Identification and Characterization of Autosomal Genes That Interact with Gkrss in the Developing Drosophila Eye

Identification and Characterization of Autosomal Genes That Interact With gkrss in the Developing Drosophila Eye

Chaoyong Ma, Hui Liu, Ying Zhou and Kevin Moses Department of Biological Sciences, University of Southern California, Los Angeles, California 90089-1340 Manuscript received September 11, 1995 Accepted for publication January 4, 1996

ABSTRACT The glass gene encodes a zinc finger, DNA-binding protein that is required for photoreceptor cell development in Drosophila melanogaster. In the developing compound eye, glass function is regulated at two points: (1) the protein is expressed in all cells’ nuclei posterior to the morphogenetic furrow and (2) the ability of the Glass protein to regulate downstream genes is largely limited to the developing photoreceptor cells. We conducted a series of genetic screens for autosomal dominant second-site modifiers of the weak allele glass3, to discover genes with products that may regulate glass function at either of these levels. Seventy-six dominant enhancer mutations were recovered (and no dominant suppressors). Most of these dominantmutations are in essential genes andare associated with recessive lethality. We have assigned these mutations to 23 complementation groups that include multiple alleles of Star and hedgehog as well as single alleles of Delta, roughened eye, glass and hairy. Mutations in 18 of the complementation groups are embryonic lethals, of and these, 13 show abnormal adult retinal phenotypes in homozygous clones, usually with altered numbers of photoreceptor cells in some of the ommatidia.

N Drosophila, the adult compound eye consists of ceptor cell development. The eyeless gene functions in I -800 facets (ommatidia), each of which contains the specification of the nervous system (QUIRINGet al. 20 cells: eight photoreceptorcells and 12 accessory cells 1994; BARINAGA1995; HALDER et al. 1995), and some (WADDINGTONand PERRY1960). The photoreceptor genes functionearly to control thegrowth of the retinal cells can be subdivided into three subtypes according cell precursors, such as sine oculis (CHEYETTEet al. 1994) to their morphology, synaptic specificity, and spectral and eyes absent (BONINIet al. 1993). Some genes func- sensitivity: the six outerphotoreceptors, Rl-R6, the tion in the specification of the regular arrays of omma- apical central photoreceptor, R7, and the basal central tidial founder cells, such as scabrous (BAKERet al. 1990; photoreceptor, R8 (reviewed by TOMLINSON1988; MLODZIKet al. 1990a; ELLISet al. 1994; BAKERand ZI- MEINERTZHAGENand HANSON 1993). The accessory TRON 1995), the Drosophila homologue of the epider- cells include lens-secreting cone cells, light-insulating mal growth factor receptor, E@, (BAKERand RUBIN pigment cells, and cells comprising a mechano-sensory 1989; BAKERand RUBIN1992; ZAK and SHILO1992), bristle. The compound eye develops from a monolayer Notch (CAGANand READY 1989; BAKER and ZITRON epithelium called the eye imaginal disc, which grows 1995) and Delta (BAKERand ZITRON 1995). Some genes by unpatterned cell divisions during early larval life, function to specify photoreceptor subtypes, such as without apparent differentiation. Cell-type differentia- rough (BASLERet al. 1990; KIMMEL et al. 1990; HEBERLEIN tion begins in the last larval instar, when a transverse and RUBIN1991), sewenless (recently reviewed by DICK- groove (the morphogenetic furrow) sweeps across the SON and HAFEN 1994) and smen up (MLODZIKet al. 1990b; HIROMIet al. 1993; BEGEMANNet al. 1995; eye imaginal disc from posterior to anterior (READY et JCRAMER et al. 1995). al. 1976; HEBERLEINand MOSES1995). Anterior to the The glass (gl)gene functionsin the terminaldifferen- furrow, cells are dividing and undifferentiated. At the tiation of photoreceptor cells. In gl mutants, the pre- furrow, cell division stops and differentiation begins, sumptive photoreceptor cells begin to develop as neu- and the ommatidial founder cells appear in a regular ronsbut fail to express photoreceptor cell-specific array. Posterior to the furrow, a final wave of cell divi- genes and die after about 60 hr of pupal development. sion occurs (READY et al. 1976; WOLFFand 1991) READY Externally, the adult compound eyes are reduced in and differentiation proceedsas photoreceptor cells and size, and the regular array of ommatidia is disrupted. accessory cells are sequentially specified (TOMLINSON Strong gl alleles also cause part of the eye to lack pig- 1988). Many genes have been found to act in photore- ment, while three independent weak alleles (gl’, glBx9 and gZBz4)produce fully pigmented eyes. gl function is Corresponding author: Kevin Moses, Department of Biological Sci- required autonomously by all the developing photore- ences, University of Southern California, SHS-172, MC1340, 825 W. 37th St., Los Angeles, CA 90089-1340. ceptor cell types and is not required by the accessory E-mail: [email protected] cells (MOSESet al. 1989). The gl gene encodes a 604

