ATX-2, the C. elegans ortholog of ataxin 2, functions in translational regulation in the germline

Human spinocerebellar ataxia type 2 (SCA2) is a neurodegenerative disorder characterized by a degeneration of Purkinje cells in the cerebellum and the progressive loss of coordination (Huynh et al., 2000). SCA2 is caused by dominant mutations in the ataxin 2 protein that are associated with expansion of a polyglutamine repeat. SCA2 probably reflects an abnormal activity of the mutant ataxin 2; the normal function of ataxin 2 remains unknown. Previous studies have shown that ataxin 2 is a cytoplasmic protein that may be enriched in the Golgi apparatus,suggesting a possible function in protein secretion(Huynh et al., 2003). Other studies have suggested that ataxin 2 might play a role in RNA metabolism. First, two-hybrid analysis identified a putative RNA-binding protein, named ataxin 2-binding protein 1 (A2BP1), as an ataxin 2-interacting protein(Shibata et al., 2000). Second, a budding yeast protein, Pbp1p, that is distantly related to ataxin 2 has been shown in a two-hybrid assay to interact with poly(A)-binding protein(Mangus et al., 1998). In vitro experiments suggest that Pbp1p might function in mRNA polyadenylation. Finally, a comparative analysis of peptides that are known to bind the C-terminal domain of human poly(A)-binding protein predicted that ataxin 2 might also bind that domain (Kozlov et al., 2001).

The yeast ataxin 2-related protein, Pbp1p, is a non-essential protein. By contrast, the C. elegans and D. melanogaster orthologs have been shown to be essential proteins. Loss or overexpression of Drosophila Datx2 leads to diverse defects such as female sterility,tissue degeneration and lethality. These abnormalities have been proposed to result from a defect in regulation of the actin cytoskeleton(Satterfield et al., 2002). Depletion of C. elegans ATX-2 was reported to cause embryonic lethality; however, the basis for this lethality has not been analyzed(Kiehl et al., 2000).

We show that ATX-2 functions in the postembryonic development of the C. elegans germline. The germline contains numerous mRNAs whose expression is temporally and/or spatially regulated(Wickens et al., 2000). For example, temporal control of TRA-2 protein translation is essential for the production of both sperm and oocytes: C. elegans is a self-fertilizing hermaphrodite that produces sperm briefly during the last larval (L4) stage of development, and then switches to produce only oocytes during adulthood. TRA-2 protein promotes female development, and repression of tra-2 mRNA translation is required to allow spermatogenesis during the L4 stage (Goodwin et al.,1993). Another example of regulated translation in the germline is the spatial control of expression of the yolk receptor RME-2. Germ cells in the hermaphrodite gonad undergo sequential developmental programs as they move proximally toward the uterus (Schedl,1997). At the distalmost end of the gonad, germ cells proliferate and remain undifferentiated. As germ cells move proximally, they enter meiosis; cells in the medial part of the gonad are in the pachytene stage of meiosis. In the proximal gonad, germ cells undergo terminal differentiation as either sperm or oocytes (Fig. 1A). While rme-2 mRNA is broadly expressed in both the medial and proximal regions of the adult gonad, RME-2 protein is expressed only in the proximal part of the gonad, where germ cells undergo oogenesis(Fig. 5A)(Grant and Hirsh, 1999; Lee and Schedl, 2001). A key regulator of mRNA translation in the gonad is the GLD-1 protein(Francis et al., 1995a). GLD-1 contains a KH-domain and is a member of the STAR (signal transduction and activation) family of RNA-binding proteins(Jones and Schedl, 1995; Vernet and Artzt, 1997). GLD-1 represses translation of several mRNAs, including tra-2 and rme-2 (Jan et al.,1999; Lee and Schedl,2001). GLD-1 has been shown to bind directly to at least some of its mRNA targets; however, the precise mechanism of GLD-1-mediated repression remains unknown.

We show here that depletion of ATX-2 by dsRNA interference (RNAi) causes several defects in the post-embryonic development of the germline. Most of these defects suggest a loss or deregulation of GLD-1-mediated repression. We provide evidence that MEX-3, a second KH-domain protein, contributes to the regulation of rme-2 mRNA expression, and that this function is disrupted in atx-2(RNAi) gonads. We also show that ATX-2 is a widely expressed cytoplasmic protein that forms a complex with PAB-1, a poly(A)-binding protein. Together, our results support a model in which ATX-2 is involved in translational regulation, and suggest that ATX-2 may directly or indirectly link translational regulators such as GLD-1 or MEX-3 with the core translation machinery through the activity of poly(A)-binding proteins.

The N2 strain of C. elegans was cultured and manipulated as described (Brenner, 1974). Mutants were as follows.

