Translational control of maternal Cyclin B mRNA by Nanos in the Drosophila germline

Nanos (Nos) represses translation of maternal hunchback(hb) mRNA in the presumptive somatic cytoplasm of the early Drosophila embryo, thereby governing abdominal segmentation(Hülskamp et al., 1989; Irish et al., 1989; Struhl, 1989). Nos is recruited into a repressor complex that contains two ubiquitous factors,Pumilio (Pum) and Brain tumor (Brat). Pum nucleates formation of the complex by recognizing Nanos Response Elements (NREs) in the 3′ UTR of hb and recruiting Nos (Sonoda and Wharton, 1999). The subsequent recruitment of Brat to the Pum-hb NRE-Nos complex (Sonoda and Wharton, 2001) results in translational inhibition via both poly(A)-dependent and poly(A)-independent mechanisms(Chagnovich and Lehmann, 2001). A transient gradient of Nos is generated by Oskar (Osk)-dependent localization and translational activation of nos mRNA in the specialized pole plasm at the posterior of the embryo(Ephrussi and Lehmann, 1992; Smith et al., 1992). For regulation of hb, Nos is thought to be the sole spatially limiting factor (Gavis and Lehmann,1992).

In addition to their role in abdominal patterning, Nos and Pum play a number of roles in the primordial germ cells (PGCs). PGCs that lack Nos or Pum function enter mitosis prematurely, fail to migrate to the somatic gonad,undergo apoptosis, and fail to maintain stem cell identity in adults(Lin and Spradling, 1997; Asaoka-Taguchi et al., 1999; Asaoka and Lin, 2004; Hayashi et al., 2004; Wang and Lin, 2004). After dissipation of the Nos gradient in the presumptive somatic cytoplasm (e.g. by nuclear division cycle 9-10), high levels of Nos are found only in the PGCs(Wang et al., 1994). The limited distribution of Nos, coupled with the study of Nos orthologs in Caenorhabditis elegans and Homo sapiens, suggests that the ancestral function of Nos is in the germline(Subramaniam and Seydoux,1999; Jaruzelska et al.,2003; Tsuda et al.,2003).

Although the regulatory targets of Nos and Pum in the PGCs have not yet been defined, one excellent candidate is maternal Cyclin B(CycB) mRNA. PGCs cease proliferating shortly after their formation at the posterior pole of the embryo, emerging from quiescence only after migrating to, and arriving in, the presumptive gonad in late embryogenesis(Su et al., 1998). At least part of this quiescence is thought to be due to Pum- and Nos-dependent repression of CycB mRNA(Asaoka-Taguchi et al., 1999). CycB accumulates prematurely in the PGCs of embryos from nos or pum mutant females (hereafter, nos and pum mutant embryos). Conversely, ectopic CycB drives otherwise wild-type PGCs into premature mitosis, consistent with the idea that repression of CycBtranslation limits proliferation. However, the ectopic CycB in these experiments was derived from a transgene that directs the maternal synthesis of a chimeric mRNA (consisting of 5′ and 3′ UTRs from nosfused to the CycB ORF) under nos transcriptional control(Asaoka-Taguchi et al., 1999);the experiment thus provides only modest support for the idea that Pum and Nos directly target native CycB mRNA.

Recent experiments have provided insight into the likely functions of two components of the repressor complex assembled on hb-Pum and Brat. Pum is a founding member of the conserved Puf domain family of RNA-binding proteins (Zhang et al., 1997). One of the budding yeast Puf proteins, MPT5, has recently been shown to bind specific mRNA targets and regulate their stability by interacting with the Pop2 subunit of the CCR4-Pop2-NOT complex(Goldstrohm et al., 2006). This complex contains deadenylation enzymes (CCR4 and Pop2) as well as factors that promote decapping (Dhh1), and thus is able to regulate either the stability or translation (or both) of mRNAs to which it is recruited. Based on the observation that Puf proteins from H. sapiens, C. elegans and Saccharomyces cerevisiae interact with orthologous Pop2 subunits,Wickens and colleagues (Goldstrohm et al.,2006) suggested that Puf proteins generally act by recruiting the deadenylase complex. Brat appears to repress translation (at least in part) by recruiting 4E-HP, which inhibits binding of the essential initiation factor eIF-4E to the mRNA cap (Cho et al.,2006). Brat is likely to have additional repressor functions,because mutants that do not bind 4E-HP exhibit relatively minor defects in hb regulation. The only function attributed to Nos to date has been to assist Pum in recruiting Brat (Sonoda and Wharton, 2001).

