argue that groBFP2 is a viable allele of groucho.

Table 1.

Summary of eye phenotypes of viable groucho allele combinations

Summary of eye phenotypes of viable groucho allele combinations
Summary of eye phenotypes of viable groucho allele combinations

groBFP2 is a pdffai loss-of-function allele

The original viable groucho allele and groBFP2 each complement the mutant phenotype of the other; gro/groBFP2 flies have normal bristles and normal eyes (data not shown). However, in trans to groBFP2, all of the lethal mutations and gro deficiencies result in viable (or semi-viable) adult flies with mutant eye phenotypes reminiscent of, but not identical to, groBFP2 homozygotes. E(spl)EI07 and E(spl)E2S, both pupal lethal, have nearly wild-type eyes in trans to groBFP2 (Fig. ID and Table 1). The five stronger lethals tested (E(spl)E48, E(spl)E, E(spl)E75, E(spl)E77and l(gro)xni ) and the embryonic lethal deficiency E(spl)BX22 have similar phenotypes in combination with groBFP2 (Fig. 1E, F and Table 1); the retinas look similar to groBFP2 homozygotes, but the defects are more severe. The eyes of trans-heterozygotes have fewer normal looking facets than groBFP2 homozygotes, and the mutant facets do not all have one or two extra outer R-cells neatly added. Instead, many ommatidia have more than two extra R-cells, some facets appear fused, some are missing inner photoreceptor cells, and some rhabdomeres are malformed.

In order to characterize the groBFP2 mutant allele further, it is important to assess the loss-of-function phenotype of groucho mutations in the eye. A deficiency that removes only the groucho gene is not available and none of the available groucho point mutations are known to be complete loss-of-function mutations. Nevertheless, when considered together, the eye phenotypes of many strong gro point mutations and the smallest deficiency should provide insight into the loss-of-function phenotype of groucho in the eye. If, as the genetic data above suggest, groBFP2 is a loss-offunction mutant, the phenotypes of the lethals and the deficiency eye clones should be similar to groBFP2 homozygotes. If the eye phenotype of groBFP2 is completely different from the other groucho alleles, then groBFP2 is likely to be performing a novel function in the eye.

Marked clones of cells in the eye, homozygous for seven different lethal gro point mutations and the small deficiency E(spl)BX22 that removes only transcripts m5 through m9/10 (see Fig. 2) were generated by X-rayinduced somatic recombination (Materials and methods). Several clones of each lethal allele were examined in tangential sections and representative results are shown in Fig. 3. None of the clones look exactly like groBFP2 eyes. The clones obtained were grouped into four phenotypic classes (wild-type, weakly mutant, moderate and severe) based on the proportion of mutant facets within the clone, which paralleled the degree of malformation of the facets. Most of the mutant facets have extra photoreceptor cells. However, sometimes facets have too few photoreceptors, and in clones of the stronger alleles, photoreceptor cells are often malformed and there are fusions of facets.

Fig. 3.

Clones of cell m the eye homozygous for lethal E(spl) alleles. Clones of cells in the eye homozygous for various lethal E(spl) alleles were obtained by X-rayinduced somatic recombination (Materials and methods). (A) Tabulation of the number of clones examined for each lethal allele and their characterization mto four phenotypic classes. E(spl)E107 and E(spl)E28 are considered the weakest alleles because homozygotes die as pupae whereas the other alleles cause earlier death (Preiss et al, 1988). See the text, Table 1 and Fig. 2 for further descriptions of the different alleles. (B-D) Tangential sections through representative clones The bar in D is 20 pm, and applies to all panels Clones are marked as w, which is seen as the absence of the pigment granules normally associated with each rhabdomere and within pigment cells The pigment granules associated with the R-cells are seen as small black dots near each rhabdomere. (B) A “weak” E(spl)E28 clone. The arrow points to the only mutant facet in the clone. (C) A “severe” E(spl)BX22 clone The arrow indicates a facet with ectopic R-cells in which every R-cell is E(spl)+ (R8 was not examined). (D) A “moderate” l(gro)x 15 clone. The arrows indicate mosaic facets with ectopic outer R-cells.

Fig. 3.

Clones of cell m the eye homozygous for lethal E(spl) alleles. Clones of cells in the eye homozygous for various lethal E(spl) alleles were obtained by X-rayinduced somatic recombination (Materials and methods). (A) Tabulation of the number of clones examined for each lethal allele and their characterization mto four phenotypic classes. E(spl)E107 and E(spl)E28 are considered the weakest alleles because homozygotes die as pupae whereas the other alleles cause earlier death (Preiss et al, 1988). See the text, Table 1 and Fig. 2 for further descriptions of the different alleles. (B-D) Tangential sections through representative clones The bar in D is 20 pm, and applies to all panels Clones are marked as w, which is seen as the absence of the pigment granules normally associated with each rhabdomere and within pigment cells The pigment granules associated with the R-cells are seen as small black dots near each rhabdomere. (B) A “weak” E(spl)E28 clone. The arrow points to the only mutant facet in the clone. (C) A “severe” E(spl)BX22 clone The arrow indicates a facet with ectopic R-cells in which every R-cell is E(spl)+ (R8 was not examined). (D) A “moderate” l(gro)x 15 clone. The arrows indicate mosaic facets with ectopic outer R-cells.

In summary, in trans to strong gro mutations or deficiencies, groBFP2 shows a stronger eye phenotype than groBFp2 homozygotes. Also, when homozygous, the strong groucho mutations and the deficiency have effects on eye development similar to but more extensive than groBFP2 homozygotes. These observations support the view that groBFP2 is a partial loss-offunction allele.

Developmental defects in groBFP2 larval eye discs

To determine when during ommatidial assembly the extra photoreceptor cells are recruited, the developing eye discs of groBFP2 mutants were stained with the neural specific antibody mAb22C10 (Fujita et al., 1982). mAb22C10 reveals the sequence of photoreceptor cell assembly (R8, R2/5, R3/4, R1/6, R7) as each R-cell begins to express the mAb22C10 antigen when it acquires neural identity (Tomlinson and Ready, 1987a). As ommatidial assembly proceeds in a postenor-toanterior wave in the eye disc, ommatidia at all stages of photoreceptor cell assembly are observed in one disc (Fig. 4A, B and F). In groBFP2 discs, one or two ectopic R-cells are first observed staining with mAb22C10 in the fifth or sixth column of assembling facets, when the R3/4 pair first stain (Fig. 4C, D, E and F). These extra cells are likely to be the mystery cells, which are normally positioned between R3 and R4 in an undifferentiated 6- to 7-cell precluster (hereafter refered to as the precluster) just posterior to the morphogenetic furrow, but then disappear into the surrounding pool of dividing cells by column 3, without expressing neural antigens (Tomlinson, 1987a; Wolff and Ready, 1991b; Fig. 4F). However, as individual R-cells can be identified only by their positions in a normally assembling facet, other explanations for the unusual mAb22C10-stainmg structures observed in groBFP2 discs are possible. For example, it is conceivable that the mystery cells are excluded appropriately in the groBFP2 discs and the ectopic cells are recruited from the surrounding epithelial cells into any position in the cluster.

