Nanos interacts with Cup in the female germline of Drosophila

Nanos (Nos) has at least five different functions during the development of Drosophila melanogaster. In the pre-cellular blastoderm embryo, maternally derived Nos is required for formation of abdominal segments (Wang and Lehmann, 1991). Subsequently, maternal Nos is required in the embryonic germline precursors, the pole cells, to inhibit their division and promote migration into the somatic gonad (Asaoka-Taguchi et al., 1999; Forbes and Lehmann, 1998; Kobayashi et al., 1996). Zygotically expressed Nos is required for maintenance of germ line stem cells in females (Forbes and Lehmann, 1998) and males (Bhat, 1999). And finally, Nos is required for viability of the adult (Spradling et al., 1999). These roles have been revealed by the analysis of various nos mutant alleles, which have different molecular lesions and attendant phenotypic consequences.

The molecular mechanism of Nos action is best understood for its role in promoting abdominal segmentation of the embryo. In collaboration with Pumilio (Pum), Nos acts to inhibit the translation of hunchback (hb) mRNA in the posterior of the embryo (Sonoda and Wharton, 1999). Pum is a site-specific RNA-binding protein that recognizes the crucial cis-acting targets in the 3′UTR of hb mRNA (Wharton et al., 1998; Zamore et al., 1997). Nos is subsequently recruited via protein-protein and protein-RNA contacts (Sonoda and Wharton, 1999), and the resulting Pum-Nos complex inhibits translation by as yet unknown mechanisms, although inhibition is accompanied by deadenylation of the mRNA (Wharton and Struhl, 1991; Wreden et al., 1997).

In contrast, little is known of how Nos acts in other cell types during development. Recent work suggests that Nos and Pum also collaborate to regulate Cyclin B (CycB) mRNA in the pole cells and thereby control their division (Asaoka-Taguchi et al., 1999; Deshpande et al., 1999); however, direct interactions of Nos and Pum with CycB mRNA have not been reported. In the germline stem cells, the targets of Nos action are not yet known. Moreover, it is unclear whether Nos and Pum act collaboratively in this tissue type. Although germ cells are rapidly lost from the ovaries of flies bearing ‘strong’ alleles of nos (nos^RC/Df(nos)) or pum (pum^ET1/pum^Msc), the mutant phenotypes are somewhat different, leading to the suggestion that Nos and Pum may act independently (Forbes and Lehmann, 1998). Finally, the discovery of lethal alleles reveals an essential function for Nos in the larva or adult, but the tissue or cell type that requires Nos activity has not been reported.

The function of Nos in germline precursor cells appears to be evolutionarily conserved, at least in part. Nos homologues have been described in Diptera, Xenopus, leech and Caenorhabditis elegans (Curtis et al., 1995; Kraemer et al., 1999; Mosquera et al., 1993; Pilon and Weisblat, 1997; Subramaniam and Seydoux, 1999). Like Drosophila Nos, each Nos homolog is expressed preferentially in the germline precursor cells, consistent with a conserved role. Moreover, recent genetic experiments suggest that Nos and Pum homologs are likely to collaborate to regulate various aspects of germline development in C. elegans. In particular, reduction of Nos1 and Nos2 activity by RNAi causes germ cell phenotypes similar to those seen in nos mutant flies (inefficient incorporation into the gonad, premature proliferation, and elimination of germ cells from the post-embryonic gonad) (Kraemer et al., 1999; Subramaniam and Seydoux, 1999).

Similar phenotypes are observed upon simultaneous disruption via RNAi of a number of the Pum homologs (Puf proteins in C. elegans). Finally, disruption of Nos and Puf function also disrupts the developmental switch from production of sperm to oocytes in the late larval hermaphrodite. This process is mediated by binding of the FBF (fem-3-binding factor) Puf domain proteins to fem-3 mRNA (Zhang et al., 1997), and perhaps by subsequent recruitment of Nos3 (Kraemer et al., 1999), much as Pum binds to hb mRNA and recruits Nos in Drosophila. Taken together, these observations support the idea that Nos and Pum play similar roles in the germline cells of Drosophila and C. elegans, and perhaps in other species.