LGI: dpy-5(e61), fer-1(hc13)(Ward and Miwa, 1978), gld-1(q485) and gld-1(q361)(Francis et al., 1995a), gld-2(q497) (Kadyk and Kimble,1998), mex-3(or20)(Draper et al., 1996), unc-13(e51)

LGII: tra-2(q122gf) (Schedl and Kimble, 1988)

LGIII: glp-1(oz112gf) (Berry et al., 1997), unc-32(e189)

LGIV: fem-1(hc17) (Nelson et al., 1978), fem-3(q20gf)(Barton et al., 1987), him-8(e1489), rme-2(b1008) (Grant and Hirsh, 1999)

LGV: fog-2(q71) (Schedl and Kimble, 1988).

Unless noted, mutations are referenced by Hodgkin(Hodgkin, 1997). Transgenic strains were lag-2::GFP JK2003(qEx233) (kindly provided by J. Kimble), vit-2::GFP DH1033(Grant and Hirsh, 1999) and atx-2::3xmyc; this construct (atx-2 genomic fragment with a triple-myc sequence inserted at the 3′ end, plus 1 kb of the 5′UTR and 0.7 kb of the 3′ UTR) was integrated into the genome by microparticle bombardment as described by Praitis et al.(Praitis et al., 2001).

To assure effective depletion of ATX-2, animals were subjected to atx-2(RNAi) from the fourth larval stage onwards. As adults, these initially produced some viable progeny that, with continued exposure to atx-2 dsRNA, grew into sterile atx-2(RNAi) adults. We obtained comparable results by injecting, soaking or feeding dsRNA; the feeding technique (Timmons and Fire,1998) was used for the experiments here. atx-2 RNAi was not efficient at 20°C; unless described otherwise, all RNAi experiments were performed at 25°C (24.8-25.2°C). Temperature-sensitive (ts) fem-1(hc17) and fem-3(q20gf) mutants were exposed to atx-2 RNAi as L4 animals at 20 or 15°C, respectively, and the L1 progeny were transferred to 25°C.

Unc progeny from gld-1(q485)unc-13(e51) heterozygotes [strain BS3156] were analyzed.

One-third of the non-Dpy, non-Unc progeny of gld-1(q361)/dpy-5(e61)unc-13(e51) animals [strain BS87] subjected to atx-2(RNAi) had well proliferated germlines that contained very few or no sperm. We presume these were gld-1(q361);atx-2(RNAi)animals.

One-third of the non-Dpy, non-Unc progeny of gld-2(q497)/dpy-5(e61)unc-13(e51) animals subjected to atx-2(RNAi) lacked pachytene nuclei in their germlines. We presume these were gld-2(q497);atx-2(RNAi) animals.

Unc progeny from glp-1(oz112gf)unc-32(e189) heterozygotes [strain BS860] were analyzed.

Feminized tra-2(q122gf) hermaphrodites were mated with tra-2(q122gf) males on atx-2 dsRNA feeding plates, and the hermaphrodite progeny were analyzed for the presence of sperm/oocytes. In control experiments, sperm-defective fer-1(hc13) ts hermaphrodites grown at restrictive temperature were mated with N2 males on atx-2(RNAi) plates, and the hermaphrodite progeny were examined for sperm/oocytes.

Anti-ATX-2 mouse monoconal antibodies mAbP1G12 and mAbP6D2 were generated following published procedures (Wayner et al., 1989) using His-tagged N-terminal half of ATX-2. Other antibodies/antisera used in this study were 1CB4 mAb(Okamoto and Thomson, 1985),anti-ph-H3 (Crittenden et al.,2002; Hendzel et al.,1997), anti-HIM-3 (Zetka et al., 1999), anti-GLP-1(Crittenden et al., 1994),anti-RME-2”INT” (Grant and Hirsh, 1999), anti-MEX-3(Draper et al., 1996),anti-MEX-5 (Schubert et al.,2000) and anti-GLD-1 (Jones et al., 1996). Worms were prepared for immunolocalization according to Lin et al. (Lin et al.,1998). In situ hybridization was performed essentially as described by Lee and Schedl (Lee and Schedl, 2001). The fragment of rme-2 cDNA used to create an antisense probe corresponds to amino acids 10-482(Grant and Hirsh, 1999).

One-step IP was performed as described by Ciosk et al.(Ciosk et al., 1998). Following in-gel digestion (Shevchenko et al.,1996), samples were desalted, dried, resuspended in 5 μl of 0.1% TFA and analyzed by LC MS/MS with a ThermoFinnigan LCQ mass spectrometer(Gatlin et al., 1998). Data were collected in the data-dependent mode in which a MS scan was followed by MS/MS scans of the three most abundant ions from the preceding MS scan. The MS/MS data were searched against the NCBI non-redundant and the C. elegans protein databases. The matches were scored by SEQUEST and identifications were considered valid if the identified protein contained at least two peptides with Xcorr scores above 2.0 and did not appear in a control sample.