Fig. 4.

Ectopic R-cells observed in groBFP2 larval eye discs stained with mAb22C10. Third instar larval eye discs were stained with the neural specific antibody mAb22C10 (Fujita et al., 1982) as described in Materials and methods. mAb22C10 reveals the assembly sequence of R-cells (see F below). The antigen is cytoplasmic, and the stained structures shown are the apical tips of the R-cells The bar is 10 μm in B and D and 20 μm in A and C. In all panels, the morphogenetic furrow is at the top. (A, C) Wild-type and groBFP2eye discs, respectively, showing that the rows of developing facets are evenly spaced in groBFP2 discs. (B, D) Close-up views of wild-type and groBFP2 discs, respectively. D is a montage so that certain facets are simultaneously in focus. The numbered arrows indicate the facets schematized m E. (El) A wild-type cluster in —row 8 of the wild-type disc. The stained apical tips of R8 and R1-R6 are visible. (E2-4) Clusters in —row 8-9 of the groBFP2 disc. The black cells in E2 and E4 represent the ectopic cells separating R3 and R4 As expected, an additional ectopic cell can sometimes be observed (not shown). E3 appears to be one of the normally assembling facets, which are expected to be present as —one-third of facets in the groBF/>2 adult eye have the normal number of photoreceptors. The extra cells, due to their positions, are likely to be the mystery cells (see F below). Based solely on the mAb22C10 staining pattern observed, we cannot assign the identities to R-cells in the facets shown m E2 and E4. The labels shown m E are based on enhancer trap marker experiments (Figs 5 and 6) which show that the extra cells in the groBFF2 disc are of the R3/4 subtype and that the other R-cells are appropriately determined. R3, R4 and the ectopic cell, as they are all R3/4 subtype cells, were labeled somewhat arbitrarily. R3 and R4 were labeled as such because of their positions next to R2 and R5, and the ectopic cell so labeled because of the position of the “mystery cells” between R3 and R4 in the undifferentiated precluster (see F). (F) A summary of normal R-cell assembly based on the mAb22C10 staming pattern (Tomlinson and Ready, 1987a). The model for the groBFP2 mutant is based on the mAb22C10 staining pattern and also on the enhancer trap experiments (Figs 5 and 6). The cells are shaded in the order that they express the mAb22C10 antigen The 6- to 7-cell preclusters contain R8/2/5/3/4 and one or two mystery cells. Only one mystery cell is shown. In groBFP2 eye discs, the mystery cells do not leave the precluster, but become ectopic R-cells (black) of the R3/4 subtype adjacent to R3/4 (Figs 5 and 6). As explained above, the assignments of R3, R4 and the ectopic cells were somewhat arbitrary. A very small number (—1%) of facets in groBFP2 adult eyes have three ectopic R-cells. The third cell probably originates from an additional mystery cell.

Fig. 4.

Ectopic R-cells observed in groBFP2 larval eye discs stained with mAb22C10. Third instar larval eye discs were stained with the neural specific antibody mAb22C10 (Fujita et al., 1982) as described in Materials and methods. mAb22C10 reveals the assembly sequence of R-cells (see F below). The antigen is cytoplasmic, and the stained structures shown are the apical tips of the R-cells The bar is 10 μm in B and D and 20 μm in A and C. In all panels, the morphogenetic furrow is at the top. (A, C) Wild-type and groBFP2eye discs, respectively, showing that the rows of developing facets are evenly spaced in groBFP2 discs. (B, D) Close-up views of wild-type and groBFP2 discs, respectively. D is a montage so that certain facets are simultaneously in focus. The numbered arrows indicate the facets schematized m E. (El) A wild-type cluster in —row 8 of the wild-type disc. The stained apical tips of R8 and R1-R6 are visible. (E2-4) Clusters in —row 8-9 of the groBFP2 disc. The black cells in E2 and E4 represent the ectopic cells separating R3 and R4 As expected, an additional ectopic cell can sometimes be observed (not shown). E3 appears to be one of the normally assembling facets, which are expected to be present as —one-third of facets in the groBF/>2 adult eye have the normal number of photoreceptors. The extra cells, due to their positions, are likely to be the mystery cells (see F below). Based solely on the mAb22C10 staining pattern observed, we cannot assign the identities to R-cells in the facets shown m E2 and E4. The labels shown m E are based on enhancer trap marker experiments (Figs 5 and 6) which show that the extra cells in the groBFF2 disc are of the R3/4 subtype and that the other R-cells are appropriately determined. R3, R4 and the ectopic cell, as they are all R3/4 subtype cells, were labeled somewhat arbitrarily. R3 and R4 were labeled as such because of their positions next to R2 and R5, and the ectopic cell so labeled because of the position of the “mystery cells” between R3 and R4 in the undifferentiated precluster (see F). (F) A summary of normal R-cell assembly based on the mAb22C10 staming pattern (Tomlinson and Ready, 1987a). The model for the groBFP2 mutant is based on the mAb22C10 staining pattern and also on the enhancer trap experiments (Figs 5 and 6). The cells are shaded in the order that they express the mAb22C10 antigen The 6- to 7-cell preclusters contain R8/2/5/3/4 and one or two mystery cells. Only one mystery cell is shown. In groBFP2 eye discs, the mystery cells do not leave the precluster, but become ectopic R-cells (black) of the R3/4 subtype adjacent to R3/4 (Figs 5 and 6). As explained above, the assignments of R3, R4 and the ectopic cells were somewhat arbitrary. A very small number (—1%) of facets in groBFP2 adult eyes have three ectopic R-cells. The third cell probably originates from an additional mystery cell.

By the fifteenth column, assembling facets have normally gone through 90° rotation with respect to a central equator (Tomlinson and Ready, 1987a). The facets in the groBFP2 disc appear to rotate properly, so the orientation abnormalities apparent in the adult retina must occur in the pupal eye disc.

