The STAR/Maxi-KH domain protein GLD-1 mediates a developmental switch in the translational control of C. elegans PAL-1

The regulation of mRNA translation is an essential mechanism of gene control utilized in diverse processes ranging from metabolism to cell growth and differentiation. Translational control offers clear advantages in enucleate and quiescent cells, but it is equally valuable in transcriptionally active cells, where it affords precise spatial and temporal control of protein expression. Some of the best-characterized examples of translational regulation come from studies on Drosophila anterior-posterior axis formation, which is controlled by morphogen gradients established through spatially restricted translational activation and repression. For example, bicoid mRNA is localized to the anterior pole of the oocyte and is translated only after fertilization to yield an anterior-to-posterior protein gradient (Berleth et al., 1988; Driever and Nusslein-Volhard,1988). In contrast, caudal mRNA is uniformly localized,and Caudal protein is present in a reciprocal posterior-to-anterior gradient as a result of translational repression by Bicoid(Mlodzik and Gehring, 1987; Dubnau and Struhl, 1996; Rivera Pomar et al.,1996).

Studies on translational control mechanisms have repeatedly converged on regulatory elements in the 5′ and 3′ untranslated regions (UTRs). Through both genetic and biochemical studies, numerous examples of UTR regulatory elements and their associated trans-acting factors have been discovered (reviewed by Wickens et al.,1996). One of the best-understood examples of 5′UTR-mediated repression is iron response element (IRE)-mediated repression of ferritin synthesis. Under low cellular iron levels, the iron regulatory protein binds the IRE in the ferritin mRNA 5′ UTR and is thought to sterically inhibit recruitment of the small ribosomal subunit(Gray and Hentze, 1994; Muckenthaler et al., 1998). 3′ UTR-mediated control is more prevalent than 5′ UTR-mediated control in developmentally regulated genes(Wickens et al., 1996). The key role of the 3′ UTR is not surprising, given that the 3′ UTR is less evolutionarily constrained than both the coding sequence and the 5′UTR; the coding sequence must code for a functional polypeptide, while the 5′ UTR must have a secondary structure that is amenable to ribosome scanning. Recent data suggest factors bound to 3′ UTRs may control translation in part, by regulating the interaction of eIF4E, the 5′ cap binding protein, with eIF4G, the adaptor protein that binds eIF4E and recruits the small ribosomal subunit via its interaction with eIF3. For example, in Xenopus, many maternal mRNAs are regulated by a 3′ UTR element called the CPE (cytoplasmic polyadenylation element) and the CPE binding protein, CPEB (Mendez and Richter,2001). A ternary complex of CPEB, Maskin and eIF4E is hypothesized to circularize mRNAs and repress translation initiation by masking the eIF4G binding site on eIF4E (Stebbins-Boaz et al., 1999). Similarly, in Drosophila, Bicoid is hypothesized to simultaneously bind eIF4E and the caudal 3′UTR, thereby preventing recruitment of eIF4G to caudal mRNA(Niessing et al., 2002).

Patterning of the C. elegans embryo relies upon the asymmetric distribution of maternal regulatory proteins among early blastomeres. Translational control is implicated in asymmetric expression of maternal PAL-1, the Caudal homolog. Like Caudal, PAL-1 is expressed in the posterior of the embryo where it functions in posterior patterning(Hunter and Kenyon, 1996; Edgar et al., 2001). PAL-1 protein is first detected in posterior blastomeres beginning at the four-cell stage, even though pal-1 mRNA is present in all cells of the early embryo (Hunter and Kenyon,1996). Translational control of PAL-1 expression is suggested by the ability of the pal-1 3′ UTR to restrict the expression of a lacZ reporter RNA to posterior blastomeres; lacZ RNA injected into hermaphrodite gonads is expressed in all blastomeres, whereas lacZ::pal-1 3′ UTR RNA is expressed primarily in posterior blastomeres (Hunter and Kenyon,1996).

Control of maternal PAL-1 expression must begin in the germline to ensure that embryos do not inherit PAL-1 protein. The gonad consists of two U-shaped arms in which presumptive oocyte nuclei progressively mature while moving from the distal to the proximal region of each arm(Fig. 1A). In the distal region, syncytial nuclei progress from mitosis to meiotic pachytene, while in the proximal region, nuclei complete meiosis I and become cellularized as they mature into oocytes. pal-1 mRNA is present throughout the germline(Nematode Expression Pattern DataBase (NEXTDB), http://nematode.lab.nig.ac.jp),yet PAL-1 protein is observed only transiently; low levels are observed in immature oocytes found in the bend of the gonad arm(Fig. 1A) (D.M. and C.P.H.,unpublished). mex-3 is hypothesized to repress pal-1translation in mature oocytes and early embryos. PAL-1 is present at high levels in oocytes and all cells of early embryos that lack maternal mex-3, and expression of lacZ::pal-1 3′ UTR RNA is not restricted to posterior blastomeres in mex-3 mutant embryos(Hunter and Kenyon, 1996). Interestingly, mex-3 encodes a putative RNA binding protein,containing two KH (K homology) domains, that is expressed cytoplasmically in a pattern complementary to that of PAL-1(Draper et al., 1996). Hence,MEX-3 may directly mediate the translational repression of pal-1 in oocytes and early embryos.

