An isoform of eIF4E is a component of germ granules and is required for spermatogenesis in C. elegans

One of the levels at which gene expression is controlled is translation of mRNA. Translational control is crucial for normal embryonic development, proper regulation of the cell cycle, tissue induction and differentiation, and germline development (Morris, 1995; Pain, 1996; Hake and Richter, 1997; Clemens and Bommer, 1999; Saffman and Lasko, 1999). Translational regulation can be achieved via cis-acting elements in the 5′ and 3′ untranslated regions (UTRs) of mRNAs, trans-acting factors that interact with these UTRs, and the binding of core translational components to mRNAs (Jackson and Wickens, 1997).

Recruitment of mRNAs to ribosomes to initiate translation is mediated by initiation factors of the eIF4 group and the poly(A)-binding protein (PABP). The eIF4 group includes eIF4A, an RNA helicase; eIF4B, an RNA-binding protein that stimulates eIF4A; eIF4E, a cap-binding protein; and eIF4G, the central organizing protein that colocalizes eIF4E, eIF4A, eIF3, PABP, the eIF4E kinase Mnk and RNA in the 48S initiation complex.

Intracellular levels of eIF4E strongly affect the rate of translation, specifically of mRNAs that are more strongly dependent on the cap (Altmann et al., 1989; De Benedetti et al., 1991), and thus eIF4E is an attractive target for regulation of translation. In fact, the level, availability, and activity of eIF4E are regulated by several processes. First, the level of eIF4E is transcriptionally regulated; e.g. transcription of the eIF4E gene is increased in fibroblasts in response to growth factor treatment (Rosenwald et al., 1993; Jones et al., 1996). Second, the availability of eIF4E is regulated by its association with eIF4E-binding proteins (4E-BPs), which, when bound to eIF4E, prevent its binding to eIF4G, thereby inhibiting translation initiation (Sonenberg, 1996). Third, the activity of eIF4E is regulated by phosphorylation. Phosphorylation of eIF4E by Mnk in response to extracellular stimuli, such as hormones, growth factors and mitogens, is generally correlated with an increase in translation rate (Rhoads, 1993; Gingras et al., 1999).

The importance of regulating the level of eIF4E is underscored by the findings that moderate overexpression of eIF4E can cause deregulated cell proliferation and malignant transformation (Lazaris-Karatzas et al., 1990; De Benedetti and Rhoads, 1990; Lazaris-Karatazas and Sonenberg, 1992). Conversely, depletion of eIF4E by antisense RNA slows growth rate (De Benedetti et al., 1991) and partially reverses oncogenic transformation in cancer cells (Rinker-Schaeffer et al., 1993). Furthermore, a direct correlation between the amount of eIF4E and malignant transformation has been reported in many cell lines and tumors (De Benedetti and Harris, 1999). The recurrence of head and neck carcinomas after surgery is strongly correlated with eIF4E levels in the tumor margins (Nathan et al., 1997). Normal development also depends on proper levels or availability of eIF4E. In immature oocytes of Xenopus, the availability of eIF4E is regulated by its association with the CPEB-Maskin complex (Stebbins-Boaz et al., 1999). Maskin contains an eIF4E-binding motif similar to that found in eIF4G and 4E-BPs. The binding of eIF4E by the CPEB-Maskin complex precludes its interaction with eIF4G. During progesterone-induced oocyte maturation the Maskin-eIF4E interaction is weakened, allowing eIF4E and PABP to form a complex with eIF4G; complex formation is required for the recruitment of polyadenylated maternal mRNAs to ribosomes (Stebbins-Boaz et al., 1999; Keiper and Rhoads, 1999).

In Caenorhabditis elegans, there are five isoforms of eIF4E, called IFE-1, IFE-2, IFE-3, IFE-4 and IFE-5 (Jankowska-Anyszka et al., 1998; Keiper et al., 2000). Based on their primary sequences, cap-binding specificity and requirement for viability, IFE proteins have been grouped into three classes (Keiper et al., 2000). Class A contains IFE-3, which is most similar to human eIF4E-1 and is essential for viability. Class B includes IFE-1, IFE-2 and IFE-5. Depletion of any individual class B member by RNA-mediated interference (RNAi) did not affect viability. However, blocking the expression of all three proteins caused 99% embryonic lethality, suggesting that they function redundantly in embryonic development. Class C contains IFE-4, which is the most divergent isoform of eIF4E in C. elegans and is completely dispensable. Class A and C members bind m⁷GTP-containing caps exclusively, while class B members bind both m⁷GTP- and m₃^2,2,7GTP-containing caps.

A potential role for IFE-1 in germline development was brought to our attention by its interaction with the germ-granule component, PGL-1, in a yeast two-hybrid screen. Germ granules (also called P granules in C. elegans and polar granules in Drosophila) are non-membrane-bound organelles that contain RNAs and proteins and are candidate ‘determinants’ of the germline. Several proteins have been found to be associated with P granules either constitutively or transiently. These include PGL-1 (Kawasaki et al., 1998), GLH-1, GLH-2 (Gruidl et al., 1996), GLH-3, GLH-4 (Kuznicki et al., 2000), GLD-1 (Jones and Schedl, 1995; Jan et al., 1999), PIE-1 (Mello et al., 1996; Tenenhaus et al., 1998), MEX-1 (Guedes and Priess, 1997), MEX-3 (Draper et al., 1996) and POS-1 (Tabara et al., 1999b). Interestingly, all of these proteins have one or more RNA-binding motif(s) and therefore might have RNA-related functions. The presence of these proteins as well as of RNAs (Subramaniam and Seydoux, 1999; Schisa et al., 2001) in these granules suggests that P granules are involved in some aspect of RNA metabolism or translation.

In this study, we show that ife-1 is expressed primarily in the germline, that IFE-1 protein associates with P granules, and that this association is dependent on PGL-1 protein. IFE-1 appears to be distinct from the other IFEs in being required specifically for the normal execution of spermatogenesis. This requirement is especially pronounced at elevated temperature. Depletion of IFE-1 causes a delay in spermatogenesis and production of defective sperm.

