A Genetic Screen for Maternal-Effect Suppressors of decapentaplegic Identifies the eukaryotic translation initiation factor 4A in Drosophila

Abstract

The Dpp signaling pathway is essential for many developmental processes in Drosophila and its activity is tightly regulated. To identify additional regulators of Dpp signaling, we conducted a genetic screen for maternal-effect suppressors of dpp haplo-insufficiency. We screened ∼7000 EMS-mutagenized genomes and isolated and mapped seven independent dominant suppressors of dpp, Su(dpp), which were recovered as second-site mutations that resulted in viable flies in trans-heterozygous with dpp^H46, a dpp null allele. Most of the Su(dpp) mutants exhibited increased cell numbers of the amnioserosa, a cell type specified by the Dpp pathway, suggesting that these mutations may augment Dpp signaling activity. Here we report the unexpected identification of one of the Su(dpp) mutations as an allele of the eukaryotic translation initiation factor 4A (eIF4A). We show that Su(dpp)^YE9 maps to eIF4A and that this allele is associated with a substitution, arginine 321 to histidine, at a well-conserved amino acid and behaves genetically as a dominant-negative mutation. This result provides an intriguing link between a component of the translation machinery and Dpp signaling.

THE TGF-β signaling pathway is an evolutionarily conserved mechanism for cell-cell communication and plays many essential roles in the growth and development of vertebrates and invertebrates. Ligands of the TGF-β superfamily signal through a receptor complex consisting of type I and type II receptors, which are serine-threonine kinases. Smads are the immediate downstream phosphorylation targets of the type I receptor and are responsible in large part for transducing the signal elicited by the extracellular ligands to the nucleus (for reviews, see Padgett et al. 1998; Raftery and Sutherland 1999; Massague 2000).

The best genetically characterized member of the TGF-β family is the decapentaplegic (dpp) gene product of Drosophila melanogaster (Padgett et al. 1987). Dpp, most homologous to mammalian bone morphogenetic proteins 2 and 4, participates in a variety of biological events during Drosophila development. The earliest zygotic requirement for dpp function is patterning along the embryonic dorsal-ventral (D/V) axis (St Johnston et al. 1990; Ferguson and Anderson 1992b). In wild-type animals, dpp mRNA is expressed in the dorsal 40% of the early embryo (St Johnston and Gelbart 1987; Ray et al. 1991). Mutants lacking dpp function at this early stage of embryogenesis are completely ventralized, and partial loss-of-function mutants show progressive loss of dorsal markers according to their allelic strength (Irish and Gelbart 1987; Wharton et al. 1993). This suggests that a DPP activity gradient is required for the specification of the dorsal cells of the embryo (Wharton et al. 1993). The dpp locus is haplo-insufficient; i.e., the animal is unable to survive as a heterozygote. Embryos of the genotype dpp^null/dpp⁺ do not undergo proper D/V patterning and are missing the dorsal-most cells, the amnioserosa (Wharton et al. 1993). This observation demonstrates that the animal is extremely sensitive to the levels of the dpp gene product (Irish and Gelbart 1987; Wharton et al. 1993). Indeed, the early embryo exhibits exquisite dosage sensitivity to Dpp, and it has been shown that Dpp acts as a morphogen that patterns the embryo, with the highest levels of dpp activity leading to the specification of the dorsal-most cell fates, and lower levels leading to more lateral cell fates (Ferguson and Anderson 1992a; Wharton et al. 1993).

Similar to mammalian TGF-β signaling, Dpp signals through the type I receptors thick veins (tkv) and saxophone (sax) and the type II receptor punt (Brummel et al. 1994; Nellen et al. 1994; Penton et al. 1994; Letsou et al. 1995; Ruberte et al. 1995). The R-Smad homolog Mothers against dpp (Mad) and the Co-Smad Medea are the intracellular transducers that mediate Dpp signaling (Raftery et al. 1995; Sekelsky et al. 1995). In addition to the Smad proteins that positively transduce Dpp signals, negative regulators that act in feedback mechanisms have also been identified. For example, the product of the short gastrulation (sog) gene is produced ventrolaterally during early embryogenesis and antagonizes the function of Dpp by forming an inactive complex with it, thus contributing to the formation of the Dpp activity gradient (Francois et al. 1994; Holley et al. 1995; Neul and Ferguson 1998; Nguyen et al. 1998), although it also plays a positive role by transporting Dpp ligand to the dorsal regions of the embryo (Decotto and Ferguson 2001). It has been shown in the imaginal disc that Daughters against dpp (Dad) functions as an inhibitory Smad, which is induced by Dpp signals and in turn blocks Dpp signaling by negative feedback (Tsuneizumi et al. 1997; Inoue et al. 1998). In addition, a transcription repressor, encoded by the gene brinker (brk), functions in the nucleus as a negative regulator of Dpp signaling by inhibiting the transcription of Dpp target genes (Campbell and Tomlinson 1999; Jazwinska et al. 1999). Finally, the ubiquitin ligase DSmurf has been shown to negatively regulate Dpp signaling by promoting Mad degradation (Podos et al. 2001).

Mutagenesis screens have been used to identify new components of the Dpp pathway and have yielded lesions in novel genes of interest or interesting new mutations in previously known genes (Raftery et al. 1995; Sekelsky et al. 1995; Y. Chen et al. 1998). These previous screens utilized enhancement of a hypomorphic dpp phenotype (Raftery et al. 1995; Sekelsky et al. 1995) or suppression of a gain-of-function phenotype (Y. Chen et al. 1998). The components of the Dpp signaling pathway that were successfully identified in these enhancer and suppressor screens include gbb-60A, screw, Mad, and Medea (Raftery et al. 1995; Sekelsky et al. 1995; Y. Chen et al. 1998). Despite the discovery of these components, how Dpp signaling is precisely regulated and what biological processes it controls are still not completely understood.

To gain further insight into the mechanisms that govern Dpp signaling and to better understand the biological functions of the Dpp morphogen gradient in the early embryo, we sought to genetically pursue the genes that, when mutated, could influence the early Dpp function. Here we describe a genetic screen for maternal-effect suppressors of dpp haplo-insuficiency. This screen takes advantage of the exquisite dosage sensitivity of the dpp locus. D/V patterning of the early embryo appears to be the only developmental stage that is haplo-insufficient for dpp, and animals that survive through this stage require only one wild-type copy of dpp to continue normal development (Wharton et al. 1993). Thus, suppression of dpp haplo-insufficiency in the early embryo would result in adult viability of the dpp^null/dpp⁺ heterozygous progeny. Moreover, in the early stages of embryogenesis when Dpp is required for D/V patterning, many components mediating Dpp signaling are maternally contributed. Thus we designed a genetic screen for maternal-effect mutations that actually would allow recovery of both maternal and zygotic mutations. In theory, we expect to obtain loss-of-function lesions in negative regulators of Dpp signaling or gain-of-function mutations in positive regulators of Dpp signaling. In addition, we could also identify genes that may not be integral components of the Dpp pathway, but are essential for the early developmental programs that are controlled by Dpp. We screened nearly 7000 mutagenized chromosomes and identified seven independent dominant suppressors of dpp [Su(dpp)]. We present results demonstrating that one of the suppressors was due to a point mutation in the gene eukaryotic translation initiation factor 4A (eIF4A) and genetically behaved as a dominant-negative eIF4A allele. Since eIF4A is an obligate component of the translation initiation machinery and is required cell-autonomously for growth and cell division (Galloni and Edgar 1999), the identification of eIF4A as a suppressor of dpp is unexpected. Our results suggest that eIF4A might have a novel function antagonizing Dpp signaling or that eIF4A might be critically involved in an essential biological process regulated by Dpp in the early embryo.