We depleted ATX-2 from animals by dsRNA-interference (RNAi) using a protocol that removes all immunodetectable ATX-2 (see below and Materials and methods); we refer to such animals as atx-2(RNAi) animals. The growth rate of atx-2(RNAi) larvae appeared similar to wild-type controls,and atx-2(RNAi) adults had no obvious defects in either movement,body size (0.96=average length of atx-2(RNAi) adults/wild-type adults, n=10 each), or morphology of somatic tissues. However, the size of the atx-2(RNAi) gonad was approximately half the size of the wild-type gonad (compare Fig. 1A,C). In wild-type adult hermaphrodites, the proximal region of the adult gonad contains only oocytes (Fig. 1A). By contrast, the proximal region of the adult atx-2(RNAi) gonad contained large quantities of sperm and no oocytes. These sperm resembled normal sperm in morphology, by DAPI staining(Fig. 1C,D) and by staining with the antibody 1CB4 (Fig. 1E) that recognizes membranous organelles enriched in spermatocytes and spermatids (Kelleher et al., 2000). Thus, ATX-2-depleted animals display two major postembryonic phenotypes: the reduced size of the gonad and the absence of oogenesis in the germline.

Previous studies have shown that human ataxin 2 is a cytoplasmic protein that may be enriched in the Golgi apparatus(Huynh et al., 2003), and that Datx2, the Drosophila ortholog of ataxin 2, may function in actin filament formation (Satterfield et al.,2002). To see if ATX-2 is required for actin cytoskeleton organization in C. elegans, we stained atx-2(RNAi) gonads for microfilaments but did not detect obvious abnormalities (data not shown). To determine the localization pattern of ATX-2, we generated mouse monoclonal antibodies against ATX-2. On a western blot, anti-ATX-2 Abs produced a prominent signal (Fig. 1F, left arrow). Extracts from transgenic worms with a Myc-tagged allele of atx-2 showed an additional, slightly larger band, as expected(Fig. 1F, right arrow). The ATX-2 signal was greatly reduced in comparable extracts from worms subjected to atx-2 RNAi (Fig. 1G). Thus, the antibodies specifically recognize ATX-2 in worm extracts. In whole-mount preparations of wild-type hermaphrodites and males,the anti-ATX-2 Abs stained throughout the cytoplasm in many, if not all, cell types, with slightly stronger staining in the gonad(Fig. 1H; data not shown). This staining is specific to ATX-2, as the gonads of atx-2(RNAi) worms did not stain (Fig. 1I), nor did somatic cells, with the exception of neurons; several studies by others have shown that neurons in C. elegans are particularly refractive to RNAi. We conclude that ATX-2 is a widely expressed protein enriched in the cytoplasm but not in the Golgi apparatus or in filamentous structures.

To identify proteins that might interact with ATX-2 in vivo, we immunopurified ATX-2 from extracts of adult worms. Silver-stained ATX-2 immunoprecipitates (IPs) consistently showed three major bands(Fig. 1J, arrows). Microsequencing showed that the two closely spaced top bands were ATX-2. The third, lower band was identified as PAB-1, one of the two C. eleganscytoplasmic poly(A)-binding proteins(Mangus et al., 2003). In control experiments, we found that both ATX-2 and PAB-1 were absent from an IP prepared with an unrelated antibody (Fig. 1J, lane 3), and that atx-2(RNAi) caused a significant reduction in the amount of purified PAB-1 and of the lower ATX-2 band(Fig. 1J, lane 5); the lower ATX-2 band appears to be specific to the gonad (R.C., M.D.P. and J.P.,unpublished). We conclude that ATX-2 forms a complex with PAB-1, suggesting a possible role for ATX-2 in PAB-1-mediated processes.

To address the function of PAB-1, we subjected animals to pab-1RNAi. Depletion of PAB-1 did not prevent development of most somatic tissues,except that the vulva protruded abnormally in adults(Fig. 1K, arrowhead). By contrast, development of the germline was highly abnormal; adult gonads were very small and contained few, undifferentiated germ cells(Fig. 1K). Thus, PAB-1 appears to be required primarily for development of the gonad; depletion of both PAB-1 and PAB-2, a second C. elegans cytoplasmic poly(A)-binding protein,caused additional defects in somatic development (data not shown). The observation that the phenotype of PAB-1-depleted animals (or both PAB-1 and PAB-2-depleted animals) is much more severe than the phenotype of ATX-2-depleted animals suggests that ATX-2 does not have an essential role in all activities mediated by poly(A)-binding proteins.

