An F₁ Genetic Screen for Maternal-Effect Mutations Affecting Embryonic Pattern Formation in Drosophila melanogaster

Corresponding author: Stefan Luschnig, 279 Campus Dr., Beckman Center, Rm. B453, Stanford University School of Medicine, Stanford, CA 94305-5307., luschnig{at}stanford.edu (E-mail)

Communicating editor: T. SCHÜPBACH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Large-scale screens for female-sterile mutations have revealed genes required maternally for establishment of the body axes in the Drosophila embryo. Although it is likely that the majority of components involved in axis formation have been identified by this approach, certain genes have escaped detection. This may be due to (1) incomplete saturation of the screens for female-sterile mutations and (2) genes with essential functions in zygotic development that mutate to lethality, precluding their identification as female-sterile mutations. To overcome these limitations, we performed a genetic mosaic screen aimed at identifying new maternal genes required for early embryonic patterning, including zygotically required ones. Using the Flp-FRT technique and a visible germline clone marker, we developed a system that allows efficient screening for maternal-effect phenotypes after only one generation of breeding, rather than after the three generations required for classic female-sterile screens. We identified 232 mutants showing various defects in embryonic pattern or morphogenesis. The mutants were ordered into 10 different phenotypic classes. A total of 174 mutants were assigned to 86 complementation groups with two alleles on average. Mutations in 45 complementation groups represent most previously known maternal genes, while 41 complementation groups represent new loci, including several involved in dorsoventral, anterior-posterior, and terminal patterning.

THE establishment of the major body axes of the Drosophila embryo is governed by maternal gene activity. Four groups of genes act in a largely independent manner to specify the anterior-posterior and the dorsoventral axes, as well as the terminal regions of the embryo. Localized, maternally derived RNAs provide the source for anterior and posterior determinants, while local activation of transmembrane receptors and subsequent signal transduction pathways define the dorsoventral axis, as well as cell fates at the embryonic termini. About 35 genes whose products are required maternally for embryonic pattern formation have been identified by classic screens for recessive female-sterile mutations (GANS et al. 1975 ; MOHLER 1977 ; PERRIMON et al. 1986 ; SCHUPBACH and WIESCHAUS 1986 , SCHUPBACH and WIESCHAUS 1989 , SCHUPBACH and WIESCHAUS 1991 ; reviewed in ST. JOHNSTON and NUSSLEIN-VOLHARD 1992 ). Analysis of these genes has revealed a framework of the axis-forming systems in the Drosophila embryo. However, taking into account all the known components involved in axis determination, we still do not understand several aspects. For instance, How are anterior and posterior determinants localized to opposite poles of the oocyte? How are spatially restricted ventral and terminal signals generated and maintained in the perivitelline space? Growing knowledge about the molecular properties of the pathways involved in axis formation also allows us to postulate the existence of missing factors, such as a missing link in the dorsoventral signaling cascade between the proteins Pelle and Cactus (GROSSHANS et al. 1999 ) or a missing transcriptional activator downstream from the Torso pathway (reviewed in FURRIOLS and CASANOVA 2003 ). For reasons explained below, it is likely that several components of these pathways have not been identified by the classic screens.

A large number of maternal gene products are deposited in the Drosophila egg during oogenesis (GARCIA-BELLIDO and ROBBINS 1983 ; PERRIMON et al. 1984 ; SCHUPBACH and WIESCHAUS 1991 ; reviewed in WILKINS 1986 ). Screens for female-sterile mutations led to the isolation of the majority of "strictly" maternal genes, which are required for the normal development of the egg and/or the embryo, but whose functions are dispensable for adult viability (PERRIMON et al. 1986 ; SCHUPBACH and WIESCHAUS 1986 , SCHUPBACH and WIESCHAUS 1989 , SCHUPBACH and WIESCHAUS 1991 ). A limiting factor in these screens has been the large effort associated with generating enough lines to saturate the genome for maternal-effect mutations. Moreover, identification of a gene by means of its mutant phenotype can be obscured by functional redundancy of different genes on the one hand or by multiple functions of a single gene during development on the other hand. While functional redundancy poses a general problem for forward genetic approaches, it is of less concern for Drosophila as compared to vertebrate genomes. In many cases, Drosophila possesses a single or few homologs of a corresponding multigene family in vertebrates, and often the Drosophila gene has been shown to mutate to a distinct phenotype (ADAMS et al. 2000 ; RUBIN et al. 2000 ). However, multiple requirements of a gene product at different developmental stages are rather common. For example, components involved in signal transduction [e.g., the RAS-mitogen-activated protein kinase (MAPK) pathway] or in intracellular transport (of RNA or proteins) are used repeatedly in different contexts during development. Genes encoding such multiply required components are likely to mutate to zygotic lethality, thus precluding their identification in genetic screens for female-sterile mutations. In Drosophila, it is possible to overcome the requirement of a gene for viability by studying mosaic animals that carry clones of homozygous mutant cells in an otherwise heterozygous animal. Such mosaics can be generated by mitotic recombination between homologous chromosomes, induced by X-rays (PATTERSON 1928 ; STERN 1936 ), or, more recently, by a heterologous site-specific recombination system from yeast, the Flipase-Flipase recombinase target (Flp-FRT) system (GOLIC and LINDQUIST 1989 ; GOLIC 1991 ). The Flp-FRT system can induce clones in germline and somatic tissues with high efficiency, and combining FRT chromosomes with the dominant female-sterile ovo^D mutation has allowed positive selection for recombination events in the female germline ("Flp-DFS"—dominant female sterile—technique; CHOU et al. 1993 ; CHOU and PERRIMON 1996 ). Systematic screens have been carried out using germline clonal analysis to identify genes involved in embryonic patterning. A screen of X-linked lethal mutations and screens of P-element-induced lethal mutations on the autosomes have revealed specific maternal functions of several zygotic lethal loci for embryonic patterning (PERRIMON et al. 1989 , PERRIMON et al. 1996 ; BELLOTTO et al. 2002 ). These screens have also identified embryonic lethal mutations that become phenotypically manifest only in the absence of both the maternal and zygotic contributions of a gene product.

We have carried out a large-scale germline clone (GLC) screen specifically aimed at identifying new genes involved in the four axis-forming systems. Because the early steps of axis formation are controlled by the maternal genome before the onset of zygotic transcription, the paternal contribution to the embryo's genotype is irrelevant for these processes. This situation allowed us to devise an efficient F₁ screening scheme, in which individual females carrying clones homozygous for a mutagenized chromosome are mated with wild-type males. The F₁ scheme allows large numbers of individuals to be scored in a rapid fashion, as the same animal is used to detect a mutant phenotype and to establish a line of mutant carriers. In contrast to F₂ or F₃ screens, stocks of potential mutants are established only after a phenotype has been scored, thus greatly reducing the number of lines to maintain.

In this work, we describe the technique and results of our GLC screen on the five major chromosome arms, comprising most of the euchromatic portion of the Drosophila genome. We identified 232 mutations with distinct phenotypes and ordered the mutants into 10 different phenotypic classes. We were able to assign 174 (75%) of the 232 mutants to complementation groups, with 86 complementation groups in total and an average allele frequency of 2 alleles per complementation group. A total of 41 complementation groups represent previously undescribed loci, while 45 complementation groups represent most of the previously known maternal genes. We found on average a higher number of alleles for previously known genes (2.71 alleles per locus) than for new mutations (1.27 alleles per locus), 47 of which are represented by a single allele.