MATERIALS AND METHODS

EMS mutagenesis screen for suppressors of dpp:

Drosophila males were mutagenized with ethyl methanesulfonate (EMS) according to standard method. EMS-treated isogenized Canton-S males were crossed to D³ gl³/TM3, Sb¹ Ser¹ females, and the female progeny with TM3, Sb¹ Ser¹ of this cross were collected. Virgins of these females were individually crossed to dpp^H46 Sp¹ cn¹/CyO-23 and the progeny of each cross were examined for the presence of dpp^H46 heterozygous survivors, which are recognized by the absence of CyO and presence of Sp¹ dominant markers (Figure 1). dpp^H46 is a dpp null allele associated with a deletion within the dpp gene; dpp^H46 heterozygous animals are completely lethal (St Johnston et al. 1990; Wharton et al. 1993). CyO-23 contains a second copy of the dpp^Hin region on the standard CyO balancer. This derived balancer, CyO-23, contributes sufficient additional dpp⁺ activity to rescue the haplo-insufficiency of balanced dpp null alleles (Padgett et al. 1993). Mutations were recovered by crossing the males of dpp^H46 heterozygous survivors to appropriate balancer strains. All crosses were performed at 25°.

Figure 1.

Mutagenesis screen of maternal-effect suppressors of dpp^H46. Isogenized Canton-S males were mutagenized and outcrossed to a line containing the TM3, Sb¹ Ser¹ third chromosome balancer. The progeny females with TM3, Sb¹ Ser¹ in the F₁ were individually crossed to dpp^H46 Sp¹ cn¹/CyO-23 males to test for maternal-effect suppression of dpp^H46 haplo-insufficiency. The presence of dpp^H46 survivors, which were recognized by the absence of CyO-23 and the presence of Sp¹ dominant markers in the F₂, would indicate suppression. Mutations were recovered by crossing the males of dpp^H46 survivors to appropriate balancer strains. The CyO-23 contains a second copy of the dpp^Hin region on the standard CyO balancer. A total of 6982 mutagenized chromosomes were scored.

Fly stocks:

dpp^H46 Sp¹ cn¹/CyO-23 was used to test the suppression of the dpp haplo-insufficient phenotype. The multiply marked second chromosome al¹ dp^ov1 b¹ pr¹ c¹ px¹ sp¹ or third chromosome ru¹ h¹ th¹ st¹ cu¹ sr¹ e^s ca¹ were used in meiotic mapping. The transposase stock Sp¹/CyO; Sb¹ Δ2-3/TM6, Ubx was used in P-element-induced male recombination. The “green balancers” for chromosome 2 or 3 (CyO or TM3 carrying Gal4-Kr.C, UAS-GFP.S65T, Bloomington Drosophila Stock Center) were used to balance the lethal mutations. Multiple deficiencies and P-element insertions within interesting regions, as well as various tissue-specific Gal4 lines, including the maternally expressed V37-Gal4, used in this study were from public Drosophila stock centers. The P-element-associated eIF4A alleles eIF4A¹⁰⁶⁹, eIF4A¹⁰⁰⁶, and eIF4A¹⁰¹³ are as described (Galloni and Edgar 1999). Mutant alleles of the Dpp pathway components used for testing complementation with lethal Su(dpp) mutations include tkv⁷, tkv^a12, lack^KG07014 (DSmurf), sax⁴, and Mad^k00237.

Cuticle preparations:

Embryos were collected and aged for a minimum of 24 hr. “Green balancer” chromosomes were used to facilitate the identification of homozygous embryos for a particular Su(dpp) mutation for cuticle preparation. Embryos were dechorionated in 50% chlorine bleach and mounted in Hoyer's lactate according to standard method. Embryos were visualized and photographed with a Zeiss Axiophot microscope using DIC or darkfield optics.

Antibody labeling:

Embryos were collected and dechorionated in 50% chlorine bleach and washed with plenty of water. The dechorionated embryos were transferred to a glass scintillation vial containing 50% n-heptane and 50% PEM-formaldehyde (0.1 m PIPES, 2 mm EGTA, 1 mm MgSO4, pH to 6.95 with HCl, mix 9:1 with stock solution of 37% formaldehyde) and vigorously shaken for 15–30 min at room temperature. After fixation, the embryos were devitellinized by shaking in heptane/cold methanol (1:1) mixture. For antibody staining, fixed embryos were rehydrated in 1× PBS for 10 min followed by blocking for 30 min in PBST (1× PBS, 0.1% TritonX-100) plus 1.0% BSA. The embryos were incubated with anti-Krüpple antibody (1:5000 dilution) overnight at 4°. Following extensive washing with PBST, the embryos were incubated with a secondary antibody (biotinylated anti-rabbit IgG, Vector Laboratories, Burlingame, CA) according to the manufacturer's instructions. Following the secondary labeling, the embryos were additionally incubated with the A:B mix from the Vectastain ABC kit (Vector Laboratories) and developed in diaminobenzidine (DAB) staining solution (0.5 mg/ml DAB in PBST, 0.01% H₂O₂) for 2–3 min. Stained embryos were dehydrated with ethanol and mounted in Euparal (Carolina Biological Supply) and photographed with a Zeiss Axiophot microscope using DIC optics.

Recombinational mapping:

Recombinants were made between the suppressor chromosomes and the multiply marked second chromosome al¹ dp^ov1 b¹ pr¹ c¹ px¹ sp¹ or third chromosome ru¹ h¹ th¹ st¹ cu¹ sr¹ e^s ca¹. Recombinant chromosomes were tested for suppression by crossing female recombinants to dpp^H46 Sp¹ cn¹/CyO-23 males and scoring for dpp^H46 heterozygous survivors in the progeny. P-element-induced male recombination (B. Chen et al. 1998) was employed to further map Su(dpp)YE9. The Su(dpp)YE9 chromosome was recombined to have a recessive visible marker on each side of Su(dpp)YE9 [al Su(dpp)YE9 sp in this case]. Then the transposase Δ2–3 was introduced to generate P-element-induced male recombination. F₁ recombinants were sorted by the loss of one of the recessive visible markers flanking Su(dpp)YE9. The recombinants were stocked and then tested individually for suppression of dpp^H46/+. For fine mapping against a set of P elements, Su(dpp)YE9 was recombined with a nearby P element, l(2)[k09923], with the mini-white⁺ marker to the same chromosome and this chromosome was used to recombine with a set of different P elements (mini-white⁺ was used as a visible marker). We selected for meiotic recombination events between the two P elements, the flies that had lost both copies of the mini-white⁺ gene. The recombinants were w⁻ and were tested individually with dpp^H46 Sp¹ cn¹/CyO-23 to score for the presence or absence of the dpp^H46 survivors.