To determine the cause of the small size of atx-2(RNAi) gonads, we first counted the number of mitotic germ cells in the distal region of the gonad using an antibody that recognizes a mitosis-specific form of histone H3(Hendzel et al., 1997; Crittenden et al., 2002). We found that the number of mitotic figures was approximately half that of the wild-type number (Fig. 2A). This finding suggests that atx-2(RNAi) gonads are small because mitotic proliferation in the germline is reduced. To determine whether ATX-2 is required for normal proliferation irrespective of the sexual identity of the germline, we depleted ATX-2 from hermaphrodites with either masculinized or feminized germlines, and from wild-type males. In all these cases,proliferation was reduced (Fig. 2A), suggesting that reduced proliferation and masculinization of the germline are independent consequences of depleting ATX-2.

Mitotic proliferation in the germline depends on a GLP-1/Notch signaling pathway (Kimble and Simpson,1997; Seydoux and Schedl,2001). Germ cells in the distal gonad express the receptor GLP-1/Notch, and are in contact with a somatic cell called the distal tip cell(DTC) that expresses the ligand LAG-2. A lag-2::GFP transgene was expressed normally in the DTC of atx-2(RNAi) gonads (data not shown). By contrast, we found that the levels of GLP-1 were markedly reduced in atx-2(RNAi) gonads and feminized fem-1;atx-2(RNAi) gonads(compare Fig. 2B,C; data not shown). Repression of GLP-1/Notch expression in the medial gonad has been shown to depend on the translational repressor GLD-1(Marin and Evans, 2003). Normally, GLD-1 is expressed at low levels in the distalmost, mitotic part of the germline. GLD-1 levels increase as cells enter meiosis(Jones et al., 1996). We found that GLD-1 was expressed at higher than normal levels in the distalmost germline of atx-2(RNAi) animals (compare Fig. 2E,F), suggesting that elevated levels of GLD-1 could repress GLP-1 repression. Indeed, we found that gld-1;atx-2(RNAi) animals had restored GLP-1 expression in the distal gonad (Fig. 2D). Moreover,mitotic proliferation in gld-1;atx-2(RNAi) animals was also restored(Fig. 2A); this proliferation,as in wild-type gonads, was restricted to the distal germline. These findings suggest that abnormally high expression of GLD-1 in the distalmost germline of ATX-2-depleted animals causes reduced expression of GLP-1. However, we found that many feminized fem-1;atx-2(RNAi) gonads with reduced levels of distal GLP-1 (data not shown) appeared to have approximately normal low levels of distalmost GLD-1 (Fig. 2G). This observation suggests the possibility that ATX-2 depletion alters GLD-1 activity, in addition to GLD-1 localization.

Wild-type gonads depleted of ATX-2 are masculinized and produce only sperm,while fem-1 mutant gonads depleted of ATX-2 are feminized and produce only oocytes. This suggests that ATX-2 is not essential for spermatogenesis or oogenesis per se, but rather for the switch from spermatogenesis to oogenesis. FEM-3 and TRA-2 are key regulators of this switch and promote spermatogenesis and oogenesis, respectively. The protein levels of FEM-3 and TRA-2 in the germline are under tight translational control(Goodwin and Ellis, 2002). During the last larval stage (L4), translation of tra-2 mRNA is repressed (TRA-2 Off); this allows FEM-3 protein to promote spermatogenesis so the L4 germline has a `male' identity. In adults, the germline switches into a`female' mode; this switch requires repression of fem-3 mRNA and translation of TRA-2 protein (TRA-2 On). Thus, the masculinization of ATX-2-depleted germlines could result from the failure to repress FEM-3, or from the failure to express TRA-2. Alternatively, the masculinization of ATX-2-depleted germlines could result from expression of FEM-3 and TRA-2 simultaneously, as masculinizing activity of FEM-3 is dominant over feminizing activity of TRA-2 (Barton et al.,1987; Schedl and Kimble,1988).

Translational repression of tra-2 mRNA and fem-3 mRNA is mediated by distinct proteins. Genetic and biochemical data suggest that GLD-1 participates in translational repression of tra-2 mRNA by binding two sequences in the 3′UTR called TGEs(Goodwin et al., 1993; Jan et al., 1999). To see if masculinization of atx-2(RNAi) germline could be caused by a failure in translating TRA-2, we used tra-2(q122), a weak gain-of-function mutation with a deletion of one of the two TGE sequences(Goodwin et al., 1993; Schedl and Kimble, 1988). If tra-2 mRNA was present but translationally repressed in atx-2(RNAi) animals, the tra-2(q122gf) mutation should allow TRA-2 protein expression, thus promoting oogenesis at the expense of spermatogenesis. Indeed, we found that the gonads of tra-2(q122gf);atx-2(RNAi) animals contained oocyte-like germ cells(compare Fig. 3B,D), and did not contain sperm (Fig. 3A,C). Because the germline is fully feminized upon forced translation of TRA-2 in tra-2(q122gf);atx-2(RNAi) gonads, FEM-3 is likely to be repressed in these gonads. This observation suggests that the masculinization of ATX-2-depleted germlines represents a failure in TRA-2 translation.