Genetics 142 1199-1213 (April, 1996) 1200 C. Ma et al. amino acid protein with five zinc finger motifs that can notypes as heterozygotes, vital genes affecting eye devel- bind to sequences within the enhancer regions of the opment can be identified.This is important since many Rh1 (rhodopsin) gene and regions within its own pro- genes have pleiotropic effects. It has been estimated moter (MOSES and RUBIN1991). At least three of the that about two-thirds of the vital genes may be essential five zinc finger domains are required forDNA binding, for eye development (THAKERand WKEL1992), how- and there is also an activation domain (O’NEILLet al. ever, because most genetic screens for geneswith func- 1995). Sequences resembling G1-binding sitesare found tion in eye development have required mutations to be near the promoterof the chicken rhodopsin gene, and viable (BAKERet al. 1992),only a few vitalgenes affecting they are bound by a retinal-specific, chicken nuclear eye development have been identified, such as Notch protein (SHESHBERADARANand TAKAHASHI 1994). g1 ap- (CAGAN and READY 1989), Delta (BAKERand ZITRON pears to be the last regulatory gene in its hierarchy, 1995) and EdrE (BAKERand RUBIN 1989).Some new asit acts directly on terminal “realizator”functions screening strategies, such as FLP-mediated somatic re- (GARCIA-BELLIDOet al. 1975), such as Rhl. Thus g1 may combination (XU and RUBIN1993) and the “enhancer be a master regulatory gene that controls photorecep- trap” technique (O’KANE and GEHRING 1987),also fa- tor differentiation, and the regulation of g1 activity may cilitate the identification of such vital genes, such as be a critical step in the specification of the photorecep- seven-up (MLODZIKet al. 1990b). tor cell type. We report here the results of a screen for autosomal In Drosophila, both transcription and translation of genes that interact with gl, as second-site dominant en- the g1 gene begin at the morphogenetic furrow (in all hancers or suppressors. All known gl alleles (including cells) and then persist to the posterior margin of the protein nulls) are completely recessive in trans to wild disc, although at a lower level (MOSES et al. 1989; ELLIS type. This implies that 50% of the normal quantity of et al. 1993). G1 protein is detected in the nuclei of all G1 protein is sufficient for normal development.Homo- cells posterior to the furrow, but its ability to activate zygotes of weak g1 alleles (glj, gl’”’ and gl”‘) have an transcription differs between cell types: a G1-regulated intermediate, rough eye phenotype (Figure 1B). Unlike P-galactosidase reporter construct is expressed in the null alleles (Figure lC), their phenotype can be made developing photoreceptor cells but not in the accessory stronger by placing them in trans to a deficiency, and as cells. In the developing R7 cell g1 activity is dependent such they are likely to be hypomorphic alleles (MULLER on thesevenless tyrosine kinase signal (MOSESand RUBIN 1932). Modification of this weak gl phenotype may be 1991). There are atleast two classes ofmodel to account caused by mutations in genes that areupstream regula- for this restriction. (1) There could be photoreceptor tors of gl and that act in a stoichiometric manner. A cell-specific, positive regulators of gl function. G1 pro- loss-of-function mutation in a gene thatnormally acts to tein may be modified in the developing photoreceptor positively regulate gl could be recovered as a dominant cells: an anti-Gl monoclonal antibody detects multiple enhancer of the weak g13 phenotype. Similarly, a loss- forms of the protein on blots (ELLIS et al. 1993). (2) of-function mutation in a gene that normally acts to Alternatively, there could be negative regulators of gl negatively regulate gl could be recovered as a dominant function, whichwould be present in allcells in the suppressor of the weak gl’ phenotype. Modification of morphogenetic furrow and then later specifically elimi- this weak gl phenotype is less likely to be caused by nated from the developing photoreceptor cells. This mutations in genes that aredownstream of 91.91 appears second possibility appears to be more likely, as a se- to be the last regulatory gene in its pathway and acts quence adjacent to (and overlapping with) a gl-binding on many terminal “realizators.” Mutations in any indi- site in an Rh1 enhancer is responsible for the repression vidual terminal gene are unlikely to have a detectable of gl response in nonphotoreceptor cells (ELLISet al. effect on the phenotype tested in this screen. For in- 1993). However g1 activity is regulated, genes that act stance, mutationsin the Rh1 opsin gene (nindalleles) in the gl pathway might be identified as dominant, sec- do notaffect external morphology to cause a rough eye ond-site modifiers of 91. and are unlikely to enhance the phenotype ofgl’ as F, Screens for second-site modifiers have been used to heterozygotes. identify functionally linked genes inDrosophila (BOTAS We report here the results from a series of screens et al. 1982; BRANDand CAMPOS-ORTEGA1990; ROGGEet for dominantmodifiers of gl’. We used three mutagens al. 1991; SIMONet al. 1991; CARTHEWet al. 1994; HAR- (EMS, y-rays and hybrid dysgenesis) to treat males and DING et al. 1995; WEMMERand KLAMBT 1995). In such examined -455,000 F, progeny. We report theisolation a screen, the activity of a gene (in our case 81) is re- and characterization of 76 dominant enhancer of gl duced, such that an intermediate, or weak, phenotype mutations in 23 complementation groups. is seen (Figure 1). In this sensitized condition, a recessive loss-of-function mutation in a functionally related MATERIALS AND METHODS gene may cause a visible phenotype although only one Drosophila stocks: Drosophila cultures were carried out copy of the gene is inactivated. Thus the new mutation at 25” on standardcornmeal-molasses-agar medium. glalleles may be revealed as dominant, while otherwise it would used were as follows: g13 is a weak mutation caused by a be recessive. Sincethe new mutations cause visible phe- spontaneous insertionof -2.5 kb, g16” is a null allele caused Enhancers of glass 1201 by a spontaneous insertion of -30 kb, glBx9is a weak muta- tosomal genes for their ability to act as dominant enhancers tion induced by X-rays, glBZ4is a weak mutation induced by of g1’. The genes tested were armadillo, cubitus interuptus, en- EMS (MOSESet al. 1989). The stock used for mutagenesis grailed, even skipped, fused, fushi tarazu, giant, ghs, goosebeny, was isogenic for the second and third chromosomes and is haiy, hunchback,knirps, K~ppel,naked, odd paired, paired, of the genotypewlll*; P[(w,y)D] 3 gl’ e, where P[(w,y)D] 3 is patched,roughened gre, rotund, runt, sloppY paired,smooth, Star, one cM proximal to gl (R. LEVIS and G. M. RUBIN, unpub- tailless and wingless (see alleles listed above and RESULTS sec- lished results). The white gene carried by this P element tion below). renders the eye red and was used as a reference marker to Linkage and complementation tests: The mutations were monitor the efficiency of the mutagenesis by scoring the assigned to linkage groups by their patterns of inheritance white-eyed flies in the screens. The following P-element in- relative to the dominant markers on the balancer chromo- sertions wereused inmapping experiments and mosaic somes, and on whetherthey were linked to the third chromc- clone analyses (kindly provided by G. M. RUBINand T. LA- some P[ (w, y)D] 3 element. TheE(g1) mutations that areassoci- VER~,see Table 1):P[ (w,y)A]N22 (in polytene region 30C, ated withrecessive lethality were assigned to lethal 2-34 onthe meiotic map), P[(w,y)A] 1-1 (47A, 2-60), complementation groups by complementation tests (within P[(~,y)A-R]012(66E,3-17),P[~]21(89A,3-58),P[(~,~)A]4- each linkage group). Inthose cases where map position and/ 4 (100F, 3-108),all from (HAZELRIGGet al. 1984);P[ (w,ly)Fl4- or mutant phenotype suggested that a particular enhancer 2 (21D, 2-0), P[(~,y)Dll(25C, 2-15), P[(w,?)q4-1 (57B, 2- of gl might be an allele of a known gene, we tested this by 95), P[(w,y)Wl (78CD, 3-46), P[(w,y)D]3 (90E, 3-62) and complementation. P[(w,y)El7 (94D, 3-81),from R. LEVIS and G. M. RUBIN (un- Mapping: We determined the approximate meiotic map published results). The markerstocks for mosaic clone analy- position of one allele of each previously unidentified comple- ses were as follows: wl”*; P[(w,y)A]N22 (for mutations on mentation group obtained in the screens. We determined 2L), w’”*; P[(w,y)A]l-l (for mutations on ZR), w’”~; P[gfl- the position of both the dominant E(g1) phenotype and the 30B; P[(w,y)W 1 (for mutations on X), w’”~; P[gfl30B; recessive lethality associated with it (when applicable) in inde- P[w] 21 (for mutations on X).P[gfl30B is a gl+ transgenic pendent experiments. We mapped the positions of the domi- insertion in polytene region 30B that fully rescues gl mu- nant enhancers of gl by scoring the enhancer and eye color tant’s phenotype (MOSES et al. 1989). gl deletion chromo- phenotypes, of the pro eny of the outcross of heterozygous somes were Df(3R)glBm, Df(3R)glBx7 and Df(3R)glBx8 (MOSES females (genotype: w1’I5; E(gl)/E(gl)+ P[w’], in a g13 homozy- et al. 1989). The balancer chromosomesused wereIn(2L40, gous background) crossed to males of the wl”*; g13 stock. We TM3 Sb and TM6B Tb Hu (LINDSLEYand ZIMM 1992). A hc’ mapped the positions of the recessive lethality associated with stock was a gift of N. L. BROWNand S. B. CARROLL.A hhAC most of the dominant enhancers of gl by back-crossing F1 stock was a gift of J. J. LEE and P. A. BEACHY.A pdNstock heterozygous females (genotype: wl”*; E(gl)/E(gl)’ P[w’], in was a gift of J. E. HOOPER.Stocks containing m5 and mzo a g1’ homozygous background) to males carrying a mutation were gifts of R. GRIFFIN-SHEA.Stocks carrying the following in the same lethal E(gZ) complementation group (using a dif- mutations were obtained from the Bloomin ton and Bowl- ferent allele, when available) of the genotype w’”’; E(gl)/ ing Green stock centers: armx2, ciD, Dlx, en’“, eveiDl9, jiz”’, balancer and then scoring the eye color phenotypes of the f~”~~,Df(2R)gsb, gt”’, h5H07,hb14F, kni5F’07, KT2, nkd7E,opa5~97, nonbalancer progeny. The scale of our mappingexperiments roe’, roe3, roe4, runE9,SI, sip"", ~mo@~,til'.'' and wgcx4. and the distribution of the white’ markers were such that the Mutant screens: For chemical and radiation mutagenesis, resulting positions are accurate to only 25 cM. males of the genotype ,I1’*; P[(w,y)D]3 gl’ e were treated Mosaic clone analpix Mutants were crossed to the marker with 25 mM EMS (Sigma #M0880) as described in ASHBURNER stocks described above, as appropriate to their map positions. (1989) or y-rays (4000 rads from a I3’Cs source) and crossed To induce mitotic recombination, heterozygous Fl progeny to w””; gl’ virgin females. About 395,000 F, progeny were in the early first instar were treated with y-rays from a I3’Cs inspected under dissecting microscope for modified eye phe- source (1000 Rad) or X-rays (1000 Rad at 120 kv, 5 mA). notypes. Mutations of the white gene on the P[ (w,?)D] 3 ele- Adults were then screened for mosaic clonesin the eye, ment were also scored as a monitor of the efficiency of muta- marked by a lack of retinal pigment. Heads of such flies were genesis. As asecond method to monitor the mutagenesis fixed and embedded in resin, and 3-pm retinal sections were efficiency, for each mutagenesis run we measured the fre- cut as described by TOMLINSONand Rwny (1987a). Thesec- quency of induced X-linked lethals, by crossing mutagenized tions were examined and photographedusing phase contrast males en masse to FM6K y w B virgin females. The FI progeny lenses. were allowed to mate inter se, and then 100 single females Electron microscopy, cuticle preparation, and histochem- were bred in individual vials. We scored for the presence or istry: Scanningelectron microscopy was performed as de- absence of B+ males in the F2 progeny of each female. Each scribed (MOSESet al. 1989).Embryonic cuticles were prepared case of the absence of such males was scored as one X-linked as described in WIESCHAUSand NUSSLEIN-VOLHARD(1986). lethal. For Pelement transposon mutagenesis, Bim-2/+; Ki Embryos were stained as described by DEQUINet al. (1984). A2-3 y 506/gl’males were crossed to w””; gl’virgin females. The antibody used for embryonic central nervous system stain- The A2-3 element is a stably integrated source of P-element ing wasmAb BP102, gift of C. Goonm (HORTSCHet al. transposase (at polytene band 99B) and Bim-2 is a second 1990). The antibody used for embryonic peripheral nervous chromosome isolated from the Birmingham (Bim) stock of D. system staining was mAb 22C10, gift of S. L. ZIPUEKY (FUJITA melanogaster, which has a largenumber of defective Pelements et al. 1982). Eye imaginal discs were stained as described by and no completeones. The A2-3; Bim-%combined genotype TOMLINSONand READY (1987b) as modified by h4~et al. results in a relatively high level of Pelement transposition, (1993). Antibodies used were anti-Elav, from the University which can then be stabilized by selecting against Ki (and thus of Iowa developmental biology hybridoma bank (BIERet al. A2-3) in the next generation (LASKI et al. 1986; ROBERTSON 1988), anti-Hairy, gift of S. CARROLL(CARROLL and WHV~E et al. 1988). We scored -60,000 F1 Ki+ progeny for modified 1989), anti-Scabrous, gift of N. E. BAKER(MLonzIK et al. eye phenotypes. The mutants obtainedwere crossed to w””; 1990a), and anti-Dpp, gift of F. M. HOFFMANN(PANGANIBAN gl’ flies. If mutations from any of the screens bred true, they et al. 1990). were crossed tothe following balancer stocks, In(ZLR)O/ Genomic DNA blots: DNA was isolated from adults as de- Sco, gl’ and wfll*; TM3 Sb e glBz/TM6B Tb Hu e glET1,and scribed (MOSESet al. 1989). The DNA was digested with six maintained as balanced stocks. We also tested 25 known au- restrictionendonucleases, BamHI, EcoRI, HindIII, PstI,Sac1 1202 C. Ma et al.