Mapping with single nucleotide polymorphisms:

To use single nucleotide polymorphisms (SNPs) as recombinational markers, we first identified SNPs among isogenized tester strains by sequencing. The tester strains used were vri[k05901], l(2)k06502), and Canton-S [from which Su(dpp)YE9 is derived]. We utilized the Drosophila genome sequence to generate pairs of PCR primers within the chromosomal interval of interest and sequenced the PCR products. By comparing sequences of PCR products from different strains, we were able to determine the location of single nucleotide differences among these strains. Multiple SNPs were identified. SNP-JL2 was used in recombinant mapping as it is the most informative. A PCR-based technique, modified from Drenkard et al. (2000), was used to genotype SNPs in recombinants. Primers that contain a “3′-mismatched” nucleotide according to the single-nucleotide difference were used. A genotyping PCR contains four primers: two “3′-mismatched” primers at the SNP locus, each matching to only one tester strain but not to the other at the 3′ terminal nucleotide, and two genomic primers. Primers were designed to amplify different sizes of fragments from genomic DNA. For example, in the case of SNP-JL2, the two “3′-mismatched” primers are JL2mis5(A) (5′-ATAAACCCATTCGCACCCCA-3′) and JL2mis3(C) (5′-TGGAGTGAGTTGGTCAAGCGCG-3′). Underlining indicates the mismatched single nucleotide. The two genomic primers are JL26F2 (5′-TCATATATGAACGTAAGATTGCAGC-3′) and JL26R2 (5′-GCGGCAAGAACAGAAATTTTAA-3′). The PCR product of JL26F2 and JL2mis3(C) is ∼380 bp in size, the PCR product of JL2mis5(A) and JL26R2 is 180 bp in size. They are easily distinguished by running an agarose gel. Only one band is expected from a genome homozygous at the SNP locus and this band should have the same size as one of the tester strains (Figure 5A).

Figure 5.

Identification of Su(dpp)YE9 as a point mutation of gene eIF4A. (A) Structure of eIF4A protein is shown. The eight highly conserved amino acid sequence motifs are indicated. The N-terminal, helicase, and RNA-recognition domains are indicated. Amino acid sequence is shown for motif V and its flanking regions. The Su(dpp)YE9 locus had a single amino acid change at position 321 of the conceptual eIF4A protein from arginine (R) to histidine (H). This amino acid is next to the highly conserved motif V of DEAD-box-containing proteins. (B) Alignment of the highly conserved surrounding region of motif V of eIF4A protein from human to yeast. The numbers indicate the position in each primary amino acid sequence. The position of R-to-H change is indicated at the top.

Overexpression experiments:

A full length of wild-type eIF4A cDNA, GM14109 (GenBank accession no. AY069283), was purchased from Research Genetics (Huntsville, AL). A mutant eIF4A cDNA containing the R321H mutation (eIF4A^R321H) found in the Su(dpp)YE9 chromosome was isolated from Su(dpp)YE9 homozygous embryos by RT-PCR. GM14109 and eIF4A^R321H were cloned into the pUAST plasmid (Brand and Perrimon 1993), respectively, and transformed into y¹w^67c23 flies by standard techniques. To test whether overexpression of eIF4A can neutralize the ability of Su(dpp)YE9 to suppress dpp haplo-insufficiency, strains carrying the UAS-eIF4A on the X and the second chromosomes were obtained. Females of the genotype UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4/+ were heat-shocked at 37° for 2 hr/day for 3 consecutive days before they were individually crossed to dpp^H46 Sp¹ cn¹/CyO-23. The progeny of these crosses were examined for the presence of dpp^H46 survivors. Rescue of the lethality of Su(dpp)YE9 was carried out by similar means to examine the presence of Su(dpp)YE9/Su(dpp)YE9 homozygous or Su(dpp)YE9/Df survivors after inducing eIF4A expression from a transgene throughout embryonic and larval development by heat shock for 2 hr/day at 37°.

Hatching rate and lethal-phase studies:

In each cross, ∼500 embryos were collected on an apple juice plate with fresh yeast paste after 2 hr of egg lay at 25°. Then the plates were continuously cultured at 25°. The hatching rate and larval viability were checked every 8 hr thereafter.

DNA sequencing:

To determine whether Su(dpp)YE9 is associated with mutations in eIF4A, genomic DNA from Su(dpp)YE9 homozygous embryos was isolated. To identify Su(dpp)YE9 homozygous embryos, the Su(dpp)YE9 chromosome was balanced over a “green balancer” chromosome (see above). By using a fluorescent microscope, the embryos without GFP [homozygous for Su(dpp)YE9] were collected. The genomic DNA was amplified by using several sets of primers for the eIF4A coding region. The primer pairs that yielded the point mutation have the following sequences: 5′-GGCACCCTTTGCGATCTGTA-3′ and 5′-GCGGCTCAGCAGAAAAAATT-3′. Several separate PCR fragments were sequenced to confirm identified nucleotide change. The parental chromosome was sequenced as a control to confirm the mutation.

RESULTS

Isolation of maternal-effect suppressors of dpp haplo-insufficiency:

To screen for mutations that suppress the haplo-insufficiency for a dpp null allele, dpp^H46, we designed a scheme as shown in Figure 1. The mutagenized chromosomes were borne by the mother in each F₁ cross so that both maternal and zygotic-effect suppressors could be identified. The mechanics of the screen were straightforward, as the assay was adult viability of the dpp^H46/dpp⁺ class, which is normally lethal (see materials and methods). The presence of this class of adults in F₂ would indicate suppression of the lethality due to the haplo-insufficiency of dpp^H46.

In this screen, 6982 fertile females carrying a set of EMS-mutagenized chromosomes were individually crossed and tested first for dominant maternal-effect suppression of dpp^H46. We recovered seven independent suppressors of dpp [Su(dpp)], which, in their presence in the genome, gave rise to viable dpp^H46/dpp⁺ adults. In a control experiment (unmutagenized mothers), none of the 500 females tested produced any progeny with the genotype dpp^H46/dpp⁺ (>50,000 progeny flies were examined), indicating that the suppression of haplo-insufficiency of dpp^H46 was a result of mutagenesis rather than a genetic background effect. Among these seven Su(dpp) mutants, suppression rates varied in strength, showing a surviving dpp^H46/dpp⁺ progeny class ranging from 27.7% to 55.6% of their siblings bearing two doses of dpp⁺ (Table 1).

TABLE 1.