In the normal hermaphrodite gonad, the transient repression of tra-2 mRNA requires both GLD-1 and FOG-2, an F-box protein that associates with GLD-1 (Clifford et al.,2000). In mutants with masculinized germlines, however, removal of GLD-1 is not sufficient to de-repress TRA-2 and to allow oogenesis(Francis et al., 1995b); we found similarly that the gld-1;atx-2(RNAi) germlines did not contain oocytes (Fig. 3E, enlarged in F). By contrast, depletion of FOG-2 has been shown to restore oogenesis in hermaphrodites with masculinized germline(Schedl and Kimble, 1988). Similarly, we found that most fog-2;atx-2(RNAi) animals produced oocytes (compare Fig. 3G,H,enlarged in I). Together, our results suggest that atx-2(RNAi)germlines are masculinized because translation of tra-2 mRNA is inappropriately `locked' in the repressed state.

In addition to roles in the regulation of glp-1 mRNA and tra-2 mRNA, GLD-1 has been shown to be involved in the entry of germ cells into pachytene, although the mRNA target(s) for this process have not been identified. Genetic studies have shown that the requirement for GLD-1 in pachytene entry is redundant with GLD-2, a poly(A)-polymerase(Kadyk and Kimble, 1998). In wild-type gonads, the presence of pachytene nuclei can be determined by the distinctive thread-like appearance of DAPI-stained chromosomes(Francis et al., 1995a). Although gld-1 or gld-2 single mutants contain pachytene nuclei (Francis et al.,1995a), no pachytene nuclei are present in gld-1;gld-2double mutants (Kadyk and Kimble,1998).

To test if GLD-1 was able to promote pachytene in gld-2 mutants depleted of ATX-2, we asked if gld-2;atx-2(RNAi) germlines contained pachytene nuclei by DAPI. We also examined expression of HIM-3, a core component of meiotic chromosomes (Fig. 4A) (Zetka et al.,1999). In control experiments, we found that pachytene nuclei were present in the gonads of atx-2(RNAi), gld-2/+;atx-2(RNAi),gld-1;atx-2(RNAi) and gld-2 adults(Fig. 4B-D; data not shown)(Hansen et al., 2004). By contrast, we found that atx-2;gld-2 gonads did not contain any pachytene nuclei. Germ cell nuclei in atx-2;gld-2 gonads did not resemble pachytene nuclei when stained by DAPI(Fig. 4E), and contained only low levels of HIM-3 that did not associate with chromosomes(Fig. 4F). This result is consistent with the hypothesis that the function(s) of GLD-1 in promoting pachytene entry is compromised in the absence of ATX-2.

GLD-1 has been shown to be directly required for the translational repression of rme-2 mRNA (Lee and Schedl, 2001). RME-2 is a yolk receptor that is expressed only in late pachytene germ cells and in maturing oocytes in the proximal gonad(Fig. 5A)(Grant and Hirsh, 1999; Lee and Schedl, 2001). In gld-1 mutants, RME-2 expression extends throughout the pachytene region (Lee and Schedl, 2001). Because GLD-1 represses expression of RME-2 in the pachytene germ cells destined to become oocytes, we examined RME-2 expression in feminized fem-1;atx-2(RNAi) animals. We found two abnormalities. First, fem-1;atx-2(RNAi) gonads expressed RME-2 throughout the pachytene,medial region of the gonad (Fig. 5B). Because GLD-1 was present in the pachytene region of fem-1;atx-2(RNAi) gonads (Fig. 2G), this suggests that GLD-1 is unable to repress rme-2translation in the absence of ATX-2.