We found several new loci involved in dorsoventral, anterior-posterior, and terminal patterning, as well as mutants affecting other processes, such as embryonic segmentation, epidermis development, and morphogenesis. Interestingly, a significant fraction of the new loci isolated in this work represent viable mutations, suggesting that these genes had been missed in earlier screens due to incomplete saturation, rather than to a zygotic requirement of these genes.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Fly stocks:
Balancer chromosomes and marker mutations used in this study are listed in LINDSLEY and ZIMM 1992 and in FLYBASE 1999 . The following FRT and hs-Flp chromosomes were used:

• hs-Flp22 (X): P[ry⁺, hs-Flp]22 (CHOU and PERRIMON 1996 )

• hs-Flp122 (X): P[ry⁺, hs-Flp]122 (STRUHL and BASLER 1993 )

• FRT9-2 (X): P[>w⁺>, FRT]18E (CHOU and PERRIMON 1996 )

• FRT40A (2L): P[ry⁺, hs-neo, FRT]40A (XU and RUBIN 1993 )

• FRTG13 (2R): P[>w⁺>, FRT]42B (CHOU and PERRIMON 1996 )

• FRT2A (3L): P[>w⁺>, FRT]79D-F (CHOU and PERRIMON 1996 )

• FRT82B (3R): P[ry⁺, hs-neo, FRT]82B (XU and RUBIN 1993 ).

hs-Flp22 was used for the third chromosome screens; hs-Flp122 was used for the first and second chromosome screens. hs-Flp122 showed considerably stronger Flipase activity than hs-Flp22 when we tested for the frequency of germline clones or flip-out clones in the eye (data not shown). The Flp-DFS technique and FRT chromosomes containing ovo^D mutations are described in CHOU and PERRIMON 1996 . Conditional dominant temperature-sensitive balancer chromosomes (CyO, P[hs-hid, w⁺] and TM3, P[hs-hid, w⁺]) have been described (GRETHER et al. 1995 ; MOORE et al. 1998 ). The marked FRT chromosomes listed in Table 1 were constructed by recombining proximal FRT sites (XU and RUBIN 1993 ; CHOU and PERRIMON 1996 ) with distal visible markers or with P[ubi-nlsGFP, w⁺] insertions (DAVIS et al. 1995 ). All recombinant FRT chromosomes generated in this work are available from the Bloomington Stock Center (http://flystocks.bio.indiana.edu/).

Generation of P[ubi-nlsGFP, w⁺] insertions:
To generate P[ubi-nlsGFP, w⁺] insertions on all major chromosome arms, we first combined two insertions of this construct on the third chromosome (P[ubi-nlsGFP, w⁺]34N, P[ubi-nlsGFP, w⁺]34A3; DAVIS et al. 1995 ) to make a "jumpstarter." We then mobilized the P-element construct by crossing these flies to a Delta2-3 transposase source on the TM2 balancer chromosome. New P[ubi-nlsGFP, w⁺] insertions were outcrossed and analyzed for segregation of the insertion with the second or the X chromosome. New third-chromosomal insertions were generated separately in a similar scheme. Several homozygous viable and strongly expressing insertions were mapped genetically to a chromosome arm (visible genetic markers indicated in Fig 3 were used for mapping). A single or two different insertions were recombined to proximal FRT sites to make the stocks listed in Table 1.

View larger version (40K): In this window In a new window Download PPT slide   Figure 1. Crossing scheme exemplified for chromosome arm 2R. In the shaded box, the different maternal genotypes of eggs laid by a mosaic female are indicated; the corresponding intensity of GFP fluorescence is shown schematically. The asterisk indicates an EMS-induced mutation. Strains that were used to screen the other arms are listed in Table 1. The scheme for the X chromosome differs from the autosomal schemes in that F2 females rather than males were used to recover the mutagenized chromosome. The retest procedure is not shown in this scheme (see MATERIALS AND METHODS for details).

View larger version (70K): In this window In a new window Download PPT slide   Figure 2. ubi-nlsGFP in egg chambers and embryos. (A) Confocal micrographs of an egg chamber (top) and cellular blastoderm embryo (bottom) from 2x GFP FRT2A females; the father of the embryo is wild type (GFP negative). Anterior is to the left. (B and C) Undechorionated embryos from a cross of mosaic y w hs-Flp22/w; 2x GFP FRT2A/ru h th st FRT2A females and wild-type (GFP-negative) males. The photo was taken from embryos on an agar plate covered with Voltalef oil. (B) GFP-negative GLCs. The embryos show weak yellow autofluorescence. (C) GFP-positive siblings from the same egglay.

View larger version (14K): In this window In a new window Download PPT slide   Figure 3. Graphic representation of marked chromosomes used for the screen. Horizontal bars correspond to approximate genetic map positions of P[ubi-nlsGFP, w+] insertions that were recombined to an FRT site on the respective chromosome arm. Visible markers that were recombined to FRT sites are indicated with their positions (numbered cytological bands).

Screening procedure:
The crossing scheme, exemplified for chromosome arm 2R, is illustrated in Fig 1. Strains that were used to screen the other chromosome arms in the same fashion are listed in Table 1. The crossing scheme for the X chromosome differed from the scheme for the autosomes in that isogenic lines were started from single F₂ females (rather than males) balanced with FM7c.

Mutagenesis:
We carried out several rounds of mutagenesis for each chromosome arm. For one experiment, 600 males homozygous for a marked FRT chromosome were starved on water-saturated Kleenex paper for 4 hr before they were fed for 14 hr with ethyl methanesulfonate (EMS; Sigma, St. Louis) in 1% sucrose according to standard methods (LEWIS and BACHER 1968 ). We used different EMS concentrations in the range between 25 and 45 mM, the chosen dose depending on how well a particular strain tolerated the EMS treatment. We noticed that males carrying a y w hs-Flp X chromosome showed greatly increased sensitivity to EMS as compared to males carrying a w chromosome in an otherwise genetically identical background. Per 200-ml food bottle, 50 mutagenized males and 80 females were mated and cultured at 25°. Male parents were removed after 3 days and females were transferred to fresh medium every 2 days. To estimate the efficiency of the EMS treatment for inducing lethal mutations, we determined the fraction of lethal chromosomes among a sample of randomly picked F₁ males from mutagenized fathers. In a test for the screen on chromosome arm 3L, we found that 82% (90/110) of mutagenized (30 mM EMS) ru h th st FRT2A chromosomes were lethal. Assuming a Poisson distribution for the frequency of lethal hits, this corresponds to an average of 1.7 lethal hits per chromosome. Similar lethality rates were observed for other chromosome arms tested (data not shown).