FIGUREl."External retinal morphologies of E(g1) mutants. All panels are SEMs of adult retinas. Anterior is to the right and dorsal is up. All panels are to the same scale, bar in A, 100 prn. (A) Wild type. (B) gl' homozygote. (C) g1"O homozygote, a null allele. D-F are examples of E(g1) heterozygotes in a g13 homozygous background. (D) E(gl)W, a weak enhancer, (E) h""', an intermediate strength enhancer, (F) E(gl)3C, a strong enhancer. and XhoI, then run on agarose gels and transferred to nitrocel- meiotic map position of each E(gl) complementation lulose. Blots were hybridized to R2Pnick-translated probes cov- group was determined relative to several transgenic p ering the genomic region of the hh gene from a Hind111 site elements carrying a white+ gene (see Tables 1 and 2, at -5.5 kb to the EcoRI site at +16.4 kb, position numbers as LEE et al. ( 1992). Figure 2 and MATERIALSMETHODS). AND Mutations in genes that are not functionally related RESULTS to gl may be recovered in this screen, and secondary testswere required to eliminate such falsepositives. Mutant Screens: The compound eyes of g1" homozy- Such extraneous genes can be recovered through the gotes have a phenotype that is intermediate between neomorphic or antimorphic mutation of genes with that of wild type and that of a @ null (compare Figure products that do not normally act in the gl pathway. 1B with 1A and IC). We recovered 76 dominant en- These mutations are inherently rare and are usually hancers of gl' in a total of --455,000 FI flies: 49 EMS recovered as the sole allele in a complementation induced mutants (in -282,000), 24 y-ray-induced mu- group. Indeed, thesingle haiq allele appears to be anti- tants (in -113,000), and three dysgenic mutants (in morphic (see below). All E(g1) mutations werecom- -60,000) (see MATERIALS AND METHODS). No dominant pared to knownloss-of-function alleles of the same suppressors were found. Seventy of the mutations are gene, whenever possible. Also haplo-insufficient, domi- homozygous lethal when cultured at 25". One mutation nant rough eye mutations (such as Star) may have a was found to be temperature sensitive, and thehomozy- simple additive effect with the rough eye mutation gl" gotes are viable at 18". Pair-wise complementation tests and thus appearin our screen. To test for such nonspe- were used to assign the E(@) mutations to 23 comple- cific interactions, members of each complementation mentation groups. Figure 1 shows external views of the group were crossed to wild type (Canton-S), and their dominant modified-eye phenotypes of one weak, one progeny were examined for dominant eye phenotypes intermediate and one strong E(@) complementation in this gl+/gl" background. Mutations in seven of the group (shown in a gl' homozygous background). The E(g1) loci show this phenotype and thus are notconsid- Enhancers of glass 1203

TABLE 1 Pelement insertion markers

A bbreviation Polytene Map Reference" MapName Polytene Abbreviation

P[(w,?)fil4-2 2-1 21D 2-0 1 P[(w,ly)DIl 2-11 25c 2-1 5 1 Pt(w,ly)AlN22 2-111 30C 2-34 2 P[( w,q)AI 1-1 2-Iv 47A 2-60 2 P[(w,ly)d4-1 2-v 57B 2-95 1 P[(w,ly)A- 3-1 66E 3-17 2 P[(w,ly)W 1 3-11 78CD 3-46 1 plwl21 3-111 89A 3-58 2 P[(w,ly)D13 3-Iv 90E 3-62 1 P[(w,ly)fl7 3-v 94D 3-8 1 1 P[( w, ly)AI 4-4 3-VI l00F 3-108 2 1, R. LEVISand G. M. RUBIN,unpublished results; 2, HAZELRICCet al. (1984). ered to be specific enhancers of gl (see "Specific En- ration and by immunohistochemical staining of the ner- hancer" column in Table 2). vous system (see below). The gl' mutation was spontaneously induced (CSIK Characterization of identified loci: Six complemen- 1929) and is associated with an insertion of 2.5 kb near tation groups were identified based on their mutant the 5' end of the gene (which may be a transposable phenotypes, map positions, and complementation tests element) (MOSES et al. 1989). It may be possible for a with mutations of known genes. They are Delta, glass, second-site enhancer mutation to identify a gene that ha@, hedgehog, roughened eye and Star (Table 2). interacts with g13 through regulatory effects on the in- Delta (DZ): Dl encodesa protein with transmem- sertion, and that itself does not have any normal func- brane domain and EGF-like repeats (KOP~ZYNSKIet al. tion in eye development. An example of such an inter- 1988;ALTON et al. 1989). In theeye, ommatidia formed action is that between a gypsyinduced allele of Hairy by cells homozygous for weak Dl loss-of-function alleles wing and mutationsat the suppressor of Hairy wing contain an increased number of rhabdomere-bearing [su(Hw)]locus (PARKHURSTand CORCES1986; PARKH- cells (which are likely to be photoreceptor cells) (DIE- URST et al. 1988; HOOVERet al. 1992; HARRISON et al. TRICH and CAMPOS-ORTEGA1984), and Dl plays a role 1993; SHENet al. 1994; SMITHand CORCES1995). To (with Notch and scabrous) in spacing the array of omma- rule out such allele-specific interactions, members of tidial preclusters in the morphogenetic furrow (BAKER each complementation groupwere crossed to two other and ZITRON 1995). weak g1 alleles: glHX9(induced by X-rays) and glDZ4(in- gZms (gZ): glS""' is associated with a lethal deletion duced by EMS) and that areunlikely to harbor thesame between polytene bands 90F and 91F (data not shown). transposon as g13. By this test we found that noneof the gl is located in 91A1-2, and glSCG'fails to complement mutations discovered in our screen interactsspecifically the lethality associated with three extant lethal gl dele- with the g13 allele alone. As the homozygous hypomor- tions: Df(3R)gloX3,Df(3R)glnX7, and Df(3R)gl"Xx(MOSES phic phenotype of g13 is stronger when placed in trans et al. 1989). to a null allele, we could (and did) recover a new gl hairy (h): Homozygotes of h"':';' show a typical h cu- mutation in this screen (see below). ticular pair-rule phenotype (NUSSLEIN-VOLHARDand The most direct test of the function of a gene in eye WIESCHAUS1980; INCHAMet nl. 1985) and fail to com- development is to examine thehomozygous phenotype plement thelethality associated with two known h alleles of null mutations of that gene alone, in an otherwise (h5H"7and h"'). Ommatidia that are homozygous for wild-type genetic background. For homozygous lethal h"""' (in retinal mosaic clones, Figure 3G) are often mutations, this can only be done through the examina- missing rhabdomere-bearing cells (20% of such omma- tion of homozygous somatic mosaic clones. To examine tidia). h encodes abasic helix-loop-helix (bHLH) family the retinal phenotypesof the recessive lethal mutations transcription factor (MURREet al. 1989; RUSHLOWet al. (in a glf genetic background), we made retinal mosaic 1989). H protein is normally expressed anterior to the clones of members of each complementation group morphogenetic furrow in the developing eye, and ec- that were marked with white and examined them in topic expression of h can alter retinal cell fates (CAR- adults by sectioning and microscopy (see MATERIALS ROLL and WHY~E 1989).However, homozygous h retinal AND METHODS). Of the 18 genes that can mutate to mosaic clones show no mutant phenotype (BROWNet recessive lethality, 13 showed retinal phenotypes in this al. 1991), and it is likely that H acts redundantly in test (see below and Table 2). We also examined the the developing eye to regulate other bHLH proteins embryonic phenotypes of members of each recessive (BROWNet al. 1995; HEBERLEIN andMOSES 1995). We lethal complementationgroup by embryo cuticle prepa- showed that a h null mutation is not a dominant en- 1204 C. Ma et al.

TABLE 2 Summary of the Enhancer-of+s mutations

Distance from Distance from Map leftflanking rightflanking Specific G ene Allele(s)" positionLAllele(s)" Gene W' Notemarker' Lethal enhancer w' marker'