The second abnormality in fem-1:atx-2(RNAi) animals was the presence of RME-2 in the distalmost gonad(Fig. 5B). This distalmost RME-2 also was present in masculinized atx-2(RNAi) gonads(Fig. 5D) and male atx-2(RNAi) gonads (data not shown). Because rme-2 mRNA has not been reported to be present in the distalmost region of wild-type hermaphrodite or male gonads, we addressed this issue using in situ hybridization. We often detected a very low level of staining with probes for rme-2 mRNA in the distalmost regions of wild-type hermaphrodite gonads, but never in the distalmost regions of control rme-2(RNAi)gonads (data not shown). Because the gonad is a syncytium, we do not know whether rme-2 mRNA is transcribed by the distalmost germ nuclei, or originates from more proximal nuclei. GLD-1 is not thought to repress RME-2 in the distalmost hermaphrodite gonad or in the male gonad. GLD-1-depleted hermaphrodites do not express RME-2 in the distalmost gonad(Lee and Schedl, 2001), and RME-2 is not expressed in GLD-1-depleted males (data not shown), even though low levels of rme-2 mRNA are present in the male germline (T. Schedl,personal communication). These observations suggest that the distalmost germ cells, irrespective of the sexual identity of the germline, have a GLD-1-independent mechanism to prevent expression of RME-2, and that this mechanism requires ATX-2.

A possible candidate for a role in the distalmost repression of RME-2 is the KH-domain protein MEX-3. MEX-3 has been shown to repress translation of pal-1 mRNA in the proximal germline and in the embryo(Draper et al., 1996; Hunter et al., 1996). Interestingly, MEX-3 also is present at low levels in the distalmost part of the germline where its function has not been investigated(Fig. 5E) (B. Draper and J.P.,unpublished). We examined RME-2 expression in mex-3 mutants and in animals depleted of MEX-3 by RNAi. In both hermaphrodites and males, we found that RME-2 was expressed inappropriately in the distalmost region of the gonad(Fig. 5F) (data not shown). We then examined rme-2 mRNA expression by in situ hybridization, and could detect no differences in the levels of distalmost rme-2 mRNA between mex-3 hermaphrodites and wild-type hermaphrodites(Fig. 5G). Thus, MEX-3 appears to have a role in regulating rme-2 mRNA translation, rather than transcription, in the distalmost gonad. Because the distalmost gonad consists of mitotic germ cells, we would predict that glp-1(gf) gonads that consist entirely of mitotic cells (Berry et al., 1997) should have widespread MEX-3 expression and lack RME-2 expression. As expected, we found that glp-1(gf) gonads expressed MEX-3 but not RME-2 (data not shown). Removing MEX-3 from these gonads should allow expression of RME-2; indeed, we found that RME-2 was widely expressed in the gonads of glp-1(gf);mex-3(RNAi) animals(Fig. 5H,I).

Because MEX-3 is expressed in the distalmost region of the wild-type gonad,where GLD-1 is scarce, but not in more proximal regions that contain high levels of GLD-1, we considered the possibility that GLD-1 might regulate MEX-3 expression, or that MEX-3 might regulate GLD-1 expression. We found that gld-1 animals expressed MEX-3 throughout the gonad(Fig. 6B). By contrast, GLD-1 expression appeared normal in mex-3 mutant gonads (data not shown). These results suggest that GLD-1 activity in the medial, pachytene region of a wild-type gonad prevents expression of MEX-3, thus limiting MEX-3 to the distalmost region of the gonad (Fig. 6E). Although we do not know if GLD-1 directly binds to and represses translation of mex-3 mRNA, we found several possible GLD-1-binding sites in the 3′UTR of mex-3(Ryder et al., 2004) (data not shown).

Because depletion of ATX-2 results in abnormally high levels of distalmost GLD-1, we examined whether depleting ATX-2 affected the expression of MEX-3. Indeed, the distalmost levels of MEX-3 in atx-2(RNAi),fem-1;atx-2(RNAi) and gld-1/+;atx-2(RNAi) animals were reduced compared with wild type (compare Fig. 6A,C; data not shown). GLD-1 activity appeared to be responsible for the reduced levels of MEX-3, because gld-1;atx-2(RNAi) animals had approximately wild-type levels of MEX-3 in the distalmost germline(Fig. 6D). Thus, in the absence of ATX-2, GLD-1 may repress expression of MEX-3 in the distalmost germline,allowing RME-2 to be expressed inappropriately. However, the ability of MEX-3 to repress the expression of RME-2 may also be defective in the absence of ATX-2. gld-1;atx-2(RNAi) germlines have a relatively normal distal mitotic zone that appears to contain wild-type levels of MEX-3(Fig. 6D). Nevertheless, we found that RME-2 is expressed in this zone(Fig. 6F). Using in situ hybridization, we found that this distal expression of RME-2 was not due to increased levels of rme-2 mRNA(Fig. 6G). In a separate experiment, we found that MEX-3 is present, but not able to repress expression of RME-2, in the fully mitotic germline of glp-1(gf) animals depleted of ATX-2 (Fig. 6H-J). These results suggest that ATX-2 is required for the MEX-3-dependent repression of rme-2 translation in the distalmost, mitotic part of the germline. Taken together, RME-2 appears to be inappropriately expressed in the distalmost germline of ATX-2-depleted animals for two reasons. First,expression of the translational repressor MEX-3 is down-regulated, owing to high levels/increased activity of GLD-1 in the distalmost gonad. Second, MEX-3 itself cannot repress translation of RME-2.