Production of germline clones:
To induce clones in the F₁ females that were tested in the screen, third instar larvae in 200-ml food bottles were heat-shocked twice (one heat shock per day on two consecutive days) for 2 hr at 37° in a circulating water bath. For retesting mutant candidates, virgin females of a balanced candidate line were crossed to males of the general genotype hs-Flp/Y; FRT ovo^D/Balancer, P[hs-hid]. Larval progeny of these crosses were heat-shocked and mosaic females of the genotype hs-Flp; FRT mutation/FRT ovo^D were collected in egg-laying blocks and examined for unhatched embryos. A single 1-hr heat shock at the third instar larval stage was sufficient to induce germline clones, as well as, at the same time, to eliminate all unwanted progeny. The only surviving genotype was mosaic hs-Flp; FRT mutation/FRT ovo^D flies, thus eliminating the need to manually sort flies from the retest cross (note that in the retest the males are therefore heterozygous for the mutagenized chromosome).

Detection of embryonic phenotypes:
We used the block-agar method for large-scale collection of eggs from single females (NUSSLEIN-VOLHARD 1977 ). Techniques for collection and observation of embryos are described in WIESCHAUS and NUSSLEIN-VOLHARD 1998 . Undechorionated live embryos, 0 to 18 hr old, on apple juice agar plates were covered with Voltalef 3S oil and sorted using a Leica MZ12 epifluorescence stereomicroscope equipped with a green fluorescent protein (GFP)1 filter set. GFP-negative embryos were picked and transferred to fresh apple juice agar plates, allowed to develop for 24 hr at 25°, and then examined for abnormalities in the embryonic cuticle. Cuticle preparations were done according to standard procedures (WIESCHAUS and NUSSLEIN-VOLHARD 1998 ).

Mapping of mutations:
Autosomal mutations were mapped genetically by testing for noncomplementation of lethality or female sterility, using a set of chromosomal deficiencies that uncovers most of the respective chromosome arm. X-linked mutations were mapped using a set of duplications that covers most of the X chromosome. Deficiency and duplication kits were obtained from the Bloomington Stock Center. In parallel to deficiency or duplication mapping, mutations were mapped by meiotic recombination to verify cytological map positions, as well as to clean the mutagenized chromosome from associated lethal mutations. All mutations were induced on recessively marked FRT chromosomes; for mapping, the marked mutagenized FRT chromosome was allowed to recombine with the corresponding unmarked parental FRT chromosome. Approximately 50 independent recombinants were analyzed for segregation of the mutant phenotype with visible markers. In the case of w⁺-marked FRTs (FRT9-2, FRTG13, and FRT2A), the FRT site itself could be used as an additional visible marker (due to our screening procedure, the FRT-flanked w⁺ marker on the mutagenized P[>w⁺>, FRT] chromosomes is excised by Flipase, resulting in a single remaining FRT site lacking the w⁺ marker).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Overview of the screening procedure:
To screen on a large scale for maternal-effect lethal mutations, we adapted the Flp-FRT system (CHOU and PERRIMON 1992 ; XU and RUBIN 1993 ) to devise an F₁ screening scheme (see Fig 1). In this scheme, mutant phenotypes are detected among the eggs from single females, which are obtained as F₁ progeny of mutagenized EMS-treated males, thus allowing large numbers of individuals to be scored in a rapid fashion. The use of GFP as a vital marker allowed us to distinguish embryos derived from homozygous germline clones. The polyubiquitin (ubi) promoter directs strong maternal GFP expression in eggs, which is detectable up to late stages of embryogenesis (see Fig 2; DAVIS et al. 1995 ). We constructed a set of ubi-GFP-marked FRT chromosomes that allow marking of clones in germline and somatic tissues by the absence of the GFP marker from the clones (see Fig 1 and Fig 2B and Fig C). A list of these strains is in Table 1.

We carried out a separate screen for each of the five major chromosome arms. The crossing scheme, exemplified for chromosome arm 2R, is shown in Fig 1. Corresponding schemes were applied for the other autosome arms and, slightly modified, for the X chromosome (see Table 1). In general, males carrying an isogenized, recessively marked FRT chromosome were treated with EMS and crossed en masse to females carrying an hs-Flp source and an FRT chromosome with distal P[ubi-nlsGFP, w⁺] insertions. Larval progeny of this cross were heat-shocked to induce Flp-mediated mitotic recombination. The emerging adults carry clones (in somatic and germline tissues) that are homozygous for the mutagenized FRT-based chromosome arm. F₁ females were crossed individually to males with a marked tester chromosome and embryos from the F₁ females were genotyped on the basis of the presence or absence of maternally contributed GFP: embryos derived from GLCs homozygous for the mutagenized chromosome arm lack GFP, while the remaining embryos derived from heterozygous germline cells or from GFP-homozygous twin spots fluoresce green. The intensity of fluorescence depends on the copy number of GFP provided both maternally and zygotically (see Fig 2B and Fig C).

We determined the efficiency of GLC induction by scoring the proportion of GLC-derived eggs (marked either by a mutation or by the absence of GFP) among the eggs laid by single females. In control experiments without mutagenesis, nearly 100% of the mosaic females laid GLC-derived eggs; the average proportion of GLCs among the total number of eggs laid by a single female was between 5 and 10%. If the fathers were mutagenized, only between 49 and 63% (depending on the chromosome arm) of the F₁ females laid GLC-derived eggs (see Table 2). The remaining females did not produce GLCs, presumably due to the presence of cell-lethal mutations on the FRT-tagged chromosome arm. The average number of eggs laid by a female over 1 day was 30, and thus the number of GLC-derived eggs that can be collected per day was small (in most cases 1–3 eggs). In the screen we therefore collected eggs from most of the females on two consecutive days. The use of a strongly expressing hs-Flp source (hs-Flp122 instead of hs-Flp22; see MATERIALS AND METHODS) increased the proportion of GLCs to >10% on average (data not shown).

We sorted out 0- to 24-hr-old GFP-negative embryos from each single female, transferred these embryos to an agar plate where they were allowed to complete embryonic development, and examined them after 24 hr. The majority of the embryos were phenotypically normal and had hatched by 24 hr; embryos that did not hatch were examined for patterning defects in larval cuticle preparations. To establish a line from a mutant candidate, the female was recovered and allowed to lay eggs in a food vial. Three to six isogenic lines balanced for the mutagenized chromosome were established from single males in the F₂ generation (in the case of the X chromosome screen, lines were established from single F₂ females carrying an X chromosome balancer chromosome). The presence of recessive markers (see Fig 3) on the mutagenized FRT chromosome was used to avoid the isolation of recombinants that may have lost the mutation. Stocks of mutant candidates were retested to confirm the initially scored phenotype. We used the Flp-DFS (ovo^D) technique (CHOU and PERRIMON 1996 ), aided by conditional temperature-sensitive balancer chromosomes (see MATERIALS AND METHODS; GRETHER et al. 1995 ; MOORE et al. 1998 ) for the retest, as a higher number of mutant embryos can be obtained by the ovo^D method than by sorting according to GFP fluorescence. We discarded all lines that produced a high percentage of hatching larvae in the retest. A total of 353 (48.6%) of the 726 potential mutants initially isolated gave embryos that failed to hatch in the retest. These lines were kept for further analysis.

In this work, we describe 232 of these lines, which show distinct and, in most cases, completely penetrant phenotypes. Table 2 summarizes the five screens that were carried out. For each screen, we counted the number of females from which we collected GLCs as the number of screened chromosome arms. On the basis of lethality tests, most of these chromosomes contained more than one lethal mutation (see MATERIALS AND METHODS).

