The product of the oskar gene directs posterior patterning in the Drosophila oocyte, where it must be deployed specifically at the posterior pole. Proper expression relies on the coordinated localization and translational control of the oskar mRNA. Translational repression prior to localization of the transcript is mediated, in part, by the Bruno protein, which binds to discrete sites in the 3′ untranslated region of the oskar mRNA. To begin to understand how Bruno acts in translational repression, we performed a yeast two-hybrid screen to identify Bruno-interacting proteins. One interactor, described here, is the product of the apontic gene. Coimmunoprecipitation experiments lend biochemical support to the idea that Bruno and Apontic proteins physically interact in Drosophila. Genetic experiments using mutants defective in apontic and bruno reveal a functional interaction between these genes. Given this interaction, Apontic is likely to act together with Bruno in translational repression of oskar mRNA. Interestingly, Apontic, like Bruno, is an RNA-binding protein and specifically binds certain regions of the oskar mRNA 3′ untranslated region.
A wide variety of mechanisms serve to regulate gene expression. Some of these act after transcription and control the stability, distribution or translation of existing mRNAs. Post-transcriptional control is especially prominent during late oogenesis and early embryogenesis, when transcriptional activity is low or not detectable, and the vast majority of proteins are synthesized from maternally contributed mRNAs. Control mechanisms are often tailored for individual mRNAs, allowing for specialized patterns of expression. This is particularly true for mRNAs encoding proteins that direct important developmental events and must appear only at the appropriate times and, in some cases, only at certain positions in the egg or embryo (reviewed by St Johnston, 1995; Curtis et al., 1995; Macdonald and Smibert, 1996; Wickens et al., 1996). Subcellular positioning is commonly achieved through a combination of mRNA localization and translational regulation, while the latter alone is usually sufficient for temporal control. Both of these forms of post-transcriptional control are often mediated by cis-acting regulatory elements in the 3′ untranslated region (UTR) of the mRNA, and several of the factors that bind such elements have been identified (Macdonald et al., 1995; Kim-Ha et al., 1995; Smibert et al., 1996; Rivera-Pomar et al., 1996; Chan and Struhl, 1997; Kelley et al., 1997; Bashaw and Baker, 1997; Gebauer et al., 1998; Paillard et al., 1998; Hake and Richter, 1994; Murata and Wharton, 1995; Zhang et al., 1997; Deshler et al., 1997; Ross et al., 1997; Ostareck et al.,1997). A central issue in understanding the biochemical mechanisms responsible for such control events involves the roles played by these RNA binding proteins. Of particular interest is the question of how translation can be influenced by the binding of a protein to a region of the mRNA, the 3′ UTR, which is not traversed by ribosomes during protein synthesis.
Several mechanisms contribute to specialized forms of translational regulation. The best understood, and probably the simplest, involves the binding of a protein to regulatory sequences in the 5′ UTR, as exemplified by the regulation of ferritin mRNA translation by IRE-BP (reviewed in Klausner et al., 1993). In this case translation is repressed because the bound protein interferes with movement of the preinitiation complex from the 5′ cap to the start codon. This type of mechanism does not appear to require a specific interaction between the RNA binding protein and other factors (Stripecke et al., 1994). Another mechanism of translational regulation, which is mediated through the 3′ UTR, involves changes in poly(A)-tail length (Richter, 1996; Wickens et al., 1996). The logic underlying this type of control is simple, although many of the details are unknown. In brief, mRNAs with long or growing poly(A) tails tend to be more efficiently translated than those with short tails. Thus, enzymes that modify poly(A)-tail length can control translation. 3′ UTR binding factors could either possess such activities or recruit the appropriate enzymes to specific mRNAs. It is less certain how the binding of a protein to the 3′ UTR can affect translation in the absence of a change in poly(A)-tail length, as has been observed for the lipoxygenase mRNA in reticulocytes (Ostareck-Lederer et al., 1994; Ostareck et al., 1997). However, it does seem likely that this type of mechanism will involve interaction of the binding protein with other protein factors, and so identification and characterization of those factors should lead to mechanistic insights.