To better understand the mechanism of Nos action in the germ line, we have undertaken a screen for interacting molecules that either mediate or modulate its activity. In this report, we describe an interaction between the product of fs(2)cup (Cup) and the non-conserved N-terminal domain of Nos. Nos and Cup are coincidentally expressed in the germarium within the ovary. The interaction with Cup within these cells apparently inhibits Nos activity, since reducing the level of Cup suppresses the oogenesis defects associated with the nos^RC allele. We show that ovaries from nos^RC mutant flies bear a very low level of residual Nos, which evidently is only barely inadequate to support normal germline development.

Df(Dl)^FX1nos^RC, nos^RD, nos^BN, nos^RW are described by Arrizabalaga and Lehmann (1999); nos^P7117 is described by Spradling et al. (1999); pum^Msc, pum^FC8, pum^ET1 are described by Forbes and Lehmann (1998); and cup¹, cup³, cup⁸, cup¹³, cup¹⁵, cup²⁰, cup²¹, cup²⁴, cup²⁶ are described by Keyes and Spradling (1997). Transgenic flies were obtained by injecting pCaSpeR derivatives into w¹¹¹⁸ embryos by standard methods. Antibodies were raised against two different portions of Cup (encoded by nucleotides 785-1655 and 1655-2729), expressed in bacteria with N-terminal hexa-histidine tags. The Myc epitope tag was detected with mAb9E10 (Santa Cruz Biotechnology). The nos-myc transgene was prepared by introduction of a BamHI site immediately after the penultimate nos codon in a 5.7 kb genomic fragment containing nos⁺ by PCR mutagenesis. Into this site was inserted a 260 bp fragment encoding 6 copies of a Myc epitope tag derived from a plasmid that was a gift from S. K. Chan and G. Struhl.

Clones encoding fragments of Cup were isolated using the host strain PJ69-4A (James et al., 1996) and a GAL4 transcriptional activation domain fusion library prepared from 0-4 hours embryonic mRNA (Dahanukar et al., 1999). In all subsequent experiments, nos or cup cDNA fragments were inserted into pACT2 to create plasmids encoding activation domain fusions. Nos fragments were tested against a DBD-Cup fusion bearing residues 342-1132 and Cup fragments against a DBD-Nos fusion bearing all 401 residues of Nos. Residues deleted in the Nos derivatives of Fig. 3 are as follows: Δ1 (43-171), Δ2 (172-287), Δ3 (288-401), Δ4 (43-287), Δ5 (172-401) and Δ6 (1-287). The pAct2-encoded epitope present in fusion proteins was detected by western blot of 100 μg samples using the HA-probe(F-7) mAb (Santa Cruz Biotechnology). In most experiments, protein-protein interaction was assessed via monitoring activation of HIS3 transcription by streaking transformants on selective medium lacking His and containing 3-10 mM 3-aminotriazole. In the experiments reported in Fig. 3, activation of the lacZ reporter as a result of protein-protein interaction was tested using the host strain Y190 (Harper et al., 1993) and standard methods.

Protein extracts from 0-4 hour embryos were prepared by douncing in 25 mM Hepes (pH 7.5), 10% sucrose, 5% glycerol, 0.1 mM EDTA, 1 mM DTT and 1× protease inhibitors. Binding reactions were performed by incubating embryonic extracts with either 5 μg GST-Cup or GST bound to glutathione agarose beads in 50 mM Hepes (pH 7.5), 150 mM NaCl, 2.5 mM EGTA, 1 mM EDTA and 0.1% Tween-20 for 3 hours at 4°C. Beads were washed five times, and bound protein eluted by boiling in SDS sample buffer. The plasmid encoding GST-Cup was prepared by inserting a XbaI-NarI fragment (encoding amino acids 593-962) into pGEX3X. Plasmids encoding Nos deletions were derivatives of pNB40; these were used to prepare labeled protein by coupled in vitro transcription and translation in the presence of ³⁵S-Met (Promega) by standard methods. Each binding reaction was performed as described above, except incubation was for 2 hours at room temperature, and the binding buffer contained 25 Mm Hepes (pH 7.5), 50 mM NaCl, 2.5 mM MgCl₂, 0.2 mM EDTA, 0.5% Triton X-100 and 0.5 mM DTT. Bound protein was displayed by gel electrophoresis and quantitated by phosphoimager analysis (Molecular Dynamics).