General methods for maintaining C. elegans are described in Brenner (Brenner, 1974). Strains used were wild-type variety Bristol, strain N2; LGI, glp-4(bn2); LGIII, unc-32(e189), fem-2(b245); LGIV, pgl-1(ct131, bn102), him-3(e1147), fem-3(q20gf); LGV, rde-1(ne219). Strains were maintained at 16°C, except for the temperature-sensitive mutants glp-4, fem-3 and pgl-1, which were maintained at 16°C and analyzed at 25°C (for glp-4 and fem-3) and 26°C (for pgl-1).

ife-1 cDNA was isolated in a yeast two-hybrid screen using pgl-1 cDNA as ‘bait’ and oligo(dT)-primed C. elegans cDNAs (Barstead and Waterston, 1989) as ‘prey’ (I. K., A. A., Y. F., T. Karashima, Y. K. and S. S., unpublished). ife-1 cDNA was subcloned from the yeast vector into the XhoI site in Bluescript KS⁺ vector to form pBS-ife-1. The cDNA sequence was verified using an ABI PRISM DNA sequencing kit and ABI PRISM 310 Genetic Analyzer (PE Applied Biosynthesis). pBS-ife-1 contains the full-length ife-1 ORF.

Northern hybridization analysis was performed as in Holdeman et al. (Holdeman et al., 1998). An ife-1-specific RNA probe was made using Strip-EZ™ T7/T3 Kit (Ambion). The ife-1 probe (160 bp) corresponds to nucleotides 620-780 of ife-1 cDNA. The rpp-1 transcript, which encodes a ribosomal protein (Evans et al., 1997), was used as a loading control. In situ hybridization was carried out as in Tabara et al. (Tabara et al., 1996) using the cDNA clone yk504h9 as a probe.

Full-length pgl-1 cDNA (2190 bp) was amplified by PCR from pBS-pgl-1 (Kawasaki et al., 1998), using primers 5′-CTCGAGATGGAGGCTAACAAGCGAGAA-3′ and 5′-GCGGCCGCTTAGAAACCTCCGCGTCCAC-3′. The PCR product was digested with XhoI and NotI and ligated to a glutathione-S-transferase (GST) gene fusion vector, pGEX-5X-3 (Amersham Pharmacia Biotech), pre-linearized by digestion with SalI and NotI, to form pGEX-pgl-1.

Escherichia coli strain BL21 (DE3) harboring pGEX-pgl-1 plasmid or a derivative of pGEX-2T (Pharmacia) were grown at 37°C in Luria-Bertani medium plus ampicillin. Expression of GST-PGL-1 or GST were induced by growth in 0.2 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 4 hours. Cells were lysed by sonication in lysis buffer (20 mM Hepes, pH 7.4, 0.5 M NaCl, 0.5 mM dithiothreitol (DTT), and 1 Complete™ Protease Inhibitor tablet per 10 ml (Boehringer-Mannheim)). Lysates were incubated at 4°C in the presence of 1% Triton X-100 for 30 minutes and centrifuged at 12,000 g. Supernatants were filtered through 0.45 μm filters. GST-PGL-1 and GST were bound to Glutathione-Sepharose (Pharmacia) and washed repeatedly with lysis buffer. Protein purity and concentration were assessed by SDS-PAGE and staining with Coomassie Blue. Before use in GST pull-down assays, the protein concentrations were equalized by dilution with unbound beads. Protein-bound beads were washed three times in binding buffer (100 mM NaCl, 5 mM DTT, 20 mM Hepes, pH 7.4 and 0.5% non-fat dry milk) and used directly in pull-down assays (Kraemer et al., 1999).

Radiolabeled IFE proteins were synthesized by in vitro transcription-translation with the TNT-coupled reticulocyte lysate system (Promega), pBS-ife-1, pBS-ife-2, pBS-ife-3, pBS-ife-4, pBS-ife-5 and [³⁵S]Met. For binding assays, 20 μl of in vitro-translated reaction product were incubated with 20 μl of bead-bound GST-PGL-1 or GST in 1 ml of binding buffer (see above) for 2 hours at 4°C on a rotator. After five washes with 1 ml of binding buffer, bead-bound radioactive proteins were resuspended in SDS sample buffer and separated by SDS-PAGE. Radiolabeled proteins were detected by autoradiography or by PhosphorImager. For RNase A treatment, [³⁵S]IFE-1 was diluted twofold in binding buffer containing RNase A (5 mg/ml) or mock-treated with water for 1 hour at 30°C. RNA was extracted from an aliquot of the RNase A-treated or mock-treated samples, electrophoresed and stained with Ethidium Bromide (Luitjens et al., 2000). Binding assays were performed as described above.

Radiolabeled proteins were synthesized in reticulocyte lysate as described above from the plasmids pBS-ife-1 and pBS-pgl-1. Reticulocyte lysate (20 μl) was diluted into 0.3 ml buffer B (20 mM Hepes, pH 7.6, 50 mM KCl, 1 mM EDTA, 0.5 mM DTT, 5% (v/v) glycerol, 1 mM PMSF and 10 μg/ml leupeptin) and passed three times over a 0.2-ml column of m⁷GTP-Sepharose. After a 0.5-ml wash, cap-bound proteins were eluted with 0.3 ml of 200 μM m⁷GTP. Lysate, flow-through, and eluted protein were resolved by SDS-PAGE and radiolabeled proteins detected by autoradiography or by PhosphorImager.