Eye discs from larvae carrying groBFP2 in trans to several lethal gro alleles were also stained with mAb22C10 (data not shown). As expected, E(spl)EI07/ groBFP2 and E(spl)E28/groBFP2 eye discs appeared normal. The discs of groBFP2 in trans to the stronger lethal alleles or E(spl)BX22 looked very similar to groBFP2 homozygous eye discs. Thus, the ectopic R-cells are likely to have the same origin in the transheterozygotes as in groBFP2 homozygotes, as they are first observed at the same time during ommatidial assembly. The additional defects apparent in the adult eyes of these genotypes as compared with groBFP2 homozygotes must occur during pupal eye development.

Photoreceptor cell identities in groBFP2 eye discs

To ascertain the subtype and position of the extra photoreceptors in the groBFF2 mutant eye disc, and also to investigate whether the other R-cells in the groBFP2 disc acquire their normal identities, the groBFP2 mutation was combined with seven different enhancer trap lines. Each enhancer trap line expresses β-galactosidase in the nuclei of different subsets of photoreceptor cells, thus allowing the identification of every R-cell in the developing disc by staining with antibodies to β galactosidase. The results are shown in Figs 5 and 6. In groBFP2 mutant eye discs, all seven enhancer trap fines express j5-galactosidase in their normal patterns, except that in the four lines that express βgalactosidase in the R3/4 pair, an extra nucleus is often observed next to the R3/4 cells (Figs 5 and 6). No ectopic nuclei stain in fines A2-6, X81 or N30, which express j3-galactosidase in R8, R8/2/5 and Rl/6/7, respectively (data not shown). In addition, groBFp2 discs were stained with an antibody to the rough protein, which, behind the morphogenetic furrow, is expressed in the nuclei of R2/5/3/4 (Kimmel et al., 1990). Staining was observed in the four R-cells and also in an ectopic cell between R3 and R4 (data not shown). We conclude, as summarized in Fig. 4F, that the extra R-cells in the groBFP2 mutant arise between or next to the normal R3/4 pair, and thus they are very likely to be the mystery cells. In addition, the ectopic cells are of the R3/4 subtype, and the other R-cells in the groBFP2 mutant eye disc attain their usual identities.

Fig. 5.

Photoreceptor cell identities in groBFP2 larval eye discs. Seven different enhancer trap lines, identified at the left, that express βgalactosidase in the subsets of R-cell nuclei indicated in parentheses, were stained with antiβgalactosidase antibodies (Materials and methods) in wildtype and groBFP2 backgrounds. The enhancer trap lines are described in detail in Materials and methods. The morphogenetic furrow is at the top in all panels. Shown are the four enhancer trap fines that normally express βgalactosidase in the R3/4 pair and also in adjacent ectopic cells in groBFP2 discs. The staining patterns are identical in wild-type and groBFP2 discs except for the appearance of the extra R-cells in the mutant discs. The arrows indicate some of the ectopic cells (see Fig. 6). The bar in the lower right-hand panel is 10 μm and 15 μm in all of the other panels except for the two at the lower left in which it is 20 μm See Fig. 6 for enlargements of individual assembling facets

Fig. 5.

Photoreceptor cell identities in groBFP2 larval eye discs. Seven different enhancer trap lines, identified at the left, that express βgalactosidase in the subsets of R-cell nuclei indicated in parentheses, were stained with antiβgalactosidase antibodies (Materials and methods) in wildtype and groBFP2 backgrounds. The enhancer trap lines are described in detail in Materials and methods. The morphogenetic furrow is at the top in all panels. Shown are the four enhancer trap fines that normally express βgalactosidase in the R3/4 pair and also in adjacent ectopic cells in groBFP2 discs. The staining patterns are identical in wild-type and groBFP2 discs except for the appearance of the extra R-cells in the mutant discs. The arrows indicate some of the ectopic cells (see Fig. 6). The bar in the lower right-hand panel is 10 μm and 15 μm in all of the other panels except for the two at the lower left in which it is 20 μm See Fig. 6 for enlargements of individual assembling facets

Fig. 6.

Ectopic R3/4 subtype photoreceptor cells in groBFP2 eye discs. Shown are enlarged images of individual facets from Fig. 5A. The R3/4 pair are the only R-cells in which all four enhancer trap lines, indicated at the left, express μgalactosidase (see Fig. 5A). μgalactosidase is also expressed in the ectopic R-cell nuclei of groBFP2 mutant eye discs. The identities of the nuclei are as shown and asterisks (*) indicate the ectopic nuclei seen in groBFP2 discs. As there are three R3/4 subtype nuclei shown in each panel of facets from groBFP2 discs, the labeling of cells as R3, R4 or * is somewhat arbitrary. R3 and R4 were labeled according to their positions adjacent to R2 and R5, and the ectopicnucleus was labeled between R3 and R4 as that is the normal position of the mystery cells in the undifferentiated precluster (see Fig. 4F).

Fig. 6.

Ectopic R3/4 subtype photoreceptor cells in groBFP2 eye discs. Shown are enlarged images of individual facets from Fig. 5A. The R3/4 pair are the only R-cells in which all four enhancer trap lines, indicated at the left, express μgalactosidase (see Fig. 5A). μgalactosidase is also expressed in the ectopic R-cell nuclei of groBFP2 mutant eye discs. The identities of the nuclei are as shown and asterisks (*) indicate the ectopic nuclei seen in groBFP2 discs. As there are three R3/4 subtype nuclei shown in each panel of facets from groBFP2 discs, the labeling of cells as R3, R4 or * is somewhat arbitrary. R3 and R4 were labeled according to their positions adjacent to R2 and R5, and the ectopicnucleus was labeled between R3 and R4 as that is the normal position of the mystery cells in the undifferentiated precluster (see Fig. 4F).

The neural determination of the mystery cells is independent of their genotype or the genotype of any other photoreceptor cell in groBFP2-wild-type mosaics In order to determine which cells in the groBFP2 mutant eye disc are responsible for the inappropriate recruitment of the mystery cells as photoreceptors, we generated marked clones of homozygous mutant cells (w–groBFP2–) in wild-type w+groBFP2+) eyes (Materials and methods and Fig. 7). Within patches of w–groBFP2 cells, the retina looks like that of groBFP2 homozygotes and outside of the clones the retina appears wild-type (Fig. 7A and legend). Therefore, the effect of the groBFP2 mutation, as is the case for the other E(spl) alleles (Fig. 3), is local.

Fig. 7.