Many of the best characterized mRNAs under elaborate post-transcriptional control come from Drosophila, where the protein products of maternal mRNAs dictate pattern along the dorsoventral and anteroposterior body axes. One of these mRNAs is oskar (osk), which encodes a spatial determinant required for posterior patterning (Lehmann and Nüsslein-Volhard, 1986; Ephrussi and Lehmann, 1992; Smith et al., 1992). The osk mRNA is transcribed in the nurse cells of the ovary, rapidly transported into the oocyte, and eventually localized to the posterior pole of the oocyte (Kim-Ha et al., 1991; Ephrussi et al., 1991). Translation of the osk mRNA is repressed during the early stages of oogenesis and activated coincident with posterior localization in the oocyte (Kim-Ha et al., 1995; Rongo et al., 1995; Markussen et al., 1995). Translational repression requires cis-acting elements within the osk mRNA 3′ UTR called BREs (Bruno response elements), and the Bruno (Bru) protein, which specifically binds these elements. Disruption of the Bru-osk mRNA interaction in vivo allows translation prior to localization of the mRNA, and Osk protein accumulates throughout the oocyte. As a consequence, the entire oocyte and embryo acquire posterior positional values, a lethal condition (Kim-Ha et al., 1995; Webster et al., 1997). At present there is no evidence to suggest that Bru-mediated translational repression involves changes in polyadenylation. The poly(A) tail of the osk mRNA in ovaries is short, and its length does not appear to vary with the translational status of the mRNA (Webster et al., 1997). Thus the binding of Bru protein to osk mRNA is likely to interfere with translation by an as yet undefined mechanism. Here we describe the results of a yeast two-hybrid screen to identify proteins that interact with Bru and potentially contribute to translational regulation of osk mRNA. One gene characterized in detail, apontic (apt), has properties indicating that it encodes a protein that acts as a corepressor of osk translation.
MATERIALS AND METHODS
Yeast two-hybrid interaction trap
We used the system of Gyuris et al. (1993). In this system protein-protein interactions between bait (Bru fused to the LexA DNA binding domain) and prey (unknown cDNA fused to a transcriptional activation domain) constitute a protein assembly that binds to LexA operator sequences and can function as a transcriptional activator. In the appropriate yeast strain in which the leu2 gene has been placed under control of LexA operator sequences, the bait-prey interaction allows for growth on medium lacking leucine, thereby allowing selection of positive prey cDNAs. A bait plasmid (pY3) was constructed that contained the complete bru cDNA fused to the LexA DNA-binding domain in the 2μ HIS3+ plasmid pEG202. Expression of protein from this plasmid is under control of the ADH1 promoter. In preliminary tests we found that this plasmid conferred substantial transcriptional activation in the absence of any prey plasmid, making it difficult to use this selection scheme to identify prey cDNAs encoding Bru-interacting proteins. To address the problem of transcriptional activation by bait alone, Brent and coworkers have constructed lacZ reporter plasmids that can be used as an alternative method of detecting bait-prey interactions. These plasmids contain the lacZ gene under the control of variable numbers of LexA operator sequences, setting different threshold levels for detection of transcriptional activation. We tested several such reporter plasmids and found that in the yeast strain EGY48 (MATa,ura3, his3, trp1, LEU2::LexAop6-LEU2) containing plasmid pRB1840 (a 2μ URA3+ plasmid that contains a single LexA operator upstream of the lacZ gene), transcriptional activation conferred by the Bru bait plasmid was below the level of detection by colorimetric plate assay. The prey library was RFLY3, a gift from R. Brent. This library was made from Drosophila melanogaster ovarian cDNAs in the 2μ TRP1+ plasmid pJG4-5. Expression of the library cDNA fusion from this vector is under control of the GAL1 promoter. The yeast strain was transformed with pRB1840 and pY3 and then subsequently transformed with RFLY3. 246,000 yeast transformants containing all three plasmids were plated on media lacking histidine, tryptophan and uracil, and containing 2% galactose and the substrate X-gal. Colonies were visually screened for β-galactosidase activity. 89 blue colonies were selected. In order to determine whether this activity was dependent upon expression of the library cDNA, each colony was subsequently retested for dependence of β-galactosidase activity on galactose-containing medium. 56 colonies retained activity. All of these were then tested for interaction specificity. Library plasmids were individually transformed into several yeast strains, each containing a different bait plasmid as well as the reporter plasmid pRB1840. True interacting proteins were identified as those that activated lacZ expression in the presence of the Bru bait plasmid and not in the presence of Bcd (pRFHM-1) or Exu bait plasmids. 12 of the original lacZ+ clones were specific in their interactions with Bru. DNA sequencing indicated that these clones could be grouped into five different classes. The apt class contained four clones with identical apt cDNA sequence and variable numbers of EcoRI adaptors at their 5′ ends. These clones encode amino acids 429-490 of the Apt protein. All were further tested for specificity of interaction with Bru using bait plasmids expressing Tra, Tra-2, Rbp-1, Vasa or the N-terminal portion of Vasa; activation was only observed with the Bru bait plasmid.