C. elegans wild-type N2 and glp-4 mutant worms were cultured at 25°C on agar medium supplemented with chicken egg yolk and E. coli strain OP50 (Brenner, 1974; Sulston and Hodgkin, 1988; McDermott et al., 1996). Animals were harvested, cleaned by sucrose flotation (Sulston and Hodgkin, 1988), pelleted, resuspended in an equal volume of water and drop-frozen in liquid N₂. Frozen worms were crushed with a mortar and pestle under liquid N₂ and thawed in the presence of buffer components to yield the following final concentrations: 20 mM MOPS, pH 7.5, 1 mM EDTA, 2 mM EGTA, 100 mM KCl, 0.5 mM DTT, 80 μg/ml each of leupeptin and pepstatin, 10 μg/ml E-64 (Boehringer Mannheim), 1 mg/ml TAME (Sigma Chemical), 50 mM NaF and 10 mM β-glycerophosphate. Homogenates were centrifuged at 20,000 g for 15 minutes at 4°C, and the supernatants were applied immediately to affinity chromatography columns as described previously (Jankowska-Anyszka et al., 1998; Keiper et al., 2000). Briefly, cleared C. elegans extracts (1-3 ml) were applied to 0.2-ml columns of m⁷GTP-Sepharose equilibrated in buffer A (20 mM MOPS, pH 7.5, 1 mM EDTA, 100 mM KCl, 10% (v/v) glycerol and 0.5 mM DTT). To assay the specific retention of PGL-1 on m⁷GTP-Sepharose, GTP or m⁷GTP were added to the cleared extract to a final concentration of 100 μM before chromatography to compete guanine nucleotide or cap-binding interactions, respectively. Columns were washed with 10 ml of buffer A. Proteins were eluted with 1 ml of buffer A containing 100 μM m⁷GTP and quantitated by Bradford assay (Maniatis et al., 1989). Procedures for the preparation of cap-affinity purified protein from ife-1(RNAi) worms were identical, except that protein was eluted with 200 μM m⁷GTP and precipitated with 10% (w/v) trichloroacetic acid in the presence of 120 μg linear acrylamide before electrophoresis.

Cleared worm lysates or cap-binding column elution fractions derived from the same amount of total protein were resolved on 6%, 10% or 12% gels by SDS-PAGE. After electrophoresis, proteins were transferred to PVDF membranes and immunostained using affinity-purified monospecific antibodies against each IFE (Jankowska-Anyszka et al., 1998; Keiper et al., 2000) or antiserum to PGL-1 (Kawasaki et al., 1998). Alternatively, gels were stained with silver nitrate to visualize proteins (Maniatis et al., 1989).

The histone cDNA fragment was excised from pJH4.52 (pie-1::GFP::histone H2B in pBS) (Strome et al., 2001) by SpeI digestion and replaced with ife-1 cDNA. ife-1 cDNA was amplified by PCR from pBS-ife-1 using primers 5′-TGAACTAGTGAAACGGAGCAAACGACG-3′ and 5′-TGAACTAGTGGATTTCTCGGCGACTGG-3′. The PCR product was digested with SpeI and subcloned into pJH4.52 (minus the histone insert) to form pSS4.52.1. Sequencing revealed a PCR-introduced amino acid change (A161→T) in IFE-1. Worms were transformed with pSS4.52.1, and GFP-expressing worms were maintained as described in Strome et al. (Strome et al., 2001).

Whole worm staining was carried out as described in Kawasaki et al. (Kawasaki et al., 1998). Immunostaining of extruded germlines or embryos was carried out as in Strome and Wood (Strome and Wood, 1983). We used the following primary antibodies: rabbit anti-GFP (Clontech) diluted 1:250; rabbit anti-PGL-1 (Kawasaki et al., 1998) diluted 1:30,000; chicken anti-GLH-2 (Gruidl et al., 1996) diluted 1:100; mouse monoclonal anti-α-tubulin (DM1-α) (Amersham); and mouse monoclonal SP56 (Ward et al., 1986). Secondary antibodies (from Jackson ImmunoResearch) were rhodamine-conjugated goat anti-rabbit IgG, rhodamine-conjugated goat anti-mouse IgG, and FITC-conjugated donkey anti-chicken IgY (IgG), all diluted 1:100. Samples were examined by fluorescence microscopy as described in Kawasaki et al. (Kawasaki et al., 1998).

To make double-stranded RNA (dsRNA), sense and antisense transcripts were generated from pBS-ife-1 using the MEGAscript kit (Ambion). The two strands were denatured at 85°C for 5 minutes and annealed by cooling slowly to room temperature (Subramaniam and Seydoux, 1999). dsRNA (200 ng/μl) was injected into young adult hermaphrodites. Injected animals were allowed to purge their embryos for 8-16 hours at 16°C. Injected animals were then picked to individual plates every 24 hours at either 16°C, 20°C or 25°C. F1 progeny of the injected mothers were picked to individual plates and examined for sterility/fertility by counting their progeny. As a control, the F1 progeny of animals injected with water were analyzed. For mating experiments, injected mothers were mated with N2 males 8-12 hours after injection. The F1 males were then crossed with unc-32 or fem-2 mutant hermaphrodites.

To address whether IFE-1 functions during larval development, we used an unc-32; rde-1 strain (Tabara et al., 1999a). Injected unc-32; rde-1 mothers were mated to N2 males 12 hours after injection and transferred to fresh plates every 24 hours. ife-1(RNAi); rde-1/+ F1 progeny were analyzed for sterility/fertility as above.

To prepare ife-1(RNAi) worms for western analysis, we used the feeding method of delivering dsRNA to worms (Kamath et al., 2000). Briefly, full-length ife-1 cDNA was subcloned into the XhoI site of the feeding vector L4440 (Timmons and Fire, 1998) to form L4440-ife-1. E. coli strain H115 (DE3) harboring L4440-ife-1 plasmid or L4440 vector alone were grown at 37°C in LB medium plus ampicillin. Bacteria were grown on NGM plates containing 80 μg/ml IPTG and ampicillin at 25°C. Worms were transferred to fresh plates every day for 2 days, and the progeny from the second and third plates were analyzed for sterility/fertility as above.

The interaction of IFE-1 with a P-granule-specific protein, PGL-1, was originally observed in a yeast two-hybrid screen using full-length PGL-1 fused to the GAL4 DNA-binding domain as ‘bait’ and oligo-(dT)-primed C. elegans cDNAs fused to the GAL4 activation domain as ‘prey’ (I. K., A. A., Y. F., T. K., Y. K. and S. S., unpublished). To verify that this interaction was authentic, we performed in vitro binding assays. Full-length PGL-1 fused to GST was expressed and purified from E. coli, immobilized on glutathione-agarose beads, and tested for binding to radiolabeled full-length IFE-1 protein produced by in vitro translation in a rabbit reticulocyte lysate. ³⁵S-labeled IFE-1 was incubated with equal amounts of GST-PGL-1 or GST attached to beads. After extensive washes, binding was assessed by SDS-PAGE analysis of the bead samples. Radiolabeled IFE-1 (26 kDa) bound to GST-PGL-1 but not to GST alone (Fig. 1A), suggesting that IFE-1 and PGL-1 are able to bind to each other in vitro in the absence of other C. elegans proteins.