Complementation analysis and mapping:
To establish complementation groups, all mutants on a given chromosome arm that showed a similar phenotype were crossed to each other and to mutants in known candidate genes located on the relevant chromosome arm. We tested for viability and female fertility of the trans-heterozygous progeny of these crosses. Mutations that did not fall into any of the previously known complementation groups were subsequently mapped by meiotic recombination and/or complementation tests, using a set of overlapping deficiencies on the relevant chromosome arm. Due to the time-consuming X chromosomal genetics, mapping and complementation of many X chromosomal mutants is still in process and has not been included in this work. The number of newly found alleles at previously known loci was used to estimate the mutation rate achieved using our screening procedure. Overall, we isolated 2.7 alleles on average for previously known loci (122 mutants falling into 45 complementation groups), corresponding to a mutation rate of approximately one hit per locus in every 1000–2000 chromosome arms screened (see Table 2 and Table 4).

Phenotypic classification:
Mutations were initially grouped into the following phenotypic categories: defects in dorsoventral, anterior, posterior, or terminal patterning; segmentation defects; defects in morphogenetic processes (such as germband retraction or dorsal closure); defects in epidermis development; and alterations of eggshell morphology. Eighteen lines that did not fit any of the above categories were grouped as "other phenotypes." Although our screening scheme was specifically designed to isolate maternal mutations that are not zygotically rescuable, we also found several zygotic embryonic lethal mutations in the X chromosome screen, where 50% of the GLC-derived embryos are hemizygous males. In addition, some autosomal zygotic mutations were identified coincidentally in the retest, where the fathers are heterozygous for the mutagenized chromosome (see MATERIALS AND METHODS). The mode of action of a mutation (maternal effect, paternally rescuable maternal effect, or zygotic embryonic lethal) is indicated in supplemental Table 1 at http://www.genetics.org/supplemental/. We also isolated several mutants that produce unfertilized eggs or show an early arrest in embryonic development. However, we did not systematically screen for these phenotypes and excluded them from description in this work. Supplemental Table 1 shows a list of the 232 mutants that displayed distinct phenotypes. In addition, 121 lines, which showed less clear or incompletely penetrant phenotypes, were kept for further analysis; these lines are not described in this work. In the following, we describe the different phenotypic classes and emphasize new loci that were identified by this study.

Dorsoventral patterning:
Dorsalized embryos: Twenty-nine mutants falling into nine complementation groups show dorsalized phenotypes. We identified new alleles of all previously known dorsal group genes (dorsal, easter, gastrulation defective, pelle, snake, spaetzle, and toll), except for those that map proximal to an FRT site (tube) and those whose products act in the somatic follicle cells (pipe, nudel, and windbeutel). Three new loci were discovered, krapfen, seele, and weckle, which define new dorsal group genes. krapfen (kra, one allele) and seele (sel, two alleles) mutants show partially dorsalized [D1–D2, according to the terminology in ANDERSON and NUSSLEIN-VOLHARD 1986 ] phenotypes. After cleaning the original mutagenized chromosomes from associated lethal mutations, we found that both kra and sel mutants are homozygous viable and female sterile. We also showed that the original kra allele isolated in the screen is a hypomorphic allele and that subsequently generated null mutations at the kra locus show a completely dorsalized phenotype (CHARATSI et al. 2003 ). The third new locus, weckle (wek, three alleles) gives rise to strongly (D0–D1) dorsalized embryos and corresponds to a zygotically lethal complementation group. wek is allelic to l(2)35Ea, which encodes a putative zinc-finger transcription factor (ASHBURNER et al. 1999 ). We have identified point mutations in the l(2)35Ea gene in all three wek alleles, confirming that wek is indeed an allele of l(2)35Ea. The role of wek in establishing the dorsoventral axis is currently under investigation and will be published elsewhere (B. PRIESTER, unpublished results).

Ventralized embryos: Twenty-six mutations falling into 10 complementation groups show ventralized phenotypes. We isolated new alleles of the previously known genes cactus (cact, eight alleles), medea (med, four alleles), mothers against dpp (mad, four alleles), saxophone (sax, two alleles), thick veins (tkv, two alleles), and Toll (one ventralizing allele, as well as three dorsalizing alleles). Interestingly, we found that two of the four medea alleles give rise to a partially penetrant bicaudal phenotype in addition to ventralization of the embryo, suggesting a potential role of DPP/SMAD signaling in anterior-posterior polarity determination. Five cact alleles were identified in the screen on chromosome arm 2R by virtue of the haplo-insufficient dominant phenotype of the cact locus, while three cact alleles were found on the basis of their recessive phenotype in the 2L screen. Four new loci were identified, which show ventralized phenotypes in homozygous mutant GLCs. eclair (eca) and baiser (bai), each defined by a single allele, give rise to weakly ventralized embryonic phenotypes. Both eca and bai mutants are adult semilethal, and surviving homozygous females do not lay eggs, while only mosaic females produce embryos that show the eca or bai phenotype. The roles of the eca and bai genes in dorsoventral patterning will be described elsewhere (S. BARTOSZEWSKI, unpublished results). The third new locus in this group, croissant (cst; two alleles), corresponds to a viable and female-sterile complementation group. Embryos derived from cst GLCs or from homozygous mutant females show a U-shaped phenotype similar to saxophone (sax) mutants. The fourth new locus in this group, hoernchen (hrn), is defined by a single X chromosomal mutation, which is viable and female sterile. Embryos derived from hrn GLCs or homozygous mothers show a head-open and U-shaped embryonic phenotype, as well as gastrulation defects characteristic of ventralized mutants (data not shown).

Anterior-posterior patterning:
Anterior defects: We identified 12 mutants falling into 11 complementation groups that show either distinct anterior (head) defects or a bicaudal phenotype (mirror-symmetric duplications of the abdomen). Five alleles of staufen (stau), which show a distinct head defect, as well as posterior abdominal defects, were grouped in the "posterior defects" class. The anterior group mutants include two alleles of bicoid (bcd). Two new mutants falling into separate complementation groups, krake (krk, one allele) and tintenfisch (ttf, one allele), show strong anterior (thoracic) defects reminiscent of hunchback (hb) mutants. A third line, ziehharmonika (zih, one allele), shows a distinct head defect and is homozygous viable. The zih mutation was mapped to the cytological interval 43F–44D3-8. Six mutations that display either partially or completely penetrant bicaudal phenotypes were found. One line, 3R-103-30, shows a partially penetrant bicaudal phenotype and is an allele of bullwinkle (bwk; RITTENHOUSE and BERG 1995 ). A new locus, named alice (ali, one allele), shows a fully penetrant, recessive bicaudal phenotype (mirror-symmetric embryos with extended Filzkörper at both ends; F. SCHNORRER, unpublished data). gl(2L)179-26 GLCs give rise to short, small eggs, and embryos show a variable bicaudal phenotype. gl(2L)343-11 embryos either are bicaudal or show head defects. gl(3R)66-35 is a homozygous viable mutation, and embryos show a partially penetrant bicaudal phenotype and a weakly ventralized eggshell (fused dorsal appendages). gl(3R)70-22 GLCs show variable head defects, while a small fraction of the embryos are bicaudal. Anterior localization of bcd mRNA appeared normal in these embryos (F. SCHNORRER, unpublished data).