Analysis of clones of groBFP2 mutant cells in wildtype eyes. Clones of homozygous groBFP2 mutant cells were generated by X-ray-induced somatic recombination and sectioned as described in Materials and methods The mutant cells are marked by the absence of the white gene, which results in the absence of the granules normally associated with each photoreceptor and pigment cell. The pigment granules of the R-cells are seen as small black dots near the rhabdomeres. (A) Tangential section through a portion of a clone at the level of R7. The bar is 20 pm. The arrow indicates a mutant facet (it has an ectopic outer R-cell) in which each R-cell has pigment granules associated with it and is thus genotypically wild-type (w+groBFP2+). (R8 is not visible in this plane of section). 35 clones were examined for such facets, and 26 examples were found within 15 different clones. 14 of these facets were on the border of the clone, like the example shown in A, and 12 appeared to be outside of the clone, one or two facets away from the border. These facets appear to be separated from the mutant clone probably because the wgroBFP2 epithelial cells responsible for the mutant phenotype are no longer adjacent to the mutant facets as they were in the larval disc (see Karpilow et al., 1989). For technical reasons, in the phenotypically mutant facets in which all of the apical R-cells appear to be w+groBFP2+, not all of the R8 cells could be scored as w+or w– However, all 12 of such facets just outside of the clone border and 6 of the facets at the clone edge could be analyzed definitively and these had w+ R8 cells. (B) The normally constructed facets in 10 different clones were analyzed cell by cell and the frequency with which each R-cell was w+groBFP2+ was tabulated. Facets were considered normally constructed if they had 8 R-cells in the appropriate arrangement; the orientation or trapezoidal shape of a facet was not considered. The frequency of individual R-cells being w+groBFP2+ is nearly random (50%) in all cases The slight deviations upwards from 50% are not surprising. In similar analyses of mosaic ommatidia where strict requirements for gene products were found in specific R-cells, other R-cells related by lineage to the required cells showed upwards deviations from random far greater than those observed here (Tomlinson et al., 1988, Carthew and Rubin, 1990; Mlodzik et al., 1990b; Reinke and Zipursky, 1988). Thus, the deviations observed are likely to reflect the close proximity of the R-cells to the cells within which the groBFI2 mutation acts. If there were a strict requirement for any particular R-cell to be groBFP2+, taking into account that ∼33% of the facets in a groBFP2 mutant eye are normally constructed, it would be expected that 100%-(50%)(∼33%) = ∼84% of those specific R-cells would be w+groBFP2+ in the mosaic normally constructed facets The number of wild-type and mutant mosaic facets were counted in the same 10 clones. Approximately 50% of the mosaic facets are wild-type (see text).

Fig. 7.

Analysis of clones of groBFP2 mutant cells in wildtype eyes. Clones of homozygous groBFP2 mutant cells were generated by X-ray-induced somatic recombination and sectioned as described in Materials and methods The mutant cells are marked by the absence of the white gene, which results in the absence of the granules normally associated with each photoreceptor and pigment cell. The pigment granules of the R-cells are seen as small black dots near the rhabdomeres. (A) Tangential section through a portion of a clone at the level of R7. The bar is 20 pm. The arrow indicates a mutant facet (it has an ectopic outer R-cell) in which each R-cell has pigment granules associated with it and is thus genotypically wild-type (w+groBFP2+). (R8 is not visible in this plane of section). 35 clones were examined for such facets, and 26 examples were found within 15 different clones. 14 of these facets were on the border of the clone, like the example shown in A, and 12 appeared to be outside of the clone, one or two facets away from the border. These facets appear to be separated from the mutant clone probably because the wgroBFP2 epithelial cells responsible for the mutant phenotype are no longer adjacent to the mutant facets as they were in the larval disc (see Karpilow et al., 1989). For technical reasons, in the phenotypically mutant facets in which all of the apical R-cells appear to be w+groBFP2+, not all of the R8 cells could be scored as w+or w– However, all 12 of such facets just outside of the clone border and 6 of the facets at the clone edge could be analyzed definitively and these had w+ R8 cells. (B) The normally constructed facets in 10 different clones were analyzed cell by cell and the frequency with which each R-cell was w+groBFP2+ was tabulated. Facets were considered normally constructed if they had 8 R-cells in the appropriate arrangement; the orientation or trapezoidal shape of a facet was not considered. The frequency of individual R-cells being w+groBFP2+ is nearly random (50%) in all cases The slight deviations upwards from 50% are not surprising. In similar analyses of mosaic ommatidia where strict requirements for gene products were found in specific R-cells, other R-cells related by lineage to the required cells showed upwards deviations from random far greater than those observed here (Tomlinson et al., 1988, Carthew and Rubin, 1990; Mlodzik et al., 1990b; Reinke and Zipursky, 1988). Thus, the deviations observed are likely to reflect the close proximity of the R-cells to the cells within which the groBFI2 mutation acts. If there were a strict requirement for any particular R-cell to be groBFP2+, taking into account that ∼33% of the facets in a groBFP2 mutant eye are normally constructed, it would be expected that 100%-(50%)(∼33%) = ∼84% of those specific R-cells would be w+groBFP2+ in the mosaic normally constructed facets The number of wild-type and mutant mosaic facets were counted in the same 10 clones. Approximately 50% of the mosaic facets are wild-type (see text).

At the clone borders, ommatidia mosaic for wgroBFP2 and w+groBFP2+ cells were observed (Fig. 7A). As expected, these genetically mosaic facets were sometimes normally constructed (with 8 R-cells) and sometimes abnormal (with 9, 10 or 11 R-cells.) Approximately 50% of the mosaic ommatidia are normally constructed (Fig. 7B), in contrast with —33% of the isogenic groBFP2 mutant facets (Fig. IB). Thus, —17% of the mosaic facets are “rescued” to wild-type (no extra R-cells) by wild-type cells at the borders of the clones.

The genotypes of the different R-cells in 161 normally constructed mosaic ommatidia in 10 different clones were scored to determine if there is a tendency for particular R-cells to be w+groBFP2+ The distribution of the R-cell genotypes is nearly random (Fig. 7B). Thus, no particular R-cells in a facet need to be groBFP2+ in order to exclude the mystery cells from the ommatidial cluster. Remarkably, we often observed at the clone borders phenotypically mutant facets (9 R-cells; note that only 8 are visible in Fig. 7A) in which all of the R-cells, including the ectopic one, are genotypically wildtype (w+groBFP2+) (Fig. 7A). In 35 clones examined, 26 examples of such facets were observed in 15 different clones (see legend to Fig. 7).

A similar detailed analysis of mosaic facets in the eye clones of lethal groucho mutations was not attempted because the eye phenotypes of these alleles were either too weak or too complex (Fig. 3). However, several examples of facets with ectopic R-cells were observed at the clone borders that appeared to be composed of genotypically wild-type R-cells (Fig. 3 and legend).