Suppressors of dpp haplo-insufficiency

Suppressor	Suppression rate (%)a	Homozygous viability/fertility	Requirement for suppression	Chromosomal location
Su(dpp)YE2	32.7 ± 5.3	Lethal	Maternal and zygotic	2R, between c and px
Su(dpp)YE5	27.7 ± 7.2	Lethal	Maternal and zygotic	2R, between c and sp
Su(dpp)YE9	30.0 ± 8.1	Lethal	Maternal and zygotic	2L, 26B1-2
Su(dpp)YE10	30.6 ± 8.7	Lethal	Maternal	3L, between h and th
Su(dpp)YE12	53.7 ± 11.1	Viable, male sterile	Maternal and zygotic	2L, between dp and b
Su(dpp)YE28	42.0 ± 9.6	Viable, female sterile	Maternal	2L, between dp and pr
Su(dpp)YE31	55.6 ± 9.3	Lethal	Zygotic	2L, between dp and b

^aFor a cross between a female carrying mutagenized chromosomes and dpp^H46 Sp¹ cn¹/CyO-23 males, the suppression rate was calculated by the number of dpp^H46/+ progeny class (Sp¹ Cy⁺) compared with their CyO-23 siblings that did not inherit dpp^H46 (Sp⁺ Cy). The suppression rate was derived from at least 20 independent crosses for each suppressor, each cross producing at least 200 progeny.

Most the Su(dpp) mutations, except Su(dpp)YE10 and Su(dpp)YE28, were required both maternally and zygotically, as all the adult survivors of the dpp^H46/dpp⁺ class also contain the Su(dpp) mutation in the genome. In the case of Su(dpp)YE10 and Su(dpp)YE28, some survivors did not inherit the Su(dpp) mutations zygotically. This indicates that the mother's genotype determines the survival of the progeny, and thus Su(dpp)YE10 and Su(dpp)YE28 can act as maternal-effect suppressors (Table 1).

To determine whether these Su(dpp) mutations can also suppress dpp^H46/+ only when introduced zygotically, we performed the reciprocal cross, with males bearing the suppressor mutation and mated to dpp^H46 Sp¹ cn¹/CyO-23 females. In all the crosses in which the father bears a Su(dpp) mutation, only Su(dpp)YE31 produced surviving dpp^H46/dpp⁺ progeny, indicating that Su(dpp)YE31 is a zygotic suppressor (Table 1).

To identify the genes responsible for these Su(dpp) mutations, we first determined the chromosomal locations of these mutations by standard meiotic mapping (see materials and methods). Interestingly, all except Su(dpp)YE10 were located on the second chromosome (see Table 1). For the five Su(dpp) mutations associated with recessive lethality, meiotic mapping indicated that the lethality cosegregates with the suppressor mutation. Complementation tests among these suppressors indicated that none of them carried the same recessive (lethal) mutation. We also tested complementation with mutant alleles of known Dpp pathway components located on the second chromosome, such as of dpp, Mad, sax, tkv, and DSmurf. None of the lethal Su(dpp) mutations mapped to the second chromosome were allelic to the known Dpp pathway genes tested (data not shown).

Phenotypes of Su(dpp) mutations:

To understand the functions of the Su(dpp) mutations, we sought to examine their mutant phenotypes. Since these mutations suppressed the lethality associated with halving the dpp gene dosage, we expect that the Su(dpp) mutations would cause an increase in Dpp signaling. Dpp is required for embryonic D/V patterning. Embryos lacking early Dpp signaling (e.g., zygotic null for dpp or maternal null for Mad) exhibit completely ventralized cuticles (Das et al. 1998). Interestingly, embryos with as many as four dosages of the zygotic dpp gene are viable and produce normal-looking cuticles (Wharton et al. 1993). Therefore, cuticular morphology may not reflect increases in Dpp signaling. However, it has been shown that cells that originate from the dorsal-most region, the amnioserosa, are specified by early Dpp signaling and that the number of amnioserosa cells correlates with the strength of Dpp signaling (Wharton et al. 1993). Indeed, while wild-type embryos have an average of 130 amnioserosa cells, embryos with a single copy of dpp⁺ contain <90 amnioserosa cells, and those with four doses of dpp⁺ have as many as an average of 325 amnioserosa cells (Wharton et al. 1993).

Among these suppressors, Su(dpp)YE2, Su(dpp)YE5, Su(dpp)YE9, Su(dpp)YE10, and Su(dpp)YE31 were associated with embryonic lethality when homozygous. The other two, Su(dpp)YE12 and 28, are homozygous viable and without discernible morphological defects. We examined the cuticles and the amnioserosa cell number for the five lethal Su(dpp) mutations. The cuticles of these embryos exhibited a range of defects, including fusion or missing ventral denticle bands [Su(dpp)YE2, Su(dpp)YE5, and Su(dpp)YE10], defective head skeletons [Su(dpp)YE9], and dorsal holes [Su(dpp)YE31] (Figure 2). Most of these cuticle defects are not interpretable with regard to Dpp signaling except for the dorsal holes in Su(dpp)YE31 homozygous embryos, which are consistent with a disruption of Dpp signaling. However, holes in the dorsal epidermis are reminiscent of decreased rather than increased Dpp signaling in late embryogenesis during dorsal closure (Penton et al. 1994), which runs contrary to the expected nature of the Su(dpp) mutations (see below).

Figure 2.

Cuticle phenotypes and amnioserosa cells of Su(dpp) mutants. Larval cuticles (A, C, E, G, I, and K) and amnioserosa cells of stage 10 embryos as revealed by immunostaining with anti-Krüpple antibody (B, D, F, H, J, and L). All the embryos are shown as lateral views with anterior to the left and dorsal up except in K, which was a dorsal view. (A and B) Wild-type cuticle and amnioserosa cells. (C and D) Su(dpp)YE2 homozygous embryos with segmentation defects in cuticle (C) and an increased number of amnioserosa cells. (E and F) Su(dpp)YE5 homozygous embryos with mild segmentation defects in cuticle (E) and an increased number of amnioserosa cells. (G and H) Su(dpp)YE9 homozygous embryos show severely defective head skeletons and mild segmentation defects in cuticle (G) and an increased number of amnioserosa cells. (I and J) Su(dpp)YE10 homozygous embryos most commonly exhibit a deletion of the A4 denticle band (I) and an increased the number of amnioserosa cells. (K and L) Su(dpp)YE31 homozygous embryos exhibit defects in the dorsal epidermis ranging from a dorsal hole (K) to completely dorsally open (not shown). These embryos have a decreased number of amnioserosa cells (L).

To determine whether the Su(dpp) mutations caused increased Dpp signaling, we examined cell numbers of the amnioserosa, which more accurately reflect the strength of Dpp signaling. Indeed, we found that many of the Su(dpp) mutant embryos exhibited increased amnioserosa cell numbers, which is often observed when Dpp signaling is increased (Wharton et al. 1993). We examined 10 embryos for each genotype and found 170 ± 10.2 amnioserosa cells in wild-type embryos. In contrast, the number of amnioserosa cells was increased in Su(dpp)YE2, Su(dpp)YE5, Su(dpp)YE9, and Su(dpp)YE10 homozygous embryos to ∼267.3 ± 40.5, 236.0 ± 12.5, 236.0 ± 10.6, and 284.7 ± 32.7, respectively (Figure 2). Consistent with our interpretation that the dorsal hole phenotype of Su(dpp)YE31 reflects a decrease in Dpp signaling, these embryos also had fewer amnioserosa cells, ∼128.7 ± 13.3 in total. However, as with all the other Su(dpp) mutants, we found more amnioserosa cells in Su(dpp)YE31 heterozygous embryos than in wild-type control embryos (not shown). This could be explained if the wild-type product of Su(dpp)YE31 functions as a dimer (see discussion). These results are thus consistent with the interpretation that the Su(dpp) mutations increased Dpp signaling.