Recombinant Apt protein
The apt cDNA beginning at the internal SalI site was blunt-ended and cloned into the blunt-ended BamHI site of the pET-3b vector (Novagen), and this plasmid was transformed into the bacterial strain pLysS. The resulting protein lacks the N-terminal 37 amino acids of Apt. Expression was induced by the addition of 0.5 mM IPTG to log phase cultures and subsequent growth for 2.5 hours at 37°C. Bacterial protein was prepared by centrifugation of the culture at 4,000 rpm for 10 minutes at 4°C to pellet the cells. The pellet was resuspended in cold TNE solution (50 mM Tris-Cl, pH 8.0, 250 mM NaCl, 2 mM EDTA) and pelleted a second time by centrifugation. The pellet was frozen at −80°C, then thawed in cold TNE with 2 mM β-mercaptoethanol, 1 mM PMSF and 10 mM benzamidine. Lysozyme was added to about 0.2 mg/ml, and after a 10 minute incubation on ice Triton X-100 was added to 1% and the mixture incubated for another 10 minutes on ice. The lysate was then sonicated for 3×10-second pulses and this solution was centrifuged through a 40% sucrose cushion (40% sucrose, 10 mM Tris-Cl, pH 8.0, 200 mM NaCl, 1 mM EDTA) for 30 minutes at 12,000 rpm at 4°C. The resulting pellet was recovered and resolubilized in 8 M urea, 50 mM Tris-Cl, pH 8.0, 500 mM NaCl, 1 mM EDTA, 1 mM PMSF. This solution was dialyzed overnight at 4°C into 50 mM Tris-Cl, pH 8.0, 500 mM NaCl, 1 mM PMSF and 10% glycerol. Soluble protein, in which Apt was the major component, was recovered in the supernatant after centrifugation for 15 minutes at 12,000 rpm at 4°C to pellet insoluble proteins. This Apt protein was used both for UV cross-linking assays and for injection into rats for preparation of antiserum (Josman Laboratories). The rat polyclonal serum specifically detects Apt on western blots (data not shown) and was used for all immunoprecipitations and whole-mount staining.
For immunoprecipitations, protein extracts were prepared by homogenizing hand-dissected ovaries in extraction buffer (50 mM Hepes, pH 7.9, 150 mM NaCl, 1 mM EDTA and 0.5 mM PMSF) and clearing the extract by centrifugation at 15,000 rpm for 12 minutes at 4°C. Glycerol was added to a final concentration of 20%. 25 μl extract diluted with 75 μl of extraction buffer was preincubated with 5 μl serum for 1 hour at 4°C. Samples were centrifuged 10 minutes at 15,000 rpm at 4°C to pellet aggregates. After transferal of the supernatant to a fresh microfuge tube, 50 μl of a 1:1 slurry of equilibrated protein G-agarose beads (Boehringer-Mannheim) to extraction buffer was added. These samples were then rotated for 2 more hours at 4°C. Beads were washed four times with cold 50 mM Tris-Cl, pH 8.0, 250 mM NaCl and 1% Tween-20. 50 μl 2× protein sample loading buffer was added to the washed beads; samples were heated to 100°C for 5 minutes and electrophoresed through 9% SDS-polyacrylamide gels. Electrophoresed samples were transferred to nitrocellulose (Idea Scientific), and subsequent analysis was by western blotting. Proteins were detected by chemiluminescence (Western Light, Tropix) using the primary antibody at a 1:2000 dilution for anti-BruA antiserum (Webster et al., 1997) and the secondary goat anti-rabbit alkaline phosphatase-conjugated antibody as per the manufacturer’s protocol.
RNA and protein analysis
In situ hybridizations were performed as described by Tautz and Pfeifle (1989) using RNA probes. Whole-mount antibody stains were done with anti-Apt antibodies at a 1:100 dilution. Secondary antibodies for signal detection were a goat anti-rat horseradish peroxidase conjugate or a goat anti-rat Cy3 conjugate (Jackson ImmunoResearch Laboratories). To visualize nuclei, fixed tissues were stained for 5 minutes with 1 μg/ml DAPI and washed thoroughly in 1× PBS, 0.1% Tween-20.
Fly stocks and manipulations
apt41 and apt167 both have missense mutations in the third exon of the apt gene. The apt P element stock l(2)09049 contains a P element insertion just upstream of the first intron of the apt gene (Gellon et al., 1997). tdfPΔ3 and tdfPΔ4, which we refer to in the text as aptPΔ3 and aptPΔ4, are both deletion mutants lacking 5′ regions of apt (Eulenberg and Schuh, 1997). The aretQB72 allele was used in all experiments described and contains a nonsense mutation resulting in a stop codon in the protein at amino acid 404 (Webster et al., 1997). w1118 flies were used for immunohistochemistry, in situ hybridization and protein extracts.
