The Caenorhabditis elegans sex determination gene, tra‐2, is translationally regulated by elements in the 3′‐untranslated region called TGEs. TGEs govern the translation of mRNAs in both invertebrates and vertebrates, indicating that this is a highly conserved mechanism for controlling gene activity. A factor called DRF, found in worm extracts binds the TGEs and may be a repressor of translation. Using the yeast three‐hybrid screen and RNA gel shift analysis, we have found that the protein GLD‐1, a germline‐specific protein and a member of the STAR family of RNA‐binding proteins, specifically binds to the TGEs. GLD‐1 is essential for oogenesis, and is also necessary for spermatogenesis and inhibition of germ cell proliferation. Several lines of evidence demonstrate that GLD‐1 is a translational repressor acting through the TGEs to repress tra‐2 translation. GLD‐1 can repress the translation of reporter RNAs via the TGEs both in vitro and in vivo, and is required to maintain low TRA‐2A protein levels in the germline. Genetic analysis indicates that GLD‐1 acts upstream of the TGE control. Finally, we show that endogenous GLD‐1 is a component of DRF. The conservation of the TGE control and the STAR family suggests that at least a subset of STAR proteins may work through the TGEs to control translation.
The precise temporal and spatial expression of key regulatory genes is crucial for normal development. It is now apparent that translational control by elements in the 3′‐untranslated region (3′‐UTR) play major roles in regulating developmentally important genes (Wickens et al., 1996). For example, elements in the 3′‐UTR of the Drosophila hunchback and oskar mRNAs are necessary for repressing translation and hence controlling anterior–posterior axis formation (Wharton and Struhl, 1991; Kim‐Ha et al., 1995; Rongo et al., 1995). While many 3′‐UTR cis‐acting elements are known, only a few trans‐acting factors have been identified. As a result, the mechanisms underlying 3′‐UTR translational controls are poorly understood. To comprehend these mechanisms better, we sought trans‐acting factors that interact with the translational regulatory elements in the 3′‐UTR of the Caenorhabditis elegans sex‐determining gene, tra‐2.
Caenorhabditis elegans has two sexes: hermaphrodite and male. Hermaphrodites are essentially female in the soma but make both sperm and oocytes in the germline. The primary signal for sex determination is the ratio of the number of X chromosomes to sets of autosomes, such that animals with two X chromosomes (XX) develop as hermaphrodites while animals with a single X chromosome (XO) develop as males (for a review see Meyer, 1997). The X to autosomal ratio controls the activity of a number of genes that act in a cascade to regulate sexual cell identity (Figure 1; see Meyer, 1997).
The sex‐determining gene, tra‐2, is required for female cell fates (Hodgkin and Brenner, 1977). tra‐2 is predicted to encode a large transmembrane protein, called TRA‐2A, that is necessary to inhibit downstream male determinants (Kuwabara et al., 1992). TRA‐2A is thought to be part of a signal transduction pathway that is important in ensuring that all the cells in an animal adopt the same sexual fate (Kuwabara et al., 1992). In males, TRA‐2A activity is low and male development ensues (Hodgkin, 1980).
Proper male development requires that tra‐2 activity is repressed. Dominant gain‐of‐function (gf) mutations have been identified that result in excessive tra‐2 activity causing inappropriate female development in both XX and XO animals. XX animals develop as females (they make no sperm) and XO animals produce oocytes in the germline and yolk in the intestine (Doniach, 1986). The gf mutations all map to a direct repeat located in the tra‐2 3′‐UTR (Goodwin et al., 1993). This direct repeat consists of two regulatory elements, called TGEs (for tra‐2 and GLI elements), which control tra‐2 activity by repressing the translation of the tra‐2 mRNA (Goodwin et al., 1993; Jan et al., 1997). In addition, the TGEs bind a factor, called DRF (direct repeat factor), that is present in crude worm extracts. Previous analyses suggest that DRF is a repressor of translation (Goodwin et al., 1993; Jan et al., 1997). Our working model is that the binding of DRF to the TGEs results in translational repression of tra‐2.