Dl SCI 3-66 No Yes gl SCGl 3-63 Yes Yes h SCGl 3-26 Yes Yes Antimorph hh SCI-SC21 581 Yes Yes One viable allele, one temperature-sensitive SCGl-SCG5 roe SCI 3-47 Yes No S scI-sc12 2-1.3 No Yes SCGlSCGl2 SCPI-scP3 E(gO 2A SCI 2-24 2-11, 6 cM 2-111, 10 cM Yes Yes No obvious retinal" or embryonic' defects. (16/271) (22/212) w SCI 2-50 2-111,21 cM 2-N, 10 cM No Yes No obvious embryonic' defects. Honleotic (W87) (9/93) phenotypes in some retinal" clones (Figure 7). Other retinal clones contain ommatidia with both increased and decreased numbers of rhabdomeres (average number 6.8, Figure 3B) E(gl)2B SCl 265 2-V, 25 cM Yes Yes No obvious retinal" or embryonic' defects. (75/305) E(gO2C SCI 2-65 2-V, 33 cM Yes Yes No obvious embryonic" defects. Retinal (15/45) clones" contain ommatidia with both increased and decreased numbers of rhabdomeres (average number 8.1, Figure X:) 2-70 2-N, 11 cM 2-V, 38 cM Yes Yes No obvious embryonic' defects. Retinal (46/408) (76/200) clones' contain ommatidia with both increased and decreased numbers of rhabdomeres (average number 7.9). Alsu 67% of the ommatidia are miss-orirntrd Figure 3D) SCI 2-70 2-N, 10 cM 2-V, 23 cM Yes Yes No obkious retinal'' or embryonic-' delects. (18/174) (44/190) SCI 2-70 2-N, 11 cM 2-V, 30 cM Yes Yes No obvious embryonic' defects. Retinal (25/227) (80,268) cloned contain rare ommatidia with decreased numbers of rhabdomeres (average number 7.9, Figure 3E) SCI 2-70 2-N, 11 cM 2-V, 31 cM No Yes No obvions retinal" or embryonic' deferw. (15/141) (80/255) SCI 2-97/ 2-N, 40 cM 2-V, 2 CM Yes No No obvious retinal" defects. (721'182) (4/189) SCI 3-0' 3-1, 26 cM 3-11, 44 cM No Yes Embryonic nervous system defects, (62/239) (131,298) dominant retinal" defects (see text). SCI, sc2 3-42 3-1, 15 cM 3-11, 4 cM Yes Yes No obvious embryonic' defects, required by (44/258) (14/320) all retinald cells (see text). SC1 3-43 3-11, 43 cM 3-11, 3 cM Yes No No obvious retinal'' defects. (74/173) (10/333) SCGl 345 311, 1 cM 3-111, 14 cM No Yes No obvious embryonic' defects, dominant (1/166) (32/229) retinal'' defects (see text). SCGl 346 3-11, 0 cM 3-111, 11 cM No Yes No obvious embryonic' defects, dominant (0/W (4/38) retinald defects (see text). SCGl 3-81 3-V, 0 cM %VI, 2 cM Yes Yes No obvious embryonic' defects. Retinal (0/69) (3/185) clones" contain rare ommatidia with decreased numbers of rhabdomeres (average number 7.9, Figure 31.) E(g0 3c SCGl 3-99 3-V, 19 cM %VI, 24 cM Yes No Homozygous adult retinas lack 2% of the (50/267) (52/213) ommatidia, and the surviving ommatidia occasionally have both increased and decreased numbers of rhabdomeres (average number 6.5, Fignre 30). E( go 30 SCl, SCGl 3-108 Yes Yes Embryonic nervous system defect9 (Figure 8). No obvious retinal" defects. EMS alleles are listed as SC numbers. y-ray induced alleles are listed as SCG numbers. Hybrid-dysgenesisinduced alleles are listed as SCP numbers. 'The map positions of known genes are taken as published in LINDSLEYand ZIMM (1992). 'In each case, the abbreviated name of the marker is given first, then the estimated map distance to it in cM, and then the number of recombinants per total number in parentheses. The retinal phenotypes of lethal genes were examined in adults in negatively marked mosaic clones, see MATERIALS ANI) METHODS. e The embryonic cuticular and nervous system phenotypes of homozygous embryos were examined, see MATERIAIS AND METH- ODS. 'Gene is distal to the outer marker, thus the two closest proximal markers were used. Enhancers of glass 1205

SC2 SCG4 SC3 SCGS SC4 SCG6 SC5 SCG7 SC6SCG8 E(g02D SC7 SCGS SC8 SCGIO SC9 SCGl1 E02E SClO SCGl2 SCll SCPl E(gl)2B E(gl)2F I scii SCP3 E(ql)2A Opl E(gl)2C E(gl)2G E(gl)2H FIGURE 2.-Geneticmap of E(gZ) mutations andmapping markers. The two heavylines 4 4 4 4 marked 2 and 3 represent the second and third 2-111 2-IV 2-v 2-1 2-11 2-1 chromosomes of D.melanogaster. Below each chromosome are the positions of the transgenic w+ mapping insertions used (see Table1 and MATERI- ALS AND METHODS). Above each chromosome are shown the positions and names of all autosomal . I hh E(gZ) mutations recovered (see Table 2). Genes with multiple alleles are boxed.

-+ 4 3-1 3-11 3-1 3-V I 3-IV 3-VI 3-111 hancer of gl' (by examining the genotype h"'/h+; g13/ etic furrow, and that it is required for the progression gl'), and it appears that hSc"' is a rare antimorphic of the furrow (MA et al. 1993), where it appears to act allele ( hs""' homozygous embryos have a more extreme throughthe TGFP homologue decapentaplegzc (dpP). phenotypethan that of a null mutation) (NADEAN These results are consistent with the effects of hh in BROWNand SEAN CARROLL,personal communication). retinal mosaic clones (HEBERLEINet al. 1993b; MA et al. hedgehog (hh): We recovered 26 alleles of hh as domi- 1993), and ectopic Hh expression anterior to the fur- nant enhancers of gl' (21 induced byEMS, and five row is sufficient for furrow initiation (HEBERLEINand induced by y-rays, see Table 2), as confirmed by their MOSES 1995; HEBERLEINet al. 1995). embryonic segment polarity phenotype, map position Another of the y-induced hh alleles ( hhf""" enr'r or and failure to complement thelethality associated with hhf", is homozygous viable with no embryonic defects. a known hh null allele, hhA" (LEEet al. 1992). hh is a Homozygous hhf" animals have a kidney-shaped eye that zygotic segment polarity gene (N~SSLEIN-VOLHARD andis very similar to the first hh mutation discovered, bar- WIESCHAUS1980), whichhas nonautonomous effects 3 (now known as hh') . We examined developing eye on neighboring cells in the developing embryo and imaginal discs from hhf" homozygous (Figure 4) and elsewhere (MOHLER1988). hh encodes a secreted and found that the phenotypeis indistinguishable from that cleaved protein, which comprises an extracellular signal of bar-3 the furrow stops early (HEBERLEINet al. 1993b). (LEE et al. 1992; MOHLERand VANI1992; TABATAet al. The bar-3 mutation is associated with a 1.7-kb deletion 1992; TASHIROet al. 1993; LEE et al. 1994;BUMCROT in the first intron of the hh transcription unit (LEEet et al. 1995; MARTI et al. 1995; PORTERet al. 1995). hh al. 1992). We examined genomic DNA from hhfi' and homologues have been found to act in several aspects compared it to its progenitor chromosome and to bar- of vertebrate development, including the eye (SMITH 3 by genomic DNA gel blot. Four of six restriction en- 1994; CONCORDETand INGHAM1995; EKKERet al. 1995). zymes used reveal the presence of a new 0.8- to l-kb One of the y-induced hh alleles ( hhts2)is temperature deletion associated with hhfie (data not shown). This sensitive. The coding sequence of hh" has been deter- new deletion overlaps with that of bar-3 (Figure 5). mined, and it shows a replacement of seven amino acid roughmd eye (roe): roesc' homozygotes are viable and residues by five others (PORTERet al. 1995). We have show defects in the eyes, wings, and legs. The eyes are found that hh is expressed posterior to the morphogen- rough and reduced in size (Figure 6A) with variable and 1206 C. Ma ~t nl.

FK;L,KI-X--Ketin;d sections of E(g/) mutants in a gI’/gI’ genetic background. All panels arc shown as the dorsal half of the right eye. Anterior is to the right and dorsal is up. All panels are to the same scale, bar in A, 5 pm. (A) Wild type. B-L are mosaic clones that are homozygous for I?(@) mutations and are marked with u-. (B) @l, (C)E(g1)2C, (D)E(g1)2D, (E) E(gl)2F, (F) Pas, (G) hSf.’”’, (H) dogSf:‘, (I) Ron, anterior clone, (J) Ron, posterior clone, (K) Onh, (L) E(g1)3B. (M) Ron heterozygote (anterior region of the eye), (N) Onh heterozygote, (0)E(g1)3C homozygote. Arrowheads show abnormally oriented ommatidia, open arrows show ommatidia with extra rhabdomere-bearing cells, arrows show ommatidia with reduced numbers of rhabdomere- bearing cells, white asterisks show the positions of missing ommatidia. reduced numbers of photoreceptors in almost all omma- adjacent to rn (ACNELet al. 1989). We found that roe’, tidia (Figure6B). The wings are short, with the distal part m3,and m4fail to complement the small and rough eye deleted (compare Figure 6, C and D), and the legs are phenotype of rot””. Therefore, roe.“,“ lesioninactivates also short due to the fusion of tarsal segments (compare both rn and roe. The m mutation is responsible for the Figure 6, E and F). These phenotypes are similar to those enhancement of g1 phenotype, since while m’,roe3, roe4 of some mutations in d# (SPENCERd al. 1982) and &ch- and the deletion of the region (rn’”)are all dominant hund (dac) (MARDON ~t (11. 1994), which are known to enhancers of the phenotype of gd’, the strong rn allele affect the morphogenetic furrow (HERERLEIN MOSESand (rn’) is not. Thus roe and rn show complex complementa- 1995). m mutations have also been recovered as domi- tion, and it is possible that they are two classes ofmutation nant enhancers of W”‘ (BRANDand CAMPOSORTECA of the same gene, each of themdisrupting a subfunction: 1990). roes(:’ maps close to the rotund (rn) and m loci class I represented by m’,roe’ and m“and class I1 repre- (KERRIDGE and THOMAS-CAVAL.I.IN1988;ACNEL et al. sented by rn’, and several more alleles in these classes 1989, 1992). Complementation tests were carried out be- exist (LINDSLEYand ZIMM1992). msc’and the rn2‘’dele- tween me,”:’ and rn5 (a strong rotund mutation), and rn2” tion disrupt both classes. There are severaltranscripts (a deletion in the rotund region that also includesm) (R. from this genomic location (AGNELet al. 1992). However, GRIFFINSHFA, personal communication). r0e,”,”/rn5flies some roe and rn alleles are separable by meiotic recombi- show the same wing and leg defect as described above. nation, and we found 0.023% recombinants between roe’ ms(:l/rn 20 flies show small and rough eye phenotype in and rn5 (1/4267 recombinant/total progeny). addition to the wing and leg defect. This eye phenotype Star (S): S mutations are haplo-insufficient domi- is the same as that caused by m mutations, which are nants with a rough eye phenotype (RENFRANZ and BE- Enhancers of glass 1207

FIGURE 4.--hhf" mutant phenotypes. All panels are third instar eye-imaginal discs. Anterior is to the right and dorsal is up. All panels are to the same scale, bar in A, 100 pm. V's show the position of the furrow. A-C are wild type. D-F are hhPe homozygotes. A and D are stained to show the expression of Elav (black) and Hairy (brown) (see MATWSAND METHODS). B and E are stained to show Elav (black) and Scabrous (brown) (see MATERIALS AND METHODS). C and F are stained to show Dpp (see MATERLALS AND METHODS). Note, in hhrscmutants, overly mature clusters are seen at the furrow, and Scabrous and Dpp expression are repressed.