Germline clones were induced using the FLP-DFS method (Chou and Perrimon, 1992). apt alleles were recombined onto the P[FRT-G13, w+] chromosome. The aptPΔ3, P[FRT-G13, w+] recombinant fly stock was a gift from R. Schuh. Female flies carrying a recombinant chromosome were crossed to y, w, P[hs-Flp-1]; P[FRT-G13, w+], P[ovoD1, w+] males. Second and third instar larvae were heat-shocked at 37°C for 30 minutes, allowed to recover at room temperature for 30 minutes, and then heat-shocked a second time at 37°C for 30 minutes. Flies were maintained at 25°C. Upon eclosion, virgin females of the desired genotype were selected and placed in cages with wild-type males to determine egg-laying ability. Ovaries were hand-dissected, fixed and stained with DAPI as described above.
UV cross-linking assay
0.5 μg bacterially expressed Apt protein was preincubated with 10 μg yeast tRNA, and labeled probe (1×106 cpm) was added (along with competitor RNA for competition experiments). Radiolabeled probes were made by in vitro transcription using different regions of the osk 3′ UTR as templates (Kim-Ha et al., 1995). Other RNA probes used in this assay were transcribed from the following templates: C78, the 5′-most 185 nucleotides (nt) of the nanos (nos) 3′ UTR; C84, the same 185 nt fragment with point mutations in the SREs (smaug recognition elements); C88, multimerized 3× SRE (89 nt); C90, multimerized point mutated 3× SRE (89 nt); pY17, the 3′ end of vitelline membrane protein 32E (approx. 900 nt); pY84 (approx. 800 nt) and pY86 (approx. 800 nt), two uncharacterized cDNAs isolated in the two hybrid interaction trap; and the pSP73 (Promega) polylinker (105 nt). Reactions were incubated for 5 minutes at room temperature and subsequently irradiated on ice with 105 erg/mm2 of UV light in a Stratagene UV cross-linker. Following UV cross-linking, RNA was digested with 30 μg RNase A for 15 minutes at room temperature. Protein sample loading buffer was added, samples were heated to 100°C for 4 minutes, and proteins were resolved by SDS-PAGE on 9% gels followed by autoradiography. Apt bound specifically to regions of the osk 3′ UTR (see Fig. 5) and also bound all the other unrelated RNAs tested (listed above) except for C88 and C90.
Bru-mediated repression of osk mRNA translation presumably requires the interaction of Bru with other protein factors. To identify candidate Bru binding partners, we performed a yeast two-hybrid screen using the interaction trap system (Gyuris et al., 1993). 12 Drosophila ovarian cDNAs passed initial tests for evidence of a specific interaction of the encoded protein with Bru and could be grouped into five classes (see Materials and methods). For the class of cDNA (consisting of four nearly identical clones) described in this report, we performed additional tests in the two-hybrid system using a panel of other proteins (see Materials and methods), which further demonstrated specificity in the interaction of the cDNA-encoded protein with Bru.
The cDNAs recovered from the two-hybrid screen were not full length. Consequently, additional cDNAs were isolated from an ovarian library and the longest one was sequenced. This cDNA corresponds to a gene which was recently identified independently in three different laboratories and has been called apontic (apt) (Gellon et al., 1997; Su et al., 1999) or tracheae defective (tdf) (Eulenberg and Schuh, 1997). We refer to the gene and mutants as apt. Our cDNA (GenBank AF027123) has minor differences relative to the other cDNA sequences. These include several single nucleotide changes as well as small insertions or deletions that do not change the reading frame; all are likely to represent sequence polymorphisms or errors introduced by reverse transcription. The sequences also diverge in the 5′ region, resulting in an additional six amino acids in the protein defined by our cDNA. This divergence can be simply explained by use of an alternative promoter and 5′ exon, an interpretation that is consistent with the appearance in northern blot analysis of a slightly smaller form of the mRNA during oogenesis and early embryogenesis (data not shown; see also Fig. 4B of Gellon et al., 1997).
Identification of apt in the two-hybrid screen provides suggestive evidence that Bru and Apt proteins interact. In the following sections we first describe the expression of apt in ovaries, which shows that Bru and Apt are found in the same subcellular compartments and thus may interact in vivo. We then present biochemical and genetic evidence strongly supporting the existence of a Bru-Apt interaction and suggesting that Apt, like Bru, acts in regulation of osk mRNA translation.
Patterns of apt expression are consistent with an interaction with Bru
Previous reports on apt have focused primarily on its expression in the embryo, and details of the ovarian expression have not been described. We examined the mRNA expression pattern of apt during oogenesis by in situ hybridization (Fig. 1A). Expression occurs in both the somatic follicle cells and the germline nurse cells and oocyte. apt transcripts are detected as early as stage 2A at low levels in the germarium and at higher levels in the follicle cells. The amount of apt mRNA in the soma decreases during the remainder of oogenesis, while the level in the germline increases. apt mRNA becomes concentrated in the oocyte and also accumulates in the nurse cells at about stage 6. apt transcripts continue to be found in both the oocyte and nurse cells throughout oogenesis.