It is unclear how PAL-1 expression is repressed in the hermaphrodite gonad prior to the appearance of MEX-3. MEX-3 is detected only in the proximal gonad arm (Draper et al., 1996), and no ectopic PAL-1 is detected in the distal gonad arm of mex-3 mutant hermaphrodites (D.M. and C.P.H., unpublished). We have investigated PAL-1 repression in the distal gonad arm and here we provide evidence that a member of the conserved STAR/Maxi-KH domain family, GLD-1, binds a minimal element in the pal-1 3′ UTR and represses pal-1 translation. We also show that GLD-1 represses the distal germline expression of MEX-3, thus allowing MEX-3 to accumulate and take over the task of PAL-1 repression in the proximal gonad arm where GLD-1 is absent. Finally, we provide data suggesting that GLD-1 may repress translation after initiation.

Standard culture techniques were used(Brenner, 1974). The following strains were used: Bristol N2 (wild-type), JK1466 [gld-1(q485)/dpy-5(e61)unc-13(e51)], BS1077 [ozEx54 [GLD-1::GFP-FLAG, unc-119(+)]; unc-13(e51) gld-1(q485)], AZ212 [unc-119(ed3)ruIs32[unc-119(+) pie-1::GFP::H2B] and HC93 [unc-119 (ed3)qtIS5[unc-119(+) pie-1::GFP::H2B::2X GRE]. Worms for sedimentation and RNase protection experiments were synchronized as L1 larvae by allowing embryos to hatch in the absence of food. Worms of the appropriate developmental stages (verified at 250× magnification) were cleaned by sucrose flotation, washed 3× with M9 Buffer and 1× with water, and flash-frozen.

All pal-1 3′ UTR sequence was cloned into pJK370 which bears an NLS::lacZ coding sequence and a poly(A) tract of 30 residues. Residue 1 in Fig. 2 corresponds to the third residue following the pal-1 stop codon. Capped RNAs were synthesized essentially as described previously(Evans et al., 1994) and poly(A)+ purified using Oligotex kits (Qiagen) according to manufacturer's instructions, with the exception that RNA was not denatured before incubation with Oligotex resin. RNA concentration was determined spectrophotometrically,and RNA integrity was verified by electrophoresis. Two or more preparations of RNA were injected for each construct. RNAs (50 nM) were injected into the distal gonad arm, and worms were stained with X-gal(Fire et al., 1990) after a 6-hour recovery at 25°C. In situ hybridizations were performed 75-90 minutes following injections using a lacZ antisense probe as described previously (Seydoux and Fire,1995). Strong signal in the in situ hybridization experiments was defined as being visible through a Nomarski filter at 100×.

The GH::2X GRE construct was created by inserting two copies of the GRE and the pal-1 cleavage and polyadenylation signal after the stop codon of pKS114 (kindly provided by K. Subramanian and G. Seydoux), which is derived from pJH4.52 (G. Seydoux, personal communication), yet lacks the pie-1 3′ UTR. The extra-chromosomal array line was generated using the complex array method (Kelly et al., 1997) and the integrated line used in RNase protection assays was created using microparticle bombardment(Praitis et al., 2001). RNAi was performed using bacteria expressing RNA hairpins(Winston et al., 2002). Worms were placed on RNAi bacteria as L4 larvae and the F₁ adults exhibited highly penetrant gld(–) and mex-3(–)phenotypes. For the double RNAi, bacteria were mixed together in equal volumes.