In the experiments of Fig. 4, all nos transgenes are derivatives of a 5.7 kb genomic fragment that completely rescues the abdominal segmentation defects in embryos derived from nos^BN females, as well as the oogenesis defects in nos^RC/Df(nos) females. Derivatives encode proteins with the following amino acids deleted: ΔA, 5-44; ΔB, 43-116; ΔC, 117-150; ΔD, 172-217; ΔE, 218-257; ΔF, 259-305; and ΔG, 5-150. At least three independent lines of each transgene were tested. The following list contains the genotypes for which no genetic interaction was observed (in addition to those reported in the text): cup⁻/+; nos^RW/Df(nos) (abdominal segmentation and oogenesis); cup⁻/+; nos^BN (abdominal segmentation); cup⁻/nos(ΔBX); nos^BN (abdominal segmentation) (Dahanukar and Wharton, 1996). In addition, no modification of the cup phenotype was observed in cup⁻; nos^RC/+ ovaries, and no modification of the nos^RC phenotype was observed in cup⁻; nos^RC/Df(nos) ovaries.

Total RNA was prepared by crushing 3-5-day-old females in Trizol (Gibco-BRL). Subsequently, poly(A)⁺ RNA was obtained using the PolyAT tract mRNA isolation kit (Promega). cDNA was prepared from 500 ng poly(A)⁺ RNA using SuperScriptII, and 50 ng of the resulting product were used in each PCR reaction using the following primers. For nos, CTCAACATTCTGGGCCTGCAGG and GTTG-CCGCCATTGGTCTGCAGC; and for actin88F, CCAGCCCTCGT-TCCTGGGC and GATCCAGACGGAGTACTTCC. In each case, the primer pair flanks intron-coding sequence, such that amplification from contaminating genomic DNA yields a larger product (543 versus 1091 bp for nos and 237 versus 297 bp for actin88F). 50 μl PCR reactions were performed using standard conditions with the following protocol: 3 minutes at 98°C; 40 cycles of 1 minute at 93°C, 1 minute at 60°C, 1 minute at 72°C; and 7 minutes at 72°C. 10 μl aliquots were removed every 10 cycles and visualized by ethidium bromide staining following electrophoresis through agarose. Reaction products were cleaved with PstI (which cuts within each primer), subcloned into pSP73, and sequenced by standard methods. Unambiguous sequence was obtained for six clones that represent four different RNA species. The nos sequences immediately flanking the cryptic splice sites (‘/’ in the list), and the net number of amino acids deleted (as appropriate) are as follows: CTTTGCGCAG/G-ACAAGGTAA (+4); CATTACGCCG/GACAAGGTAA (−76); GCCACTTTGA/CTATTCCCAG (−92); CTTTGCGCAG/CTATTC-CCAG (out of frame).

To identify proteins that interact with Nos, we performed a yeast two-hybrid screen using full-length Nos fused to the GAL4 DNA-binding domain as the bait. Two of the interactors proved to be fragments of Cup, a cytoplasmic protein of unknown biochemical function that is required for normal oogenesis (Keyes and Spradling, 1997). In the ovaries of cup mutant females, egg chamber maturation arrests between stages 5 and 14, and the nurse cells have aberrant nuclear morphology. However, all of the extant cup alleles encode detectable protein, and thus the null phenotype may be stronger. The interaction with Nos appears to be specific, as Cup fails to interact with a variety of other baits in yeast. In particular, Cup did not interact with the RNA-binding domain of Pum (Fig. 1A).

To test the interaction between Nos and Cup in vitro, we first defined the minimum region of Cup required for interaction in yeast by deletion analysis. As shown in Fig. 1B,C, residues 593-963 constituted the smallest fragment of Cup tested that interacted with Nos in yeast. We next prepared a GST fusion protein bearing these residues of Cup in bacteria and incubated it with embryonic extracts from either wild-type or transgenic flies that produced a Myc epitope-tagged Nos (see Materials and Methods) that was fully functional and rescued the defects in otherwise nos⁻ embryos and ovaries. Fig. 2 shows that approximately 10% of the Nos-Myc from the extract is retained by GST-Cup under the reaction conditions, whereas a negligible amount of Nos-Myc is retained in a control reaction with GST. As a further control, the blot was reprobed with antibodies to two other cytoplasmic proteins – Pum and Smaug; neither of these is appreciably retained by GST-Cup (data not shown). In summary, Nos appears to interact specifically with Cup in yeast and in vitro.