Because IFE-1 is a cap-binding protein and PGL-1 has an RNA-binding motif (RGG box) (Kawasaki et al., 1998), there was a possibility that PGL-1 and IFE-1 interact indirectly via an RNA ‘bridge’ in both yeast two-hybrid and GST pull-down assays. To test this, reticulocyte lysate containing radiolabeled IFE-1 was treated with RNase A before incubation with GST-PGL-1. RNA extraction from the RNase A-treated lysate showed that all detectable RNAs were eliminated (Fig. 1B). RNase A treatment did not inhibit the binding of [³⁵S]IFE-1 and GST-PGL-1 (Fig. 1B), suggesting that the interaction between IFE-1 and PGL-1 is not dependent on RNA and therefore is direct.

We used GST pull-down assays to test whether PGL-1 also binds to the other four IFEs. Radiolabeled IFE-2, IFE-3, IFE-4 and IFE-5 were synthesized in vitro and incubated with GST or GST-PGL-1. None of the other IFEs bound to GST (not shown) or GST-PGL-1 (Fig. 1A), suggesting that PGL-1 interacts with a motif or region unique to IFE-1.

To investigate whether native IFE-1 and PGL-1 proteins interact in vivo, we determined whether PGL-1 is present in cap-binding fractions from adult hermaphrodites. Immunoblotting revealed that roughly 15% of PGL-1 was specifically retained on m⁷GTP-Sepharose and eluted with m⁷GTP (Fig. 2A). Since relatively few proteins remain bound to the affinity resin (<0.1% of soluble protein; see also Fig. 4A), PGL-1 was actually enriched approximately 180-fold in the bound fraction. Preincubation of worm extracts with 100 μM m⁷GTP (m⁷G), but not 100 μM GTP (G), greatly reduced the retention of PGL-1. As a control, we tested whether PGL-1 has the ability to bind to the cap column by itself. In vitro synthesized [³⁵S]PGL-1 was not retained on m⁷GTP-Sepharose, unless ³⁵S-IFE-1 was also present (Fig. 2B), suggesting that PGL-1 does not bind directly to m⁷GTP. Furthermore, PGL-1 was absent in the cap-bound fraction from ife-1(RNAi) worm lysates (Fig. 2C). Thus, our cap-binding results suggest that PGL-1 in worm lysates is retained on m⁷GTP-Sepharose via its association with an authentic cap-binding protein, IFE-1.

A role for IFE-1 in germline development was suggested by its interaction with the P-granule component PGL-1. To determine whether ife-1 is expressed primarily in the germline, an RNA probe to the unique 3′ end of the ife-1 cDNA was used to probe northern blots of RNA from adult wild-type and glp-4(bn2) mutant worms. glp-4 is a conditional mutant that makes very few germ cells at the restrictive temperature (Beanan and Strome, 1992). ife-1 RNA was 5.5-fold more abundant in wild-type than in glp-4 mutant worms, suggesting that ife-1 transcript is enriched in the germline (Fig. 3A).

To analyze the temporal and spatial distribution of ife-1 mRNA, we performed in situ hybridization on wild-type worms at different developmental stages. In hermaphrodites, newly synthesized ife-1 mRNA was first detected in the germline of L3 stage larvae (Fig. 3C). At this stage germ cells in the distal region of the gonad are mitotically proliferating and those in the proximal region are entering meiosis. As germline development progressed, the hybridization signal intensified in the regions of the gonad undergoing spermatogenesis (Fig. 3C, L4). The signal persisted after spermatogenesis was complete; a reduced level of ife-1 mRNA was detectable in adult germlines undergoing oogenesis.

In embryos, the distribution of ife-1 mRNA was similar to a pattern described previously as class I maternal mRNAs (Seydoux and Fire, 1994). These mRNAs show a uniform distribution in all cells during early cleavage stages and disappear at later stages. ife-1 mRNA was detected in all cells in embryos until the 100-200-cell stage and gradually disappeared by the 300-400-cell stage (Fig. 3B). These in situ hybridization results suggest that ife-1 mRNA in embryos is due to maternal transcription and that new ife-1 expression does not start until the L3 stage.

Northern and in situ hybridization results predicted that IFE-1 protein should be enriched in the germline. To assess the abundance of IFE-1 and other IFEs in the germline, we carried out a combination of affinity chromatography and western analysis. Proteins that bind to mRNA caps were enriched from extracts of wild-type and glp-4 mutant worms by m⁷GTP-Sepharose affinity chromatography. Interestingly, PGL-1 appears to be among the most abundant high molecular weight components among the cap-associated proteins (Fig. 4A, silver stained gel), supporting the notion that PGL-1 becomes enriched by association with IFE-1. The bound fractions were subjected to immunoblotting with anti-IFE antibodies (Fig. 4B). IFE-1, IFE-3 and IFE-5 signals were detectable in eluates from wild type but not in eluates from glp-4 mutants, which are deficient in germline tissue, suggesting that these proteins are predominantly present in the germline. IFE-2 and IFE-4, on the other hand, are at least as abundant in glp-4 worms as in wild type, suggesting that they are predominantly present in somatic cells.

The isoform-specific anti-peptide antibodies to IFE-1, which specifically recognize IFE-1 in western blot analyses (Jankowska-Anyszka et al., 1998; Keiper et al., 2000) (Fig. 4), did not yield a specific signal in immunostaining. Therefore, to examine the subcellular localization of IFE-1 in the germline, a construct expressing IFE-1 fused to green fluorescent protein (GFP) was expressed in worms. The construct uses the pie-1 promoter to drive expression of IFE-1::GFP in the germline and in early embryos (Strome et al., 2001). IFE-1::GFP was distributed throughout the cytoplasm and also localized to distinct particles in the germline and in embryos (n>1000 embryos examined; Fig. 5A). The size and distribution of these particles were very similar to those of P granules (Strome and Wood, 1982). To verify this assignment, worms were stained simultaneously with an anti-GFP antibody (Fig. 5B) and an antibody to a known P-granule protein, GLH-2 (Gruidl et al., 1996) (Fig. 5C). The particles stained with anti-GFP coincided with those stained by anti-GLH-2 (Fig. 5D), indicating that at least some IFE-1::GFP is associated with P granules.