Sucrose gradient buffer contained 50 mM Tris-HCl (pH 8.5), 25 mM KCl, 10 mM MgCl₂, and 1% protease inhibitor cocktail (Sigma, P8215) (EDTA was added to 0.1 M to some gradients). Linear gradients were made by overlaying 5.5 mls 0% sucrose gradient buffer on 5.5 mls 56% sucrose gradient buffer in a 14×89 mm tube. Tubes were first placed horizontally for 2 hours at room temperature, and then placed vertically for 30-45 minutes at 4°C (the linearity of the gradients made by this method was confirmed by measuring the refractive index of multiple fractions). 0.5 ml was removed from the gradient tops before sample loading. Polysome extracts were prepared by grinding frozen worms to a fine powder using a mortar and pestle on dry ice. The resulting worm powder was added to 3-4 volumes of ice-cold buffer consisting of 0.2 M Tris-HCl (pH 8.5), 50 mM KCl, 25 mM MgCl₂, 1 mM DTT, 2 mM EGTA, 0.2 mg/ml heparin, 500 U/ml RNasin, 2% PTE, and 1% protease inhibitor cocktail(EDTA was added to 0.1 M in some samples). Samples were centrifuged for 10 minutes at 16,000 g and supernatant was layered on a sucrose gradient. Gradients were centrifuged at 4°C for 1.5 hours at 40,000 rpm in a SW-41 rotor. 0.2 or 0.5 ml fractions were collected manually from the top of the gradients and RNA was extracted using Trizol (Gibco). Absorbance (260 nm)measurements were made on an aliquot of each purified RNA fraction. Pooled fractions were analyzed for pal-1 or mex-3 mRNA using Ambion's RPAIII kit. Metrizamide gradient analysis was performed as described by Olsen and Ambros (Olsen and Ambros,1999). Ribosomal fractions were identified by the index of refraction as described by Olsen and Ambros(Olsen and Ambros, 1999). Only the top 90% of the gradient was analyzed because of the high metrizamide viscosity at the bottom. Fractions analyzed by western blot were concentrated by TCA precipitation.

Antibody staining was performed as described in Huang et al.(Huang et al., 2002). The biotin pull-down assays were performed essentially as described by Lee and Schedl (Lee and Schedl, 2001),except that 4 picomoles of each RNA were used and worms were fragmented using a mortar and pestle as described above.

To determine which regions of pal-1 mRNA are required for germline repression, we used a reporter RNA assay in which lacZ RNAs containing pal-1 sequence were injected into the distal gonad arm. The pal-1 3′ UTR was sufficient to repress distal germline expression, since lacZ::3′ UTR RNA was never expressed in the distal germline (n=62), while control lacZ RNA was always expressed throughout the distal germline (n=80)(Fig. 1B,C,E). However, unlike endogenous PAL-1, lacZ::3′ UTR RNA was expressed throughout the proximal germline, perhaps reflecting perdurance of β-galactosidase transiently translated in immature oocytes and/or the need for additional pal-1 sequence to promote subsequent repression.

Several lines of evidence suggest that the absence of distal germline expression observed following injection of lacZ::3′ UTR RNA is due to translational repression, rather than RNA degradation or transport into the proximal germline. First, similarly strong β-galactosidase signal was detected in oocytes and embryos following injection of lacZ::3′UTR RNA and lacZ RNA, suggesting that the lacZ::3′ UTR RNA was not subject to significant degradation. Second, β-galactosidase was detected in oocytes as late as 8 hours post-injection of lacZ::3′ UTR RNA (signal in oocytes is first detected 70 minutes post-injection). Since germline nuclei are continuously passing from the distal to the proximal gonad arm, this result suggests that the RNA translated in the proximal germline at the end of the 8-hour period was present in the distal germline throughout much of the incubation period. Finally, lacZ and lacZ::3′ UTR RNA were detected in the distal germline by in situ hybridization in a comparable percentage of gonad arms (90%, n=31 and 79%, n=63, respectively)(Fig. 1B,C, insets).

It is possible that the detected lacZ::3′ UTR RNA was degraded and not competent for translation, therefore we determined the efficiency with which unstable RNA can be detected. Since uncapped RNAs are translated inefficiently and degraded rapidly in multiple systems (reviewed by Mitchell and Tollervy, 2000),we injected uncapped lacZ RNA and found that it was not expressed in the germline (n=16) and was inefficiently detected by in situ hybridization. Only 33% (n=27) of injected gonad arms exhibited signal, and of these, only one (4% of total) exhibited strong signal. This is in contrast to the high percentage of gonad arms that exhibited a strong in situ signal following injection of lacZ or lacZ::3′UTR RNA (89% and 62%, respectively). These data indicate that degraded RNA yields a minimal in situ signal, and suggest that the lacZ::3′UTR RNA is intact, but translationally repressed.

To identify regions of the pal-1 3′ UTR responsible for this translational repression activity, we performed deletion analysis of the lacZ::3′ UTR RNA. For ease of discussion, the first, middle,and last thirds of the 3′ UTR have been designated A, B and C,respectively. These thirds average 175 nucleotides in length and they overlap by approximately 18 nucleotides (Fig. 1E). Region C has minimal germline repression activity, as full germline repression is maintained when region C is deleted (construct AB, Fig. 1E), and lacZ RNA bearing region C alone is expressed in the distal germline of 92%(n=53) of injected gonad arms.