To determine when and where Apt protein is expressed during oogenesis, antisera directed against a recombinant Apt protein were prepared and used for protein detection in whole-mount ovaries by confocal microscopy (Fig. 1B,C). Apt protein appears in both the germline and somatic cells of the ovary throughout all stages of oogenesis. In the germline, Apt protein is present in both cytoplasm and nuclei. Within the nurse cells the protein is more concentrated in the cytoplasm, while in the oocyte more protein is found in the nucleus. The protein, however, is not localized to any subdomain within the cytoplasm of either the nurse cells or the oocyte.
Although Apt protein is not strictly nuclear or cytoplasmic in cells of the female germline, the protein is highly concentrated in nuclei of the ovarian follicle cells (Fig. 1B,C) and in post-cellularization-stage embryos (Eulenberg and Schuh, 1997). The developmental differences in subcellular location suggest that Apt may have functions, perhaps different, in both nuclei and cytoplasm. Nuclear proteins expressed from maternal mRNAs are sometimes present at high levels in the cytoplasm of early embryos. Examples include the Bicoid, Caudal and Hunchback proteins, which appear in both nuclei and cytoplasm shortly after egg laying. As nuclear divisions progress and the density of nuclei increases, nuclear localization of these proteins remains strong while the fraction of protein in the cytoplasm diminishes (Driever and Nüsslein-Volhard, 1988; Macdonald and Struhl, 1986; Mlodzik and Gehring, 1987; Tautz, 1988). Thus there appears to be no early impediment to nuclear localization, simply a paucity of nuclei. In contrast, the subcellular distribution of Apt protein appears to be actively controlled in early development. We monitored Apt protein in early embryos, using DAPI staining of nuclei to define developmental stages. Even after migration of nuclei to the surface of the embryo, Apt protein remains evenly distributed between nuclei and cytoplasm (Fig. 1D), unlike any of the other examples described above. This unusual persistence of Apt protein in the cytoplasm suggests the existence of a mechanism to control its distribution, reinforcing the notion of roles for Apt in both cytoplasm and nuclei.
Biochemical evidence for the Bru-Apt interaction
To support the idea that a physical interaction between Bru and Apt occurs in Drosophila, we performed immunoprecipitation experiments. Dissected ovaries were homogenized and incubated with Protein G-agarose beads and either preimmune serum or anti-Apt antiserum. Immunoprecipitated proteins were then assayed by western blot analysis for the presence of Bru protein (Fig. 2). Notably, Bru was coimmunoprecipitated with anti-Apt antiserum but not with preimmune serum, demonstrating that Bru and Apt interact in the ovarian extract. Although we also performed the complementary immunoprecipitations with anti-Bru antiserum, we were unable to detect Apt protein in the Bru immunoprecipitates (data not shown). It is possible that the epitope recognized by the anti-Bru antibody is masked by the Bru-Apt interaction, which would interfere with coimmunoprecipitation. In support of this interpretation, we note that while much of Bru protein is readily detectable by immunohistochemistry in whole mount ovary preparations, the fraction of Bru protein colocalized with osk mRNA at the posterior pole of the oocyte can be detected only if the ovaries are pretreated with protease (Webster et al., 1997). Thus, certain populations of Bru protein appear to have epitopes not readily accessible to the anti-Bru antibodies.
apt and aret interact genetically
Physical interactions between proteins suggest but do not prove that the proteins function together in vivo. We therefore looked for a genetic interaction between the bru and apt genes, which would argue that the physical interaction is important for function. The arrest (aret) mutants are defective for Bru (i.e. aret and bru are the same gene; Webster et al., 1997) and lead to a developmental arrest early in oogenesis (Schüpbach and Wieschaus, 1991; Castrillon et al., 1993; Webster et al., 1997). For our analysis we used aretQB72, a strong allele that has an internal stop codon (Webster et al., 1997). Mutants in apt are zygotic lethal (Gellon et al., 1997; Eulenberg and Schuh, 1997), and some alleles also cause arrested oogenesis (below). We used several different apt alleles for all analyses (see Materials and methods), as the genetics of apt are complex and different alleles have different effects (see below and Discussion). To test for a genetic interaction between the aret and apt mutants, we asked if reducing the dosage of both genes would cause a phenotype. Females heterozygous for aretQB72, heterozygous for any of the five apt alleles, or transheterozygous for both aretQB72 and an apt allele, were crossed to wild-type males, and the progeny embryos were then examined for cuticular defects. In these crosses mothers heterozygous for aretQB72 or for any allele of apt produce only embryos with wild-type cuticles. In contrast, females transheterozygous for aretQB72 and apt41, apt167, aptl(2)09049, aptPΔ3 or aptPΔ4 produce a fraction of embryos with head defects (Fig. 3B, Table 1).