Fine mapping of Su(dpp)YE9:

To identify the genes mutated by the Su(dpp) mutations, we carried out more detailed mapping for Su(dpp)YE9, because it was associated with highly penetrant embryonic lethality as homozygotes and initial meiotic mapping indicated that the Su(dpp)YE9 mutation and lethality cosegregate. Although Su(dpp)YE9 is associated with homozygous lethality, we mapped the suppression of dpp^H46/+ instead of the lethal mutation, because it is possible that the lethal mutation is separate from the suppressor mutation. Both the lethality and the suppressor activity associated with Su(dpp)YE9 mapped to the same region/locus (see below), suggesting that the same mutation caused both the lethality and the suppression of dpp^H46 haplo-insufficiency.

To identify the gene mutated in Su(dpp)YE9, we used the following methods to further determine the exact chromosomal location of Su(dpp)YE9, which has been mapped to between dp and b on the left arm of the second chromosome by meiotic mapping. First, we generated a set of P-element-induced male recombination products (see materials and methods) to analyze the location of Su(dpp)YE9 relative to known P elements in the genomic region between dp and b. The five P elements used in this analysis and the results of their suppression test are summarized in Table 2.

TABLE 2.

Mapping of Su(dpp)YE9 by P-element-induced male recombination

P element	Cytological location	Male recombinant (N)a	Suppressionb	Location of Su(dpp)YE9 relative to the P element
vri [k05901]	25D4–5	al+sp− (4)	Yes	Proximal
al−sp+ (3)	No
l(2) [k09923]	26D1–2	al+sp− (9)	No	Distal
al−sp+ (2)	Yes
l(2) [k09022]	27C1–2	al+sp− (5)	No	Distal
al−sp+ (1)	Yes
l(2) [k05404]	28C7–9	al+sp− (5)	No	Distal
al−sp+ (1)	Yes
hoip [k07104]	30C1–2	al+sp− (5)	No	Distal
		al−sp+ (1)	Yes

Su(dpp)YE9 was recombined to have a recessive visible marker, al and sp, on each side. Five P elements between dp and b were used to generated a set of P-element-induced male recombination products, identified as al⁺sp⁻ or al⁻sp⁺. The recombinant flies were stocked and then tested for suppression of dpp^H46/+ individually. Analysis of the recombinants (see text) allowed us to deduce that Su(dpp)YE9 lies between two P elements: vri[k05901] and l(2)[k09923].

^aN, the number of recombinants.

^bSuppression of dpp^H46/+ lethality.

Because male recombination events supposedly happen only near the P-element insertion site (B. Chen et al. 1998), we could determine whether Su(dpp)YE9 was located near the centromere-distal or proximal side of a particular P element. Analysis of these recombinants allowed us to assign Su(dpp)YE9 to a location between two P elements: vri[k05901] and l(2)[k09923], which are ∼1.1 Mb apart in the cytological region 25D5-26D1 on the left arm of the second chromosome.

Next, we conducted fine meiotic mapping by classical recombination with a set of P elements to further narrow down the interval containing Su(dpp)YE9 to a manageable size. The reason that we could not continue to use P-element-induced male recombination was that the recombinant frequency was extremely low in our hands (<1/2000) and very often the recombinants were not healthy enough for further testing and stocking. This might be due to the deleterious lesions that might have arisen during P-element-induced male recombination. Thus, we resorted to classical meiotic recombination mapping between Su(dpp)YE9 and different P elements located within the 25D5-26D1 region. We used the mini-white⁺ gene of the P elements as a visible marker for meiotic recombination (see materials and methods). The white⁻ recombinants that had lost both copies of the mini-white⁺ genes were tested for suppression of dpp^H46/+ and lethality in trans-heterozygous with Su(dpp)YE9. We found that the suppression always cosegregated with lethality. The six P elements used in this analysis and the results of their suppression test are summarized in Figure 3A. Such mapping led us to narrow the Su(dpp)YE9 mutation to an ∼74-kb region between chic[k13321] and EP(2)2273 in 26A1-B3 (Figure 3B).

Figure 3.

Fine meiotic mapping of the Su(dpp)YE9 by P-element recombination. The Su(dpp)YE9 was recombined with l(2)[k09923] and crossed to a set of different P elements. Female flies with the genotype of Su(dpp)YE9 l(2)[k09923]/P element were crossed with w⁻ males. We scored the w⁻ recombinant flies in the next generation for those that had lost both copies of the mini-white⁺ gene, resulting from a recombination event between l(2)[k09923] and the P element. These w⁻ recombinants were then individually crossed with dpp^H46 Sp¹ cn¹/CyO-23 males to test for suppression and lethality. (A) P elements utilized for the fine meiotic mapping of Su(dpp)YE9. They are within the 1.1-Mb region that we had narrowed down by male recombination mapping. The number of w⁻ recombinants obtained, the results of suppression of dpp^H46/+ lethality, and the location of Su(dpp) YE9 relative to each P element are indicated. In the case of chic[k13321], 34 of the 37 w⁻ recombinants maintained the ability to suppress dpp^H46/+ and the remaining 3 did not. These data indicate that Su(dpp)YE9 was proximal to chic[k13321]. In the case of EP(2)2273, all the w⁻ recombinants (120 in total) retained the suppression, suggesting that Su(dpp)YE9 is distal to EP(2)2273. (B) A scheme to illustrate the location of the P elements and the distance between them. The large arrows indicate the mapping data for Su(dpp)YE9, which was narrowed down to a 74-kb region in 26A1-B3.

Finally, we used a SNP to conduct fine mapping (Drenkard et al. 2000) (see materials and methods). We identified a SNP (SNP-JL2) in the middle of the sequence between chic[k13321] and EP(2)2273. As shown in Figure 4A, a set of four PCR primers consisting of two “3′-mismatched” primers at the SNP locus and two genomic primers were used (see materials and methods). Only one band was amplified in each reaction that reflected the genotype of the fly strain that we tested. By analyzing PCR bands amplified from the white⁻ recombinants obtained from the fine meiotic mapping described above, we were able to determine that the genomic location of Su(dpp)YE9 was distal to SNP-JL2. As summarized in Figure 4B, three types of recombination events occurred in the white⁻ recombinants, namely type I, II, and III. In type I recombination, the white⁻ recombinants lost the suppression and had the sequence of the P element at SNP-JL2. In type II recombination, the white⁻ recombinants retained the suppression but still had the sequence of the P element at SNP-JL2. In type III recombination, the white⁻ recombinants retained the suppression and had the sequence of Canton-S, from which Su(dpp)YE9 is derived. For example, in the vri[k05901] group of white⁻ recombinants, 13 of them had the vri[k05901] sequence at SNP-JL2, 12 of them lost the suppression (type I), and only 1 of them retained it (type II). Nine of them had the Canton-S sequence at SNP-JL2 and all of them retained the suppression (type III). Similar results were found in the l(2)[k06502] group of white⁻ recombinants. Such mapping allowed us to further narrow down Su(dpp)YE9 to within a 33-kb genomic fragment between chic[k13321] and SNP-JL2 in the cytogenetic location 26B1-2 (Figure 4B).