We next wished to determine which portion of Nos mediated the interaction with Cup. Nos contains two regions – a well-conserved C-terminal Zn²⁺-binding domain (Curtis et al., 1997) that mediates the interaction with Pum and hb mRNA (Sonoda and Wharton, 1999), and a poorly conserved N-terminal region. As shown in Fig. 3, the N-terminal region of Nos mediates interaction with Cup. No biochemical function has previously been ascribed to this portion of Nos, which is very poorly conserved even among closely related Dipteran homologs (Curtis et al., 1995).

Further deletion analysis of the N-terminal region reveals that it contains at least two redundant sub-domains that can interact with Cup (Fig. 3A,B).

As an independent test of these results, the same Nos fragments tested in yeast were prepared by coupled transcription/translation in vitro using ³⁵S-Met. Each fragment was first incubated with GST-Cup, and then bound protein was eluted and visualized by autoradiography. As shown in Fig. 3C, the N-terminal region of Nos bound to GST-Cup in vitro, whereas the C-terminal Pum-interaction domain of Nos did not (consistent with results obtained in the yeast assay).

The function of the N-terminal region of Nos in vivo is not clear. A recent analysis of 60 nos⁻ alleles revealed that mis-sense mutations that eliminated both ovarian and embryonic function altered residues in the conserved C-terminal domain, consistent with the idea that it is required for function in both tissues (Arrizabalaga and Lehmann, 1999). In contrast, no such mis-sense mutations were found in the coding region for the poorly conserved N-terminal domain. Microinjection of mRNAs encoding deletion derivatives of Nos into nos⁻ embryos suggested that no single part of the N-terminal region was essential for regulation of hb mRNA (Curtis et al., 1997). No comparable analysis of Nos residues required for ovarian function has been reported.

To identify residues of Nos essential for its activity in the ovary, we prepared nos transgenes that encode the deletion derivatives shown in Fig. 4. Collectively, almost every amino acid between residue 4 and the C-terminal Pum-binding domain (residue 304) was deleted in these derivatives. The only residues not deleted in one of the derivatives of Fig. 4 were the 21 amino acids encoded by bases flanking the first intron in the pre-mRNA. At least three independent lines of each transgene were then crossed into homozygous nos^BN and hemizygous nos^RC backgrounds to assess function in the embryo and the ovary, respectively.

As summarized in Fig. 4, no part of the N-terminal region of Nos is essential for its function in either the embryo or the ovary. Maternal expression from a single transgene encoding each deletion derivative rescued the ovarian morphology and egg-laying defects associated with nos^RC. Each deletion derivative also rescued the abdominal segmentation defects associated with nos^BN either completely (a full complement of 8 abdominal segments) or nearly completely (6-8 abdominal segments). One derivative, ΔG, appears to have somewhat less activity than the others; however, expression from two maternal copies of the ΔG transgene completely rescues abdominal segmentation in 100% of embryos. Thus, we conclude that the N-terminal region of Nos contains no unique sequence that is essential for its activity in either the embryo or the ovary. This latter observation is consistent with the finding that interaction with Cup is mediated by redundant elements in the N-terminal region of Nos (Fig. 3).

To determine whether the interaction between Nos and Cup is functionally significant, we asked whether lowering the level of Cup modified any of the ovarian or embryonic phenotypes associated with altered Nos function. In one case, we observed a strong genetic interaction: introduction of a single cup allele substantially suppressed the oogenesis defects in hemizygous nos^RC mutant females (Fig. 5). As described by Forbes and Lehmann (1998), the cystoblasts that give rise to the germline components of the egg chamber do not develop normally in nos^RC mutant ovaries, and germline stem cells that give rise to cystoblasts are not maintained. As a result, nos^RC mutant ovaries contained only rare mature egg chambers (Fig. 5). In contrast, in cup⁻/+; nos^RC/Df(nos) ovaries, many of the egg chambers appeared normal and matured into oocytes that were fertilized and oviposited (Fig. 5). (The resulting embryos developed no abdominal segments, presumably because they lacked sufficient Nos activity to repress hb translation.) The cup⁻/+; nos^RC/Df(nos) females lay eggs for at least 3 weeks, suggesting that germline stem cells are maintained and function normally. Thus, reducing the level of Cup appears to specifically suppress the oogenesis defects associated with the nos^RC allele, but not the embryonic defects.