In contrast, both region A and region B are necessary for complete distal germline repression. lacZ::3′ UTR RNA is never expressed in the distal germline, whereas lacZ::3′ UTR RNAs lacking region A or B (constructs BC and AC) are expressed in the distal germline in 31% and 60%of gonad arms, respectively. Importantly, like the lacZ::3′ UTR RNA, these two constructs (and all other deletion constructs) were expressed in the proximal germline, thus showing that functional reporter RNAs were injected. The germline repression activities of region A and B are further supported by the observation that constructs that contain either region A or B in isolation were sufficient to repress distal germline expression of lacZ reporter RNA in approximately two-thirds of gonad arms.

Deletion of region B from the lacZ::3′ UTR construct impaired repression more significantly than deletion of region A, therefore we further mapped repression activity within region B. We found that the second half of B (B2) lacks repression activity, while the first half (B1) robustly inhibits germline expression; lacZ::B1 RNA was expressed in the distal germline in only 2% (n=63) of gonad arms(Fig. 1D,E). Hence, B1, a 107-nucleotide element (Fig. 1F), is both necessary and sufficient for robust germline repression and hereafter B1 will be referred to as the germline repression element, or GRE. Importantly, lacZ RNA bearing two copies of the GRE is efficiently detected in the distal germline by in situ hybridization,suggesting that the GRE represses translation, as opposed to regulating RNA stability or localization (Fig. 1D, inset).

To understand how the GRE promotes translational repression, we sought to identify trans-acting factors that mediate its activity. A clear candidate is GLD-1, a cytoplasmic RNA binding protein (STAR/Maxi-KH domain protein) that represses the expression of at least four genes(Jones and Schedl, 1995; Jones et al., 1996; Jan et al., 1999; Lee and Schedl, 2001; Xu et al., 2001; Marin and Evans, 2003). GLD-1 is expressed in the distal, but not the proximal germline, and thus is expressed at the right time and place to be a pal-1 repressor(Jones et al., 1996). To assess the genetic interaction between GLD-1 and the GRE, we constructed a transgenic line in which expression of a GFP::histone H2B fusion protein is driven by a germline promoter (pie-1) and under the regulation of two tandem copies of the GRE (GH::2X GRE). The expression pattern of this reporter recapitulated that of the lacZ::GRE reporter, as nuclear GFP was observed throughout the proximal germline, but not the distal germline(Fig. 2B). Importantly, a similar transgene that differs only by the presence of the pie-13′ UTR instead of the pal-1 GREs is expressed in nuclei throughout both the distal and proximal germline(Fig. 2A), indicating that the lack of distal germline expression of the GFP::2X GRE reporter is due to the repression activity of the GRE. In addition, upon long exposures (several seconds) of dissected GH::2X GRE distal gonad arms using a CCD camera, we noted weak cytoplasmic GFP expression (Fig. 2D), which is discussed below.

Following bacterially mediated RNA interference (RNAi) of gld-1 in GH::2X GRE worms, GFP was detected in virtually all nuclei of the distal germline (Fig. 2C), indicating that GLD-1 acts through the GRE to inhibit distal germline expression. Because gld-1 null adults have a proximal germline tumor which is formed by nuclei that exit meiotic pachytene and proliferate mitotically(Francis et al., 1995a), we considered the possibility that the ectopic GH::2X GRE expression in the distal germline was a secondary consequence of the aberrant proximal germline development. However, we eliminated this explanation because we observed ectopic reporter expression at the extreme distal ends of the adult gonad,where germline development appears normal, and because we observed ectopic expression in gld-1 L4 larvae, which do not yet have a tumorous germline (Francis et al.,1995a) (data not shown). To determine whether gld-1controls the stability or the translation of GH::2X GRE RNA, we quantified the amount of GH::2X GRE in wild type and gld-1 RNAi worms. Duplicate RNase protection assays indicated that the amount of reporter RNA and endogenous RNA was not significantly changed following gld-1 RNAi,suggesting that GLD-1 represses translation as opposed to destabilizing the RNA (Fig. 2E).