The results shown in Fig. 1 and Fig. 2 suggested that IFE-1 may associate with P granules via a direct interaction with PGL-1. To test this possibility, we introduced the IFE-1::GFP-expressing construct into pgl-1(ct131) null mutants. In 99% of the pgl-1 embryos examined (n>160), IFE-1::GFP signal was not detectably associated with P granules but was still present in the cytoplasm (compare Fig. 5E and Fig. 5H). Co-staining embryos with anti-GFP and anti-GLH-2 antibodies confirmed that pgl-1 mutants contain P granules but that those P granules lack detectable IFE-1::GFP (Fig. 5F,G). These results suggest that PGL-1 is required for IFE-1 localization to P granules in wild-type animals.

Two lines of evidence support the assumption that the localization of GFP-tagged IFE-1 reflects the pattern of native IFE-1. First, the same pie-1-based vector has been used to express several other GFP-tagged proteins (γ-tubulin, β-tubulin and histone); these other proteins showed the expected localization (to centrosomes, microtubules and chromosomes, respectively) (Strome et al., 2001) and did not associate with P granules. Thus, the association of IFE-1::GFP with P granules is not due to the GFP part of the fusion protein but appears to be specific to the IFE-1 portion. Second, the association of IFE-1::GFP with P granules is dependent upon PGL-1, which was shown to bind IFE-1 in independent assays (Fig. 1 and Fig. 2).

The role of IFE-1 in germline development was examined by RNA-mediated interference (RNAi). Injection of double-stranded RNA (dsRNA) into hermaphrodites has been shown to result in a gene-specific, loss-of-function phenotype in the injected mothers and their progeny (Rocheleau et al., 1997; Fire et al., 1998; Montgomery et al., 1998). RNAi can also be achieved by feeding worms E. coli that produce dsRNA (Timmons and Fire, 1998; Kamath et al., 2000). Disruption of ife-1 expression has been shown previously to have no lethal effects on injected worms or their F1 progeny (Keiper et al., 2000). To verify that RNAi effectively depleted IFE-1 protein, we examined GFP fluorescence in ife-1::gfp worms injected with dsRNA prepared from full-length ife-1 cDNA. IFE-1::GFP signal was no longer detected in the embryos of injected hermaphrodites (Fig. 5H). Furthermore, IFE-1 protein was not detectable in lysates prepared from worms fed E. coli that produce ife-1 dsRNA but was present in control lysates prepared from worms fed E. coli harboring only the feeding vector (Fig. 2C). Wild-type hermaphrodites were then injected with ife-1 dsRNA, allowed to develop and reproduce at 16°C, 20°C and 25°C, and the F1 worms were examined. ife-1(RNAi) F1 hermaphrodites displayed temperature-sensitive sterility (Fig. 6A-C; Table 1). At 25°C, 80% of the F1 progeny of injected hermaphrodites were sterile (i.e. produced no progeny). A similar level of sterility was induced by the feeding method of RNAi (data not shown). The remaining fertile animals showed greatly reduced brood sizes (an average of 36 progeny from fertile ife-1(RNAi) worms compared with 221 progeny from control worms) (Fig. 6C). Unlike pgl-1 mutants, which show 75-85% sterility at 25°C and 100% sterility at 26°C (Kawasaki et al., 1998), ife-1(RNAi) worms did not display higher sterility at 26°C (data not shown). At lower temperatures (16° and 20°C), most ife-1(RNAi) F1 hermaphrodites were fertile, but their brood sizes were considerably lower than wild type. Control hermaphrodites produced an average of 266 and 280 progeny at 16°C and 20°C, while ife-1(RNAi) F1 hermaphrodites produced an average of 139 and 45 at those temperatures, respectively (Fig. 6A,B). Thus, depletion of IFE-1 protein caused temperature-sensitive sterility and reduction of brood size in hermaphrodites.

The requirement for IFE-1 in male germline development was similarly investigated by blocking the expression of ife-1 in males. ife-1(RNAi) males were generated by mating dsRNA-injected hermaphrodites with wild-type males. The ability of the resulting F1 males to produce outcross progeny when mated was then tested at 25°C. To enable scoring of outcross, the mating partners were homozygous unc-32 or fem-2. unc-32 hermaphrodites produce uncoordinated selfcross progeny, which can be easily distinguished from heterozygous progeny derived from mating. fem-2 hermaphrodites produce only oocytes and thus no selfcross progeny at 25°C. ife-1(RNAi) males displayed higher sterility (85% did not produce outcross progeny, compared with 22% for control males) and when fertile produced fewer outcross progeny than control males (Fig. 6D). These results show that IFE-1 is required for normal germline development in both hermaphrodites and males.

To investigate whether RNAi-induced sterility is due to depletion of the maternal load of IFE-1 or of larvally synthesized IFE-1, we used an rde-1 mutant (Tabara et al., 1999a). Homozygous rde-1 hermaphrodites are resistant to RNAi, but heterozygous outcross progeny (i.e. rde-1/+) are not resistant to RNAi. unc-32; rde-1 hermaphrodites were injected with ife-1 dsRNA and later mated with wild-type males to generate ife-1(RNAi); unc-32/+; rde-1/+ F1 hermaphrodites. The injected rde-1/rde-1 mothers were resistant to RNAi and therefore provided a maternal load of IFE-1 to their progeny. The heterozygous rde-1/+ offspring were no longer resistant to RNAi and therefore susceptible to inhibition of larval ife-1 expression. We observed that 49% of the ife-1(RNAi); unc-32/+; rde-1/+ F1 hermaphrodites were sterile at 25°C (n=38). Thus, inhibiting only larval synthesis of IFE-1 led to the same phenotype (sterility) as that caused by inhibiting both maternal and larval synthesis. This is consistent with IFE-1 functioning during larval development. The fact that inhibiting larval ife-1 expression resulted in fewer sterile F1s (49%) than after inhibiting both maternal and larval expression (80% sterile F1s) suggests that maternally provided IFE-1 is also important for proper germline development in progeny worms.