TGEs control translation not only in C.elegans, but also in the nematode Caenorhabditis briggsae and mammalian cells. Furthermore, the translation of at least three genes [C.briggsae tra‐2, human GLI (Jan et al., 1997) and C.elegans tra‐1 (E.Jan, Y.Yoo and E.B.Goodwin, unpublished results)] is controlled by TGEs. These results indicate the TGE control is a conserved mechanism that may regulate the translation of a number of mRNAs.
To explore further the mechanism of how TGEs control sexual development by regulating the translation of tra‐2, we screened for TGE‐binding factors using the yeast three‐hybrid system. We found that the protein GLD‐1 (defective in germline development) specifically binds to the TGEs. GLD‐1 is a member of the STAR (signal transduction and activation of RNA) family of RNA‐binding proteins which are present in both invertebrates and vertebrates and are essential for many developmental decisions (Vernet and Artzt, 1997). The RNA targets of STAR proteins and how STAR proteins regulate RNA activity are poorly understood. Here, we show that GLD‐1 is a translational repressor that acts through the TGEs to inhibit tra‐2 translation. The finding that the TGE control is a conserved mechanism raises the possibility that other STAR family members may act via TGEs to regulate translation.
Identification of GLD‐1 as a TGE‐binding factor
To understand better the mechanism of the TGE control, we sought factors that bind to the C.elegans tra‐2 TGEs. Recently, a yeast three‐hybrid screen was developed to identify RNA‐binding proteins (SenGupta et al., 1996). The three‐hybrid system selects for proteins that bind to specific RNA sequences. A diagram of the yeast three‐hybrid screen is shown in Figure 2A. Briefly, a hybrid RNA is expressed that contains the MS2 coat protein‐binding site, fused to an RNA ‘bait’, in our case the tra‐2 TGEs. For the ‘bait’, both tra‐2 TGEs arranged in tandem were used, which is a total of 60 nucleotides (see Materials and methods for sequences). This arrangement is precisely how the TGEs are found in the tra‐2(+) 3′‐UTR (Goodwin et al., 1993). A fusion protein consisting of the MS2 coat protein and the LexA DNA‐binding domain anchors the hybrid RNA to the promoter of either the lacZ or HIS3 reporter gene. The binding of a protein expressed from the cDNA library to the TGEs results in the formation of a tripartite complex that activates the transcription of the reporter genes.
We used a C.elegans cDNA library to screen for proteins that interact with the tra‐2 TGEs. The library was transfected into yeast, and colonies that expressed β‐gal and that grew on plates lacking histidine were selected. From a screen of 6×105 transformants, 87 positive colonies were isolated. Of these 87 positives, 20 were dependent on the presence of the hybrid MS2–tra‐2 TGE RNA. From these 20, we screened for clones that specifically required the TGEs to activate lacZ and HIS3 reporter genes. Toward this end, we tested the ability of the 20 clones to activate transcription of lacZ and HIS3 when the hybrid RNA contains the TGEs but not when it contains other 3′‐UTR elements. Four target hybrid RNAs were used: the MS2‐binding site alone or the MS2‐binding site fused to the TGEs, to a poly(A)30 or to an iron response element (IRE) (Figure 2B). Of the 20 positives, only one activated transcription when the hybrid bait contained the TGEs but failed to activate transcription when other RNA baits were used (Figure 2B).
Sequence analysis of the single positive clone revealed that it coded for the protein GLD‐1. GLD‐1 is a germline‐specific cytoplasmic protein (Jones and Schedl, 1995), and is part of a family of RNA‐binding proteins called the STAR family (for review see Vernet and Artzt, 1997). The STAR proteins are thought to link signal transduction pathways and RNA metabolism. The hallmarks of the STAR family are a single KH RNA‐binding domain and conserved QUA1 and QUA2 domains. The STAR family includes the murine and human SAM68 (Darnell et al., 1994; Fumagalli et al., 1994; Taylor and Shalloway, 1994) and SF1 (Kramer, 1992; Toda et al., 1994; Agger and Freimuth, 1995; Arning et al., 1996), the murine, Xenopus, Zebrafish and human QUAKINGs (Ebersole et al., 1996; Vernet and Artzt, 1997; Zorn et al., 1997) and the Drosophila HOW/WHO proteins (Baehrecke, 1997; Zaffran et al., 1997). STAR proteins play important roles in a number of developmental events. They are necessary for embryogenesis and myelination in mice, as well as notochord differentiation in Xenopus embryos and muscle development in Drosophila (Ebersole et al., 1996; Zaffran et al., 1997; Zorn and Krieg, 1997). How STAR proteins perform these roles is still poorly understood.