GLD-1 also represses the expression of endogenous PAL-1 in the distal germline, as ectopic PAL-1 immunofluorescence was detected in the distal germline of adult gld-1 null mutants (q485) and RNAi worms(Fig. 3B). This ectopic PAL-1 signal was restricted to small clusters of nuclei instead of being distributed throughout the distal germline, like the ectopic GFP expression of GH::2X GRE reporter following gld-1 RNAi. We hypothesized that derepression of endogenous pal-1 translation was incomplete in gld-1 mutants because of additional repression elements, and asked whether mex-3,which functions to repress PAL-1 expression in the proximal germline, is ectopically active in the distal germline of gld-1 mutants. Indeed,we observed ectopic MEX-3 throughout the distal germline of gld-1null mutants (q485) and RNAi worms (n=59), and we observed a clear increase in ectopic PAL-1 in the distal germline of mex-3(RNAi);gld-1(RNAi) worms (Fig. 3C,E). In 100% of gld-1 (q485) and RNAi worms, between 0-50 distal germline nuclei express PAL-1 (n=63), whereas in 71% of gld-1(RNAi); mex-3(RNAi) worms, between 50-200+distal germline nuclei express PAL-1 (n=41). Hence, in wild-type worms, gld-1 inhibits both PAL-1 and MEX-3 expression in the distal germline, while MEX-3 inhibits PAL-1 expression in the proximal germline. Interestingly, a leaky switch from GLD-1 to MEX-3-mediated repression may account for the transient PAL-1 expression in the bend of the gonad arm(Fig. 1A), which is precisely where GLD-1 levels decline and MEX-3 levels rise(Draper et al., 1996; Jones et al., 1996).

GLD-1 could repress pal-1 translation directly by binding the GRE or indirectly by controlling the expression of a pal-1 regulator that acts through the GRE. To distinguish between these possibilities, we tested the ability of biotinylated GRE RNA to capture a rescuing GLD-1::GFP::FLAG fusion protein (GGF) from adult extracts. Biotinylated RNA was isolated using streptavidin magnetic beads, and bound proteins were subjected to SDS-PAGE and western analysis with anti-FLAG antibody. As shown in Fig. 4 (lanes 7 and 8) and by Lee and Schedl (Lee and Schedl,2001), this assay shows the expected specificity as GGF is pulled-down by a tra-2 3′ UTR RNA that bears GLD-1 binding sites, but not by antisense tra-2 3′ UTR RNA(Jan et al., 1999). We found that sense, but not antisense 2× GRE RNA captured GGF(Fig. 4, lanes 2 and 3). Taken with the genetic interactions between GLD-1 and the GRE, this result strongly suggests that GLD-1 binds the GRE in vivo, either by physically contacting the GRE, or by contacting protein(s) that physically contact the GRE. We next tested the GLD-1 binding activity of other 3′ UTR elements and found that binding activity in extracts correlates with in vivo repression activity;regions A and B both efficiently captured GGF, while region C did not(Fig. 4, lanes 4, 5 and 6),suggesting that GLD-1 is a major regulator of pal-1 translation.

How does GLD-1 repress pal-1 translation? GLD-1 may interfere with translation initiation or it may interact with the translation machinery after initiation to impair elongation or termination. The weak cytoplasmic expression of the GH::2X GRE reporter in the distal germline(Fig. 2D) is consistent with GLD-1 repressing translation after initiation. We hypothesize that this cytoplasmic GFP signal may represent mature GFP tethered to ribosomes by the incompletely translated histone H2B (C-terminal domain of the fusion protein),as there is precedence for tethered proteins exhibiting activity when their C-termini are extended to allow protrusion from the ribosome exit tunnel(reviewed by Kramer et al.,2001). To test the idea that PAL-1 expression may be inhibited after translational initiation, we asked whether pal-1 mRNA co-fractionates with polysomes using the following two independent methods of sedimentation: (1) sucrose density sedimentation, which separates cellular components based on size, weight and shape (sedimentation coefficient), and therefore resolves free mRNA, ribosomal subunits, and monosomes, disomes,etc.; and (2) metrizamide equilibrium sedimentation, which separates complexes based on buoyant density and therefore causes monosomes and polysomes of all sizes to co-fractionate. The distribution of ribosomes (monosomes and polysomes) was determined by measuring the amount of total RNA (primarily ribosomal RNA) in fractions taken manually from the gradient tops. The fractions were then pooled and RNase protection assays were performed to determine the distribution of pal-1 mRNA relative to the ribosomes.

For these experiments, we used worm samples that contained abundant germline pal-1 mRNA, but no PAL-1 protein – hermaphrodites of the last larval stage (L4), which synthesize maternal mRNAs but do not yet produce oocytes, and young adult hermaphrodites, which have just completed the L4 molt and do not yet produce oocytes. Germline enrichment of pal-1mRNA in L4s and adults is supported by RNase protection experiments indicating that very little pal-1 mRNA is present in glp-4 (bn2ts) L4s and adults, which have a greatly reduced germline(Beanan and Strome, 1992)(Fig. 5A). Duplicate comparisons of wild-type and glp-4 worms indicate that approximately 75% of pal-1 mRNA from wild-type L4s and 95% of pal-1 mRNA from adults is associated with the germline.