We used RNAi to determine whether ife-2, ife-4, and ife-5 also serve an essential role in germline development. ife-3 was excluded from our analysis, as all hermaphrodites injected with ife-3 dsRNA produce dead embryos (Keiper et al., 2000). ife-4(RNAi) and ife-5(RNAi) F1 worms were fertile at both 20°C and 25°C, and ife-2(RNAi) F1 worms showed negligible sterility (Table 1). These results suggest that when other IFEs are present, IFE-2, IFE-4 and IFE-5 are not required for germline development.

A previous study of worms depleted for multiple IFEs suggested that ife-1, ife-2 and ife-5 function redundantly during embryogenesis; one of the three proteins must be present to ensure embryo survival, and ife-2 appears to be the most important of the three proteins (Keiper et al., 2000). To test whether these genes also function redundantly in germline development, ife-1(RNAi);ife-2(RNAi), ife-1(RNAi);ife-5(RNAi) and ife-1(RNAi);ife-2(RNAi);ife-5(RNAi) F1 worms were analyzed for enhancement of sterility at 20°C and 25°C. These RNAi worms did not display higher sterility than ife-1(RNAi) worms at either temperature (Table 1), suggesting that neither IFE-2 nor IFE-5 functions redundantly with IFE-1 in the germline.

We observed that although ife-1(RNAi) F1 hermaphrodites failed to produce embryos, they laid oocytes. This phenotype could reflect production of defective oocytes, failure to make sperm, production of defective sperm, or some combination of these defects. To distinguish between these possibilities, sterile ife-1(RNAi) adult hermaphrodites were mated with wild-type males. When provided with wild-type sperm, ife-1(RNAi) animals became fertile (n=6), producing viable embryos that developed to adulthood, suggesting that IFE-1 is required for spermatogenesis but not for oogenesis.

The defect in spermatogenesis was characterized in the germline of ife-1(RNAi) F1 hermaphrodites. In wild-type worms, spermatogenesis initiates during the L4 larval stage (Fig. 7). Spermatogonial nuclei proliferate mitotically in the distal region of the gonad, entering meiosis as the cells move toward the loop region. Pachytene-stage primary spermatocytes separate from the central cytoplasmic core, the rachis, and divide to form secondary spermatocytes, and then haploid spermatids, which mature into motile spermatozoa. In males, spermatogenesis continues throughout adulthood. In hermaphrodites, the germline switches from spermatogenesis to oogenesis during young adulthood; oocytes arrested in prophase of meiosis I are fertilized by sperm stored in the spermatheca. The stages of spermatogenesis and oogenesis can be recognized by their characteristic chromatin morphology after staining with a DNA dye (Fig. 8A). Sterile ife-1(RNAi) adult hermaphrodites contained oocyte nuclei, but lacked nuclei typical of haploid sperm (n>100 worms examined) (Fig. 8B). The few ife-1(RNAi) adult hermaphrodites that were fertile contained some sperm nuclei (not shown). These results suggest that ife-1(RNAi) worms produce no or few progeny because they produce no or few normal sperm.

To test whether germ cells in ife-1(RNAi) hermaphrodites fail to enter spermatogenesis or else enter but fail to complete spermatogenesis, we stained L4 and young adult hermaphrodites with the spermatogenesis-specific antibody SP56 and a DNA dye (Fig. 8C-H). In wild type, SP56 binds to the plasma membrane, membranous organelles, and pseudopodial cytoplasm in primary spermatocytes and at later stages of spermatogenesis (Roberts et al., 1986). We observed SP56 staining in all ife-1(RNAi) gonad arms examined (n>100) (Fig. 8F,H), demonstrating that germ cells initiated spermatogenesis. The most mature spermatogenic cells observed in sterile RNAi hermaphrodites resembled late primary or secondary spermatocytes, based on the degree of DNA condensation (Fig. 8G) and on the occasional presence of meiotic spindles (not shown). Therefore, spermatogenesis and oogenesis appear to be initiated in proper sequence, but spermatogenesis is not completed in most ife-1(RNAi) hermaphrodites.

As described above, IFE-1 depletion also causes males to be sterile. To determine whether ife-1(RNAi) males also lack sperm, we stained RNAi males with a DNA dye. All ife-1(RNAi) males examined (n>20) contained highly condensed sperm nuclei, suggesting that at least some spermatocytes completed meiosis (Fig. 9A). Therefore, in contrast to ife-1(RNAi) hermaphrodites, spermatogenesis generally proceeds beyond the spermatocyte stage in ife-1(RNAi) males. Nevertheless, the sperm produced are apparently defective (Fig. 6D).

The presence of mature-looking sperm in ife-1(RNAi) males raised the question of whether the requirement for IFE-1 in spermatogenesis is different in males and hermaphrodites, or whether the requirement is similar and the absence of mature sperm in hermaphrodites is due to the onset of oogenesis. In wild-type hermaphrodites, by the time that mature oocytes are produced, most sperm are mature and motile. As oocytes pass through the narrow spermatheca, sperm are often swept into the uterus. However, these sperm are able to move back to the spermatheca (Singson, 2001). Immotile spermatocytes, present in ife-1(RNAi) hermaphrodites, would be unable to migrate back to the spermatheca and would be lost. To test whether preventing the onset of oogenesis would enable ife-1(RNAi) hermaphrodites to produce mature-looking sperm, as in males, we used a fem-3(gf) mutant. fem-3(gf) hermaphrodites do not undergo the sperm/oocyte switch and as a result produce sperm throughout adulthood (Barton et al., 1987). As expected, after staining with a DNA dye, we observed that L4 and young adult fem-3 mutants contained hundreds of sperm. As in ife-1(RNAi) hermaphrodites, 85% of L4 and young adult ife-1(RNAi); fem-3 hermaphrodites contained only spermatocytes at 25°C (n=47). However, after allowing ife-1(RNAi); fem-3 hermaphrodites to age (2 days beyond mid-L4 stage), 88% (n=26) contained mature-looking sperm, although the number of sperm was reduced compared to fem-3 controls (Fig. 9B). Thus, progression past the spermatocyte stage can occur in RNAi hermaphrodites if oogenesis is prevented and the period of spermatogenesis is extended. This result suggests that depleting IFE-1 causes a delay but not an arrest in spermatogenesis in hermaphrodites.