Head defects can result from ectopic or excessive posterior body patterning activity, as this activity interferes with expression of the anterior body patterning morphogen, Bicoid (Wharton and Struhl, 1991). Consequently, the observed head defects could be explained if both Bru and Apt contribute to repression of osk mRNA translation. Alternatively, the head defects could result from a more direct effect on anterior development, a possibility suggested by the genetic interaction between apt and Dfd, a gene involved in head development, and the fact that homozygous apt− embryos have head defects (Gellon et al., 1997). To distinguish between these possibilities we determined the consequences of reducing nanos (nos) gene dosage in mothers transheterozygous for aret and apt. nos encodes a limiting component of the posterior patterning activity (Lehmann and Nüsslein-Volhard, 1991; Wang and Lehmann, 1991), and reduction of the nos gene dosage should only affect anterior patterning defects arising from misexpression of posterior patterning molecules. For transheterozygotes of aretQB72 and four of the five apt alleles, reducing nos dosage largely suppressed the head defects phenotype (Table 1). Thus the head defects phenotype of these transheterozygotes can be attributed, at least in large part, to ectopic or excessive posterior body patterning activity, a finding consistent with a role for apt in control of osk mRNA translation. The fact that the phenotype of the aptl(2)09049/aretQB72 transheterozygotes is not suppressed by reducing nos dosage indicates that this allele affects apt function differently than the other alleles (see Discussion).
apt ovarian phenotype
Earlier genetic analyses of apt have concentrated on the zygotic phenotype (Gellon et al., 1997; Eulenberg and Schuh, 1997). To define more completely the role of apt in the female germline, we created females with apt− germline clones using the FLP/DFS method (Chou and Perrimon, 1992). Ovaries containing germline clones were dissected, stained with DAPI to highlight nuclei, and examined for phenotype. Different apt mutants display dramatically different ovarian phenotypes. One allele, aptl(2)09049, is indistinguishable from wild type, as females with aptl(2)09049 germline clones had phenotypically wild-type ovaries (Fig. 4A) and laid eggs that developed into fertile adults. Females with apt167 germline clones also had phenotypically wild-type ovaries, but a small fraction of the eggs laid developed into embryos with head defects (Fig. 4B). In contrast, ovaries from females with apt41 or aptPΔ3 germline clones have phenotypes that are similar to one another and severe: development is arrested in early oogenesis (approximately stage 6), and the oocyte fails to differentiate with all nuclei becoming polyploid. In addition, some of the egg chambers have an abnormal number of nuclei (Fig. 4C-E). This phenotype is highly unlikely to result from a background mutation, as the apt41 and aptPΔ3 alleles were induced with different mutagens on different parental chromosomes (Gellon et al., 1997; Eulenberg and Schuh, 1997). We conclude that apt is necessary for oogenesis and that loss of apt activity leads to a developmental arrest during oogenesis. Just as for aret mutants, the arrest occurs too early to allow us to examine the ovaries for defects in osk mRNA translation.
Apt binds specifically to regions of the osk mRNA 3′ UTR
The biochemical and genetic data presented here demonstrate that Bru and Apt interact with one another and suggest that Apt contributes to Bru-mediated translational repression of osk. To begin to examine the mechanism of Apt function, we asked if Apt, like Bru, can bind RNA. Recombinant Apt was expressed in E. coli and tested for RNA binding activity in a UV cross-linking assay. In initial experiments we used substrate RNAs corresponding to different parts of the osk mRNA 3′ UTR. Apt binding to certain RNAs is easily detectable, while RNAs from other parts of the 3′ UTR do not support binding (results summarized in Fig. 5A). We more rigorously confirmed the apparent differences in binding affinities using competition binding experiments. Binding of Apt to the osk AB region RNA was tested in the presence of increasing amounts of unlabeled competitor RNAs (Fig. 5B). Although the osk AB RNA competes effectively, the other RNAs tested compete only weakly. Thus Apt is an RNA binding protein, and it displays substantial specificity in its binding activity. Remarkably, the regions of the osk 3′ UTR bound by Apt, the AB and C regions, are precisely those bound by Bru. To determine if Bru and Apt have the same RNA binding specificity, we tested Apt binding to a series of RNAs used to map the Bru binding sites, called BREs, within the osk C region. Three of these RNAs retain the BREs and are bound by Bru, while a fourth RNA, CΔ4, lacks the BREs and fails to bind Bru (Kim-Ha et al., 1995). Apt binds all four RNAs, including CΔ4 (Fig. 5C), indicating that Apt can bind to sites other than BREs.