Posterior defects: Forty-three mutants falling into 16 complementation groups show posterior (abdominal) patterning defects. In addition to alleles of the known genes cappuccino (capu, four alleles), oskar (osk, four alleles), pumilio (pum, three alleles), spire (spir, seven alleles), staufen (stau, five alleles), tudor (tud, three alleles), and valois (vls, one allele), we identified at least two new loci involved in posterior patterning. An X-linked mutation, which we named napoleon (nap, one allele), affects formation of the abdomen, but not of the pole cells, suggesting that nap acts downstream of tudor at the level of pumilio and nanos in the posterior patterning system (H. KNAUT, personal communication). Mutants in a second new locus, which we named shorty (soy; five alleles), lack pole cells and show defects in abdominal segmentation. soy is a lethal complementation group. Another lethal complementation group composed of three alleles (2L-150-11, 2L-192-9, and 2L-257-19) shows variable deletions of abdominal segments and turned out to be allelic to brain tumor (brat), a factor that has been implicated in the translational control of maternal hunchback mRNA (CHAGNOVICH and LEHMANN 2001 ; SONODA and WHARTON 2001 ). Six mutations [gl(1)82-33, gl(2L)50-27, gl(2L)137-8, gl(2R)2-11, gl(2R)62-37, and gl(3R)60-38] show weak or partially penetrant posterior segmentation defects and complement each other, as well as all known posterior group genes located on the relevant chromosome arms.

Terminal defects: Twenty-one mutants falling into 10 complementation groups show patterning defects at the embryonic termini, the acron and the telson. We identified new alleles of the previously known genes kinase suppressor of ras (ksr, one allele; THERRIEN et al. 1995 ), ras1 (one allele; SIMON et al. 1991 ), l(1)polehole/draf (three alleles; AMBROSIO et al. 1989 ), son of sevenless (sos, two alleles; SIMON et al. 1991 ), torso (tor, five alleles; SPRENGER et al. 1989 ), and trunk (trk, four alleles; CASANOVA et al. 1995 ). We also isolated a mutation at the dshc locus, whose effects on the Torso and epidermal growth factor receptor signaling pathways we described previously (LUSCHNIG et al. 2000 ). Interestingly, dshc mutants are semilethal and homozygous females lay no eggs, while only germline mosaics produce embryos that show the terminal group phenotype. We isolated a new locus involved in terminal patterning, which we termed rumpf (rum, one allele). rum is a viable mutation and embryos derived from GLCs or from homozygous mothers display the amorphic Torso pathway phenotype. Mapping and complementation analysis of three X-linked mutations with terminal group phenotypes is currently in progress. Two of these mutations, X-145-32 and X-197-36, are allelic.

Segmentation defects: The largest number of mutants was grouped in this class. Forty-six mutants show various defects in segmentation, including gap-like, pair-rule-like, and segment polarity phenotypes and homeotic transformations, as well as other, less easily classified deviations from the normal segmental pattern. Due to the variety of different phenotypes, the classification as "segmentation defects" is somewhat artificial. We were able to assign 30 of the mutants in this class to a total of 17 complementation groups, including the previously known genes eyelid/osa (eld, five alleles; TREISMAN et al. 1997 ; VAZQUEZ et al. 1999 ), ftzf1 (four alleles; GUICHET et al. 1997 ), hopscotch (hop, three alleles; BINARI and PERRIMON 1994 ), marelle/stat92E (mrl, two alleles; HOU et al. 1996 ), and kismet (kis, one allele; DAUBRESSE et al. 1999 ). Two mutations, 3R-92-38 and 3R-112-6, which show a severe segmentation defect, are allelic to the gene encoding the 140-kD subunit of RNA polymerase II (rplII140/wimp). A mutation in this gene has previously been shown to reduce the transcription of a subset of segmentation genes and to cause a specific segmentation defect (PARKHURST and ISH-HOROWICZ 1991 ). A lethal complementation group with two alleles (2L-75-1 and 2L-193-35) shows a maternal-effect segmentation defect similar to that seen in amorphic even-skipped (eve) mutants. This complementation group turned out to be allelic to a gene that was recently described by three other groups and was named lilliputian (lilli) on the basis of growth defects seen in lilli mutant clones in imaginal tissues (SU et al. 2001 ; TANG et al. 2001 ; WITTWER et al. 2001 ). A new locus with two lethal alleles, named wollknäuel (wol), shows deletions and fusions of abdominal denticle belts, as well as a short head skeleton and Filzkörper. A locus defined by a single mutation (3L-22-3) shows a specific pattern of segmental deletions in the abdomen. We mapped the 3L-22-3 mutation to the cytological region 66C and identified five P-element-induced alleles at the locus (see supplemental Table 1). We cloned genomic DNA fragments and cDNAs adjacent to the P-element insertion sites (H. KNAUT, F. SCHNORRER, S. LUSCHNIG and C. NÜSSLEIN-VOLHARD, unpublished data). The gene affected by the mutations shows homology to the Caenorhabditis elegans gene egl-27 and to the human atrophin-1/drpla gene, the locus mutated in patients suffering from dentatorubral pallidoluysian atrophy (DRPLA), a rare hereditary neurodegenerative disorder (UENO et al. 1995 ). This gene has been recently described as a transcriptional corepressor involved in gap gene regulation in the early embryo and in regulation of teashirt expression in imaginal discs, and the gene has been named Drosophila atrophin (atro) or grunge (gug; ERKNER et al. 2002 ; ZHANG et al. 2002 ). Another mutation, gl(3R)53-6, shows abdominal segment fusions and head defects. The mutation interacts genetically with certain eld/osa alleles: gl(3R)53-6 is semilethal in trans to one eyelid/osa allele (3R-108-31). Combinations with other eld/osa alleles [eld(308), eld(3R-68-15), and eld(3R-51-26)] are viable, but adults display wing vein defects or wing blisters (data not shown). Five mutants [gl(2R)18-14, 2R-183-20, X-100-6, X-63-3, and X-97-26] show segment polarity phenotypes; among those, one line, 2R-183-20, is a costal-2 (cos) allele. Certain mutations in the segmentation class show a stronger phenotype when embryos lack both the maternal and zygotic contributions of the gene product (these mutants are indicated as "MZ" in supplemental Table 1): mutants in caudal (cad, one allele; MLODZIK and GEHRING 1987 ), costal-2 (cos, one allele; GRAU and SIMPSON 1987 ), and extra sex combs (esc, one allele; STRUHL 1981 ) were isolated on the basis of their weak maternal phenotypes. We also isolated an allele of even-skipped (eve, one allele; NUSSLEIN-VOLHARD and WIESCHAUS 1980 ), which has no maternal effect, but was found in the screen because of its partially penetrant dominant segmentation defect.