Nine different cup alleles tested suppress the defects associated with nos^RC, suggesting that it is simply a reduction of Cup activity that suppresses the oogenesis phenotype. In contrast, the genetic interaction appears to be specific to the RC allele; the nos^RD mutant encodes an unstable protein bearing a substitution at one of the conserved Cys residues in the C-terminal domain (Curtis et al., 1997). This allele exhibits oogenesis defects similar to nos^RC (Wang et al., 1994), but these defects are not ameliorated by lowering the level of Cup (not shown), presumably because the level of active Nos protein is insufficient. In addition, reducing the level of Cup has no effect on the oogenesis defects associated with two different allelic combinations of pum (pum^ET1/pum^Msc and pum^FC8/pum^Msc). Thus, reduction of Cup activity does not appear to globally suppress oogenesis defects resulting from alterations in Nos or Pum activity, but specifically suppresses the defects associated with nos^RC.

The genetic interaction between nos and cup suggests that expression of the protein encoded by each gene coincides, and previous reports support this idea (Keyes and Spradling, 1997; Wang et al., 1994). However, we wished to visualize the distributions of Nos and Cup simultaneously to determine whether the spatiotemporal distribution of the proteins was consistent with the genetic interaction we observed. Using our anti-Nos antibodies, we could not reliably detect Nos in the germarium. Therefore, we examined the localization of Cup and Myc-tagged Nos in the transgenic flies described above that carry a fully functional nos⁺ transgene altered to encode a Myc epitope tag at the C terminus of the protein.

As shown in Fig. 6, Cup was present throughout the cytoplasm of all the germ cells in the germarium, the terminal region of the ovary that contains the stem cells, cystoblasts and most immature egg chambers (consistent with the previous report of Keyes and Spradling, 1997). In contrast, the Nos-Myc distribution was not uniform. It was present in the germline stem cells and cystoblasts in region 1 of the germarium, fell beneath the level of detection during the early cystoblast cleavages in regions 1 and 2, rose to relatively high levels in the germline cysts in region 2, and fell to somewhat lower levels in the maturing cysts of region 3 (consistent with the findings of Forbes and Lehmann, 1998). The significant finding was that Nos and Cup co-localized to the cytoplasm of the stem cells, the cystoblasts and the cysts, consistent with the genetic interaction described above.

The data reported above support the idea that Cup interacts with the N-terminal region of Nos and thereby lowers its activity, perhaps titrating it away from regulatory targets. This follows from the physical interaction described in Figs 1-3 and the genetic interaction between nos^RC and cup described in Fig. 5. However, nos^RC, which bears a mutation in the splice donor of intron 1 in the pre-mRNA, has been described as a null allele, and no mature mRNA is detectable by either in situ hybridization or northern blot (Arrizabalaga and Lehmann, 1999; Wang et al., 1994). How then can a physical interaction between Cup and Nos account for the genetic interaction between cup and nos^RC?

To address this question, we asked whether nos^RC actually encoded a very low level of functional protein with two approaches. First, semi-quantitative RT-PCR was used to determine whether nos^RC flies contained low levels of mRNA. Using primers that flanked intron 1, a low level of nos mRNA was detected in extracts prepared from whole flies (Fig. 7). Since the major site of transcription in adult females is the ovary, we assume that most of this mRNA is derived from the rudimentary nos^RC ovaries. Two major cDNA species were detected by ethidium bromide staining of the PCR product following electrophoresis. To further characterize these cDNAs, the PCR products were subcloned and six individual clones were sequenced. The six clones appear to represent mRNAs generated by processing from cryptic splice sites; the open reading frame is preserved in three different clones, and one of these encodes a Nos derivative that is four amino acids larger than wild type (see Materials and Methods). This cDNA clone plausibly represents the mRNA species that gives rise to the nos^RC-encoded protein described below.