Figure 4.

SNP mapping of the Su(dpp)YE9. (A) The P-element strains such as vri[k05901] and l(2)k06502) contain A at SNP-JL2 and are designated as T1. T2 strains such as Canton-S [from which Su(dpp)YE9 is derived] contain C at this position. The sequences surrounding SNP-JL2 are shown, and the single different nucleotide between the these two stains is underlined. The two “3′-mismatched” primers at the SNP-JL2 locus are JL2mis5(A) and JL2mis3(C), as indicated by long arrows. Each primer matches to only one strain, but not the other at the 3′ terminal nucleotide. When combined with two genomic primers, JL26F2 and JL26R2, DNA fragments of 380 or 180 bp will be amplified (see materials and methods). The first two lanes in the agarose gel are PCR amplification from Canton-S (T2) and vri[k05901] (T1) genomic DNA. Only one band was amplified from a genome homozygous at the SNP-JL2 locus and this band had the same size with either T1 or T2, reflecting the polymorphic variant of the original chromosomes. The rest of the lanes are examples of the w⁻ recombinants. (B) Summary of the SNP mapping data. Three types of recombination events assumed to have occurred in the white⁻ recombinants are indicated. The data from the suppression test and the corresponding genotype of these white⁻ recombinants are shown in the table (see text). These results indicate that Su(dpp)YE9 is located to the distal side of SNP-JL2. (Bottom) Su(dpp)YE9 was narrowed down to a 33-kb region as indicated.

Su(dpp)YE9 is associated with an R321H substitution in eIF4A:

The 33-kb genomic region in 26B1-2 encompasses six genes according to the FlyBase Consortium (2003) (http://flybase.bio.indiana.edu/). Among these, we determined from the following experiments that Su(dpp)YE9 is associated with a mutation in eIF4A. First, Su(dpp)YE9 failed to complement all the existing lethal mutant alleles of eIF4A. These alleles include eIF4A¹⁰⁶⁹, eIF4A¹⁰⁰⁶, eIF4A¹⁰¹³, eIF4A⁰²⁴³⁹ (Galloni and Edgar 1999), eIF4A^k01501 (Bloomington Drosophila Stock Center), and all the new eIF4A alleles that we recently generated, including eIF4A^24A03, eIF4A^10A07, eIF4A^10A12, and eIF4A^09A09 (Huet et al. 2002; F. Huet and W. M. Gelbart, unpublished data). All of the above alleles are P-element alleles and most of them are associated with an insertion in the first intron of the eIF4A gene except eIF4A¹⁰⁰⁶, which is in the second intron. Moreover, the deficiencies that disrupt eIF4A, Df(2L)E110, and Df(2L)GpdhA were also lethal in trans with Su(dpp)YE9. Second, Su(dpp)YE9 is associated with a point mutation in the eIF4A coding region. We amplified genomic DNA containing the eIF4A coding region from Su(dpp)YE9 homozygous embryos and sequenced several independent clones (see materials and methods). We found a single nucleotide change from G to A at the position 1027 of the eIF4A cDNA coding sequence in the Su(dpp)YE9 chromosome, which was not found in the parental chromosome. This mutation would cause an amino acid change at position 321 from arginine (R) to histidine (H) (Figure 5A). Finally, we generated transgenes that express a wild-type eIF4A cDNA under the control of the UAS promoter (UAS-eIF4A). When we overexpressed UAS-eIF4A using a heat-shock-inducible Gal4 driver (see materials and methods ), we were able to rescue to a limited extent the lethality associated with Su(dpp)YE9 hemizygotes (Table 3, row 3). Although the percentage of rescue was low (1.4%), this number was significant, as no Su(dpp)YE9/Df survivors were recovered in the control cross (Table 3, row 2). These results strongly suggest that the lethal mutation associated with Su(dpp)YE9 is due to disruption of eIF4A gene function.

TABLE 3.

Effects of expressing eIF4A transgenes on suppression of dpp^H46 haplo-lethality or Su(dpp)YE9 viability

Parental genotype			Viability of F1 progeny
Mother	Father	Heat shock	dppH46/+(%) (N)a	Su(dpp)YE9/Su(dpp)YE9 or Df (%) (N)a
UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4	UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4	+b	0 (2563)
UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4	Df(2L)E110/CyO	−	0 (>5000)
UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4	Df(2L)E110/CyO	+b	1.4 (2126)d
Su(dpp)YE9/SM6a	dppH46 Sp1/CyO-23	30.0 (>4000)
UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4	dppH46 Sp1/CyO-23	−	28.6 (563)
UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4	dppH46 Sp1/CyO-23	+c	1.6 (628)d
UAS-eIF4AR321H/+	dppH46 Sp1/CyO-23	0 (>5000)
Df(2L)GpdhA/+	dppH46 Sp1/CyO-23	3.2 (968)
Df(2L)GpdhA/+; UAS-eIF4AR321H/+	dppH46 Sp1/CyO-23	3.4 (1116)
Df(2L)GpdhA/+; UAS-eIF4AR321H/V37-Gal4	dppH46 Sp1/CyO-23		9.6 (2314)d

^aN, total number of progeny flies scored from separate crosses.

^bAs in c, except the progeny were subject to the same daily heat shock for the entire larval and pupal growth stages. V37-Gal4 is a maternally expressed Gal4 line (Brand and Perrimon 1993).

^cFemales of the genotype UAS-elF4A; su(dpp)YE9/CyO; hs-Gal4/+ were heat-shocked at 37° or 2 hr/day for 3 consecutive days before they were individually crossed to dpp^H46 Sp¹ cn¹/CyO-23.

^dStatistical significant difference from the control (Student's t-test).