In summary, these observations imply that defects in cells outside of the photoreceptors and mystery cells result in the recruitment of ectopic R-cells in groBFP2 mutant eye discs. As groBFP2 is completely recessive and behaves as a partial loss-of-function mutation, the focus of action of the mutant protein is likely to be in at least a subset of the cells in which the wild-type groucho protein functions. Thus, we conclude that cells outside of the photoreceptors or mystery cells require groucho to inhibit the neuralization of the mystery cells.

We have described an unusual viable mutation in the Drosophila groucho gene, groBFP2, that results in ectopic photoreceptors in adult eyes. We show that groBFP2 is likely to be a partial loss-of-function mutation by its genetic behavior and also by comparing groBFP2 eyes to eye clones homozygous for lethal groucho alleles. There are two main conclusions from the analysis of groBFP2 eyes. First, by examining the early development of groBFP2 eye discs with neural and R-cell-specific markers, we show that many facets in groBFP2 retinas contain extra R3/4 subtype photoreceptors adjacent to the normal R3 and R4. Second, by observing the genotypes of R-cells in facets mosaic for groBFP2 and wild-type cells, we find that the focus of action of the groBFP2 mutation appears to be outside of the photoreceptors, including the ectopic ones, within a particular facet.

groucho function in the eye

The extensive eye defects in strong groucho mutant clones suggests that groucho may be involved in many different aspects of eye development. Indeed, the groucho protein is found in all cell nuclei in the eye disc, both anterior and posterior to the morphogenetic furrow (Delidakis et al., 1991; our observations). Using a temperature-sensitive allele, it has been demonstrated that another “neurogenic” gene, Notch, mediates cell interactions in all types of commitment decisions in the developing eye disc (Cagan and Ready, 1989b). Perhaps pleiotropic function will prove to be characteristic of many neurogenic genes.

Despite the apparent complexity of the role of groucho in eye development, using the groBFP2 allele, it is possible to ask where groucho is required to perform the function of preventing some mystery cells from becoming photoreceptors. The simplest interpretation of the mosaic analysis is that groucho is required outside of both the mystery cells and the photoreceptor cell precursors. First, the genotypes of the R-cells in the wild-type mosaic facets are random, consistent with the groBF12 mutation having no effect in photoreceptor cells. Second, the observation that the R1-R8 appear to be correctly determined in groBFP2 mutant discs is consistent with the groBFP2 mutant affecting cells other than photoreceptors. Finally, the frequent appearance at the groBFP2~ clone borders of facets containing an ectopic R-cell, in which every R-cell is groBFP2, is particularly compelling evidence in support of this interpretation. We performed a similar mosaic analysis with a mutant of a different gene with a null phenotype in the eye similar to groBFP2, that is, another mutant in which the mystery cells become photoreceptors. In 30 clones examined, not one example of a facet containing an extra R-cell in which all of the R-cells were genotypically wild-type was ever observed either at the clone border or outside of the clone (J.A.F.-V. and G.M.R., unpublished data). Moreover, genotypically wild-type, phenotypically mutant (with extra R-cells) facets were also observed at the borders of clones of lethal groucho mutations.

We cannot rule out more complicated models to explain the results of the mosaic experiments. For example, the phenotypically mutant (containing an extra R-cell) facets in which every R-cell is groBFP2+ observed at clone borders could be explained by proposing that groBFP2 R3 or R4 fail to signal the mystery cells to leave the facet, and are then sometimes competed out of the facet when surrounded by wildtype cells. “Sometimes” is emphasized as we often observed facets with ectopic photoreceptors in which groBFP2 cells in the positions of R3 or R4 were surrounded by wild-type cells. This model would require that a signal from R3 or R4 be propagated across more than one cell. In addition, the failure to observe a bias towards groBFP2+ R3 and R4 cells in normally constructed mosaic facets is not easily explained by this model. Another possibility is that the genotypes of cells from neighboring facets could influence the fates of the mystery cells. This interpretation is unlikely because the phenotype of Ellipse mutations shows that facets develop autonomously (Baker and Rubin, 1989).

Previous observations suggest that mystery cell fate is controlled by cells within the developing facet. The seven-up (svp) gene product is required in R3/4/1/6 cells to repress the R7 developmental pathway (Mlodzik et al., 1990b). In seven-up mutant clones, an extra outer photoreceptor cell of unknown subtype sometimes appears adjacent to svp – R3 cells, and always between svp – R3 and R4, independent of its own genotype (Mlodzik et al., 1990b). Thus, R3 and to some extent R4 influence the fate of the mystery cells. Our results with groBFP2 suggest that mystery cell fate is also controlled by cells outside of the precluster. Wolff and Ready (1991b) have shown that the first structure to emerge from the morphogenetic furrow is a rosette in which 10-15 cells, including the R8, R2/5, R3/4 precursors and the mystery cells form a ring around 4-5 core cells. The ring then opens and the 6- to 7-cell preclusters, containing the precursors to R8/2/5/3/4 and the mystery cells, are formed. Precisely when the mystery cells are determined to leave the precluster is unknown. Our results suggest that groucho mediates this process, presumably through cell contact. Thus, the cells that require groucho to prevent neurogenesis of the mystery cells could be the core cells or cells next to the mystery cells within the ring. Alternatively, if the cell communication process interrupted by the groBFP2 mutation occurs later, during the 6- to 7-cell precluster stage, epithelial cells surrounding the precluster could be involved. Unfortunately, these cells cannot be identified in the adult eye.

These results suggest that the cells requiring groucho to signal the mystery cells are uncommitted cells. All of the “neurogenic” genes, including E(spl), appear to play key roles in cell-contact-mediated neural inhibition in the embryonic neurectoderm that forms the CNS and probably also in the proneural regions of imaginai discs from which bristles arise. In these processes, cells compete for neural determination and the victor then inhibits its neighbors from also becoming neural cells. The particular role of groucho described here is different in that the cells sending the inhibitory signals appear to be uncommitted epithelial cells.

In cell transplantation experiments, cells containing E(spl) deletions behave autonomously, that is they always become neural when surrounded by wild-type cells (Technau and Campos-Ortega, 1987). This observation implies that E(spl) is required for the reception of a neural inhibition signal in embryonic cells. Our mosaic results suggest a non-autonomous role for groucho, in that cells outside of the mystery cells require groucho to influence mystery cell fate. However, our results do not necessarily contradict the previous findings, as the apparently non-autonomous role we find for groucho could be indirect. In other words, it is possible that groucho is autonomously required by the cells that direct the mystery cells to leave the precluster.