NZER 1989) and are recessive lethal, due to an essential mentation groups, and of these, 15 are represented by embryonic function in ventral midline development. only one allele (Table 2 and Figure 2). Nine of them S is one of six genes in the ''spitz" group (NUSSLEIN- are on the second chromosome, and eight are on the VOLW et al. 1984; MAYER and N~SSLEIN-VOLHARDthird (none lie on the fourth). We have named some 1988), and S encodes a possible transmembrane pro- of them according to their mutant phenotype and the tein, required for the prospective photoreceptor cells rest are identified only by the generic E(g1) with the to differentiateas neurons (HEBERLEIN and RUBIN1991; chromosome number and a letter as a specific suffix HEBEFUEINet al. 1993a; KOLODKINet al. 1994). (e.g., E(gZ)2A) (see Table 2). Characterization of unidentified loci: We were un- Opthulm@ediaZike (Opl): We examined homozygous able to assign mutations at 17 loci to known comple- OpZ tissue in negatively marked (w-) retinal mosaics, and in some cases(3/15 clones in one experiment) the bb-1 (1.7 kb deletion) mutant retinal tissue was transformed into an append- -m bb-fse (0.8 - 1.O kb deletion) age-like extrusion (Figure 7). We cut thin sections of

I I I I I five mosaicretinal clones, in whichno such transforma- -5 kb 5kb OW 10 kb 15 kb tion was seen. Of these, two show no obvious defects, 5' 3' and in three -70% of Oplhomozygous ommatidia (32/ 46 examined) show reduced numbers of photoreceptor FIGURE 5.-Map of the hh genomic locus. The heavy line cell rhabdomeres, and one ommatidium (in 46) had represents genomic DNA, coordinates in kb, and the structure of the transcript are after LEE et al. (1992). The deletions an extra cell (an overall average of 6.8 rhabdomeres, associated with bar-3 and hhrseare shown. End-point uncer- Figure 3B). tainty is shown by the open box for hh? Partial seuenless (Pas): Pas homozygotes die as late 1208 C. Ma et al.

FIGURE6.-roe phenotypes. A is an SEM of the adult retina of roesc’/m2” (n”is a deletion of the region). Anterior is to the right, dorsal is up. Bar, 100 pm. B is a section of an eye of the same genotype as A. Anterior is to the right, dorsal is up. Bar, 5 pm. Note the reduced numbers of rhabdomere-bearing cells. C and D are whole mounts of wings. Bar, 500 pm. E and F are whole mounts of first male legs. Bar, 100 Dm. Ta, tarsal segments;- V, sex combs. C and E are wild type, and D and F are roesc’ homozygotes. Note the shortevned wing and fused tarsi. embryos, with defects in their nervous systems (Figure dog is recessive cell-lethal. In the center of the mosaic 8, E and F).In a gl+/gl’genetic background, Pas shows tissue, there are noommatidia and along theperiphery a dominant effect: -40% (123/321) of the ommatidia there are abnormal ommatidia with reduced numbers in a Pas heterozygote lack the apical central photore- of rhabdomere-bearing cells. We scored 189 rhabdom- ceptor rhabdomere, appearing to be “sevenless”, and ere-bearing cells in 50 such ommatidia and found that often also one or more outercells (average is 6.1 rhab- all the rhabdomere-bearing cells carry pigment gran- domere-bearing cells per ommatidium). We examined ules, marking them as dog+. Homozygous dog retinal homozygous Pas tissue in three negatively marked (w-) clones are associated with vacuoles in the underlying retinal mosaics, and we found that theaverage number lamina region of the optic lobe of the brain (data not of rhabdomere-bearing cells per ommatidium is also shown). 6.1 (Figure 3F). Rough anteriorly (Roa): Roa heterozygotes show a dog of gluss (dog): dog heterozygous flies are sick dominant retinal mutant phenotype (in a gl+/gl’back- (sic.), and heterozygotes have much shorter life span ground): the anteriorhalf of the eye isrough, flattened than that of wild-type flies. We examined homozygous, and devoid of bristles. Sections of the mutant region negatively marked (w-) retinal mosaic clones for both show that there are only sparse ommatidia isolated by dog.qcl and dogs“ (six and two clones, respectively), and massive pigment cells (Figure 3M). Forty percent of we found thatdogis absolutely required for thedevelop the remaining ommatidia contain reduced numbersof ment of all the retinal cells, as no w- photoreceptor or rhabdomere-bearing cells (average number is 4.3). Be- pigment cells remain (Figure 3H). It is thus likely that cause the anterior partof the eye in a Roa heterozygote is mutant, we analyzed Roa homozygous clones in the anterior andposterior parts of the eye separately. Ante- rior hahomozygous clones have a phenotype that is very similar to that of the heterozygous tissue (Figure 31). Posterior homozygous Roa clones show no mutant phenotype (Figure 35). Orientation abnormal (Oab): Oab heterozygotes show a dominantretinal mutant phenotype (in gZ+/gl’back- a ground): a slightly rough eye. We cut sections of these eyes, and we found that the orientationsof -23% (56/ 245) of the mutantommatidia are randomized and the normally regular array of ommatidia is disrupted (Fig- ure 3N). We also examined two negatively marked (w) retinal mosaics, and we found similar defects: -30% (15/49) of the mutant ommatidia showed an orienta- FIGURE7.-OpL mosaic clones transform retinal tissue to tion defect (Figure 3K). another fate. A and B are SEMs of two examples of homozygous mosaic clones of 91 in a heterozygous background. Tests of mutations in candidate genes as dominant Anterior is to the right and dorsal is up. Bar, 100 pm. Note modifiers of gZ: As mutations of two segmentation bizarre objects growing from the clones. genes (hedgehog and ha@) were recovered as dominant Enhancers of glass 1209