Regulation of maternal hb and CycB mRNA differs in two important respects. First, repression of CycB is Brat-independent(Sonoda and Wharton, 2001). As the only known function of Nos for regulation of hb is in Brat recruitment, the role of Nos in regulation of CycB (and presumably other mRNAs in the PGCs) has been unclear. Second, repression of hboccurs both in the PGCs and broadly throughout the posterior of the embryo,whereas repression of CycB is strictly limited to the PGCs(Tautz, 1988; Asaoka-Taguchi et al., 1999). These different spatial domains of repression might be due to differential sensitivities of the hb and CycB mRNAs to the concentration of Nos, which persists at high levels only in the PGCs. However, this idea has not been critically tested.

In this report, we investigate the regulation of maternal CycBmRNA as a model for understanding Nos and Pum action in the germline. We first show that CycB indeed is directly regulated by binding of Nos and Pum to an element in its 3′ UTR. We then describe experiments that suggest Nos is primarily responsible for recruiting the CCR4-Pop2-NOT deadenylase complex to CycB. Finally, we show that regulation of CycB in the somatic cytoplasm is deleterious and that it is restricted to the germline by a dual requirement for high levels of Nos and another factor active only in the PGCs.

pum^ET3, In(3R)Msc, and nos^BN flies were from R. Lehmann (New York University, NY); brat^fs1,Df(2L)TE37C-7, CCR4^KG877, and Df(3R)crb-F89-4 flies were from the Bloomington Stock Center; CycB² and CycB³ flies(Jacobs et al., 1998) were from C. Lehner (University of Bayreuth, Germany). CycB, hb, nos-bcdand osk-bcd transgenes were constructed in pCasPeR derivatives and introduced into w¹¹¹⁸ flies by standard methods. The rat anti-Hb and rabbit anti-Nos antibodies were gifts from P. Macdonald(University of Texas, Austin, TX) and A. Nakamura (RIKEN, Kobe, Japan),respectively. The F2F4 anti-CycB developed by P. O'Farrell was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa. Anti-Xpress antibodies were from Invitrogen, anti-Histone antibodies from Chemicon International(Fig. 6), and chicken anti-Vas from K. Howard (University College London, UK). Secondary antibodies were from Jackson ImmunoResearch Laboratories. The distributions of CycB, Vasa, Hb and DNA were detected in fixed embryos with a Zeiss 510 confocal microscope. The distributions of Hb and Nos were detected with peroxidase-coupled secondary antibodies with a Zeiss Axiophot microscope using either brightfield or Nomarski optics and a Spot RT digital camera. CycB mRNA was detected by standard in situ hybridization methods using digoxigenin-labeled DNA probes(Roche) and visualized as above for Hb and Nos. To detect transgenic CycB gene products, transgenes were crossed into either CycB²/+,CycB³/+, or CycB²/CycB³ backgrounds so that embryos did not receive an excess of maternal gene product. For the pum^- embryos in Fig. 4, CycB mRNA or protein from the endogenous genes would interfere with the detection of transgenic gene product, and so these experiments were performed in a CycB²/CycB³ background.

RNA for all experiments was prepared by transcription of derivatives of the plasmid R4685, which contains a SpeI site (underlined) embedded in hb 3′ UTR-coding sequence(CTAAAAACTATCATAAAGACTAGTCTGGAGAAACATAAGCCCTGCA) that is inserted into Bluescript II KS-. Oligonucleotides encoding fragments of hb(ctagTATTATTTTGTTGTCGAAAATTGTACATAAGCC), CycB(ctagtcgagGCATAAAAAAGAGACGTAGACTATTTGTAATTTATATCATGTATTTCGCACATTCATACgaattcatcgatgat),or eIF4E(ctagGCATAATATAAAATCTATCCGCTTTTTGTAATCACTGTCAATAATGGATTAGACGGAAAAGTATATTAA)were inserted into the SpeI site of R4685. (The 5′ SpeI-overhang and, where relevant, polylinker derived nucleotides are in lower case.) ³²P-labeled RNA was prepared using T7 MEGAShortscript (Ambion).