To determine whether the R321H mutation in eIF4A caused the suppression of dpp^H46 haplo-insufficiency, we carried out the following experiments. First, we found that overexpressing eIF4A neutralized the ability of Su(dpp)YE9 to suppress dpp^H46/+ (Table 3). We tested UAS-eIF4A; Su(dpp)YE9/CyO; hs-Gal4/+ females for suppression of dpp^H46/+. Without heat shock, these females had a suppression rate similar to Su(dpp)YE9/SM6a (Table 3, row 5). Following heat-shock induction of eIF4A expression, the suppression rate was reduced to 1.6% (Table 3, row 6). Second, females heterozygous for a deficiency uncovering the eIF4A locus partially suppressed dpp^H46 haplo-insufficiency (Table 3, row 8), which was never observed in control crosses (Table 3, row 7; also see control for the screen). Third, we generated an eIF4A mutant transgene with the R321H mutation (UAS-eIF4A^R321H) identified in the Su(dpp)YE9 chromosome. Although we were unable to produce any phenotype that can be interpreted as due to increased Dpp signaling by overexpressing UAS-eIF4A^R321H with various Gal4 drivers (not shown), when UAS-eIF4A^R321H was expressed by a maternal Gal4 driver, we observed a threefold increase in the ability of flies hemizygous for the eIF4A region to suppress dpp^H46/+ (Table 3, bottom row). Finally, as with most of the Su(dpp) mutations, embryos zygotically mutant for eIF4A exhibited an increased number of amnioserosa cells. We tested eIF4A¹⁰⁰⁶, a strong or null allele due to a P-element insertion (Galloni and Edgar 1999), and found that eIF4A¹⁰⁰⁶/Df(2L)GpdhA and eIF4A¹⁰⁰⁶ homozygous embryos had on average 218 ± 12.7 and 204 ± 9.4 amnioserosa cells (n = 10), respectively, as compared with 170 ± 10.2 in wild type. Taken together, these results suggest that the suppression of dpp^H46/+ by Su(dpp)YE9 was due to a mutation in eIF4A and that Su(dpp)YE9 was an allele of eIF4A. We therefore changed the Su(dpp)YE9 mutant allele to eIF4A^YE9.

Su(dpp)YE9 is a dominant-negative allele of eIF4A:

In contrast to eIF4A^YE9, which dominantly suppressed the lethality associated with dpp^H46/+ flies at a rate of 30% (see Table 1), none of the P-element alleles of eIF4A that we tested suppressed dpp^H46 heterozygotes, and the deficiencies that remove eIF4A exhibited much weaker suppression (3.2%; Table 3, third row from bottom). This property of eIF4A^YE9 could be explained if eIF4A^YE9 had dominant-negative effects. To determine the nature of the eIF4A^YE9 allele, we analyzed the lethal phase of eIF4A^YE9 homozygotes and eIF4A^YE9 in trans with its deficiency or other eIF4A alleles. First, we found that the phenotypes of eIF4A^YE9 homozygotes were stronger than those of its hemizygotes. Embryos produced by eIF4A^YE9/+ flies had the lowest hatching rate, 72.3% (Table 4), which was less than the expected rate (75%) if eIF4A^YE9 mutants were embryonic lethal as homozygotes. When crossed to a deficiency, however, the hatching rate was higher than expected if eIF4A^YE9/Df animals were embryonic lethal (Table 4). The difference in hatching rates, although small, was statistically significant and reproducible (P = 0.03; Student's t-test). Judging from the low hatching rate, the lack of any hatching eIF4A^YE9 homozygotes (see materials and methods), and the defective cuticles (see Figure 2), we deduced that eIF4A^YE9 homozygotes were completely embryonic lethal. eIF4A^YE9/Df(2L)E110 animals, however, could survive until the first instar larvae stage. Embryos from a cross between eIF4A^YE9/+ and Df(2L)E110/+ had a hatching rate of 78.3% and very few of the dead embryos exhibited cuticle defects. Second, eIF4A^YE9 strongly enhanced the phenotype of eIF4A⁰²⁴³⁹ (also known as eIF4A¹⁶²), the weakest P-element-induced eIF4A allele available (Galloni and Edgar 1999). Thus eIF4A^YE9/eIF4A⁰²⁴³⁹ larvae died at first or second instar stages, while homozygous eIF4A⁰²⁴³⁹ larvae lived to the third instar stage and eIF4A⁰²⁴³⁹/Df(2L)E110 animals lived until the second or third instar stages (Table 4). Finally, eIF4A^YE9/+ larvae exhibited obvious growth delays. When eIF4A^YE9/CyO flies were mated to wild-type flies, the eIF4A^YE9/+ progeny eclosed 2 days later than their balancer-carrying siblings. This dominant growth delay was not found for any existent eIF4A alleles or its deficiencies (not shown). Therefore we concluded that eIF4A^YE9 was a dominant-negative allele of eIF4A. This is consistent with the nature of the amino acid change encoded by eIF4A^YE9 (see discussion).

TABLE 4.

Lethal phase of eIF4A allelic combinations

Mother	Father	Hatching rate (Na)	Lethal phase and growth rate (25°)	Cuticle phenotype	Mutation
eIF4AYE9/+	eIF4AYE9/+	72.3% (400)	eIF4AYE9/eIF4AYE9: embryonic lethal eIF4AYE9/+ growth delay	Anterior defect: head malformation, anterior open, mouth hook underdeveloped	R321H
eIF4AYE9/+	Df(2L)E110/+	78.3% (400)	eIF4AYE9/Df(2L)E110: late embryonic and first instar larval lethal	Very few anterior defect	Deficiency
eIF4AYE9/+	eIF4A02439/+	92.5% (400)	eIF4AYE9/eIF4A02439: embryonic viable, first and second instar larval lethal	Normal	First intron of eIF4A
eIF4A02439/+	Df(2L)E110/+	95.5% (400)	eIF4A02439/Df(2L)E110: second or third instar larval lethal	Normal
eIF4A02439/CyO	eIF4A02439/CyO	NDb	eIF4A02439/eIF4A02439: third instar larval lethalc	ND

^aTotal embryos counted.

^bNot determined.

^cFrom Galloni and Edgar (1999).

DISCUSSION

The earliest known function of Dpp is patterning along the D/V axis of the early embryo, a process that is exquisitely sensitive to the concentration of Dpp (Irish and Gelbart 1987; St Johnston et al. 1990). It has been shown that the wild-type dosage of Dpp is essential for the correct expression of a set of genes that are important for the differentiation of the dorsal-lateral cell fates (Wharton et al. 1993). However, the molecular mechanisms underlying the biological processes of cell fate determination and differentiation in the dorsal embryonic region remain largely unknown.

We report here a genetic screen based on the dosage sensitivity of the early embryo to the dpp locus and the identification of a series of mutations that were able to suppress the dpp haplo-insufficiency phenotype. The molecular identity of one of the suppressors, eIF4A, has provided new insight into the function of Dpp signaling in the patterning of the embryonic body axis of Drosophila, suggesting an involvement of the translation machinery in the biological functions of Dpp signaling in the early embryo.

Specificity of the dpp suppressor screen:

Since D/V patterning occurs very early in the embryo, we anticipated that many such genes that could mutate to become suppressors of the dpp haplo-insufficiency might act very early, having been supplied maternally to the cytoplasm of the ooctyte and early embryo. Indeed, we identified seven suppressors from ∼7000 mutagenized genomes. Among these suppressors, we recovered maternal only, maternal and zygotic, and zygotic only suppressors (Table 1). Since the dpp^H46 allele that we used for this screen is a null allele caused by a deletion within the dpp gene (St Johnston et al. 1990; Wharton et al. 1993), the suppression cannot be due to increased expression levels of a slightly functional allele. These suppressors are likely mutations in genes that function during early embryogenesis and are directly involved either in Dpp signaling or in processes that Dpp signaling controls. However, so far we have not recovered alleles of known negative regulators of dpp pathway. For instance, loss-of-function alleles of DSmurf or sog can suppress dpp^H46 haplo-lethality (Francois et al. 1994; Podos et al. 2001). Mapping and complementation tests indicated that DSmurf and sog were not among the Su(dpp) alleles that we identified, presumably because this screen was not saturating.