A n

IE F

FIGURE8.-Embryonic phenotypes of two E(g1) mutations. Late embryos are shown; all panels are to the same scale. Bar in A, 100 pm. Anterior is to the left. A, C and E are ventral views of embryos stained with mAb BP102 to show the central nervous system (CNS) (see MATERIAIS AND METHODS). B, D and F are lateral views, ventral down, of embryos stained with mAb 22C10 to show the peripheral nervous system (PNS). A and B are wild type. C and D are E(gl)3D homozygotes. Note the commisural failures in the CNS and the neural hypertrophy in the PNS. E and F are Pas homozygotes. Note the commisural failures in the " CNS and the neural disruption in the-PNS. enhancers of g1, we tested 25 known autosomal genes tect genes that act to regulate gl activity. We recovered (see MATERIALS AND METHODS) for this genetic interac- 76 mutations in 23 genes, as well as nine alleles of Notch tion. We constructed genotypes that are heterozygous (N, which were not retained), and these fall broadly for each of the 25 individual segmentation mutations into two groups: genes for which a large number of and homozygous for gl'. None of them was a dominant alleles were recovered (N, Sand hh) and genesat which modifier of gl'. However, we found that although ptc"" only one or two were found (all the rest). Enough is mutation does not modify the gl' homozygous pheno- known of the molecular biology of N, S and hh to state type, it does suppress the rough eye phenotype caused that while these genes clearly play crucial roles in eye by the hh"' mutation, which is a dominant enhancerof development, they do not interact with gl in a direct gl (MA et al. 1993; MA and MOSES 1995). In the course way. N, Sand hh are all involved in the function of the of the EMS screen, we recovered nine alleles of Notch morphogenetic furrow. N is required for the correct (N),which werenot retained.We constructed thegeno- spacing of neurons in the furrow and later functions types N""/N+;g13/g13 and N''/N+; g13/g13 and found (CAGAN and READY 1989; BAKERand ZITRON 1995). S that loss-of-function N alleles are dominant enhancers is required for the correct recruitment of those photo- of gl. While we do not know if null alleles of Dl are receptor cells that immediately follow the founding R8 enhancers of g1, this appears unlikely, as we only recov- cell (HEBERLEIN and RUBIN 1991) and islikely to be ered one Dl allele in this screen. involved in the EGF receptor pathway (KOLODKIN et al. 1994). hh encodes a diffusible signal that is produced DISCUSSION in differentiating photoreceptor cells behind the furrow and acts to induce more anteriorcells to enter the In this screen we set out to isolate autosomal second- furrow (HEBERLEIN et al. 1993b, 1995; MA et al. 1993; site dominant modifiers of the phenotype of a weak @ HEBERLEINand MOSES 1995). Thus all three of these allele. Such a screen might have been expected to de- genes are required for the normal specification and 1210 C. Ma et al. differentiation of photoreceptor cells, and in that (indi- 0 Number of mutations / rect) sense, they act upstream of gZ. It is worth noting that a screen for genes that interactwith N in the eye T recovered alleles of gZ as well as other loci identified in 80 this screen (roe and DZ) (BRAND andCAMPOS-OKTEGA 70 1990). For the most other loci detected in this screen, we do notyet have sufficient information to determine 60 their biochemical relationship to gZ. 50 It is thought that the photoreceptorcell specific activity of gZ is due to a negative factor that binds to a DNA 40 sequence adjacentto that of the G1 protein-binding site 30 (ELLISet al. 1993). If this is the case, the negative factor itself must be either be expressed only in the nonpho- 20 toreceptor cells, or it must be downregulated in the 10 photoreceptor cells as they differentiate. Because that factor has not been identified yet, these possibilities remain open. One would expect loss-of-function muta- 0 100 200 300 400 tions in this gene to be dominantsuppressor of gZ. How- Flies screened (in thousands) ever, we did not recover any dominant suppressor muta- FIGURE9.-Graph showing screen saturation. Abscissa is tions. It ispossible thatone of the single-hit E(gZ) the number of F, progeny screened. Ordinate is the number mutations is a gain-f-function mutation of such a gene. of genes (0)and mutations (0).Left of the vertical dashed line, the mutagen was EMS. Right of the vertical dashed line, Why were mutations at 21 loci detected at a 10-fold the mutagen was y-rays. Hybriddysgenic screen data are not lower frequency than N, S and hh? It could simply be shown. Two lines were fitted by eye: the solid line through that thelevel of saturation inthis screen was insufficient the open circles and the dotted line through the filled circles. to detect many genes. This is unlikely, based on three Note that these two lines intersect (see text). criteria: the number of control mutations recovered in the white gene, the numberof recessive lethals induced can yield therequired phenotypic effect (this class on the Xchromosome in control experiments, and the could include a vast number of genes). This may be curves seen when the number of loci and the number true in many cases for the 22 one- or two-hit loci we of mutations are plotted against the number of flies recovered. Indeed, the oneallele of h appears to behave screened. In this screen the mutagenized males’ third as an antimorph. chromosomes carried a transgenic copy of white, and as In some cases three other factors may produce apau- they were crossed to white mutant females, new white city of mutations. Mutations at some loci are subviable, alleles could easily be detected in the F, (and tested for even as heterozygotes. In this screen both dogmutations true-breeding in the F1 males). In the course of these are subviable asheterozygotes. Thus many potential dog screens, we recovered 70 true-breeding white alleles. mutant F1 flies may simply not have survived to be de- The white gene is not known for hypermutability with tected in the screens. Mutations at some loci have only EMS or radiation, and thus this large number argues very weak effects (low expressivity), for example E(gZ)2D strongly for a high degree of saturation. The average (Figure 1D). Such mutations are difficult to detect (by frequency ofX-linked lethalmutations in the EMS definition) and may often be missed. Similarly, some screens was 14%, and in the y-ray screens it was 5%. loci may produce variable phenotypes (low penetrance) The total number of flies inspected in those screens and thus can also be missed in the F1. were 282,000 and 113,000, respectively. Assuming that Several of the E(gZ) mutations have intriguing pheno- there are -1000 vital genes on the Xchromosome, the types. The subviability of dog heterozygotes may suggest average number of mutations per locus was 39 in the thatthe gene encodes a dose-sensitive function re- EMS screens and six in the y-ray screens. Plotting the quired forcell viability.The specificity of its interaction number of mutations recovered in the screens against with gZ might suggest that this function is associated the number of F1 flies screened (open circles and solid withzinc finger transcription factors. The homeotic line in Figure 9) yields a constantly increasing linear transformation seen in the homozygous OpZclones may relationship. When we plot thenumber of genes against be toward wing fates. This may correlate to the observa- the number of flies screened (filled circles and dotted tion from disc transplantationexperiments, that eye line in Figure 9), a much lower slope is seen and the imaginal disc tissue can most easily transform to wing lines cross. Perhaps the simplest explanation of these (WORN1968). Further studies of these, and other data are that there are two classes of potential targets loci from these screens, may produce new insights into in this (and perhaps any) genetic screen: those loci at Drosophila retinal development in the future. which a simple loss-of-function mutation will produce We thank G. LANDIS for help with the screen and N. E. BAKER, the required phenotype (in this case N, Sand hh), and N. L. BROWN,S. CARROI.I., C.GOODMAN, K. GRIFFIN-SHEA,J. E. those loci at which only a rare morecomplex mutation HOOPER,F. M. HOFFMANN,T. L.AW.RTY, G. M. KUBINand S. L. ZIPUR- Enhancers of glass 1211

SKY for materials and reagents. This research was supported by Na- CARTHEW,R. W., T. P. NEUFELDand G. M. RUBIN,1994 Identifica- tional Eye Institute/National Institutes of Health grant 1RO1-EYO- tion of genes that interact with the sina gene in Drosophila eye 9299. development. Proc. Natl. Acad. Sci. USA 91: 11689-11693. CHANG, D. T., A. LOPEZ,D. P. VON KESSLER, c. CHIANG, B. K. SIMANDI. et al., 1994 Products, genetic linkage and limb patterning activ- LITERATURECITED ity of a murine hedgehog gene. Development 120: 3339-3353. CHEYEITE,B.N. R., P. J. GREEN, K. MARTIN,H. GARREN,V. AGNEL,M., S. KERRIDGE,C. VOLA and R. GRIFFIN~HEA,1989 Two HARTENSTEINet al., 1994 The Drosophila sine oculas locus encodes transcripts from the rotund region of Drosophila show positional a homeodomaincontaining protein required for the develop specificities in imaginal disc tissues. Genes Dev. 3: 85-95. ment of the entire visual system. Neuron 12 1-20. AGNEI., M., L. RODER,R. GRIFFIN-SHEAand C. VOLA,1992 The spa- CONCORDET,J.P., and P. INGHAM,1995 Patterning goes sonic. Na- tial expression of Drosophila rotund gene reveals that the imaginal ture 375 279-280. discs are organized in domains along the proximal-distal axis. CSIK,L., 1929 Ein neuer Erbfaktor bei Drosophila melanogaster. Biol. Roux’s Arch. Dev. Biol. 201: 284-295. Zentbl. 49: 419-421. ALTON, A. R, K. FECHTEL,C. C. KOPCZWSKI,S. B. SHEPARD,P. J. DEQUIN,R., H. SAUMWEBERand J. W. SEDAT,1984 Proteins shifting KOOH et al., 1989 Molecular genetics of Delta, a locus required from the cytoplasm into the nuclei during early embryogenesis for ectodermaldifferentiation in Drosophila. Dev. Genet. 10: 261- of Drosophila melanogaster. Dev. Biol. 104 37-48. 272. DICKSON,B., and E. HAFEN,1994 Genetics of signal transduction in ASHBURNER, A,, 1989 Drosophila. A Laboratory Manual. Cold Spring invertebrates. Curr. Opin. Genet. Dev. 4 64-70. Harbor Laboratory Press, Cold Spring Harbor, NY. DIETRICH,U., and J. A.CAMPOSORTEGA, 1984 The expression of BAKER,N. E., and G. M. RUBIN,1989 Effect on eye development of neurogenic loci in imaginal epidermal cells of Drosophila melane dominant mutations in Drosophila homologue of the EGF recep gaster. J. Neurogenet. 1: 315-332. tor. Nature 340: 150-153. DIETRICH,W., 1909 Die facettenaugen der Dipteren. Z. Wiss. Zool. BAKER,N. E., and G. M. RUBIN,1992 Ellipse mutations in the Drosoph- 92: 465-539. ila homologue of the EGF receptor affect pattern formation,cell ECHEIARD,Y., D. J. EPSTEIN,B. ST. JACQUES,L. SHEN,J. MOHLERet division, and cell death in eye imaginal discs. Dev.Biol. 150 ab, 1993 Sonic hedgehog, a member of a family of putative 381-396. signaling molecules, is implicated in the regulation of CNS polar- BAKER,N. E., and A. E. ZITRON,1995 Drosophila eye development: ity. Cell 75: 1417-1430. Notch and Delta amplify a neurogenic pattern conferred on the EKKER,S. C., A. R. UNGAR,P. GREENSTEIN,D. P. VON KESSLER, J. A. morphogenetic furrow by scabrous. Mech. Dev. 49: 173-189. PORTERet al., 1995 Patterning activities of vertebrate hedgehog BAKER,N. E., M. MLODZIK andG. M. RUBIN,1990 Spacing differenti- proteins in the developing eye and brain. Curr. Biol. 5: 944- ation in the developing Drosophila eye: a fibrinogen-related lateral 955. inhibitor encoded by scabrous. Science 250: 1370-1377. ELLIS,M. C., E.M. O’NEII.~. andG. M. RUBIN,1993 Expression of BAKER,N. E., K. MOSES, D. NAKAHARA,M. C. ELLIS,R. W. CARTHEW Drosophila glassprotein andevidence for negative regulation of its et al., 1992 Mutations on the second chromosomeaffecting the activity in non-neuronal cells by another DNA-binding protein. Drosophila eye. J. Neurogenet. 8: 85-100. Development 119: 855-865. BARINAGA, M.,1995 Focusing on the eyelessgene. Science 267: 1766- ELLIS,M. C., U. WEBER,V. WIERSDORFFand M. MLODZIK,1994 Con- 1767. frontation of scabrous expressing and nonexpressing cells is es- BASLER,K., D. YEN, A. TOMLINSONand E. HAFEN, 1990 Reprogram- sential for normal ommatidial spacing in the Drosophila eye. De- ming cell fate in the developing Drosophila retina: transformation velopment 120: 1959-1969. of R7 cells by ectopic expression of rough. Genes Dev. 4 728- FUJITA,S. C., S. L. ZIPURSKY,S. BENZER, A. FERRUSand S. L. SHOTWELL, 739. 1982 Monoclonal antibodies against the Drosophila nervous sys- BEGEMANN,G., A,”. MICHON,L. VAN DE VOORN, R. WEPFand M. tem. Proc. Natl. Acad. Sci. USA 79: 7929-7933. MLODZIK,1995 The Drosophila orphan nuclear receptor Seven- GARC~A-BELLIDO,A.,P. A. LAWRENCE and G. MORATA,1979 Compart- up requires the Ras pathway for its function in photoreceptor ments in animal development. Sci. Am. 241: 102-110. determination. Development 121: 225-235. WORN,E., 1968 Transdeterminationin cells. Sci. Am. 219: 11- BIER,E., L. ACKERMAN,S. BARBEL, JANL. and Y. N. JAN,1988 Identifi- 120. cation and characterization of a neuron-specific nuclear antigen HALDER,G., P. CALIAERTS and W. J. GEHRING,1995 Induction of in Drosophila. Science 240: 913-916. ectopic eyes by targeted expression of the tyeless gene in Drosoph- BLAIR,S. S., 1995 Compartments and appendage development in ila. Science 267: 1788-1792. Drosophila. BioEssays 17: 299-309. HARDING,K. W., G. G.N. MCGINNISand W. MCGINNIS,1995 A BONINI,N. M., W. M. LEISER~ON andS. BENZER, 1993 The eyes absent gene: genetic control of cell survival and differentiation in the screen for modifiers of Deformed function in Drosophila. Genetics developing Drosophila eye. Cell 72 379-395. 140: 1339-1352. BOTAS,J., J. Moscoso DEL PRADOand A. GARCIA-BELLIDO,1982 HARRISON, D. A., D. A. GDULA,R. S. COWE and V. G. CORCES,1993 Genedose titration analysis in the searchof trans-regulatory A leucine zipper domain of the suppressor of Hairy-wing protein genes. EMBO J. 1: 307-310. mediates its repressive effect on enhancer function. Genes Dev. BRAND,M., and J. A.CAMPOS-ORTEGA, 1990 Second-site modifiers 7: 1966-1978. of the split mutation of Notch define genesinvolved in neurogen- HAZELRIGG,T., R. LEVISand G. M. RUBIN,1984 Transformation of esis in Drosophila melanogaster. Roux’s Arch. Dev. Biol. 198: 275- white locus DNA in Drosophila: dosage compensation, zeste interac- 285. tion, and position effects. Cell 36: 469-481. BROWN,N. L., C. A. SATTLER, D. R. MARKEY and S. B. CARROLL, 1991 HEBERLEIN,U., and K. MOSES,1995 Mechanisms ofDrosophilaretina1 hairy gene function in the Drosophila eye: normal expression is morphogenesis: the virtues of being progressive. Cell 81: 987- dispensable but ectopic expression alterscell fates. Development 990. 113: 1245-1256. HEBERLEIN,U., and G. M. RUBIN,1991 Star is required in a subset BROWN,N. L., C. A. SATTLER,S. W. PADDOCKand S. B.CARROLL, of photoreceptor cells in the developing Drosophila retina and 1995 Hairy and Emc negatively regulate morphogenetic furrow displays dosage sensitive interactions with rough. Dev. Biol. 144: progression in the Drosophila eye. Cell 80: 879-887. 353-361. BUMCROT,D. A,, R.TAKADA and A. P. MCMAHON,1995 Proteolytic HEBERLEIN,U., I. K. HARIHARANand G. M. RUBIN,1993a Star is processing yields two secreted forms of sonic hedgehog. Mol. required for neuronaldifferentiation in the Drosophila retina and Cell. Biol. 15: 2294-2303. displays dosage-sensitive interactions with Ras 1. Dev. Biol. 160 CAGAN,R. L., and D. F. READY, 1989 Notch is required for successive 51-63. cell decisions in the developing Drosophila retina. Genes Dev. 3: HEBERLEIN,U., T. WOLFFand G. M. RUBIN,1993b The TCFPhomc- 1099-1112. log dpp and the segmentpolarity gene hedgehog are required for CARROLL, S. B., and J. S. WHWE, 1989 The role of the hairy gene propagation of a morphogenetic wave in the Drosophila retina. during Drosophila morphogenesis: stripesin imaginal discs. Genes Cell 75 913-926. Dev. 3: 905-916. HEBERLEIN,U., C. M. SINGH,A. Y. LUKand T. J. DONOHOE,1995 1212 C. Ma et al.