To test whether males also are delayed in spermatogenesis, we compared similar stages of N2 and ife-1(RNAi) males at 25°C. At 32 hours past hatching, 83% of N2 males had sperm (n=6) and 0% of ife-1(RNAi) males had sperm (n=5). Six hours later, 71% of ife-1(RNAi) males contained a small number of sperm (n=7). Thus, similar to hermaphrodites, ife-1(RNAi) males are delayed in production of sperm. Sperm produced in ife-1(RNAi) males and in aged ife-1(RNAi); fem-3 hermaphrodites have a normal appearance by Nomarski and DAPI, and can be activated to form a pseudopod (data not shown). Their failure to generate progeny may be due to defects in motility and/or fertilization.

P granules are present in the germ cells of both hermaphrodites and males throughout most of their development. We previously observed that P granules, as detected by anti-GLH-1 and anti-GLH-2, are partitioned to the residual body during spermatogenesis and thus are absent in mature sperm (Gruidl et al., 1996) (A. A., unpublished). Interestingly, unlike GLH-1 and GLH-2, PGL-1 disappears from P granules after the pachytene stage of spermatogenesis (Fig. 10). Because the association of IFE-1 with P granules depends on PGL-1 (Fig. 5), it is likely that IFE-1 also is released from P granules after the pachytene stage. The stages of spermatogenesis known to be affected by IFE-1 depletion coincide with stages when IFE-1 is predicted to be dissociated from P granules. Based on these results, we speculate that P granules regulate the level and perhaps the function(s) of IFE-1 during spermatogenesis; association with P granules may reduce the availability of IFE-1 during germline proliferation, and release from P granules may increase the level of IFE-1 to function during post-pachytene stages of spermatogenesis.

This study shows that one of the five C. elegans isoforms of eIF4E (IFE-1) plays a unique and essential role in germline development. IFE-1 is specifically required during spermatogenesis, and this requirement is more pronounced at elevated temperatures. Interestingly, IFE-1 is a component of P granules. The association of a translation initiation factor with P granules supports the view that these granules are involved in translational regulation.

Expression of ife-1 in worms is significantly enriched in the germline, and mRNA accumulates in the region of the gonad undergoing spermatogenesis. Elevated expression of eIF4E in specific tissues and cell types has been noted in other systems. Specifically, eIF4E mRNA levels are elevated in the primordial germ cells of Drosophila (Hernandez et al., 1997), germ cells in rat testes (Miyagi et al., 1995) and oocytes in zebrafish (Fahrenkrug et al., 1999), suggesting that germ cells often, perhaps universally, require high levels of this translation initiation factor. eIF4E levels are also elevated in many solid tumors and malignant cell lines, with the highest levels being reported in breast and prostate cancers, and in head and neck squamous cell carcinomas (De Benedetti and Harris, 1999). These observations have led to speculations that eIF4E is not simply a general translation factor, but also may act as a tissue-specific translational enhancer.

Our RNAi studies suggest that IFE-1 is unique among the five IFEs in being essential for proper germline development. Blocking the expression of ife-1 causes sterility in 80-85% of both hermaphrodites and males grown at high temperature (25°C); the remaining 15-20% of worms produce a greatly reduced number of progeny. Because RNAi causes defects specifically in spermatogenesis, we infer that 80-85% of RNAi animals produce no functional sperm and the remaining worms produce only a small number of functional sperm. Worms depleted of IFE-1 and grown at lower temperature (16-20°C) also produce reduced numbers of progeny. Thus, IFE-1 appears to be required for production of a normal number of functional sperm, especially at elevated temperature.

Hermaphrodites and males display similar RNAi-induced defects in spermatogenesis. In both sexes, the initial defect is a delay in production of mature sperm. In hermaphrodites, this delay generally leads to a spermless phenotype, probably because immature sperm are not motile. In this scenario, the first oocytes produced encounter spermatocytes (instead of sperm); the oocytes are not fertilized and as they are ovulated and laid, they carry the spermatocytes out of the oviduct, spermatheca and uterus. Interestingly, preventing the onset of oogenesis and prolonging the window of spermatogenesis in RNAi hermaphrodites (in fem-3(gf) mutants) allows at least some mature sperm to be formed, as seen in males. From testing the fertility of ife-1(RNAi) males, it is apparent that the sperm formed in males at 25°C are defective. They may be defective in motility or in their ability to fertilize oocytes. In the rare cases in which a few mature sperm were observed in ife-1(RNAi) sterile hermaphrodites, the sperm also were defective, as evidenced by the presence of unfertilized oocytes in the uterus. Thus, in both hermaphrodites and males, IFE-1 appears to be required for normal entry into and/or progression through spermatogenesis and for formation of functional sperm. We do not know if IFE-1 function is required at multiple steps in spermatogenesis, or only at a single early step. If the latter, then defective or slow progression through that step must lead to formation of abnormal sperm.

Worms depleted of IFE-1 do not display obvious defects in other aspects of germline development, such as proliferation, the ability to switch from spermatogenesis to oogenesis, and oogenesis. In fact, RNAi hermaphrodites produce healthy embryos when normal sperm are provided by mating with wild-type males. Thus, it seems that blocking the expression of ife-1 in C. elegans leads to a preferential decrease in translation of a limited set of mRNAs, which are specifically involved in spermatogenesis.