Plasmids encoding GST-Pum, His6-Nos and His6-Brat are described elsewhere(Wharton et al., 1998; Sonoda and Wharton, 1999; Edwards et al., 2003). A plasmid encoding the Drosophila Pum RNA-binding domain was constructed in a derivative of pET-19b in which the thrombin site is replaced with a Tev cleavage site. Cleavage with Tev liberates the following fragment of Drosophila Pum from an N-terminal His-tag: gtRSRLL...YYlKITN(where gt are vector-encoded and l is substituted for the native M near the C-terminus) (Edwards et al.,2000). Like the RBD of human Pum, the Drosophila protein is monomeric at mM concentrations.

Gel mobility shift experiments were performed essentially as described(Murata and Wharton, 1995),with the following modifications. Each 10 μl reaction contained purified protein, reaction buffer [10 mM HEPES (pH 7.4), 20 mM KCl, 1 mM DTT, 0.2 mg/ml heparin, 0.05 mg/ml poly(U), 5% glycerol], and heat-denatured labeled RNA(10,000 cpm). Nos and Brat recruitment experiments were performed essentially as described (Sonoda and Wharton,1999; Sonoda and Wharton,2001), with the following modifications: each 40 μl reaction contained 1.5 μM GST-Pum, 0.6 μM His6-Nos, 2 μM RNA, and, for Brat recruitment, 0.4 μM His6-Brat, in reaction buffer [20 mM HEPES (pH 7.9), 5 mM MgCl₂, 5 μM ZnCl₂, 5mM DTT, 100 mM NaCl, 0.5%Tween 20, 0.1% BSA, 500 U/ml RNase Inhibitor (Roche) and 1×EDTA-free protease inhibitor (Roche)]. For the experiments shown in Fig. S3 of the supplementary material, all CycB RNAs were either the wild type sequence above (i.e. bearing 59 nt from the 3′ UTR) or mutant derivatives thereof.

CycB transgenes were derived from a modified wild-type rescuing construct (Jacobs et al.,1998) that bears an XbaI site immediately 3′ to nucleotides encoding the stop codon to facilitate plasmid construction. Oligonucleotides encoding the 59 nt CycB NRE (above), the 50 nt CycB NRE(ctagtAGAGACGTAGACTATTTGTAATTTATATCATGTATTTCGCACATTCATAC),the inactive 40 nt CycB NRE (underlined nucleotides in the 50 nt sequence deleted), two copies of the 32 nt hb NRE (above) and either one or two copies of the MS2hp (ctagAAACATGAGGATCACCCATGTA) were inserted into a SpeI site that replaces sequences deleted in the ΔI portion of the ΔI+II gene or into the deleted portion of the ΔIII gene(Fig. 1, and see Fig. S2 in the supplementary material). hb transgenes were derivatives of p2343(Murata and Wharton, 1995). Derivatives of a wild-type nos transgene that encode three different Nos-CP fusions were constructed by inserting sequences encoding wild-type MS2 CP between wild-type Nos residues 3 and 4, residues 197 and 198, and at the C-terminus. Essentially identical regulation of CycB(2x MS2hp) was observed with each fusion.

GST-pulldown experiments were performed as described(Goldstrohm et al., 2006),except that Pop2 proteins were labeled with ³⁵S-methionine during synthesis in vitro. The GST-HsPum fusion protein was as described(Goldstrohm et al., 2006); the GST-Dm-Pum protein used in these experiments contains the homologous region only (i.e. residues 1091-1433 containing the RBD but not the C-terminal tail). NOT4 clones were isolated in two different yeast interaction screens, a two-hybrid screen performed by M. Patterson and R.P.W. using DBD-Nos(full-length) as bait and a four-hybrid screen(Sonoda and Wharton, 2001). The NOT4 clones identified in the latter screen proved to interact with the Nos moiety of the three-hybrid bait. Fusions of Drosophila Pop2 to the DBD or AD were prepared from cDNA clone RH51274 and plasmids pGBT9,pGAD424,and pActII. Interaction between various protein pairs was tested by co-transformation with plasmids encoding Nos, Pum, Cup or NOT4 derivatives into PJ69-4A (James et al.,1996). Tethering of Pop2 via the DBD robustly stimulated transcription of the HIS3 and ADE2 reporters without co-expression of any AD-fusion, and thus the DBD-Pop2 fusion was not used further.