Growth and differentiation in the Drosophila eye coordinated by wlanogaster, edited byM. BATE and A. MARTINEZ-ARIAS. Cold hedgehog. Nature 373: 709-71 1. Spring Harbor Laboratory Press, Cold Spring Harbor, NY. HIROMI,Y., M. Mr.o~mti,S. R. WEST,G. M. RUBIN andC. S. GOOI)MAN, MEIAMED,J., and 0. TKUIII.I.O-CEN(~Z,1975 The fine structure of 1993 Ectopic expression of seven-up causes cell fate changes the eye imaginal disks in Muscoid flies. J.Ultrastr. Res. 51: 79- during ommatidial assembly. Development 118 1123-1135. 93. HOOVER,K-K, T. I. GERASIMOVA,A. J. CHIENand V. G. COKES, 1992 MLODZIK,M., N. E. BAKER and G. M. RURIN,199Oa Isolation and Dominant effects of suppressor of Haily-7uing mutations on gypsy- cxpression of srabrous, a gene regulating rleurogenesisin Drosoph- induced alleles of fmked and cut in lhosqbhila melanogaster. Genet- ila. G~KI~SDev. 4 1848-1861. ics 132: 691-697. MI.OL)ZIK,M., Y. HIROMI,U. WEBER,C. S. GOODMANand G. M. RURIN, HOKTSCH,M., A. J. BIEBEK,N. H. PATH.and C. S. GOODMAN,1990 1990b The Dro.rophilaseven-up gene, a member of the steriod Differential splicing generates a nervous system-specific form of receptor gene superfamily, controls photreceptor cell fates. Cell Drosophila neuroglian. Neuron 4: 697-709. 60: 211-224. IN(;IIAM,P. W., 1994 Hed,gehog points the Way. Curr. Biol. 4 347- MOIKER,J., 1988 Requirements for hedgehog, a segmental polarity 350. gene, in patterning larval and adultcuticle ofDrosophila. Genetics INCXAM, P. W., S. M. PINCHIN,K. R. HOWAKI) andD. ISH-HOROWCZ, 120: 1061-1072. 1985 Genetic analysis of the haily locus in Drosophila melanogav MoHI.k.K, J.: and K VANI, 1992 Molecular organization and embry- tw. Genetics 111: 463-486. onic expression of the hedgehoggene involved in cell-cell commu- KERKILIGE, S., and M. TIIOMAS-C,ZV~~I.I.IN,1988 Appendage morpho- nication in segmental patterning of Drosophila. Development 115: genesis in Drosophila: a development study of the rotund (rn) 957-971. gene. Roux's Arch. Dev. Bid. 197: 19-26. Mows, K., and G. M. RUBIN,1991 glass encodes a site-specific DNA- KIMMEI., B. E., U. HEBEKIXIN andG. M. RURIN,1990 The holneo hirrding protein thatis regulated in response to positional signals domain protein roughis expressed in a subset of cells in the in the developing Drosophila eye. Genes Dev. 5: 583-593. developing Drosophila eye where it canspecify photoreceptor cell Mos~s,K., M. C. EI.I.ISand G. M. RURIN,1989 The glass gene en- subtype. Genes Dev. 4 712-727. codes a zinc-finger protein required by Drosophila photoreceptor KOI.OI)KIN, A.L., A. T. PICKUP,D. M. LIN,C. S. GOINMAN andU. cells. Nature 340: 531-536. BANWJEE,1994 Characterization of Star and its interactions MUI.I.I:K,H. .J., 1932 Further studies on the nature and causes of with sevmless arid EGF receptor during photoreceptor cell devel- gene mutations, pp. 213-255 in International Congress of Genetics, opment in Drosophila. Development 120 1731- 1745. edited by D. F. JONM. Brooklyn Botanic Garden, Brooklyn, NY. Korc:~wstil, C. C., A. K. &.TON, K. FECH'I'EI.,P. J. KOOH and M.A. MUKKI'.,C., P. S. MCCAWand D. B.U.TIMOKE, 1989 A new DNA bind- MUSKAW"(:H, 1988 ZMa, a Drosophila neurogenic gene, is tran- ing and dimerization motif in immunoglobulin enhancer bind- scriptionally complex and encodes a protein related to blood ing, daughterless, MyoD, and myc proteins. Cell 56: 777-783. coagulation factors and epidermal growth factor of vertebrates. NISWANDER,L., S. JEFFREY,G. R. MAKTINand C. TICKI.E,1994 A Genes Dev. 2: 17223- 1735. positive feedback loop coordinates growth and patterning in the KKAMER, s.,s. R.WEST and Y. HIROMI,199.5 Cell fate control in the vertebrate limb. Nature 371: 609-612. Drosophila retina by the orphan receptor seven-up: its role in the NOHNO,T., Y. KAWAKAhII, H. OtIUCHI, A. FUJIWARA,H. YOSHIOKA et decisions mediated by the ras signaling pathway. Development af., 1995 Involvement of the Sonic hedgehoggene in chick feather 121: 1361-1372. formation. Biochem. Biophys. Res. Com. 206: 33-39. KKAIJSS, S., J. P. CONC:OKI)KIand P. W. INGHAM,1993 A functionally N~~SSI.~IN-\~OI.I~IAKL),C., and E. Wlesc:~Aus, 1980 Mutations affect- conserved homolog of the Dro.wphila segment polarity gene hh ing segment number andpolarity in Drosophila. Nature 287: 795- is expressed in tissues with polarizing activity in zebrafish em- 801. bryos. Cell 75: 1431-1444. NUSSI.EIY-VOI.HAIW,C., E. WIESCHAUSand H. KI.UnlNc, 1984 Muta- LASKI, F. A,, D. C. Rw and G. M. RUBIN,1986 Tissue specificity of tions affecting the pattern of the larval cuticle in Drosophila melu- Drotnphila P element tr-ansposition is regulated at the level of nogastw. I. Zygotic loci on the second chromosome. Roux's Arch. mRNA splicing. Cell 44 7-19. Dev. Biol. 183: 267-282. ~AIJFER, E., C. E. NEI.SON,R. L. JOHNSON, B.A. MORGANand C. O'KANE,C. J,,and W. J. GEHRINC:,1987 Detection in situ of genomic TABIN,1994 Sonic.hedgehog and Fd-4 act through a signaling regulatory elements in Ilrorqbhila. Proc. Natl. Acad. Sci. USA 84: cascade and feedback loop to integrate growth and patterning 9123-9127. of the developing limb bud. Cell 79: 993-1003. O'NEIIL,E. M., M. C:. EILIS,G. M. RUHIN andR. TJIAN, 1995 Func- LEE,J. J.. D.P. VON KE:SSI.EK, S. PARIGSand P. A. BEACHY, 1992 Secre- tional domain analysis of Glass, a zinc-finger-containing tran- tion and localized transcuiption suggest a role in positional signal- scriptionfactor in Drosophila Proc. Natl. Acad. Sci. USA 92: ing for products of the segmentation gene hedgehog. Cell 71: 33- ti557-6561. 50. PANGANIBAN,G., K. E. RASHKA, M. D. NEITZEI.and F. M. HOFFMANN, LEE,J. J., S. C. EKKI'R, D. P. \'ON G.SSI.EK,J. A. PORTER, B. I. SUN 1990 Biochemical characterization of the Drosophila d# pro- et al., 1994 Autoproteolysis in hedgehog protein biogenesis. tein, a memeber of the Transforming Growth Factor b family of Science 266: 1528-1537. growth factrors. Mol. Cell. Bid. 10: 2669-2677. I,INIISI.EY,D. L., and G. C;. ZIMM, 1992 TheGenome qf Drosophila PARKHVKST,S. M., and V. G. CORC:ES,1986 Retroviral elements and melanogaster. Academic Press, San Diego. suppressor genes in Drosophila. BioEssays 5: 52-57. MA, C., and K. MOSES,1995 70inglcss and patched are negative regula- PARKIII~KST,S. M., D. A. HARRISON,M. P. REMINCXON, c. SPAN, R. L. tors of the morphogenetic furrow and can affect tissue polarity UI.I.I:.YPI al., 1988 The Drosophila su(Hw) gene, which controls ill the developing Drosophila compound eye. Development 121: the phenotypic effect of the gypsy transposable clement, encodes 2279-2289. a putative DNA-binding protein. Genes Dev. 2: 1205-1215. MA, C., Y. ZHOU, P. A.BE,Z(:HY and K. MOSES,1993 The segment PEIFEK,M., and A. Bqsowx:, 1992 Knowing yourneighbors: cell polarity gene hedgehog isrequired forprogression of the morpho- interactions determine intrasegmental patterning in Drosophila. genetic furrow in the developing Drosophila eye. Cell 75: 927- Trends Genet. 8 243-249. 938. POWER,J. A,, D. P. VON KESSI.~K,S. C. EKKEK,K. E. YOUNC,J. J. LEE MALILN, M., 1994 The limb bud-part two. Nature 371: 560-561. rt al., 1995 The product of Hedgehog autoproteolytic cleavage MAKIION, G., N. M. SOI.OMONand G. M. RUBIN,1994 dachshund active in local and long-range signaling. Nature 374 363-366. encodes a nuclear protein required for normal eye and leg devel- QIIIRINC:, R., U. W~r.~.now,U. KLOTEK and W. J. GEHRING,1994 opment in Drosophila. Development 120: 3473-3486. Homology of the eyeless gene of Drosophila to the small eye gene MARrI, E., D.A. BUMGROT,R. TAKAIMand A. P. MCMAHON,19% in mice and Aniridia in humans. Science 265: 785-789. Requirement of 19K form of sonic hedgehog for induction of READY,D. F., T. E. HANSONand S. BENXEK,1976 Development of distinct ventral cell types in CNS explants. Nature 375: 322-325. the Drocophila retina, a neurocrystalline lattice. Dev. Bid. 53: MA\KK, u., and c. NirSSl.EIN-VOI.HARI), 1988 A group Of gelleSre- 217-240. quired for pattern formation in the ventral ectoderm of the RENFRWZ,P. .J., and S. BEX7.ER, 1989 Monoclonal antibody probes fh-otophila embryo. Genes Dev. 2: 1496-1511. discriminate early and late mutant defects in development of the MEINEKTLIIAGEN,I. A., and T. E. HANSON,1993 The development Dro.rophila retina. Dev. Biol. 136: 411-429. of the optic lobe, pp, 1363- 1491 in TheUevrlopmmt of Drosophila ROREKTSON,H. M., C. R. PRESTON,R. W. PHlI.I.IS, D. M. JOHNSON- Enhancers of glass 1213