In this report, we have shown that depletion of ATX-2 causes diverse defects in the development of the C. elegans germline (summarized in Table 1). Nevertheless, germ cells depleted of ATX-2 can show normal levels of proliferation[gld-1;atx-2(RNAi) animals], and can differentiate into either sperm[atx-2(RNAi) animals] or oocytes [fem-1;atx-2(RNAi)animals]. Instead, ATX-2-depleted germlines appear defective in the spatial and temporal regulation of these events. For example, GLP-1 is expressed in the distal wild-type gonad, but this expression is strongly reduced in ATX-2-depleted gonads. Wild-type hermaphrodite gonads normally switch from spermatogenesis to oogenesis. By contrast, ATX-2-depleted animals appear to be locked in spermatogenesis, apparently owing to the continuous repression of tra-2 mRNA translation.

Many of the defects observed in atx-2(RNAi) animals appear to involve processes that normally are controlled by the translational repressor GLD-1. In addition, our analysis of RME-2 expression in glp-1(gf);atx-2(RNAi) and gld-1;atx-2(RNAi) gonads suggests that at least one process involving the translational repressor MEX-3 is abnormal in the absence of ATX-2. Although it is possible that depletion of ATX-2 affects each of the described processes indirectly, an intriguing possibility is that ATX-2 normally has a role in GLD-1- and MEX-3-mediated translational regulation.

Some of the atx-2(RNAi) phenotypes suggest a simple loss of translational regulation. First, MEX-3 is expressed throughout glp-1(gf)atx-2(RNAi) gonads, but is unable to repress the expression of RME-2. Second, expression of GLD-1 and RME-2 in the pachytene germ cells of ATX-2-depleted feminized gonads [fem-1;atx-2(RNAi)animals] suggests a loss of GLD-1-dependent repression. Third, the absence of pachytene entry in gld-2;atx-2(RNAi) double mutants is consistent with a loss of GLD-1-mediated repression.

Other phenotypes of atx-2(RNAi) animals are more complex, and suggest an alteration in, rather than loss of, GLD-1-mediated repression. Although reduced expression of GLP-1 in atx-2(RNAi) gonads may be explained by the abnormally high levels of distalmost GLD-1, GLP-1 expression is also reduced in feminized fem-1;atx-2(RNAi) animals that contain only low levels of GLD-1 in the distalmost gonad. The repression of tra-2 mRNA also appears abnormal in atx-2(RNAi) gonads; tra-2 mRNA is `locked' in the repressed state. Our genetic experiments suggest that repression of tra-2 mRNA in atx-2(RNAi) gonads occurs through a mechanism that requires TGE elements in the 3′UTR of tra-2 mRNA, as in wild-type development, and that at least some proteins that contribute to normal TGE-mediated repression appear to be functioning in atx-2(RNAi)gonads. First, removal of the GLD-1-binding partner FOG-2 promotes oogenesis in atx-2(RNAi) gonads, as it does in otherwise wild-type gonads. Second, a semi-dominant mutant form of GLD-1 encoded by the gld-1(q361) allele that has weak feminizing effect on the germline(Francis et al., 1995a)markedly reduced spermatogenesis in atx-2(RNAi) germlines (R.C.,M.D.P. and J.P., unpublished). GLD-1 has been shown to be part of a larger complex called DRF that associates with the TGE elements(Goodwin et al., 1993; Jan et al., 1999). Because GLD-1 produced from the gld-1(q361) allele is unable to bind TGEs(Jan et al., 1999), the reduction of spermatogenesis in gld-1(q361);atx-2(RNAi) germlines could reflect the formation of DRF with decreased ability to bind tra-2; this would lead to expression of TRA-2 and a decrease in spermatogenesis. One interpretation of the above results is that an aberrant DRF complex assembles on the tra-2 3′UTR in the absence of ATX-2, locking tra-2 mRNA in the repressed state. Alternatively, DRF might be normal, but some feature of tra-2 mRNA (or its associated proteins) could be altered by the absence of ATX-2, resulting in an abnormal response to repressor binding.

The distalmost gonad can be considered a stem cell niche, where germ nuclei actively divide but do not commit to either meiosis or differentiation. We have shown that MEX-3 is required to prevent these mitotic germ cells from expressing RME-2. The observation that rme-2 mRNA levels in the distal gonad appear identical in wild-type gonads and mex-3 mutant gonads suggests that MEX-3 does not regulate transcription of the rme-2 gene. Instead, we favor the hypothesis that MEX-3 directly or indirectly regulates rme-2 mRNA translation, analogous to its role in regulating translation of pal-1 mRNA in the embryo. By preventing the expression of proteins such as RME-2, that are characteristic of differentiating oocytes, MEX-3 may contribute to the totipotency of the distal germ cells.