A sequence required for translational repression of CycB mRNA in the PGCs was previously mapped to nts 430-469 of the 3′ UTR (within Fragment D in Fig. 1A) by microinjection of modified mRNAs encoding epitope-tagged protein(Dalby and Glover, 1993). However, we found that neither the full-length Pum in embryonic extracts nor a purified untagged Pum fragment consisting of little more than the RBD(residues 1092-1433) bound detectably to Fragment D in gel mobility shift experiments (Fig. 1B).

We showed that nts 594-643 of the 3′ UTR comprise an NRE necessary for repression of CycB in the PGCs, using a combination of molecular genetic and biochemical experiments that are summarized below, with supporting evidence presented in Fig. 1C-Eand see Figs S1, S2 and S3 in the supplementary material. The CycBNRE contains two UGUA motifs that are present in most Pum-binding sites(Gerber et al., 2006). Two complexes were detectable in gel mobility shift experiments using a short RNA substrate bearing the wild-type NRE. Analysis of Pum binding to various mutant NREs is consistent with the idea that one Pum RBD recognizes each UGUA motif(plus flanking nts). Mutations in either motif reduced Pum binding in vitro and abrogated repression in the PGCs, suggesting that normal regulation requires binding of Pum to both UGUA motifs. Nos was recruited to Pum-CycB NRE complexes (see Fig. S3 in the supplementary material),much as it is recruited to Pum-hb NRE complexes(Sonoda and Wharton, 1999). However, Brat was recruited only to the Pum-Nos complex assembled on the hb NRE and not the corresponding complex assembled on the CycB NRE (see Fig. S3 in the supplementary material); this finding is consistent with the observation the CycB regulation is Brat-independent (Sonoda and Wharton,2001).

We next wished to determine whether the CycB NRE is sufficient to confer regulation on another maternal mRNA, and if so, whether regulation is restricted to the PGCs. To this end, we constructed a transgene encoding a chimeric hb mRNA in which the CycB NRE is substituted for the native hb NRE, and assayed the distribution of Hb in embryos from transgenic females.

As shown in Fig. 2, the CycB NRE imparted CycB-like regulation on hb mRNA:protein accumulation was blocked but only in the PGCs at the posterior extreme of the embryo. Hb protein accumulated throughout the posterior of hb(CycB NRE) embryos, a distribution that was indistinguishable at all stages examined from that in hb(ΔNRE)embryos. The ectopic Hb in hb(CycB NRE) embryos blocked all abdominal segmentation, as described previously for hb(ΔNRE)embryos (Wharton and Struhl,1991). We conclude that the CycB NRE mediates repression in the PGCs but is not normally functional in the prospective somatic cytoplasm, even immediately adjacent to the PGCs, where high levels of Nos accumulate (albeit transiently).

Although the mechanism by which CycB mRNA is repressed is not yet known, two lines of evidence suggested that deadenylation catalyzed by the CCR4-Pop2-NOT complex is likely to be involved. First, CycB mRNA is de-regulated and hyper-adenylated in ovaries from flies hemizygous for a partial loss-of-function allele of CCR4 (also known as twin- Flybase) (Morris et al.,2005). Consistent with these observations, we find that CycB was de-repressed in the PGCs of embryos from CCR4mutant females (Fig. 3A). The second line of evidence comes from a recent study of budding yeast MPT5(structurally and functionally related to Pum), which suggested that Puf domain proteins generally interact with the Pop2 subunit of the deadenylase complex to regulate mRNA stability or translation(Goldstrohm et al., 2006). Consistent with this idea, we found that the Pum RBD bound to Drosophila Pop2 in GST-pulldown experiments(Fig. 3B). As is the case for Puf proteins from H. sapiens, S. cerevisiae and C. elegans(Goldstrohm et al., 2006), Drosophila Pum also bound to heterologous Pop2 proteins(Fig. 3B).