Morphogenetic defects: Thirteen mutants were grouped on the basis of defects in various morphogenetic processes, such as germband retraction, dorsal closure, and head involution. Mutants affecting these processes are likely to be underrepresented in our collection, as we did not specifically screen for such phenotypes. We genetically characterized two of these lines in more detail. stocherkahn (stoc, one allele) is a semilethal mutation that affects embryonic dorsal closure and head involution. Interestingly, despite its late phenotypic manifestation, stoc acts purely maternally and shows no zygotic rescue. Rare surviving homozygous females produce embryos with the same phenotype as stoc GLCs. We mapped the mutation to the cytological region 98E3–99A1/2. A second mutation, schraube (sbe, one allele), complements stoc and gives rise to twisted embryos with dorsal holes. sbe is homozygous viable.

Defects in epidermis development: We isolated 16 mutants that fail to produce a coherent cuticle (Fig 4). Mutant embryos show a characteristic reduction of the embryonic cuticle to small pieces and lack body landmarks such as the head skeleton, Filzkörper, and denticle belts. This phenotype has been associated with defects in apical-basal epithelial cell polarity caused by mutations in the genes bazooka, crumbs, stardust, scribble, and others (reviewed in MULLER 2000 ). We identified two new loci with similar phenotypes. The first locus, pschur (psu, four alleles), shows a strictly maternal requirement, as the phenotype is not rescued by a wild-type allele contributed from the father. Furthermore, the homozygous offspring of heterozygous parents are viable and sterile (three alleles, 2R-69-30, 2R-141-24, and 2R-265-2) or die late at the L2 larval stage (one allele, 2R-417-13). The psu mutations are allelic to a P-element insertion, l(2)k06403, in the atypical protein kinase C (apkc) gene. aPKC has been shown to be involved in the establishment of cell polarity in epithelial cells (WODARZ et al. 2000 ). The disintegrated cuticle phenotype of the second complementation group, goldstaub (gsu, two alleles), is unlikely to be caused by a defect in cell polarity, as the localization of the polarity marker -catenin to the subapical membrane is normal (data not shown). Rather, cellularization appears to be aberrant, as preblastoderm nuclei do not migrate properly to the plasma membrane (data not shown). Four additional mutants with disintegrated cuticle phenotypes were isolated in the 2L screen; each of these four mutants defines a separate complementation group. Three of these were mapped to cytological intervals defined by deficiency breakpoints (see supplemental Table 1).

View larger version (285K): In this window In a new window Download PPT slide   Figure 4. Embryonic phenotypes of the mutants. Cuticle preparations of GLC-derived embryos (or eggs) from mutants of the different phenotypic classes are shown. Anterior is to the left throughout. Mutants shown here are indicated by boldface type in supplemental Table 1.

Other embryonic phenotypes: Sixteen mutants that did not fit any of the phenotypic classes above were grouped in this category. Among those, five X-linked mutations show a neurogenic phenotype similar to Notch mutants (X-115-40, X-210-22, X-31-10, X-75-9, and X-87-21). One of these, X-31-10, shows a penetrant maternal-effect phenotype, while the four other lines are either zygotically rescuable maternal-effect or zygotic lethal mutations. Our screening procedure also allowed us to detect alterations in the autofluorescence of embryos. While the yolk of wild-type embryos fluoresces yellow when irradiated with blue light (450–490 nm; GFP excitation filter), three X-linked mutants (X-164-5, X-167-40, and X-187-38) instead fluoresce bright orange. This orange autofluorescence may be due to the accumulation of a metabolic product. Since these embryos hatch and look externally normal, the corresponding mutations are potentially useful as a positive, recessive marker for mitotic clones. The phenotypes of the remaining 10 lines in this group are described in supplemental Table 1.

Defects in eggshell morphology: Ten mutants show altered eggshell morphology. This class includes two alleles of fs(1)K10, which give rise to dorsalization of the eggshell and the embryo. A new complementation group, brontops (bop, two alleles), is viable and homozygous mutant females produce small eggs with fused or branched dorsal appendages. Mutants at a second new locus, sahnehäubchen (sah; one allele), produce dorsalized eggs and embryos. Other mutants produce ventralized eggs with fused dorsal appendages (3L-215-13, 3R-259-18, gl(3R)66-35, and 3R-69-15) or small eggs (3R-40-7). X-158-33 shows small, collapsed eggs with abnormally broad and branched dorsal appendages. We also found several additional X chromosomal mutants that showed this phenotype in GLCs generated using an ovo^D2 FRT chromosome, but did not show a phenotype when the GLCs were marked by the absence of maternal GFP. This suggests that the phenotype might be due to perdurance of the mutant Ovo^D2 protein in the GLCs, which may result in a "dumpless" phenotype (MEVEL-NINIO et al. 1996 ). These lines were discarded.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In this work we present the results of a new screen for genes involved in patterning the early Drosophila embryo, including those genes that had been missed in previous screens due to their essential roles for viability of the adult animal. The screen makes use of chemical mutagenesis and analysis of genetic mosaics as an efficient and unbiased means to isolate mutations. We have developed a system that allows screening for maternal-effect mutations after only one generation of breeding, in contrast to the three generations required for a classic female-sterile screen (SCHUPBACH and WIESCHAUS 1989 ; reviewed in ST. JOHNSTON 2002 ) and the two generations required for a screen using the Flp-DFS (ovo^D) system (CHOU and PERRIMON 1996 ). Since in our scheme mutation-bearing females are crossed to wild-type males, the paternal contribution may zygotically rescue the lack of a maternal gene product. Hence, autosomal mutations with a zygotically fully rescuable maternal effect could not be found by our screen. Only in the case of the X chromosome were we able to score zygotic mutations in hemizygous male embryos. This explains the significant fraction of zygotic embryonic lethal mutations, which we isolated in the X chromosome screen.

The F₁ scheme is efficient, as it allows rapid screening of a large number of individuals and significantly reduces the efforts of maintaining potentially mutant lines prior to actually screening them. In only few cases (4%) was a potential mutant lost because the respective F₁ female died, was sterile, or produced an insufficient number of progeny. Since the mutation-bearing F₁ females do not carry a balancer chromosome preventing meiotic recombination, a potential mutation could also be lost if an F₁ female produced only recombinant progeny devoid of the mutation. This does not appear to pose a frequent problem, as in most cases we were able to recover the respective mutation when a clear patterning phenotype was scored in the screen. However, in certain cases it was difficult to reliably score a phenotype among a very small number of GLC-derived embryos (often a single embryo), especially for certain phenotypic classes (embryos with holes or a little cuticle or embryos with defective head skeletons). In these cases, the originally observed phenotype was frequently not reproducible in the retest, where many GLC-derived embryos were scored. The rate of positive recovery was different for each chromosome arm screened (e.g., 23% for the 3L screen and 66% for the 3R screen; see Table 2), presumably reflecting differences in the genetic background and our ability to reliably score certain phenotypes. For instance, in the screen on the 3L arm, we frequently saw partially penetrant pair-rule-like segmentation defects, which are likely to be related to the presence of the h¹ mutation on the FRT chromosome used for mutagenesis (the same phenotype was seen at a low frequency also among embryos from the unmutagenized parental ru h th st FRT2A strain); these phenotypes were not reproducible in the retest.