GLD‐1 has multiple roles in germline development (Francis et al., 1995a). GLD‐1 is essential for oogenesis. In gld‐1. lf) animals, the oocyte germline fails to progress through meiosis and re‐enters mitosis, resulting in overproliferation of germline cells and consequently a tumorous germline (Francis et al., 1995a). GLD‐1 has non‐essential roles in germline proliferation and sex determination (Francis et al., 1995a,b). With regard to sex determination, GLD‐1 is necessary for hermaphrodite spermatogenesis. gld‐1. lf) XX animals make few or no sperm. We hypothesize that spermatogenesis results from the repression of tra‐2 translation by GLD‐1.
GLD‐1 interacts specifically with TGE
To test whether GLD‐1 directly interacts with the C.elegans tra‐2 TGEs, we used RNA gel mobility assays and asked whether purified bacterially expressed GST–GLD‐1 fusion protein bound to the tra‐2 TGEs. Incubation of GST–GLD‐1 with RNA containing the wild‐type tra‐2(+) 3′‐UTR resulted in a slower migrating band, indicating complex formation (Figure 3A, compare lane 1 with 2–5; Figure 3C). GST–GLD‐1 and the C.elegans tra‐2 3′‐UTR RNA had a binding constant of ∼500 nM. Complex binding was not due to GST since GST alone did not bind RNA (data not shown). The binding of the tra‐2 3′‐UTR to GLD‐1 was dependent upon the TGEs since GST–GLD‐1 bound only weakly to mutant tra‐2 3′‐UTRs in which the TGEs were deleted (Figure 3A, compare lane 6 with 7–9; Figure 3C). We were unable to saturate binding to the mutant tra‐2 3′‐UTR RNAs, indicating that the binding constant is much greater than 500 nM and is probably due to non‐specific binding. We also performed competition experiments and found that the tra‐2(+) 3′‐UTR but not mutant 3′‐UTRs competed for GLD‐1 binding to the tra‐2(+) 3′‐UTR (data not shown). The broadness of the GLD‐1 shift with the wild‐type tra‐2(+) 3′‐UTR may be due to oligomerization of the GLD‐1 as previous studies have shown that GLD‐1 can self‐associate (Chen et al., 1997).
To explore further the binding specificity of GLD‐1, we examined the ability of GLD‐1 to bind small RNAs containing just the TGEs (Figure 3B and summarized in C). Similarly to the full‐length tra‐2 3′‐UTRs, radiolabeled RNAs containing just the TGEs (EBG‐9) bound GST–GLD‐1 but RNAs in which the TGEs had been deleted (EBG‐11) did not (Figure 3C). Previously, we identified functional TGEs in the 3′‐UTR of the C.briggsae tra‐2 and human GLl mRNAs (Jan et al., 1997). One would predict that GLD‐1 should also bind these elements. Indeed, we found that GST–GLD‐1 specifically associated with small RNAs containing the C.briggsae tra‐2 (EJ‐19) and GLI (EJ‐38) TGEs (Figure 3B, lanes 1 and 2, and data not shown; Figure 3C), but did not form a complex with RNAs that contained a mutant C.briggsae tra‐2 TGE (EJ‐32 and EJ‐35, Figure 3B, lanes 2–6). EJ‐35 contains a six nucleotide deletion within the 31 nucleotide C.briggsae tra‐2 TGE, and EJ‐32 carries the same six nucleotide deletion as well as three base substitutions (see Materials and methods for sequences). In conclusion, GLD‐1 binds specifically to TGEs.