If Drosophila Pum interacts directly with Pop2, then what is the role of Nos? Unlike regulation by the yeast Puf protein MPT5 (which is thought to act without any co-factor), regulation by Drosophila Pum is absolutely dependent on its ability to recruit Nos to the NRE. One possibility is that Nos helps recruit the deadenylase complex, stabilizing its binding to the 3′ UTR and increasing the probability of deadenylation. We did not observe binding of Pop2 to Nos in yeast two-hybrid experiments (not shown). However, we isolated four different clones encoding Cterminal fragments of NOT4 (another component of the deadenylase complex) in two different yeast interaction screens in which Nos was part of the bait(Fig. 3C). The interaction between Nos and NOT4 was somewhat more robust than the interaction between Nos and Cup, shown previously to be biologically relevant during early oogenesis(Verrotti and Wharton, 2000). Binding of NOT4 to Nos was mediated by its N-terminal region(Fig. 3C), which is essential for Nos activity, despite being very poorly conserved(Curtis et al., 1997). Taken with previous work (Sonoda and Wharton,1999), these results suggest that Nos acts as a bridge, contacting Pum with its C-terminal zinc finger and NOT4 with its N-terminal region.

In principle, the Nos-NOT4 interaction might be sufficient to regulate CycB; alternatively, contacts made by Pum (to Pop2) and Nos (to NOT4)might both be required to efficiently recruit the deadenylase complex and regulate translation. To distinguish between these models, we asked whether tethering Nos via an exogenous RNA-binding domain would repress translation of CycB mRNA. To do so, we prepared transgenic flies that express a chimera in which Nos is fused to bacteriophage MS2 coat protein (CP)(Coller and Wickens, 2002)under the control of native nos regulatory signals. When crossed into the appropriate background, these nos-CP transgenes fully rescued various nos^- phenotypes, demonstrating that the fusion protein is functional when recruited to target mRNAs via Pum. Next, we prepared transgenic flies expressing a CycB(2x MS2hp) derivative in which the NRE is replaced with two copies of a short hairpin to which CP binds with high affinity. We then asked whether tethering of Nos via CP can repress the CycB(2x MS2hp) mRNA by monitoring CycB accumulation in embryos from doubly transgenic females.

Binding of the Nos-CP chimera repressed translation of CycB(2x MS2hp) mRNA in the PGCs (Fig. 4A). CycB(2x MS2hp) mRNA was not repressed in the PGCs by wild-type Nos, presumably because its natural binding partner, the Pum-CycB NRE complex, could not form (in the absence of the NRE). However, CycB(2x MS2hp) mRNA was repressed upon co-expression of Nos-CP, even in the absence of pum function. Apparently, the sole role of Pum in regulation of CycB is to recruit Nos.

We considered the possibility that tethering Nos via the highaffinity CP-MS2hp interaction might reveal a weak intrinsic capacity of Nos for translational regulation that is normally significant only with additional contributions from bound Pum. If so, we might expect that Nos-CP would bypass the requirements for both Pum and Brat and regulate an analogous hb(2x MS2hp) chimeric mRNA (bearing a substitution of MS2hp sites for the endogenous NRE). As shown in Fig. 4B, we saw no evidence of such regulation: Hb accumulated uniformly throughout the posterior somatic cytoplasm and in the PGCs of embryos bearing maternal hb(2x MS2hp) mRNA, whether or not Nos-CP was co-expressed. The failure to observe repression of hb(2x MS2hp) even in the PGCs argues against the idea that repression of CycB(2x MS2hp)is artefactual.

In summary, the evidence described above supports the idea that, for regulation of CycB mRNA in the PGCs, the primary role of Pum is to recruit Nos, which subsequently recruits the deadenylase complex via direct interaction with NOT4.

Nos and Pum jointly repress CycB and (with the help of Brat) hb mRNAs in different regions of the embryo. CycB is repressed only in the PGCs at the posterior extreme, whereas hb is repressed broadly throughout the posterior ∼50% of the embryo. How are these very different domains established? A priori, it seemed likely that the spatiotemporal gradient of Nos would be primarily responsible for setting different domain boundaries, because the other known direct regulators of either CycB or hb, Pum and Brat, are uniform along the anteroposterior axis of the early embryo. According to this idea, the CycB NRE is relatively insensitive, tuned to respond only to the high levels of Nos that persist in the PGCs; the hb NRE is more sensitive,tuned to respond to the low levels of Nos present transiently in the somatic cytoplasm nearer the middle of the embryo. In this model, Nos is the sole spatially limiting factor that determines the domain of regulation along the anteroposterior axis of the early embryo.