Specification of photoreceptor cell subtype

Developing ommatidia display a particular sequence of determination of specific cell types and assembling clusters have characteristic structures and cell contacts (Tomlinson, 1985; Tomlinson and Ready, 1987). These observations led Tomlinson and Ready to hypothesize that local cell contacts instruct cells to acquire particular fates. An extreme version of this model would predict that in the precluster, R8 would cue R2 and R5, and those three cells would then instruct the specification of R3 and R4. In groBFP2 eye discs, the specification of the mystery cells as R3/4 subtype photoreceptors appears to break the rules for cell specification in the precluster. How much evidence is there that cells within the initial 6- to 7-cell precluster normally cue each others determination? The best evidence comes from studies of the rough gene, which encodes a homeobox protein required only in cells R2 and R5 for their appropriate differentiation (Tomlinson et al., 1988; Saint et al., 1988; Heberlein et al., 1991). In rough mutant eye discs, although the appropriate R2/5 precluster cells become outer photoreceptors, they are not properly specified as the R2/5 subtype, and presumably do not send the necessary signals to R3/4 so that these cells often fail to join the developing precluster. It is unknown whether or not cues from R8 are also necessary for R3/4 determination (see Banerjee and Zipursky, 1990). Also, mutations disrupting communication between R8 and R2/5 have not yet been identified.

How can the mystery cells become R3/4 subtype photoreceptors? Our interpretation of the analysis of groBFP2+/groBFP2 mosaic ommatidia is that the cells responsible for the extra R3/4 cells in groBFF2 mutants are outside of the R-cells in the facet, including the extra R-cells. Moreover, the appropriate expression of many markers implies that the R-cells are properly determined in groBFP2 eye discs. Therefore, it cannot be argued that, in groBFP2 eye discs, the R-cells in contact with the mystery cells (R8 and R3/4), because they are mutant cells, send inappropriate signals to the mystery cells thus recruiting them as R3/4 cells. Likewise, it is inconsistent with our data to suppose that the mystery cells, because they are groBFp2 mutant cells, inappropriately receive positional cues from R3/4 and/or R8.

More likely explanations for the determination of the mystery cells as R3/4 subtype photoreceptors in groBFP2 eye discs allow that R8 and R3/4 act normally, but the mystery cells are in an unusual environment because they remain in contact with the developing ommatidial precluster longer than they would normally. For example, the R3/4 cells may normally send positional cues similar to those of R2/5 cells. These cues from R3/4 usually would be inconsequential because there are no cells between R3 and R4 after the mystery cells leave. The R3/4 pair express rough (Kimmel et al., 1990) although no requirement for rough in cells other than R2/5 is apparent, suggesting that these subtypes may share some signalling pathways. Alternatively, the mystery cells could acquire R3/4 fate by receiving a signal solely from R8, and thus by a pathway at least partially different from that of the normal pre-R3/4 cells. The combination of groBFP2 with mutations that disrupt R8 or R3/4 differentiation may help to distinguish among these alternatives.

groucho is involved in several different neural repression pathways

The level and nuclear distribution of groucho antigen appears, at the light microscope level, to be normal in groBFP2 eye discs (data not shown). The mutation may therefore affect the structure of the protein in a manner that is critical to one of its functions in the eye. The ability to obtain groucho mutants that specifically affect a subset of its many functions reveals that groucho is likely to be involved in several different cell signaling pathways that prevent neural cell determination. The original groucho allele very specifically affects the ability of the groucho gene product to repress the formation of specific head bristles. Similarly, although E(spl)EI07 is pupal lethal due to weak neural overgrowth, it has little effect on eye development. Likewise, groBFP2 perturbs a small subset of the many roles of the normal groucho protein in the eye.

J.A.F.-V. is extremely grateful to Doug Melton for his exceptional generosity in allowing me to work in his laboratory, where much of this work was carried out. J.A.F.-V. is also extremely grateful to Ruth Lehmann for welcoming me into her laboratory and for her generous support and enthusiasm. We thank Anette Preiss, Christos Dehdakis and Spyros Artavams-Tsakonas for fly stocks, unpublished information and the m9/10 antibody, Seymour Benzer for mAb22C10, Bruce Kimmel for the rough antibody and Anne Hart and Larry Zipursky for providing fly stocks and information prior to publication J A.F.-V. is grateful to all of her friends in the Rubin lab for sending a million and one reagents, fly stocks, etc. We thank Matthew Freeman, Bruce Kimmel, Nick Baker, Joe Heilig and Marek Mlodzik for the enhancer trap lines, especially the unpublished ones. J.A F.-V thanks Anette Preiss, Christos Dehdakis, Nick Baker and Tian Xu for advice and discussion, and Ruth Lehmann, Matthew Freeman, Tian Xu, Kevin Moses, an anonymous reviewer and especially Nick Baker for their helpful comments on the manuscript. The manuscript also benefitted significantly from a discussion with Andrew Tomlinson. J.A F -V received postdoctoral fellowships from the Helen Hay Whitney Foundation and the Howard Hughes Medical Institute P D.V. was a Jane Coffin Childs postdoctoral fellow.