Subsequent binding experiments were performed with a variety of other in vitro transcribed RNAs. Apt binds detectably to 6 of 8 RNAs tested (data not shown; see Materials and methods). We have been unable to identify a sequence shared by all of the bound RNAs. Thus, despite its ability to efficiently discriminate between different parts of the osk mRNA, Apt appears to be relatively promiscuous in its binding and may recognize many sites or perhaps a structural feature common to many RNAs.
Proper control of osk mRNA translation is essential and requires strict coordination with localization of the transcript to the posterior pole of the oocyte. Not surprisingly, translational regulation of osk mRNA appears to be complex, and several factors involved in repression and activation have already been identified. The evidence for a direct role in osk translation is strongest for Bru, a protein that binds specifically to regulatory sequences in the osk mRNA 3′ UTR (Kim-Ha et al., 1995; Webster et al., 1997). Bru is also expected to bind and regulate additional mRNAs, as mutants lacking functional Bru protein display defects in both oogenesis and spermatogenesis (Schüpbach and Wieschaus, 1991; Castrillon et al., 1993) that cannot be attributed solely to problems in osk mRNA metabolism. Consistent with this idea, Bru has been shown to bind at least one other mRNA, gurken (Kim-Ha et al., 1995). A second protein that acts in repression of osk translation is Bicaudal C (Bic-C) (Saffman et al., 1998). Bic-C mutants display a dominant maternal-effect phenotype in which osk translation initiates prematurely and the embryos thus develop with defective anterior patterning. Bic-C protein has RNA binding activity but has not been shown to bind specifically to osk mRNA. Another protein suggested to act in repression is p50, which was identified by virtue of its binding to osk mRNA, but for which genetic confirmation of such a role has not been obtained (Gunkel et al., 1998). We now add apt to this roster of proteins and genes implicated in negative regulation of translation.
Apt protein in cytoplasm and nucleus
It has been suggested that Apt functions as a transcription factor during embryogenesis, perhaps acting as a cofactor for certain Hox genes (Gellon et al., 1997; Eulenberg and Schuh, 1997). Two types of evidence have been presented to support this conclusion. First, the Apt protein is highly concentrated in nuclei during most of embryogenesis, which strongly implicates a nuclear function. Second, the predicted structure of the Apt protein includes domains similar to those found in certain transcription factors (Eulenberg and Schuh, 1997). One is a short region enriched in glutamine residues, which may serve as a transcriptional activation domain. This by itself does not strongly support a role as a transcription factor, as similar glutamine-rich regions are found in a wide variety of Drosophila proteins, some of which are not involved in transcriptional regulation. The other domain is a potential bZIP motif. One part of this motif, the leucine zipper, is clearly present in Apt and may imply that the protein homodimerizes or forms a heterodimer with another protein in vivo. The second part of the bZIP motif, a flanking basic region, appears in an unusual form: certain amino acids known to be involved in DNA binding are present, but these are positioned much closer to the leucine zipper than in any other characterized bZIP domain. In addition, there are few basic amino acids (Eulenberg and Schuh, 1997). Consequently, Apt is either a rather unusual example of a bZIP transcription factor, or it may be a related protein whose function in the nucleus is less certain.
Apt is not always nuclear. Apt protein is persistently retained in the cytoplasm of early stage embryos even after other maternally provided proteins have shifted to the nuclei. This evidence for programmed control of the subcellular location of Apt suggests a requirement for Apt in the cytoplasm of early embryos. Although this type of control could serve to prevent Apt from functioning in the nucleus at this stage of development, this seems unlikely, as Apt is not excluded from the nuclei but simply is not concentrated there. In the nurse cells of the ovary Apt protein is partitioned primarily to the cytoplasm. This phenomenon – cytoplasmic distribution in nurse cells of a protein that is nuclear in most other tissues – is not unusual. Other examples include hnRNP40 (Squid) (Matunis et al., 1994) and Sex lethal (Bopp et al., 1993). Furthermore, a number of other nucleic acid binding proteins, including TfIIIA and the Y box proteins, have distinct functions in the cytoplasm and the nucleus (reviewed by Ladomery, 1997).
One possibility for how the subcellular distribution of Apt may be controlled is suggested by differences in apt mRNAs. The use of alternate 5′ exons leads to variation at the amino terminus of the protein. Exon choice appears to vary during development, with one form of the mRNA found primarily among maternal transcripts while other forms are ubiquitous or most prevalent among zygotic transcripts. This pattern correlates well with the changing distribution of Apt protein: cytoplasmic Apt protein is synthesized largely or entirely from maternal mRNAs, while nuclear protein is synthesized from both maternal and zygotic mRNAs. Thus one form of the protein could be targeted to the nucleus and the other form to the cytoplasm.