As a test of this idea we asked whether overexpression of Nos in the anterior somatic cytoplasm from a nos-bcd chimeric mRNA results in ectopic repression of CycB mRNA. To our surprise, we saw no evidence of such repression (not shown), despite the ability of the ectopic Nos in these embryos to repress hb mRNA in the anterior(Gavis and Lehmann, 1992). We considered the possibility that maternal CycB mRNA is resistant to repression in the somatic cytoplasm (e.g. by virtue of a cis-acting activation signal that overrides repressive signals). To test this idea, we constructed a transgene encoding a chimeric CycB mRNA in which two copies of the hb NRE are substituted for the native signal. When embedded in the 3′ UTR of either hb or torso mRNAs,the hb NRE mediates repression broadly throughout the posterior 50%of the embryo (Wharton and Struhl,1991).

Expression of CycB(2x hb NRE) mRNA causes no obvious dominant phenotypes in a wild-type background and rescues the oogenesis defects associated with CycB null alleles(Jacobs et al., 1998). However, we observed profound developmental defects in embryos in which the sole source of maternal CycB derives from the chimeric transgene. As shown in Fig. 5A, over 95% of CycB^-; CycB(2x hb NRE) embryos had defects in the abdominal segments, the posterior terminalia or both. In the extreme, embryos die with an open posterior hole and no segments following the third thoracic. These development defects appear to arise during nuclear cleavage cycles 9-13, when oscillations in Cyclin/Cdk2 activity first regulate nuclear divisions (Edgar et al.,1994). As shown in Fig. 5B, somatic nuclei in the posterior of CycB^-; CycB(2x hb NRE) embryos delayed progression through cycles 9-13 and ultimately lost contact with the embryonic cortex, falling into the yolky interior. Nuclear cycle defects were limited to the posterior and were not observed during cycles 1-8, when divisions are independent of fluctuation in Cyclin/Cdk2 activity (Edgar et al.,1994). Taken together, these observations suggest that CycB activity can be significantly repressed in the somatic cytoplasm via the hb NRE.

The hb NRE imparts hb-like regulation on CycBmRNA by another criterion: during the initial nuclear cleavages, CycB protein accumulation was repressed in the posterior of ∼40% of CycB^-; CycB(2x hb NRE) embryos(Fig. 5C). After nuclear division cycle 9, however, CycB was uniform along the anteroposterior axis of all such embryos, except in the PGCs, where accumulation was efficiently repressed (Fig. 5C). The apparent difference between the response of hb and CycBmRNAs to regulation mediated by the hb NRE is probably due to a number of factors (e.g. the stability of repressed CycB mRNA, nuclear sequestration of Hb, dissipation of the somatic Nos gradient after cycle 9). Nevertheless, the salient finding is that repression of CycB can be observed in the somatic cytoplasm, ruling out the possibility that the mRNA is intrinsically resistant to regulation.

As repression of CycB in the somatic cytoplasm has a much more dramatic effect on nuclear division than on CycB protein accumulation, we re-examined the question of whether ectopic Nos in the anterior of nos-bcd embryos can repress CycB, which we did by examining nuclear morphology. As shown in Fig. 6, CycB-dependent progression through syncitial nuclear cleavages 9-13 was normal, despite the presence of high levels of persistent Nos in such embryos. The obvious difference between the ectopic Nos in the somatic anterior and the endogenous Nos at the posterior of nos-bcd embryos is that the latter is accompanied by germline factors localized in the PGCs.

Therefore, we asked whether ectopic Nos can repress CycB mRNA in the presence of pole plasm components. To do so, we examined embryos with a chimeric oskar-bicoid mRNA(Ephrussi and Lehmann, 1992). Translation of this chimera generated sufficient Oskar (Osk) at the anterior pole to direct the formation of pole cells, the recruitment of nosmRNA and accumulation of anterior Nos to levels only slightly higher than those generated in the posterior under native regulatory signals(Fig. 6). Osk also accumulates to somewhat lower levels in the adjacent somatic cytoplasm at the anterior, in contrast to the endogenous Osk at the posterior, which is confined to the PGCs(Ephrussi and Lehmann, 1992). We observed significant defects in progression through nuclear cleavages 9-13 near the anterior pole in 20% of osk-bcd embryos(Fig. 6). These defects were due to inappropriate repression of CycB mRNA, as they were rescued if the embryos contained, in addition to the wild-type CycB mRNA, the unrepressible ΔI+II mRNA that lacks a functional NRE (Figs 1, 6). Evidently, in 20% of osk-bcd embryos, the level of Osk in the anterior somatic cytoplasm was sufficient to recruit or stabilize at least one essential co-repressor that acted in conjunction with Nos (and Pum) to repress CycBmRNA.