A substantial body of evidence indicates that alteration in the amount of active eIF4E exerts different effects on the translation of different mRNAs (Clemens and Bommer, 1999; De Benedetti and Harris, 1999). Overexpression of eIF4E in tissue culture cells leads to oncogenic transformation, probably owing to elevated translation of growth-promoting mRNAs whose translation is normally limited by inhibitory features in their 5′ UTRs. Such features include long, highly structured, G/C-rich sequences and upstream open reading frames (uORFs). Many mRNAs with such inhibitory features encode proteins that operate to promote cell growth and proliferation (Kozak, 1991). It has been suggested that these mRNAs are translated poorly in resting cells, owing to their inability to compete with other mRNAs (without inhibitory structures in their 5′ UTRs) for the cap-dependent unwinding machinery, for which the availability of active eIF4E may be crucial (Rhoads, 1991; Koromilas et al., 1992). Increasing the level or availability of eIF4E (e.g. after growth stimulation) would ameliorate the competition between mRNAs for the translation initiation machinery, leading to deregulated expression of these growth-promoting proteins. Similarly, it is possible that some of the mRNAs involved in spermatogenesis in C. elegans have such inhibitory structures in their 5′ UTRs. In the absence of IFE-1, these mRNAs would not be able to compete with other mRNAs for the low level of eIF4E, leading to defects in spermatogenesis.

Why is the sterility seen after blocking the expression of ife-1 sensitive to temperature? It is known that eIF4E activity is sensitive to some cellular stresses (Sonenberg and Gingras, 1998). C. elegans spermatogenesis occurs more rapidly at high temperature. The stress associated with high-temperature growth might require higher levels of eIF4E. If IFE-1 is a prevalent isoform of eIF4E, then blocking the expression of IFE-1 would likely reduce the overall rate of translation or the rate of translation of certain mRNAs (see above). Either could lead to problems in spermatogenesis. Whatever the underlying molecular mechanism, at low temperatures other IFEs provide sufficient eIF4E activity to form at least some functional sperm, but at high temperatures IFE-1 is required. It would be interesting to test whether overexpression of other IFEs eliminates the requirement for IFE-1 in spermatogenesis at high temperatures.

We have demonstrated that IFE-1 protein is associated with P granules via PGL-1. PGL-1 is a constitutive component of P granules and is required for several processes in germline development, including proliferation, meiosis and gametogenesis (Kawasaki et al., 1998). pgl-1 mutants show a variety of defects, ranging from no germline proliferation to some germline proliferation and production of defective gametes.

Are the spermatogenesis defects seen in pgl-1 mutants the result of dissociation of IFE-1 from P granules? About 19% of pgl-1 null hermaphrodites grown at 26°C and containing a relatively well proliferated germline produce mature-looking sperm (n=54; A. A., unpublished). This value is similar to the percentage (∼20%) of ife-1(RNAi) hermaphrodites that produce sperm and are fertile at 25°C. These data are consistent with, but do not prove, that the absence of sperm in most pgl-1 mutants results from dissociation of IFE-1 from P granules.

P granules contain RNA. Originally it was shown that P-granule-associated RNA contains poly(A) and SL1, a trans-spliced leader sequence found on many C. elegans mRNAs (Seydoux and Fire, 1994). More recently, it was shown that nos-2 RNA is associated transiently with P granules in embryos (Subramaniam and Seydoux, 1999) and that pos-1, skn-1, par-3, mex-1, gld-1 and nos-2 RNAs are associated with P granules in the maternal germline (Schisa et al., 2001). The presence of RNAs in P granules suggests that these granules are important for RNA delivery, stability or translation. Consistent with this view, all P-granule proteins identified to date are predicted to bind RNA. The recent discovery that P granules are associated with nuclear pores in the germline (Pitt et al., 2000) suggests that these granules have access to the majority of mRNAs as they transit from the nucleus to the cytoplasm. Certain mRNAs may be specifically retained in P granules by resident RNA-binding proteins. Or alternatively, as suggested by Schisa et al. (Schisa et al., 2001), granule-associated proteins may facilitate export of mRNAs from the nucleus out to the cytoplasm.

At present, it is not known whether there are any other translation factors in P granules. It was reported recently that VASA protein, which is the Drosophila homolog of the C. elegans GLH proteins (Gruidl et al., 1996, Kuznicki et al., 2000) and is a component of polar granules, interacts with translation initiation factor 2, dIF2 (Carrera et al., 2000). Although it has not been shown that dIF2 is itself a component of polar granules, this discovery suggests that there might be other translation factors in germ granules.

Are germ granules active sites of translation? Although, mitochondrial ribosomal RNAs are associated with germ granules in Drosophila and Xenopus embryos (Kobayashi et al., 1993, Kobayashi et al., 1998), neither mitochondrial nor nuclearly encoded ribosomal RNAs are enriched in P granules in the C. elegans maternal germline (Schisa et al., 2001). This observation argues against a scenario in which P granules are active sites of mRNA translation. One possibility is that P granules serve an inhibitory role in translational control. For example, P granules may sequester IFE-1 and thereby reduce the level of eIF4E available for initiating translation in the cytoplasm. Furthermore, the IFE-1 retained in P granules may be inactive. This prediction is based on the observation that PGL-1 contains the motif, YXXXXLφ (where φ is a hydrophobic amino acid and X is any amino acid) (Mader et al., 1995), through which eIF4G, 4E-BPs and Maskin are known to bind eIF4E. If PGL-1 interacts with IFE-1 via this motif, then PGL-1 interaction would probably prevent the association of IFE-1 with eIF4G and thereby inhibit translation initiation. Interestingly, in wild-type worms PGL-1 levels drop to below detectable after the pachytene stage of spermatogenesis. This may liberate IFE-1, perhaps to participate in promoting the completion of meiosis.

Alternatively, as was suggested for nuclearly localized eIF4E (Dostie et al, 2000), it is possible that IFE-1 in P granules serves a unique, non-translational role. For example, P-granule-associated IFE-1 may participate in nucleocytoplasmic transport of mRNA or function in regulating mRNA stability.

We thank Marvin Wickens and David Bernstein for their advice and help with in vitro binding assays; Karen Bennet for anti-GLH-2; Eric Aamodt, Barry Lamphear and Marzena Jankowska-Anyszka for advice on preparing worm extracts; Geraldine Seydoux for stimulating discussion; and Eric Polinko for help generating IFE-1::GFP worms. Some nematode strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by the NIH National Center for Research Resources (NCRR). A. A., I. K., Y. F., and S. S. were supported by NIH grant GM34059. B. D. K. and R. E. R. were supported by NIH grant GM20818. I. K. and Y. K. were supported by CREST, JST.