Baker
,
N. E.
,
Mlodzlk
,
M.
and
Rubin
,
G. M.
(
1990
)
Spacing differentiation m the developing Drosophila eye A fibnnogen related lateral inhibitor encoded by scabrous
Science
250
,
1370
1377
Baker
,
N. E.
and
Rubin
,
G. M.
(
1989
)
Effect on eye development of dommant mutations in Drosophila homologue of the EGF receptor
.
Nature
340
,
150
153
Banerjee
,
U.
,
Renfranz
,
P. J.
,
Pollock
,
J. A.
and
Benzer
,
S.
(
1987
).
Molecular charactenzation and expression of sevenless, a genemvolved in neuronal pattern formatton in the
Drosophila eye
49
,
281
291
Banerjee
,
U.
and
Zlpursky
,
S. L.
(
1990
).
The role of cell-cell mteractlons in the development of the Drosophila Visual system
Neuron
4
,
177
187
Cagan
,
R. L.
and
Ready
,
D. F.
(
1989a
).
The emergence of order m the Drosophtla pupal retina
Dev 810/
.
136
,
346
362
Cagan
,
R. L.
and
Ready
,
D. F.
(
1989b
)
Notch 1s requrred for successive cell decisions m the developmg Drosophila retina
.
Genes Dev
.
3
,
1099
1112
.
Campos-Ortega
,
J. A.
(
1991
).
Genetic and molecular bases of neurogenes1s m Drosophila melanogaster
Ann Rev Neurosc1
14
,
399
420
.
Campos-Ortega
,
J. A.
and
Hofbauer
,
A.
(
1977
).
Cell clones and pattern format10n On the hneage of photoreceptor cells in the compound eye of Drosophila Roux’s
Arch Dev 810/
181
, |p?27- 245|
Carthew
,
R. W.
and
Rubin
,
G. M.
(
1990
)
sevenless in absentia, a gene reqmred for specificat10n of R7 cell fate in the Drosophila eye
Cell
63
,
561
577
.
Dalrymple
,
M.A.
,
Petersen-Bjorn
,
S.
,
Friesen
,
J. D.
and
Beggs
,
J. D.
(
1989
)
The product of the PRP4 gene of S
.
arev1S1ae shows homology to f3 subumts of G protems Cell
58
,
499
508
.
Delldakls
,
C.
,
Preiss
,
A.
,
Hartley
,
D. A.
and
Artavanls-Tsakonas
,
S.
(
1991
)
Two genetically and molecularly distinct functions involved in early neurogenes1s reside withm the Enhancer of spill locus of Drosophtla melanogaster
Genetics
129
,
803
823
Dietrich
,
U.
and
Campos-Ortega
,
J. A.
(
1984
)
The express10n of neurogemc IOCJ m 1magmal epidermal cells of Drosophila mulanogaster I
Neurogen
1
,
315
332
.
Doe
,
C. Q.
and
Goodman
,
C. S.
(
1985
)
Early events m msect neurogenes1s II The role of cell mteract10n and cell lineages in the determmat10n of neuronal precursor cells
Dev Bwl
.
111
,
206
219
Francbeschini
,
N.
and
Kirschfeld
,
K.
(
1971
)
Les phenornenes de pseudop1lle dans l’oe1I compose de Drosophila
Kybemellk
9
,
159
182
Fujita
,
S. C.
,
Zlpursky
,
S. L.
,
Benzer
,
S.
,
Ferrus
,
A.
and
Shotwell
,
S. L.
(
1982
)
Monoclonal antJboches agamst the Drosophila nervous system
Proc natn Acad SCl USA
79
,
7929
7933
Ghysen
,
A.
and
Dambly-Chaudlere
,
C.
(
1989
).
GenesIS of the Drosophila penpheral nervous system
Trends Genet/CS
5
,
251
255
.
Hart
,
A. C.
,
Kramer
,
H.
,
Van Vactor
,
D. L.
,
Paldhungat
,
M.
and
Zipursky
,
S. L.
(
1990
)
lnduct10n of cell fate m the Drosophila retnia· the bnde-of-sevenless protem 1s predicted to contam a large extracellular domam and seven transmembrane segments
Genes Dev
4
,
1835
1847
Hartley
,
D. A.
,
Preiss
,
A.
and
Artavanls-Tsakonas
,
S.
(
1988
)
A deduced gene product from the Drosophtla neurogemc locus Enhancer of spill shows homology to mammalian G-protem subumt
.
Cell
55
,
785
795
Heberlein
,
U.
,
Mlodzlk
,
M.
and
Rubin
,
G. M.
(
1991
)
Cell fate determmat10n m the developmg Drosophtla eye: role of the rough gene
.
Development
112
,
703
712
.
Heitzler
,
P.
and
Simpson
,
P.
(
1991
).
The ch01ce of cell fate in the ep1derm1s of Drosophtla
Cell
64
,
1083
1092
Karpllow
,
J.
,
Kolodkln
,
A.
,
Bork
,
T.
and
Venkatesh
,
T.
(
1989
)
Neuronal development m the Drosophila compound eye. rap gene function 1s reqmred m photoreceptor cell RS for ommat1drnl pattern format10n
Genes Dev
3
,
1834
1844
Kidd
,
S.
,
Kelley
,
M. R.
and
Young
,
M. W.
(
1986
).
Sequence of the Notch locus of Drosophila melanogaster. relat10nsh1p of the encoded protem to mammalian clottmg and growth factors
Mo/ Cell Bwl
6
,
3094
3108
Kimmel
,
B. E.
,
Heberlein
,
U.
and
Rubin
,
G. M.
(
1990
)
The homeodomain protem rough 1s expressed in a subset of cells m the developmg Drosophila eye where ti can speetfy photoreceptor cell subtype
Genes Dev
4
,
712
727
.
Klambt
,
D.
,
Knost
,
E.
,
Tietze
,
K.
and
Campos-Ortega
,
J. A.
(
1989
).
Oosely related transcnpts encoded by the neurogeruc gene complex Enhancer of spill of Drosophila melanogaster
EM BO I
.
8
,
203
210
Knost
,
E.
,
Tietze
,
K.
and
Campos-Ortega
,
J. A.
(
1987
)
Molecular analysis of the neurogemc locus Enhancer of spill of Drosophila melanogaster
EM BO I
13
,
4113
4123
Kopczynski
,
C. C.
,
Altoo
,
A. K.
,
Fechtel
,
K.
,
Koob
,
P. J.
and
Mnskavltch
,
M.A. T.
(
1988
)
Delta, a Drosophila neurogeruc gene, JS transcnptionally complex and encodes a protem related to blood coagulatJon factors of vertebrates Genes
Dev
.
2
, |p1723-1725|
Lawrence
,
P. A.
and
Green
,
S. M.
(
1979
)
Cell lineage in the developing retina of Drosophila
Dev Biol
71
,
142
1152
Lehmann
,
R.
,
Dietrich
,
U.
,
Jimenez
,
F.
and
Campos-Ortega
,
J. A.
(
1981
)
Mutations of early neurogenes!S m Drosophila Roux’s
Arch D Biol
190
,
226
229
Lehmann
,
R.
,
Jimenez
,
F.
,
Dietrich
,
U.
and
Campos-Ortega
,
J. A.
(
1983
)
On the phenotype and development of mutants of early neurogenests m Drosophila melanogastu
Roux’s Arch Dev Bwl
192
,
62
74
Lewis
,
E. B.
and
Bacher
,
F.
(
1968
)
Method of feedmg ethane methyl sulfphonate (EMS) to Drosophila males