In summary, Nos is not the sole spatially limiting factor for regulation of CycB, as it is for regulation of hb. Repression of CycB mRNA is confined to the PGCs jointly by a requirement for high levels of Nos and by at least one other (as yet unidentified) factor that is restricted to the pole plasm.

We have shown that Nos and Pum directly regulate maternal CycBmRNA, binding to an NRE in its 3′ UTR. As discussed below, differences in the spacing and arrangement of protein-binding sites within the hband CycB NREs appear to account for the regulation of hb but not CycB by Brat. For regulation of CycB, the main function of Pum is to recruit Nos, a role that can be bypassed by tethering Nos via an exogenous RNA-binding domain. CycB-bound Nos is then likely to act,at least in part, by recruiting a deadenylase complex, interacting with its NOT4 subunit. Regulation of CycB is limited to the PGCs to avoid the deleterious consequences of repression in the presumptive somatic cytoplasm. The requirement for both Nos plus at least one additional germline-restricted factor may be part of a mechanism to ensure that CycB regulation is strictly limited to the PGCs.

The co-crystal structure of human Pum bound to a fragment of the hb NRE shows that a single Pum RBD directly contacts eight nucleotides of the RNA (Wang et al.,2002b). However, Puf proteins bind with essentially wild-type affinity to many mutant sites (Bernstein et al., 2005; Cheong and Hall,2006) (L.K., Y.H., T.L. and R.W., unpublished), suggesting that all eight nucleotides are not rigidly specified. How, then, do Puf proteins recognize specific mRNA targets in vivo?

Part of the answer appears to be that, within functional NREs, more than eight nucleotides are recognized, at least by Drosophila Pum. Mutations that simultaneously disrupt Pum binding in vitro and regulation in vivo are spread over 20 nts of the hb NRE(Wharton et al., 1998) and 18 nts of the CycB NRE (see Fig. S3 in the supplementary material). These extended Pum mutational `footprints' are too large to be accounted for by binding of a single RBD; we suggest that two or more Pum RBDs bind each NRE, an idea supported by the detection of two RNA-protein complexes in gel mobility shift experiments using both the CycB and hb NREs(not shown). This model disagrees with earlier experiments that suggested only a single Pum RBD binds to the hb NRE(Zamore et al., 1999). Further biochemical and structural studies will be required to resolve the issue.

The distribution of Pum- and Nos-binding sites within the CycB and hb NREs is different. In the former, the Nos binding site lies 5′ to the Pum-binding site(s), whereas in the latter, the Nos-binding site is flanked by nucleotides recognized by Pum(Sonoda and Wharton, 1999). We assume that the different arrangement of Nos- and Pumbinding sites is responsible for the assembly of Pum-NRE-Nos complexes with different topographies, such that Brat is recruited to hb but not to CycB. Further definition of each RNP structure will ultimately be required to understand the combinatorial assembly of different repressor complexes on each NRE.

In addition to the NRE, Pum also binds with high affinity to at least two other sites in the CycB 3′ UTR(Fig. 1); however, binding to these sites does not mediate translational repression in the PGCs, perhaps because neither supports recruitment of Nos. These sites may simply bind Pum fortuitously, or they may mediate Nosindependent regulation at other stages of development. Pum has been suggested to destabilize bcd mRNA at the anterior of the embryo in a Nos-independent manner(Gamberi et al., 2002). Another Nos-independent function of Pum is the repression of CycBtranslation throughout the prospective somatic cytoplasm during the early syncitial nuclear cleavages (Tadros et al., 2007; Vardy and Orr-Weaver, 2007). These processes might be mediated by elements in Fragments A and F of the 3′ UTR, which bind Pum but not Nos(Fig. 1, and see Fig. S1 in the supplementary material).