If GLD‐1 regulates tra‐2 activity by the TGEs, then previously identified mutations in GLD‐1, which dramatically reduce GLD‐1 activity, may disrupt the ability of GLD‐1 to bind TGEs. Previously, Schedl and colleagues identified a point mutation, called GLD‐1(q361) (Gly227 to Asp), in the KH domain that results in a strong loss‐of‐function phenotype (Jones and Schedl, 1995). gld‐1. q361) homozygous animals do not produce sperm, and gld‐1. q361)/+ heterozygous animals have a semi‐dominant germline phenotype where some of the animals make only oocytes (Francis et al., 1995a). To test whether this point mutation altered the ability of GLD‐1 to bind the TGEs, we asked whether GST–GLD‐1(q361) mutant protein bound TGEs in a gel mobility shift assay. GST–GLD‐1(q361) was not able to form a complex with the wild‐type tra‐2 3′‐UTR (Figure 3D, compare lanes 2 and 3). These results show that the loss‐of‐function phenotype of GLD‐1 correlates with loss of GLD‐1 binding to the TGEs, supporting the idea that TGE binding is required for GLD‐1 function.
GLD‐1 is a component of DRF
Previous analyses suggest that DRF is a repressor of tra‐2 translation (Goodwin et al., 1993). Hence, one would predict that GLD‐1 should be a component of DRF. To address this, we compared the migration of DRF in an RNA gel mobility assay in the presence and absence of GLD‐1 antibody. Often DRF is a doublet, possibly indicating that it consists of multiple factors. Addition of GLD‐1 antibody resulted in a reduction of DRF binding (Figure 4A, compare lanes 2 with 3). GLD‐1 is probably a component of both complexes, since addition of antibody reduces both. Pre‐absorbed GLD‐1 antibody and an antibody to GST do not significantly affect the mobility of the DRF–tra‐2 3′‐UTR complex, indicating that the inhibition by GLD‐1 antibody is specific (Figure 4A, lanes 4 and 5). The antibody results support the idea that GLD‐1 is a component of DRF.
GLD‐1 represses tra‐2 activity via the TGEs in vivo
If GLD‐1 is important in regulating translation, then it should control the activity of mRNAs that contain TGEs in vivo. To address this, we asked whether the expression of GLD‐1 could inhibit the activity of reporter transgenes that carried TGEs. Presently, it is not possible to assay transgenes in the germline of C.elegans. Consequently, we performed this analysis by ectopically expressing GLD‐1 in the soma. To express GLD‐1 in the soma, a construct containing the heat shock promotor (hsp16‐41) fused to the entire GLD‐1‐coding region (hsp::GLD‐1) was made. Four reporter transgenes were used: all coded for the lacZ gene and contained either the wild‐type tra‐2 3′‐UTR [lacZ::tra‐2(+)3′UTR], a mutant tra‐2 3′‐UTR in which one [lacZ::tra‐2(−32)3′UTR] or both TGEs [lacZ::tra‐2(−60)3′UTR] were removed, or a 108 nucleotide deletion [lacZ::tra‐2(−108)3′UTR] that removes the TGEs plus flanking sequences. The use of the lacZ::tra‐2(−32)3′UTR transgene is a particularly sensitive assay for regulation, since a single TGE is able to partially repress translation (Goodwin et al., 1997). The transgenes were controlled by the inducible heat shock promoter (hsp16‐41). Transgenic animals carrying hsp::GLD‐1 and either lacZ::tra‐2(+)3′UTR?, lacZ::tra‐2(−32)3′UTR, lacZ::tra‐2(−60)3′UTR or lacZ::tra‐2(−108)3′UTR transgenes were heat shocked and the percentage of transgenic animals with intestinal β‐gal staining were scored.
We found that ectopic expression of GLD‐1 in animals carrying TGEs resulted in a dramatic decrease in intestinal β‐gal staining. In the absence of GLD‐1, 7 and 59% of transgenic animals carrying lacZ::tra‐2(+)3′UTR and lacZ::tra‐2(−32)3′UTR, respectively, had β‐gal staining in intestinal cells (Figure 5A, Table I). In contrast, when GLD‐1 was expressed in the soma, 0% of lacZ::tra‐2(+)3′UTR and only 18% of lacZ::tra‐2(−32)3′UTR transgenic animals had intestinal β‐gal staining (Figure 5B, Table I). Ectopic expression of GLD‐1 had little or no effect on the β‐gal expression of lacZ::tra‐2(−60)3′UTR and lacZ::tra‐2(−108)3′UTR (Figure 5C and D, Table I).