We found that virtually all pal-1 mRNA from L4 larvae and young adults co-fractionated with ribosomes in sucrose gradients. In two independent L4 gradients, 91% and 97% of pal-1 mRNA was recovered in the ribosome fractions, while in the young adult gradient, 93% of pal-1 mRNA was recovered in the ribosome fractions (Fig. 5B and data not shown). Because the absorbance of only 25 gradient fractions was measured, our analysis did not resolve the ribosomal subunits or resolve monosomes from disomes, etc., so we refer to monosomes and polysomes collectively as ribosomes. When we manually took 56 fractions, a more typical pattern of RNA absorbance was observed(Fig. 5B, inset). Importantly, pal-1 mRNA was not restricted to the first ribosome fraction, but was instead distributed throughout the ribosome fractions. To test whether the co-sedimentation of pal-1 mRNA with ribosomes is consistent with pal-1 mRNA physically associating with ribosomes, we fractionated extracts in the presence of EDTA, which causes ribosomal subunits to dissociate and release mRNAs. Virtually all pal-1 mRNA (over 90%) was shifted to slower-sedimenting fractions after EDTA treatment (data not shown),suggesting that repressed pal-1 mRNA may be physically associated with ribosomes in L4s and young adults. However, it is also possible that pal-1 mRNA is associated with other heterogeneous, fast-sedimenting mRNP complexes that are EDTA sensitive.

Similar results were obtained from L4 larval extracts fractionated on a metrizamide gradient (Fig. 5C). pal-1 mRNA was found in the same fraction as monosomes and polysomes of all sizes (fraction four), whose position is inferred both from the total RNA peak (ribosomal RNAs) and from the index of refraction of C. elegans ribosomes under these conditions(Olsen and Ambros, 1999). In addition, to determine whether GLD-1 may mediate post-initiation repression,we assayed metrizamide gradient fractions by immunoblot and found that GLD-1 protein from adults was enriched in the ribosome containing fraction(Fig. 5D).

Because gld-1 inhibits the distal germline expression of the pal-1 regulator MEX-3, we asked whether gld-1 also inhibits the distal germline expression of proteins that repress PAL-1 expression in the early embryo: SPN-4, MEX-5 and MEX-6(Huang et al., 2002). SPN-4 is an RRM protein (Gomes et al.,2001), while MEX-5 and MEX-6 are 70% identical CCCH zinc finger proteins with overlapping functions(Schubert et al., 2000). Because the individual functions of MEX-5 and MEX-6 have not been defined with respect to PAL-1 regulation, we refer to these proteins collectively as MEX-5/6. SPN-4 and MEX-5/6 are expressed in the proximal, but not distal,gonad arms of wild-type adults (N. N. Huang and C.P.H., unpublished)(Schubert et al., 2000)(Fig. 6A,C). In gld-1null mutants (q485), we observed that 80% (n=25) and 100%(n=28) of gonad arms exhibited ectopic SPN-4 and MEX5/6 expression,respectively (Fig. 6B,D). Hence, gld-1 directly or indirectly represses the distal germline expression of all three proteins known to repress PAL-1 expression at later developmental stages.

To begin to understand how gld-1 represses the expression of proteins other than PAL-1, we asked whether repressed mex-3 mRNA,like pal-1 mRNA, co-purifies with ribosomes. Indeed, when extracts made from young adults (lacking oocytes) were fractionated via sucrose density gradient sedimentation, virtually all mex-3 mRNA co-fractionated with ribosomes (Fig. 5B). This sedimentation profile of mex-3 mRNA was abolished when ribosomes were disrupted by EDTA treatment (data not shown), consistent with the idea that like pal-1, the translation of mex-3 may be inhibited after translation initiation.

Our data suggest a developmental switch in which pal-1 translation is blocked in the distal germline by a broad spectrum translation repressor,the STAR/Maxi-KH domain protein GLD-1, and subsequently blocked in the proximal germline by a more specific repressor, the KH domain protein MEX-3(Fig. 7). Our classification of GLD-1 as a broad spectrum repressor derives from our current finding that GLD-1 repress the distal germline expression of four proteins – PAL-1,MEX-3, SPN-4 and MEX-5/6 – combined with previous work showing: (1) that GLD-1 represses the expression of four other proteins in the distal germline(Jan et al., 1999; Lee and Schedl, 2001; Xu et al., 2001; Marin and Evans, 2003) and (2)that 13 additional mRNAs are enriched following GLD-1 immunoprecipitation(Lee and Schedl, 2001). Hence,GLD-1 appears to function as a general repressor in the distal germline,providing a common mechanism for repressing the expression of maternal mRNAs whose products are not needed until oocyte maturation or embryogenesis. Such a broad spectrum repressor conserves limited maternal resources and may have evolutionary advantages. Interestingly, GLD-1 represses the expression of MEX-3, thus enabling a switch to MEX-3-mediated repression in the proximal gonad arm, where GLD-1 protein is absent. To our knowledge, this is the first description of a `hand-off' in which a translational regulator represses a secondary repressor of one of its targets. The switch from GLD-1 to MEX-3-mediated repression appears to be leaky, as PAL-1 expression is often transiently observed in the gonad precisely where GLD-1 levels decline and MEX-3 levels rise. Unlike GLD-1, which has multiple mRNA targets and multiple functions during development, MEX-3′s primary role appears to be PAL-1 repression, since the mex-3(–) phenotype can largely be suppressed by removing pal-1 activity(Hunter and Kenyon, 1996).