In an attempt to detect protein encoded by nos^RC directly, we prepared transgenic flies bearing a nos^RC-myc gene that encodes an epitope-tagged protein that is otherwise identical to the nos⁺-myc gene described above. Using western blots, a very low level of nearly full-length protein was detectable in nos^RC-myc ovaries from five different transgenic lines (Fig. 7B). Comparison with dilutions of extracts prepared from nos⁺-myc transgenic ovaries suggests that the level of protein encoded by nos^RC was in the order of 1-2% of wild type. By crossing the nos⁺-myc and nos^RC-myc transgenes into cup⁻/+ backgrounds and comparing the level of Nos protein in ovarian extracts, we found that reducing the level of Cup did not significantly affect the level of protein encoded by either transgene (Fig. 7C). Thus, low levels of Cup do not appear to suppress the nos^RC ovarian phenotype by stabilizing Nos. We conclude that, in the presence of reduced levels of Cup, 1-2% of the wild-type level of Nos is sufficient to promote normal maintenance of the germline stem cells and differentiation of the cysts.

In this report, we show that the poorly conserved N-terminal region of Nos mediates an interaction with Cup in the germ cells of the ovary. Genetic experiments reveal that the Nos-Cup interaction is inhibitory, restricting the level of active Nos. Our results further suggest that Nos normally is present in vast excess in the ovary, and that ∼1-2% of the wild-type amount is sufficient to promote normal oogenesis. This level of protein is supplied by the nos^RC allele, previously thought to be null. This observation leads us to reconsider the possibility that Nos and Pum work collaboratively in the ovary, as they do in the embryo.

Reducing the level of functional Cup suppresses the oogenesis defects in hemizygous nos^RC ovaries (Fig. 5). This finding suggests that Cup acts to inhibit the residual protein encoded by nos^RC and prevent it from acting on potential regulatory targets. The identities of such targets are not currently known. In addition, the sequence of Cup sheds no light on its function. A search of the current database reveals no significant homologies to proteins of known function, neither does it possess recognizable motifs using programs such as Prosite, although Keyes and Spradling (1997) have suggested that Cup may be a microtubule-associated protein. A fragment of human sequence bears high homology to a part of the fly protein, and thus it seems likely that one or more of the Cup functions are evolutionarily conserved. Further analysis of the significance of the Nos-Cup interaction awaits definition of the biochemical activities of Cup and the identification of Nos-regulated genes in the ovary.

While physiological levels of Cup are capable of inhibiting the low level of Nos activity in hemizygous nos^RC flies, we do not currently understand what role the Cup-Nos interaction plays in the ovaries of wild-type flies. Over-expression of Nos is deleterious in many different tissues – the embryo (Gavis and Lehmann, 1992; Wharton and Struhl, 1989), the eye imaginal disc (Wharton et al., 1998) and the male germline (R. P. W., data not shown) – suggesting that Nos is a potent regulator of gene expression. Consistent with this idea, we find that extremely low levels of Nos, in the order of a few percent of the wild-type amount, suffice for biological activity in the ovary. Thus, it seems possible that the interaction with Cup helps restrict Nos activity, which otherwise might interfere with normal ovarian development. Alternatively, Nos and Cup may act together to govern some aspect of germ cell development. Cup appears to play a role in early germline development, as cup and ovarian tumor interact genetically in the ovary, leading to over-proliferation of germline cells (Keyes and Spradling, 1997).

Previous studies have led to a clear understanding of the role of the C-terminal zinc-binding domain of Nos. This domain of the protein mediates interaction with Pum on hb mRNA to regulate its translation (Sonoda and Wharton, 1999). Inactivating mis-sense mutations of nos alter residues within this domain (Arrizabalaga and Lehmann, 1999), and this domain is conserved in Nos homologs from a broad variety of species (Kraemer et al., 1999; Subramaniam and Seydoux, 1999).