The GFP-FRT chromosomes described in this work also provide a useful set of tools to mark mitotic clones in the germline and in various somatic tissues. In particular, this system can also be used to mark clones carrying mutations on different chromosome arms simultaneously and to induce GLCs in the background of dominant-female-sterile mutations (such as Tor⁴⁰²¹ or Tl^10b), which cannot be combined with the dominant-female-sterile ovo^D system (LUSCHNIG et al. 2000 ).

Strategies for mapping chemically induced mutations:
Although chemical mutagenesis has proven highly efficient in terms of allowing rapid isolation of large numbers of mutants, genetic mapping of chemically induced mutations has often remained the rate-limiting step in positional cloning projects. This is in part due to the relatively low map resolution of visible genetic markers and chromosomal deficiencies. The recent introduction of single-nucleotide polymorphisms (SNPs) as markers for genetic mapping in Drosophila should provide a valuable tool for facilitating the mapping of chemically induced point mutations (BERGER et al. 2001 ; MARTIN et al. 2001 ; NAIRZ et al. 2002 ). Fine-scale SNP mapping can be used to narrow down the position of a mutation once an approximate map interval has been defined using classic meiotic and/or deficiency mapping. High-resolution SNP maps have already been generated for some of the strains that were used in this work (MARTIN et al. 2001 ). In our screen, mutations were induced on FRT chromosomes carrying visible genetic markers distal to the FRT site. Hence, meiotic mapping can be done between the marked FRT chromosome and its unmarked parental version, thus maintaining the presence of the FRT throughout the mapping scheme. Mapping of most of the X chromosomal mutants from the screen is still in progress. To be informative, complementation tests using X chromosomal duplications or transpositions require that lethal mutations on a particular X chromosome can be rescued by a single duplication. Because many of the X chromosomal lines carry multiple lethal mutations, these chromosomes have to be cleaned by meiotic recombination prior to duplication mapping.

Numbers and saturation:
We describe here a collection of 232 mutants that show distinct embryonic phenotypes in homozygous mutant GLCs. The mutants were grouped into 10 different phenotypic classes (supplemental Table 1). By crossing all mutants with similar phenotypes to each other or to mutants in candidate genes, we were able to order 174 mutants into 86 complementation groups, corresponding to an average of two alleles per locus for these 86 groups. Complementation testing of the remaining 58 lines, including most of the X chromosomal mutations, is still in process.

The only criteria for the isolation of these mutants were (1) the inability of the mutant embryos to hatch and (2) a distinct phenotype recognizable in the embryonic cuticle. Using these criteria, we have identified new alleles of most of the previously known maternal-effect lethal loci: 45 of the 86 complementation groups correspond to previously described genes involved in embryonic patterning or morphogenesis. Alleles of most of the previously known maternal genes involved in embryonic patterning were isolated, suggesting a high degree of saturation was achieved in the screen. However, we did not obtain mutants in some previously known genes that we had expected to find, including cni, dos, dpar-1, drk, exu, grk, nos, put, vas, and swa. Some of those may have been missed because even amorphic mutations give rise to only subtle embryonic defects and the mutant embryos may be able to hatch (e.g., dos and drk; HOU et al. 1995 ; RAABE et al. 1996 ; LUSCHNIG et al. 2000 ). In the case of nos, it has been shown that most mutant alleles affect an essential function of the gene during oogenesis and result in complete sterility, while only few mutations located in a C-terminal region of the protein affect exclusively embryonic abdomen formation (ARRIZABALAGA and LEHMANN 1999 ). Similarly, DPP signaling has been shown to be required for germline stem cell maintenance in the ovary (XIE and SPRADLING 1998 ), and DPP pathway mutants produce only a small number of germline clone-derived eggs, which may explain our failure to isolate a mutation in the DPP receptor put. Correspondingly, for several other genes only certain hypomorphic alleles may result in a visible maternal-effect embryonic phenotype [e.g., fs(1)Nasrat and fs(1)polehole; JIMENEZ et al. 2002 ], or, in other cases, a given gene may have phenotypically distinct classes of alleles (e.g., capicua, also known as fettucine; GOFF et al. 2001 ; ROCH et al. 2002 ). It is not clear why we failed to isolate cni, grk, exu, and swa alleles.

Forty-one of the 86 complementation groups correspond to unknown or previously undescribed loci (on the basis of phenotype, map position, and/or complementation tests). Several of the remaining mutants that are not yet mapped or assigned to a complementation group are likely to represent new loci, as well. We characterized 21 of the 41 new complementation groups at the genetic or molecular level. It is interesting to note that the average mutation rate was lower for the new loci discovered here than for previously known loci: 122 mutants falling into 45 complementation groups (2.7 alleles per gene on average) correspond to previously known genes, while the remaining 52 mutants falling into 41 complementation groups (1.27 alleles per locus on average) represent new loci or possibly unusual alleles of previously known loci (see Table 4). In fact, of the 21 new loci that we studied in more detail, 14 are represented by single alleles. Moreover, in a secondary screen for new alleles of dshc and baiser, only one new dshc allele and no baiser allele was isolated from approximately 8000 mutagenized chromosomes (S. LUSCHNIG and S. BARTOSZEWSKI, data not shown). These findings underline the notion that certain genes may represent mutational "cold spots" that have been missed in genetic screens and show the advantage of efficient F₁ type screening schemes in picking up rare mutations. However, mutations in certain genes may result in reduced fecundity, as is the case for bai, eca, gsu, and tkv germline mosaics. Such mutations have a reduced chance to be picked up by our screen.

To attempt to estimate the degree of saturation that was achieved in the screen, one would have to evaluate each chromosome arm screen as a separate experiment. However, as a first approximation, assuming an approximately equal degree of saturation for each chromosome arm, we can use the obtained allele frequencies as a measure for saturation (see Fig 5, Table 3). The average overall allele frequency was 2.02 alleles per locus (174 mutations falling into 86 complementation groups), counting all mutants that were tested for allelism against various candidates (58 mutants, including many of the X chromosomal mutations, have not yet been tested for complementation and were not included in the calculation of allele frequencies). This number does not change significantly if we exclude the 10 X-linked complementation groups from the calculation and count only the autosomal complementation groups (156 mutants falling into 76 complementation groups; mean allele frequency = 2.05). Assuming a Poisson distribution of allele frequencies with a mean value of 2.02, the zero class would make up 13%, suggesting that we have achieved 87% saturation. However, the observed distribution of allele frequencies deviates from a random Poisson distribution, as single hits are overrepresented in the observed distribution, indicating that the actual degree of saturation is below the theoretical value.

View larger version (47K): In this window In a new window Download PPT slide   Figure 5. Analysis of allele frequencies. In the graph, the observed allele frequencies are compared to a calculated Poisson distribution for the mean allele frequency. Solid bars indicate the observed distribution of allele frequencies (number of alleles per locus) from the screens on all five chromosome arms. Open bars represent the calculated Poisson distribution for the observed mean allele frequency m = 2.02. The y-axis represents absolute numbers of loci.