S(:III.ITL,W. K. BENZet al., 1988 A stable genomic source of P TARATA, T., S. EATONand T. B. KORNBLRG,1992 The Drosophilu element transposase in Drosophila mrlanogaster. Genetics 118: hedgehog gene is expressed specifically in posterior compartment 461-470. cells and is a target of mgruikd regulation. Genes Dev. 6: 263.5- ROEI.INK, H.,A. ALIC;SBURGV.R,J. HEEMSKERK, V. KoRzrr, S. NORI.INrt 2645. nl., 1994 Floor plate and motor neuron induction by uhh-1, a TAS~IIRO,S., T. MICHIUE,s.4. HIGASHIIIMA,S. ZI'NNO, S. ISHIMARI. vertebrate homolog of hedgthogexpressed by the notochord. Cell rt nl., 1993 Structure and expression of hedgehog, a Drosophila 76: 761-775. segment-polarity generequired for cell-cell communication. Rocx:~',R. D., C. A. K~IKWICI?and U. BANERIEL,1991 Genetic Gene 124: 183-189. dissection of a neurodevelopmental pathway: son of sevenless THAKER,H. M., and D. R. IDINH:~.,1992 Mosaicanalysis gives an functions downstream of the sevenless and EGF receptor tyrosine estimate of the extent of genomic involvement in the develop- kinases. Cell 64: 39-48. ment of the visual system in Drosophila mrlanogmtrr. Genetics 13: RLJSIII.OW,C. A,, A. HOGAN,S. M. PINCHIN,K. M. HOWL,M. LARDEI.L.I 883-894. rt nl., 1989 The 1)rosophila haily protein acts in both segmenta- TOMI.INSON,A,, 1988 Cellularinteractions in the developing Dro- tion and hristle patterning and shows homology to N-myc. EMBO sophila eye. Development 104: 183-193. J. 8: 309553103, TOMI.IMON,A., and D. F. READY, l987a Cell fate in the I)ro,snt,hila SFIKN,B., J. khl and D. DORSETT,1994 The enhancer-blocking sup- ommatidium. lkv. Bid. 123: 264-275. pressor of Hairywing zinc finger protein of Drosophila mrlanogas- TOMIIKSON, A,, and D. F. &Am', l987b Neuronaldifferentiation ~PTalters DNA structure. Mol. Cell. Biol. 14: 564555652, in the Drosuphikz ommatidium. Dev. Biol. 120: 366-376. S~~'sfmE:wU.w.~\,H., andJ. S. T.\K,\HASHI,1994 Characterization of WADDINGTON,C. ti., and M. M. PERRY,1960 The ultra-structure ot the chicken rhodopsin promoter:identification of retina-specific the developing eye of lhsnphiln. Proc. Roy. Soc. L.ond. B 153: and glm-like protein binding domains. Mol. Cell. Neurosci. 5: 155-178. 309-318. WEMMIIR,T., and C. KIAMBI, 1995 A genetic analysis of the L)rusoph- SIMOK,M. A, D. D. L. BOMTEI.I.,G. S. DOIXON,T. R. LAVERTY and ila closely linked interacting genes bulgr, nrgos, and sobn. Genetics G. M. RK.BIN,1991 Rasl and a putative guanine nucleotide 140: 629-641. exchange factor perform crucial steps in signaling by the sev WIF.S~:II,\LIS,E., and C. NL~SSI.EIN-V~I.HARU,I986 Looking at em- enless protein tyrosine kinase. Cell 67: 701-716. bryos, pp. 199-227 in Ihosophiln: A Prnctzral Approach, edited hy SMI?'H,J.C., 1994 Hedgehog, the floor plate, and the zone of polar- D. B. ROBI'KIS.IRL Press, Oxford. izing activity. Cell 76: 193-196. WOI.FF,T., and D. F. RunY, 1991 The beginning of pattern forma- SMIIH,P. A., and V. G. COR(:ES,1995 The suppressor of Hairy-wing tion in thc Zlrorophzla compound eye: the morphogenetic furrow protein regulates the tissue-specific expression of the Drosophila and the second mitotic wave. Development 113: 841-850. gypsy retrotransposon. Genetics 139: 215-228. X(!, T., and G. M. RL-BIN,1993 Analysis of genetic mosaics in devel- SPEN(:F.R,F. A,, F. M. HOFFMANN andW. M. GELRART,1982 Derapm- oping and adultDro.s@hila tissues. Development 117: 1223-1237. kzpbgic: a gene complex affecting morphogenesis in Drosophilu ZAR, N. B., and B.-Z. SHII.O,1992 Localization of DER and the ndznugastpr. Cell 28: 451 -461, pattern of ccll divisions in wild-typeand Ellipsreye imaginal discs. TABA'I'A, T., and T. B. KORNBI'RG,1994 Hedgehog is a signaling Dev. Biol. 149: 448-456. protein with a key role in patterning Drosophila imaginal discs. Cell 76: 89-102. Communicating editor: T. SC~I~~PRAC~I