In contrast, the function of the N-terminal region of Nos is not clear. This region is poorly conserved, and an extensive screen for new nos alleles identified no inactivating mis-sense mutants with substitutions within this domain (Arrizabalaga and Lehmann, 1999). Analysis of several truncated Nos derivatives by microinjection suggested that the N-terminal region contains no unique component essential for embryonic activity; however, the C-terminal domain on its own was inactive in these experiments, suggesting that some residues in the N-terminal region are required for Nos function (Curtis et al., 1997).

The results shown in Fig. 4 extend this analysis to test structural requirements for the ovarian function of Nos. We were unable to identify residues in the N-terminal region that are uniquely required for either ovarian or embryonic Nos function. The only residues not deleted in one of the constructs in Fig. 4 are the 21 amino acids derived from nucleotides near intron 1 in the pre-mRNA, which are very poorly conserved even among Dipteran Nos homologues. Cup appears to be one factor that modulates Nos activity in the ovary via its N-terminal region, but other proteins may also do so. Taken together, the experiments of Fig. 4 impose two restrictions on such molecules. First, as is the case for Cup, these proteins must interact with multiple, redundant subregions. Second, since the ΔG protein of Fig. 4 is fully functional, these proteins must interact with residues C-terminal to 150.

Recent genetic experiments on C. elegans have shown that homologs of Nos and Pum regulate survival and development of germ cells (Kraemer et al., 1999; Subramaniam and Seydoux, 1999). Nos and Pum appear to play similar roles in Drosophila; in addition, the mechanism by which they act to regulate abdominal segmentation in the embryo involves an intimate interaction between the conserved RNA-binding domain of Pum and the conserved C-terminal domain of Nos (Sonoda and Wharton, 1999). Taken together, these observations lead to the attractive idea that Nos and Pum act on evolutionarily conserved targets to regulate mRNAs involved in governing the behavior of germ cells.

Earlier studies of nos and pum mutant phenotypes suggest that the two proteins have different roles in the germline, perhaps interacting with different partners or regulatory targets. This idea was supported by three lines of evidence. First, the ovarian phenotype of presumptive null alleles of pum and nos is different, the former being more severe (Forbes and Lehmann, 1998). Second, the distribution of Nos and Pum in the germ cells of the germarium is somewhat different, consistent with a different site of action (Forbes and Lehmann, 1998). Third, alleles of nos exist that apparently specifically affect embryonic function of the protein (Arrizabalaga and Lehmann, 1999).

All three lines of evidence can be reinterpreted in light of the results reported above. The apparent difference between pum and nos mutant ovarian phenotypes may simply result from the residual activity of nos^RC, the allele used in these experiments. The amount of protein derived from this allele is very low; however, it apparently is sufficient to support normal germ cell function (in a cup heterozygote). This amount of Nos (i.e. 2% of the Nos-Myc protein in Fig. 6) is far below our current level of detection, and therefore the distribution of detectable Nos in the wild-type germarium does not suggest where its activity is required. Finally, mutant derivatives of Nos that selectively reduce embryonic function bear alterations in the C-terminal tail of the protein (Arrizabalaga and Lehmann, 1999), which is required for efficient interaction with Pum on hb mRNA (Sonoda and Wharton, 1999). However, if such a low level of Nos activity is sufficient in the ovary, then even an inefficient interaction with Pum may allow recruitment of these mutant Nos proteins into ternary complexes on mRNAs regulated in the germarium.

Whether Nos and Pum act separately or together in the germ cells awaits the identification of ovarian regulatory targets. However, current evidence favors the idea of Subramaniam and Seydoux (1999), that the two molecules act together to govern germ cell development in an evolutionarily conserved manner.

We thank Michelle Patterson for performing the two-hybrid screen and Tammy Lee for generating the nos transgenes and analyzing their activity in embryos. We also thank Lihsia Chen, Jonathan Davis and Jun Sonoda for help with confocal microscopy and comments on the manuscript; laboratory members for suggestions; Trudi Schüpbach for flies; Miguel Arevalo-Rodriguez, M. Cristina Cruz and Joe Heitman for yeast reagents and advice; Cary Gardner and Scott Pyle for technical help; Glenda Jackson for media and fly food preparation; and Sandy Boyles for secretarial help. A. C. V. was supported in part by an Adriano Buzzati-Traverso post-doctoral fellowship. R. P. W. is an Assistant Investigator of HHMI.