The ataxin 2-binding protein A2BP1 is an ortholog of an RNA-binding protein from C. elegans called SPN-4. Intriguingly, SPN-4 has been recently shown to associate with, and promote translation of, glp-1 mRNA in the early embryo (Ogura et al.,2003). Thus, both ATX-2 and SPN-4 appear to function in translational regulation, suggesting that the human counterparts may also have roles in translation. Moreover, SPN-4, ATX-2 and GLD-1 activities merge on the same glp-1 mRNA to regulate translation. Although we do not know if glp-1 translation in the embryo requires ATX-2, those observations suggest that translation of some mRNAs might be regulated by a conserved set of proteins: GLD-1/SPN-4/ATX-2 in worms, and a GLD-1 ortholog/A2BP1/ataxin 2 in mammals. The mouse orthologs of GLD-1, called QKI-5, -6, and QKI-7, are different splice variants encoded by the quaking (qk) gene. QKI proteins are known for their function in stability and export of mRNAs encoding proteins, such as myelin basic protein (MBP), that are required for myelination in the central and peripheral nervous systems(Vernet and Artzt, 1997; Larocque et al., 2002). Interestingly, QKI-6 was shown in vivo to functionally substitute for GLD-1 in translational repression of tra-2 mRNA(Saccomanno et al., 1999). This result led to the hypothesis that QKI proteins might function as repressors of MBP translation (Larocque et al., 2002). If true, it would be interesting to test if translational regulation of MBP required ataxin 2 and A2BP1 proteins.

Our studies provide the first in vivo evidence that an ataxin 2-related protein forms a complex with a poly(A)-binding protein (PABP). Because two-hybrid analysis showed that an ataxin-2-related protein from lower eukaryotes (Pbp1p) interacts with PABP, the association with poly(A)-binding proteins may be a conserved feature of ataxin-2 family members. PABPs influence many aspects of mRNA metabolism such as mRNA translation, transport and stability (Mangus et al.,2003). We showed that depletion of PAB-1, the binding partner of ATX-2, had much more severe consequences for germline development than had depletion of ATX-2, suggesting that ATX-2 does not play an obligate role in all PABP activities, and may instead have specialized functions.

A common role of PABPs is to promote polyadenylation by stabilizing the poly(A) tail. The length of the poly(A) tail plays an important role in regulating translation of many mRNAs(Richter, 2000). Interestingly, the presence of a poly(A) tail is essential for TGE-mediated translational repression in Xenopus(Thompson et al., 2000). However, there is no evidence that the length of the rme-2 or tra-2 mRNA poly(A) tail plays a role in GLD-1-mediated repression(Lee and Schedl, 2001). In addition to binding the poly(A) tail, PABPs throughout the animal kingdom have been shown to bind eIF4G, a component of the 5′ cap-binding translation initiation factor eIF4F that also contains eIF4A and eIF4E(Gingras et al., 1999; Sachs, 2000). This binding is thought to result in bridging the 5′ end of mRNA with its 3′poly(A) end. Such a `closed loop' conformation is thought to facilitate translation. In mammals, a protein related to eIF4G called Paip1 mimics eIF4G binding to both eIF4A and PABP, and enhances translation(Craig et al., 1998; Roy et al., 2002). Another,unrelated PABP-binding protein called Paip2 competes with Paip1 for the binding to PABP, and is able to prevent PABP association with the poly(A) tail(Khaleghpour et al., 2001a; Khaleghpour et al., 2001b). Thus, Paip1 and Paip2 have set precedence for proteins that are not part of the core translational machinery yet influence translation through their association with PABP. Because ATX-2 interacts with PAB-1 and affects translational regulation, an interesting possibility is that ATX-2 might alter the ability of PABP to interact with components of the core translational machinery. This could influence the architecture of the mRNA, for example the conformation of mRNA or composition of its associated translational factors,which might result in altered response to the UTR-binding repressors such as GLD-1 or MEX-3.

We thank Tim Schedl for strains and valuable comments; Phil Gafken for mass-spectrometry; Liz Wayner for antibody production; Barth Grant, Tim Schedl and Monique Zetka for antibodies/antisera; Jeremy Nance, Barbara Page, Susan Parkhurst and Uta Wolke for discussions, experimental advice and comments on the manuscript. Some strains were provided by the Caenorhabditis Genetics Center funded by the NIH. R.C. was supported by a long-term fellowship from the Human Frontier Science Program, and Howard Hughes Medical Institute supports R.C., M.D. and J.P.