Recent work from the Wickens lab has provided a general framework for understanding how Puf proteins act to control either the translation or stability of target mRNAs (Goldstrohm et al., 2006). The yeast Puf protein MPT5 interacts directly with Pop2, one of the catalytically active subunits of a large deadenylase complex. Subsequent deadenylation could either silence the mRNA or cause its degradation, depending on other signals in the transcript or the composition of the deadenylase complex (or both). The Puf-Pop2 interaction is conserved across species (including Drosophila, Fig. 3), supporting the idea that the mechanism uncovered for MPT5 might generally be applicable to Puf proteins.

In this context, it is surprising that Pum is dispensable if Nos is tethered to CycB via MS2 CP. We suggest that yeast Puf proteins both recognize target mRNAs and recruit the deadenylase, but that in the Drosophila germline these functions are partitioned, with Pum primarily responsible for target mRNA recognition and Nos primarily responsible for effector recruitment (Fig. 7). This model has the attraction of attributing an important role to Nos, which is essential for Puf-mediated regulation in Drosophila,and probably other metazoans as well(Subramaniam and Seydoux,1999; Jaruzelska et al.,2003). What, then, might be the role of the conserved interaction between Pum and Pop2? One possibility is that it acts cooperatively with Nos to recruit the deadenylase; unlike CycB, other mRNA targets (e.g. hb) might require recruitment by both Nos and Pum to ensure efficient deadenylation. Another possibility is that it plays an essential role for mRNAs regulated by Pum but not Nos.

Oscillations in CycB activity underlie normal cell cycle progression. During the early embryonic syncitial nuclear cleavages, degradation in the vicinity of the nuclei is thought to deplete CycB locally(Edgar et al., 1994; Su et al., 1998). Recent work has shown that Pum can inappropriately repress de novo translation of CycB mRNA during the initial nuclear cleavages if not antagonized by the PNG kinase, resulting in mitotic failure(Tadros et al., 2007; Vardy and Orr-Weaver, 2007). This early Pum-dependent repression is thought to be Nos-independent, as it occurs efficiently in the anterior, where Nos activity is undetectable.

Our results (Figs 5, 6) show that if CycBis inappropriately subjected to Pum+Nos-dependent repression via the hb NRE, CycB is locally depleted, resulting in mitotic failure during nuclear division cycles 10-13. As is thought to be the case during the early cycles (1-7), de novo synthesis of CycB apparently is required to counteract the local degradation that probably occurs during M phase of each cycle. The CycB NRE must therefore be precisely tuned to repress translation only in the PGCs and not in the presumptive somatic cytoplasm.

Osk is the limiting factor for assembly of pole plasm in the embryo(Ephrussi and Lehmann, 1992; Smith et al., 1992); our results suggest that it stimulates the accumulation or activity of at least one factor in addition to Nos that is required for repression of CycBin the PGCs. The existence of a co-factor is inferred from the finding that ectopic Nos can repress CycB in the somatic cytoplasm only in the presence of ectopic Osk (Fig. 6). Regulation of CycB may depend on more than one germline-restricted factor to ensure that potentially deleterious repression does not occur in the somatic cytoplasm.

A germline Nos co-factor might act in a variety of ways. It could bind to the CycB NRE adjacent to Pum and Nos, substituting functionally for Brat, which is recruited to the Pum-hb NRE-Nos complex. The 50 nt CycB NRE is inactivated by a truncation at both ends that leaves the Pum- and Nos-binding sites intact, consistent with the idea that another factor binds to the element (see Figs S2 and S3 in the supplementary material). Another possibility is that the co-factor is a germline-specific component of the adenylation/deadenylation machinery, as is the case for the GLD-2 cytoplasmic poly(A)-polymerase in C. elegans(Wang et al., 2002a). Distinguishing among these ideas awaits identification of the cofactor.

We particularly thank T. Edwards and A. Aggarwal for invaluable discussions and initial preparation of the Pum RBD. We thank S. Pyle, C. Gardner, E. Tsalik and L. Fee for technical help, J. Chen for media preparation and S. Curlee for administrative help. We thank the Bloomington Stock Center, the DGRC and the Hybridoma Bank at the University of Iowa for reagents. We also thank V. Bennett, J. Heitman, D. Kiehart, D. Lew and K. Nomie for critical reading of the manuscript. This work was supported by a grant from the NIH to A. Aggarwal (GM62947). R.P.W. is an Investigator of the HHMI.