lacZ reporter RNA experiments suggest that the pal-13′ UTR is sufficient to confer translational repression in the distal germline. Specifically, lacZ::3′ UTR RNA is detected in the distal germline when no β-galactosidase protein is detected, and both the intensity and the time-course of β-galactosidase expression in the proximal germline suggest translational repression. Through deletion analysis of the lacZ::3′ UTR RNA, we identified two regions, A and B,that possess germline repression activity. Within region B, we identified the GRE, a minimal element of 107 nucleotides that robustly inhibits distal germline expression. Similar to the lacZ::3′ UTR RNA, lacZ RNA under the regulation of two copies of the GRE is stable in the distal germline and is expressed strongly in the proximal germline,suggesting that the GRE is a translational repression element.

The STAR/Maxi-KH domain protein GLD-1 mediates the translational repression activity of the GRE. GLD-1 is specifically precipitated from worm extracts by GRE RNA, and GLD-1 represses the distal germline expression of both PAL-1 and a GH::2X GRE reporter, without destabilizing the respective mRNAs. GLD-1 could be the only regulator of the GRE, since the GH::2X GRE reporter is ectopically expressed in all distal germline nuclei following gld-1 RNAi, and removal of mex-3 activity in addition to gld-1 activity does not noticeably increase the level of ectopic expression (data not shown). In contrast, repression of full-length pal-1 mRNA may require regulators in addition to GLD-1, as ectopic PAL-1 is detected only in a subset of distal germline nuclei in gld-1 mutants. Moreover, even following reduction of ectopic mex-3 activity through gld-1; mex-3 double RNAi,PAL-1 is still not detected in all nuclei. This apparently incomplete derepression of PAL-1 expression could be due to the failure to eliminate all GLD-1 and MEX-3 protein via double RNAi, or due to the up-regulation of another PAL-1 repressor in the germline of gld-1 (RNAi) worms. Indeed, we found that SPN-4 and MEX-5/6, which repress PAL-1 expression in embryos, are ectopically expressed in the distal germline following gld-1 RNAi, and they may contribute to PAL-1 repression. Alternatively, there may be additional protein(s) that normally contributes to PAL-1 repression in the distal gonad arm.

GLD-1 is homologous to a sub-family of KH domain proteins known as the GSG or STAR domain family (Jones and Schedl,1995), whose members include the evolutionarily conserved Quaking protein, mammalian Sam68 and SF1 and Drosophila How(Vernet and Artzt, 1997). The∼200 amino acid STAR domain consists of an enlarged KH RNA-binding domain(maxi-KH domain) flanked by conserved residues on both sides(Vernet and Artzt, 1997). While the functions of these family members are not well understood, they have been implicated in various aspects of RNA metabolism, including mRNA splicing,nuclear export and translation (Arning et al., 1996; Larocque et al.,2002; Saccomanno et al.,1999). Other than GLD-1, only one family member, the mouse Quaking I isoform 6, has thus far been implicated as a translational regulator, and this is based on its ability to repress tra-2 expression when expressed in C. elegans(Saccomanno et al., 1999).