Dros Inform Serv
43
,
193
Lindsley
,
D. L.
and
Grell
,
E. H.
(
1968
).
Genetic Vanauon of Drosophila melanogaster Carnegie Inst pubi
. no 627,
Washmgton, D.C
Mlodzik
,
M.
,
Baker
,
N. E.
and
Robin
,
G. M.
(
1990a
).
lsolat10n and express10n of scabrous, a gene regulatmg neurogenes1s m
Drosophila Genes Dev
4
,
1848
1861
Mlodzl.k
,
M.
,
Hiromi
,
Y.
,
Weber
,
U.
,
Goodman
,
C. S.
and
Rubin
,
G. M.
(
1990b
)
The Drosophila seven-up gene, a member of the steroid receptor gene superfamtly controls photoreceptor cell fates
Cell
60
,
211
224
Moses
,
K.
(
1991
)
The role of transcnpnon factors m the developing Drosophila eye Trends
Genetics
7
,
250
257
Murre
,
C.
,
Schonleber McCaw
,
P.
and
Baltimore
,
D.
(
1989
)
A new DNA bmdmg and d1menzat10n motif m unmunoglobulm enhancer bmdmg, daughterless, MyoD and myc protems
Cell
56
, |p777-783|
Petersen-BJorn
,
S.
,
Soltyk
,
A.
,
Beggs
,
J. D.
and
Friesen
,
J. D.
(
1989
).
PRP4 (RNA4) from Saccharomyces cuevlSlae its gene product ts associated with the U4/U6 small nuclear nbonucleoprotem parncle
Mol Cell Bwl
9
,
3698
3709
Poulson
,
D. F.
(
1937
)
Chromosomal defic1encres and embryomc development of Drosophtla melanogaster
Proc nam Acad Sci USA
23
,
133
137
Preiss
,
A.
,
Hartley
,
D. A.
and
Artavanis-Tsakonas
,
S.
(
1988
)
The molecular genetics of Enhancer of spill, a gene reqmred for embryomc neural development m Drosophila
EMBO J
7
,
3917
3927
Ready
,
D. F.
(
1989
)
A multifaceted approach to neural development
Trends Neurosci
12
,
102
110
.
Ready
,
D. F.
,
Hanson
,
T. E.
and
Benzer
,
S.
(
1976
)
Development of the Drosophtla retma, a neurocrystallme lattice
Dev Biol
53
,
217
240
Reinke
,
R.
and
Zipursky
,
S. L.
(
1988
)
Cell-cell mteract10n m the Drosophila retma; the bruie of s en/ess gene 1s required m photoreceptor cell R8 for R7 cell development
.
Cell
55
,
321
330
.
Rubin
,
G. M.
(
1991
)
Signal transduct10n and the fate of the R7 photoreceptor m Drosophila Trends
Genetics
7
,
Saint
,
R.
,
Kalionls
,
B.
,
Lockett
,
T. J.
and
Ellzur
,
A.
(
1988
)
Pattern format10n m the developmg eye of Drosophila nu/anogaster 1s regulated by the homeobox gene, rough
Nature
334
,
151
154
Shepard
,
S. B.
,
Broverman
,
S. A.
and
Muskavftch
,
M.A. T.
(
1989
)
tnparnte mteractton among alleles of Notch, Delta and Enhancer of split dunng 1magmal development of Drosophtla melanogaster
Genetics
122
,
429
438
Simpson
,
P.
(
1990
)
Lateral mh1b1tton and the development of the sensory bnstles of the adult penphetal nervous system of Drosophila
Development
109
,
509
519
Taghert
,
P. H.
,
Doe
,
C. Q.
and
Goodman
,
C. S.
(
1984
)
Cell detennmat1on and regulation dunng development of neuroblasts and neurones m grasshopper embryos
Nature
307
,
163
165
Technau
,
G. M.
,
Becker
,
T.
and
Campos-Ortega
,
J. A.
(
1988
)
Reversible cornn11tment of neural and epidermal progemtor cells dunng embryogenes1s of Drosophila melanogastu
Roux’s Arch Dev Bwl
197
,
413
418
Technau
,
G. M.
and
Campos-Ortega
,
J. A.
(
1986
)
Lmeage analysts of transplanted md1v1dual cells in embryos Drosophila melanogastu II Comrrutment and prohferattve capab1ht1es of neural and epidermal cell progemtors
Roux’s Arch Dev Bwl
195
,
445
454
Technau
,
G. M.
and
Campos-Ortega
,
J. A.
(
1987
)
Cell autonomy of expression of neurogemc genes of Drosophila melanogaster
Proc nam Acad Sci USA
84
,
4500
4504
.
Tomiinson
,
A.
(
1985
)
The cellular dynarrucs of pattern formation m the eye of Drosophila
J Embryo/ Exp Morph
89
,
313
331
Tomlinson
,
A.
(
1988
)
Cellular mteracttons in the developmg Drosophila eye
Development
104
,
fn-193
Tomlinson
,
A.
,
Kimmel
,
B. E.
and
Rubin
,
G. M.
(
1988
)
rough, a Drosophila homeobox gene reqmred m photoreceptors R2 and R5 for mduct1ve mteract10ns m the developmg eye
.
Cell
55
,
771
784
Tomlinson
,
A.
and
Ready
,
D. F.
(
1987a
)
Neuronal dtfferentJation m the Drosophila ommatidmm
.
Dev Bwl
120
,
366
376
Tomlinson
,
A.
and
Ready
,
D. F.
(
1987b
)
Cell fate m the Drosophila ommahdtum
Dev Biol
123
,
264
275
Vassln
,
H.
,
Bremer
,
K. A.
,
Knust
,
E.
and
Campos-Ortega
,
J. A.
(
1987
)
The neurogemc locus Delta of Drosophila melanogaster IS expressed m neurogemc temtones and encodes a putative membrane protem with EGF-hke repeats
EMBO J
6
,
3431
3440
Welshon.s
,
W. J.
(
1956
).
Dosage expenments wtth split mutants m the presence of an enhancer of split Dros
.
Inform Serv
30
,
157
158
Wharton
,
K. A.
,
Johansen
,
K. M.
,
Xu
,
T.
and
Artavanis-Tsakonas
,
S.
(
1985
)
Nucleotide sequence from the neurogemc locus Notch 1mphes a gene product that shares homology with protems contammg EGF-like repeats
Cell
43
,
567
581
Wolff
,
T.
and
Ready
,
D. F.
(
1991a
)
Cell death m normal and rough eye mutants of Drosophila
.
Development
113
,
825
839
Wolff
,
T.
and
Ready
,
D. F.
(
1991b
)
The begtnnmg of pattern formation m the Drosophila compound eye the morphogenetlc furrow and the second mitotic wave
Development
113
,
841
850
Yochem
,
J.
and
Byers
,
B.
(
1987
)
Structural companson of the yeast cell d1vts10n cycle gene CDC4 and a related pseudogene
J Mo/ Bwl
195
,
233
245
.
Ziemer
,
A.
,
Tietze
,
K.
,
Knust
,
E.
and
Campos-Ortega
,
J. A.
(
1988
)
Genetic analysis of Enhancer of spill, a locus mvolved m neurogenes1s m Drosophila nulanogaster
GenetlCS
119
,
63
74
Zipursky
,
S. L.
(
1989
)
Molecular and genetic analysis of Drosophila eye development sevenless, bnde-of-sevenless and rough Trends
Neurosc1
12
,
fn-189