It is of interest to compare our results with previous germline clone screens. Perrimon and coworkers (PERRIMON et al. 1989 , PERRIMON et al. 1996 ) carried out screens for maternal effects of X chromosomal and autosomal lethal loci. These mutations were induced by EMS (X chromosome screen) or by P-element insertions (autosomal screen). The percentage of cell-lethal mutations reported in these studies is similar to our results in the case of the X chromosome (40% in their study vs. 41.8% in our study; see Table 2), while autosomal cell lethals appear to be more frequent in our screens (31% in their study vs. between 36.8 and 50.8% in our study, depending on the autosomal arm screened; see Table 2). This may reflect the different nature of the mutagen used (P elements vs. EMS), as well as a difference in the screening procedure: Perrimon et al. used the DFS (ovo^D) technique to select for GLC-derived eggs, and they scored several females each carrying the same mutation. In our F₁ screen, GLC-derived eggs were not selected for, and only a single female could be scored for GLC-derived eggs. We may therefore overestimate the number of cell-lethal mutations by counting all females in which we fail to detect GLC-derived eggs. For the same reason, it is difficult to compare the number of mutants that show no maternal effect in their study with that in our study. Perrimon et al. analyzed a collection of 496 independent P-element lethal mutations on the autosomes for maternal effects. In total, they found a maternal effect in 25% of these mutations, and in many of these cases the maternal effect was fully or partially paternally rescuable (PERRIMON et al. 1996 ). A similarly high percentage of lines with paternally rescuable maternal effects was found in an independent screen of P-element-induced lethal mutations on the third chromosome (BELLOTTO et al. 2002 ). In our screen, a much smaller fraction of screened individuals showed a maternal effect. This is mostly due to the design of our screen, which was focused on genes acting before the onset of zygotic transcription, and which did not permit the isolation of fully paternally rescuable mutations on the autosomes. It was suggested that several genes involved in embryonic patterning had escaped from detection in the classic female-sterile screens because of their essential functions for viability. This holds true for many components of intracellular signaling pathways (e.g., the RAS-MAPK or JAK-STAT pathway). At least five (ali, gsu, soy, wek, and wol) of the 21 new loci studied in detail are essential for viability. For four loci (krk, nap, sah, and ttf) that are represented by a single allele only, we cannot exclude that lethality may be due to second-site mutations. However, half of the 21 loci characterized in detail are not strictly essential for adult viability: mutants in 11 (bai, cst, dshc, eca, bop, sbe, kra, rum, sel, stoc, and zih) of the 21 loci are homozygous viable or show only partially reduced viability, although some of these mutations may be hypomorphic alleles of essential genes. Overall, the percentage of nonessential loci is similar among the previously known loci (46.6% nonessential for viability; 21 of 45 groups) as compared to the newly identified loci (39% nonessential for viability; 16 of 41 groups; see Table 4). These arguments suggest that for certain genes their low mutation rate, rather than an essential role for viability, has precluded their isolation in earlier genetic screens. Interestingly, homozygous bai, eca, dshc, or psu females do not lay eggs, even though at least bai, eca, and dshc homozygotes produce mature oocytes. These genes appear to be required for egg laying due to functions in the somatic follicle cells or in the egg-laying apparatus. Thus, the embryonic phenotype seen in GLCs is "masked" in homozygous females due to earlier somatic functions of those genes, thus precluding their identification in classic screens for maternal-effect embryonic patterning mutants. Multiple requirements of a gene in the soma and the germline might be a rather common phenomenon, although only a few examples have been documented so far (e.g., cic; GOFF et al. 2001 ). The analysis of éclair showed that when the wild-type gene is expressed specifically in the follicle cells of homozygous éclair females, the egg-laying defect, but not the embryonic patterning defect, is rescued (S. BARTOSZEWSKI, unpublished results). It will be interesting to carry out a systematic mosaic analysis of egg-laying-defective mutants in order to separate somatic from germline functions of the respective genes. Several of the mutants from our screen display somewhat variable phenotypes or different classes of phenotypes. For example, dshc is required for patterning events in both the germline and somatic follicle cells, in addition to a later role in zygotic patterning (LUSCHNIG et al. 2000 ). Phenotypic variability seen in embryos from mosaic mothers may be due to variable occurrence (number, size, and position) of mutant follicle cell clones. In other cases, the paternal contribution accounts for phenotypic differences between zygotically mutant and zygotically partially "rescued" embryos, such as in the case of dshc, sos, draf, ras, and others.

Morphological screens allow the identification of genes on the basis of their earliest requirement during development. This implies that genes required rather broadly and at early stages, such as genes encoding cytoskeletal components, will mutate to early embryonic lethality, revealing relatively little functional information. Also, it is likely that proteins involved in the localization of bcd RNA are generally required for RNA transport; such genes are likely to mutate to cell lethality or to an early arrest phenotype. Examples include dynein, kinesin, and ncd (ENDOW et al. 1994 ; MCGRAIL and HAYS 1997 ). Using the embryonic cuticle as a readout of earlier patterning events, our screen did not pick up new components involved in bcd mRNA localization. However, a secondary screen using an in situ hybridization assay was performed to directly examine localization of bcd and osk mRNAs during oogenesis in mutants that showed early embryonic arrest phenotypes in the original screen (F. SCHNORRER, unpublished data). This approach revealed the roles of components of the microtubule cytoskeleton for bcd RNA localization during oogenesis (SCHNORRER et al. 2002 ). Thus, screening early arrest mutants using a direct visualization assay is an efficient way to identify genes with essential functions for RNA or protein localization.

What is the contribution of this work to the understanding of embryonic pattern formation? The screen led to the isolation of new components involved in the posterior (nap, soy), dorsoventral (sel, kra, wek, eca, bai, cst, and hrn), and terminal patterning (dshc, rum) systems. The genetic and molecular characterization of these genes is likely to reveal important new insights into all four maternal axial patterning systems.

Three new dorsal group genes, kra, sel, and wek, were identified by this study. We cloned the krapfen gene and found that it encodes an adaptor protein homologous to mammalian myd88. Kra/dMyd88 is required downstream of the Toll receptor to transmit the signal to the cytoplasmic protein Tube (CHARATSI et al. 2003 ). It will be of great interest to characterize the remaining two new dorsal group mutants, sel and wek, molecularly and to fit them into the dorsoventral signaling cascade. On the basis of genetic epistasis experiments, seele is a new component of the Toll pathway required upstream of the Toll receptor, while krapfen and weckle act downstream of Toll (CHARATSI et al. 2003 and data not shown). Two genes (gamma-tubulin37C and dgrip75) not previously known to be involved in localization of bcd RNA were identified in a secondary screen of the mutant collection using an in situ hybridization assay (SCHNORRER et al. 2002 ). Two new components (nap, soy) involved in posterior patterning have been isolated, and six others that complement known genes await further characterization. New components of the terminal system will allow studying aspects of receptor tyrosine kinase signaling. We previously reported the identification of mutants in the adaptor protein DSHC and its role in the Torso and EGFR signaling pathways (LUSCHNIG et al. 2000 ). We are currently investigating the role of a new component, rumpf, in terminal patterning. On the basis of genetic epistasis experiments, rum is required downstream of the Torso receptor (S. LUSCHNIG, unpublished results). Two new genes, bai and eca, act in the DPP pathway and appear to be required for secretion of TGF-ß receptors to the plasma membrane (S. BARTOSZEWSKI, unpublished results). These mutants will allow studies of new aspects of the cell