GLD-1 was previously shown to repress the expression of tra-2, rme-2,mes-3 and glp-1 mRNAs (Jan et al., 1999; Lee and Schedl,2001; Xu et al.,2001; Marin and Evans,2003). GLD-1 can directly bind the 3′ UTRs of all these mRNAs, yet minimal binding sites with translational repression activity in vivo have been described only for tra-2 and glp-1(Goodwin et al., 1993; Marin and Evans, 2003). The GRE is the third element that mediates gld-1 repression in vivo, yet it does not share obvious primary or secondary structures with the previously identified elements, and a comparison of all known and putative binding sites reveals only limited homology between two targets, raising the question of how GLD-1 regulates many disparate mRNAs. GLD-1 may specifically bind mRNAs in vivo in cooperation with co-factor(s). For example, the novel F-box-containing protein, FOG-2, is implicated as a co-factor specifically in the regulation of tra-2 translation. FOG-2 forms a ternary complex with GLD-1 and the tra-2 3′ UTR (Clifford et al., 2000). fog-2 is not required for pal-1repression, as no ectopic PAL-1 was observed in the germline of fog-2mutants (data not shown). Alternatively, GLD-1 may be guided to its targets by small non-coding RNAs or microRNAs (miRNAs). The C. elegans miRNA lin-4 is hypothesized to repress lin-14 and lin-28expression after translation initiation by binding imperfect complementary elements in 3′ UTRs of these mRNAs(Olsen and Ambros, 1999; Seggerson et al., 2002). In addition, the Drosophila and mouse homologs of the Fragile X translational repressor associate with the miRNA mir2b and the non-coding RNA BC1, respectively(Caudy et al., 2002; Ishizuka et al., 2002; Zalfa et al., 2003). In support of the hypothesis that miRNAs may participate in GLD-1-mediated repression, we have observed ectopic PAL-1 expression in the distal germline of worms with impaired miRNA processing (data not shown).

There are four explanations for our observation that translationally repressed pal-1 mRNA co-fractionates with ribosomes: (1) pal-1 mRNA could be a component of large heterogeneous ribonucleoprotein particles (mRNPs) that have the density of ribosomes as well as the sedimentation coefficients of polysomes of various sizes; (2) pal-1 mRNA could be in a mRNP associated with ribosomes actively translating other messages; (3) PAL-1 protein could be degraded co-translationally in a manner dependent on the pal-1 3′ UTR but independent of PAL-1 protein sequence (since lacZ and gfp reporters are accurately regulated); or (4) pal-1 mRNA may be associated with ribosomes that have slowed or stalled while translating it. The post-initiation repression mechanism is the simplest explanation, and it is the only mechanism that also explains the cytoplasmic GFP signal we observed in the distal germline of animals expressing the GH::2X GRE transgene(Fig. 2D). Reversing the order of the two protein coding regions in this transgene never resulted in GFP signal in the distal germline, but these constructs were poorly expressed in the proximal germline (data not shown). Like pal-1 mRNA, GLD-1 also co-fractionates with ribosomes, and our working hypothesis is that GLD-1 is a component of a post-initiation repression complex in the distal germline. This derives from the fact that GLD-1 directly or indirectly represses the distal germline expression of at least eight proteins (four described in this paper)and that for at least two of these, PAL-1 and MEX-3, post-initiation repression is implicated.

Although polysome analysis has been performed on one other GLD-1 target, tra-2 (Goodwin et al.,1993), it is unclear as to whether GLD-1 represses TRA-2 expression before or after translation initiation. For example, the polysome analysis was performed on adults, yet GLD-1 represses tra-2translation in larvae (Jan et al.,1999). Also, the interpretation of the data is hindered by the fact that tra-2 activity is regulated both in the soma and germline of males and hermaphrodites (reviewed by Kuwabara and Perry, 2001). All these populations of mRNA were analyzed together in the polysome analysis, but GLD-1 is required only for the repression of tra-2 mRNA in the hermaphrodite germline (Francis et al.,1995a; Francis et al.,1995b; Jan et al.,1999). The trans-acting factor(s) required for repression in the male soma and germline are unknown and the mechanism of repression may be different.

There is a small but growing list of mRNAs subject to post-initiation repression, yet GLD-1 is the one of the first trans-acting proteins to be implicated in this process. For example, the mRNAs encoding C. elegans LIN-14 and LIN-28 co-fractionate with polysomes when the proteins are not detectable, but no trans-acting proteins involved in repression have been identified (Olsen and Ambros,1999; Seggerson et al.,2002). Similarly, roughly half of Drososphila nanos mRNA is found associated with polysomes in embryos with no detectable protein, and although the RNA-binding protein Smaug can repress Nanos expression, it is not known whether it represses the polysomal or sub-polysomal population of nanos mRNA (Simbert et al.,1996; Dahanukar et al.,1999; Clark et al.,2000). Despite the growing number of examples, the mechanisms of post-initiation repression remain unknown. The identification of GLD-1 as a putative regulator of post-initiation repression may provide an inroad for the molecular dissection of this process.

We thank T. Evans and J. Kimble for pJK370, T. Schedl for the BS1077 strain, N. Huang for SPN-4 and MEX5/6 antibodies, K. Subramanian and G. Seydoux for pKS114, B. Goodwin for the tra-2 3′ UTR plasmid and anti-GLD-1 antibody, and the Caenorhabditis Genetics Center for strains. We also thank V. Ambros for helpful discussions and K. Carniol and C. Molodowitch for critical comments on the manuscript. This work was supported by an NIH grant to C.P.H. (GM616677).