Role of apt in osk mRNA translation
Evidence implicating apt in the control of osk translation is indirect. Biochemical experiments indicate that Bru and Apt proteins interact physically but provide no insight into the significance of the association. The genetic evidence – head defects among progeny of mothers transheterozygous for apt and aret mutations – reveals a functionally significant interaction between the apt and aret genes but does not specify the exact nature of the interaction. Nevertheless, given the established role for Bru in repression of osk mRNA translation, one likely explanation is that Bru and Apt both act in this process. Consequently, reducing the amount or activity of both Bru and Apt proteins could lead to a modest derepression of osk translation. This interpretation is supported by the sensitivity of the phenotype to reduction of nos gene dosage. Curiously, for one of the apt mutants, the P element insertion allele aptl(2)09049, the genetic interaction with aret is not suppressed by reduction of nos gene dosage. The same allele has no ovarian phenotype when tested by germline clonal analysis. We know of no simple explanation for the behavior of this allele, although it seems possible that insertion of the P element affects only one form of the apt transcripts, perhaps leaving the ovarian-enriched transcript intact. Notably, aptl(2)09049 is lethal when homozygous or in combination with other apt alleles, but is viable in trans to a deficiency that removes the apt gene (Gellon et al., 1997; W. McGinnis, personal communication; data not shown), further indicating that it is an unusual allele.
Although the genetic interaction of aret and apt supports a role for apt in repression of osk mRNA translation, this function may not be essential. One of the apt mutants that shows a nos-sensitive interaction with aret, apt167, has only a modest phenotype in germline clonal analysis: a small fraction of the embryos from the homozygous mutant germlines display head defects. While this phenotype is consistent with a partial relaxation of the controls on osk activity, it is inconsistent with a complete derepression of osk mRNA translation. (Note that although homozygous apt− embryos have defects in head development, the embryos obtained from the homozygous mutant germlines were fertilized by wild-type males and are thus heterozygous for apt167.) Could apt167 be a weak allele? This seems somewhat unlikely (but not impossible) as it displays a stronger genetic interaction with aret than does apt41, which has a strong arrested oogenesis phenotype in homozygous mutant germlines. Another possibility is that apt performs a redundant or partially redundant role in repression of osk mRNA translation. An appealing feature of this model is that a candidate exists for a protein with overlapping function. Gunkel et al. (1998) recently described a protein, p50, that also binds to the regions of the osk mRNA bound by Apt; Apt and p50 could have similar roles in regulation of osk expression. The gene encoding p50 has not been identified, so genetic tests of this model are not yet possible.
Our demonstration that Apt is an RNA binding protein is somewhat unexpected, as none of the well-characterized RNA binding motifs (Burd and Dreyfuss, 1994) appear in the predicted protein sequence. The ability of Apt to discriminate in its binding to certain regions of the osk mRNA 3′ UTR is striking, but its significance is uncertain, especially given the binding of Apt to a wide variety of other RNAs. In further characterization of apt function, it will be of interest to determine whether Apt RNA binding activity is important for proper regulation of osk mRNA translation, or if the interaction of Apt and Bru proteins is sufficient. Notably, Apt protein does not colocalize with Bru and osk mRNA to the posterior pole of the oocyte, raising the possibility that displacement of Apt from Bru may allow translational activation.
We thank R. Brent and R. Finley for yeast strains, two-hybrid libraries and plasmids, and helpful advice; V. Heinrichs and A. Nakamura for bait plasmids; A. Harris, L. Luo, W. McGinnis, R. Schuh, T. Schüpbach and the Berkeley Drosophila Genome Project for fly stocks; W. McGinnis for discussions and sharing unpublished information; B. Holley for assistance with confocal imaging; D. Guarnieri and T. Lee for discussions regarding generation of germline clones; C. Smibert for UV cross-linking probes; and P. Lasko for BruA antibodies. In addition, we thank E. Arn, A. Harris, B. Holley, R. Mancebo, C. Smibert and P. Webster for discussions and criticism of the manuscript. This work was supported by NIH grant GM54409 to P. M. M.
Note added in proof
Genomic sequence of the region containing apt (GenBank accession number AC005639, Berkeley Drosophila Genome Project, unpublished) confirms that there are alternative 5′ exons for this gene. The ovarian cDNA described in this work contains 5′ sequences provided by an exon located approximately 16 kB upstream of the 5′ exon contained in the embryonic transcript.