To understand the nature of the suppressor mutations, we examined the amnioserosa in embryos homozygous for these mutations. It has been shown that increasing the dpp gene dosage results in transformation of the dorsal epidermis to amnioserosa, leading to an increase in the number of amnioserosa cells and reducing the dpp gene dosage results in deletion of the amnioserosa (Wharton et al. 1993). Su(dpp)YE2, Su(dpp)YE5, eIF4A^YE9, and Su(dpp)YE10 had increased amnioserosa cell numbers, with the highest number of 284.7 ± 32.7 in Su(dpp)YE10, which almost doubled the number in wild-type embryos. This amount of amnioserosa cells was usually observed in embryos carrying three copies of dpp⁺. Thus, in embryos homozygous for most of the Su(dpp) mutations, more cells adopted the dorsal-most cell fate, consistent with an increase of Dpp signaling in these embryos. One exception was Su(dpp)YE31 homozygous embryos, which exhibited a decreased number of amnioserosa cells. This cannot be easily explained if the mutation simply results in increased Dpp signaling. However, we found that Su(dpp)YE31 heterozygous embryos had more amnioserosa cells than wild-type embryos. This was consistent with the suppression of dpp^H46 haplo-insufficiency by Su(dpp)YE31 heterozygotes. Although we could not test whether Su(dpp)YE31 homozygotes could suppress dpp^H46 because of their lethality, it appeared that the ability of Su(dpp)YE31 to increase Dpp signaling may depend on the presence of Su(dpp)YE31⁺. We speculated that Su(dpp)YE31 might encode a gain-of-function mutant Dpp signaling component that can increase Dpp signaling only when it is associated with its wild-type counterpart (as in heterozygotes), but this mutant molecule will not be able to transduce Dpp signals in the absence of any wild-type molecules (as in homozygotes). This property of the Su(dpp)YE31 mutation could be explained if the protein encoded by Su(dpp)YE31⁺ normally functions as a homodimer or a multimer to mediate Dpp signaling. Taken together, these data indicate that the genes mutated in these suppressor mutations may be involved in transducing or regulating Dpp signals.

Identification of eIF4A as a suppressor of dpp:

More interestingly, we determined that one of the suppressors, Su(dpp)YE9, which behaved as a dominant-negative mutation, is a new allele of the gene eIF4A. We thus renamed it eIF4A^YE9. eIF4A is the prototypical member of the DEAD box family of ATP-dependent RNA helicases (de la Cruz et al. 1999; Tanner and Linder 2001). DEAD box (and related DEXH box) proteins share eight highly conserved amino acid sequence motifs (see Figure 5A) and are involved in almost all aspects of RNA metabolism, including transcription, ribosomal biogenesis, pre-mRNA splicing, RNA export, translation, and RNA degradation (de la Cruz et al. 1999). eIF4A works in conjunction with eIF4B, eIF4H, and as a subunit of eIF4F in an important regulatory step of translation initiation (Gingras et al. 1999). The eIF4A^YE9 mutation contains a substitution of the residue arginine by histidine at position 321. At physiological pH, the side chain of arginine is fully ionized and positively charged. In contrast, histidine is largely uncharged. In addition, this arginine is conserved among all the eIF4A proteins from yeast to humans and is only one amino acid away from motif V, which is conserved among all the DEAD-box-containing RNA helicases (Figure 5B). Motif V is believed to be responsible for substrate binding (Tanner and Linder 2001). Therefore we assumed that a mutation in this region could interfere with the function of the RNA-recognition domain and disrupt eIF4A-RNA interaction or block the function of the translation initiation complex eIF4F.

Drosophila eIF4A has previously been identified on the basis of homology, and genetic studies indicate that it is an essential gene (Dorn et al. 1993). The Drosophila eIF4A protein shows 73% identity and 92% homology to the mouse eIF4A and 67% identity and 86% homology to the yeast TIF protein. It is abundantly present in early embryos, possibly as a maternal gene product (Dorn et al. 1993). Interestingly, a genetic screen for larval growth-defective mutants in Drosophila revealed an allelic series for the eIF4A mutations that are associated with different larval growth rates in a cell-autonomous manner, suggesting that eIF4A could be a dose-dependent growth regulator (Galloni and Edgar 1999). These eIF4A mutations exhibited inhibitory effects on DNA replication, which can be bypassed by overexpression of the E2F transcription factor. Therefore it has been proposed that the DNA replication block associated with the eIF4A mutations was possibly due to a defect in the translation of a specific set of cell cycle or growth regulatory products (Galloni and Edgar 1999). It has been demonstrated that eIF4E can be regulated by phosphorylation (Lachance et al. 2002). The finding that cellular growth is sensitive to eIF4A led to the proposal that eIF4A, in addition to eIF4E, may be rate limiting in the translation of specific cell cycle or growth-related mRNAs (Galloni and Edgar 1999).

Is eIF4A an integral component of the Dpp pathway?

The identification of a dominant-negative mutation in eIF4A as a suppressor of dpp is surprising. In the screen that we designed, the suppressor mutations should increase the amount of Dpp or its signaling intensity to compensate for the low dosage of dpp in dpp^H46/dpp⁺ animals. It is unlikely that the eIF4A^YE9 mutation could increase the translation efficiency of the Dpp protein because it is a dominant-negative mutation. On the contrary, we infer such a result to suggest that Dpp itself and other positive signaling components are insensitive to the levels of eIF4A. How is a dominant-negative mutation in eIF4A able to rescue lower levels of Dpp? We provide the following possible explanations.

First, negative regulators of Dpp signaling may be more sensitive to eIF4A levels. A selective reduction of these negative regulators may therefore allow dpp heterozygous embryos to overcome the shortage of Dpp and survive. Second, eIF4A itself may have a novel cellular function as a negative regulator of the Dpp pathway. Recently, a second Drosophila eIF4A isoform, CG7483, was discovered in the genome project (Lasko 2000). These two isoforms share 67% identity and 81% homology in protein sequences. Since existing eIF4A mutations are recessive lethal, eIF4A and CG7483 therefore are not completely functionally redundant. Thus it cannot be ruled out that eIF4A has functions in addition to a role in translation initiation. Third, eIF4A might be critically required in an essential biological process in the early embryo that is negatively regulated by Dpp. In this scenario, Dpp signaling might negatively regulate eIF4A levels, such that reducing eIF4A could have the same effects as increasing Dpp signaling on such a biological process. Alternatively, eIF4A could function completely independently of Dpp signaling, yet the two pathways converge on the same targets with opposing effects. We are currently investigating the functions of eIF4A with regard to Dpp signaling during Drosophila development.