If GLD‐1 represses tra‐2 translation, then expression of GLD‐1 in the soma may result in masculinization of somatic structures. We analyzed the phenotype of animals that carried hsp::GLD‐1 that were heat shocked once a day from embryogenesis to adults. Approximately 12% of animals in which GLD‐1 was expressed in the soma had truncated tails, indicative of somatic masculinization (Figure 6B, Table II). However, no masculinization was detected when the mutant GLD‐1(q361) protein was expressed (Table II). These results correlate with GLD‐1 repressing tra‐2 activity.
If somatic GLD‐1 expression is promoting male development by repressing tra‐2, then GLD‐1 should act upstream of the tra‐2 TGE control in a genetic hierarchy. Toward this end, we overexpressed GLD‐1 using the hsp::GLD‐1 construct in XX and XO animals that were loss of function for her‐1 or fem‐3 (Figure 1, Table II). her‐1 and fem‐3 are required for male development, and loss of their activities causes XO animals to be feminized (Meyer, 1997). Genetic analysis indicates that tra‐2 3′‐UTR regulation acts downstream of her‐1 and upstream of fem‐3 to regulate sexual identity (see Figure 1). If GLD‐1 is repressing tra‐2 translation, then overexpression of GLD‐1 would masculinize her‐1. lf) animals but would not affect fem‐3. lf) animals. We found that similarly to wild‐type animals, 23% of XX and XO her‐1. lf) animals in which GLD‐1 was overexpressed developed truncated tails (Table II). In contrast, GLD‐1 expression did not alter the sexual development of the somas of XX fem‐3. lf) animals (Table II). These data are consistent with GLD‐1 functioning between her‐1 and fem‐3 in a genetic hierarchy to inhibit male development.
As discussed above, GLD‐1 is required for hermaphrodite spermatogenesis. XX animals that lack GLD‐1 activity make no or a few sperm (Francis et al., 1995a). In contrast, a subset of gld‐1. gf) mutations, called gld‐1. mog) (masculinization of the germline), cause XX animals to make only sperm and no oocytes (Francis et al., 1995a). It is possible that the Mog phenotype results from an increase in the ability of GLD‐1 to repress tra‐2 translation in the germline. To address this possibility, we used double mutant analysis to ask what is the germline phenotype of XX gld‐1. mog); tra‐2(gf) double mutant animals. Earlier work had shown that gld‐1. q93mog); tra‐2. q122gf) animals made sperm, suggesting that GLD‐1 is not regulating spermatogenesis by the TGEs (Francis et al., 1995b). However, tra‐2. q122gf) still retains a single TGE that is capable of partially regulating translation (Goodwin et al., 1997). Therefore, we made the double mutant between gld‐1. q93Mog) and tra‐2. e2020gf). tra‐2. e2020gf) removes both TGEs and does not retain TGE control (Goodwin et al., 1993). The gld‐1. q93Mog); tra‐2. e2020gf) double mutant animals do not make sperm (n = 53), agreeing with GLD‐1 acting via the TGEs to control spermatogenesis. In conclusion, the observations that GLD‐1 requires fem‐3 activity to masculinize somatic tissue, and that gld‐1. q93Mog); tra‐2. e2020gf) animals do not make sperm, are consistent with GLD‐1 acting upstream of the TGEs to control sexual cell identity by repressing tra‐2 translation.
TRA‐2 expression is affected in a gld‐1(lf) background
To address whether GLD‐1 governs tra‐2 translation in vivo, we examined whether TRA‐2A protein levels were altered in a gld‐1. lf) background, using immunofluorescence analyses. TRA‐2A polyclonal antibodies that recognize the N‐terminal portion of the protein were used to compare levels of TRA‐2A in wild‐type and gld‐1(null) mutations. These antibodies detect TRA‐2A since mutations that dramatically reduce tra‐2 mRNA also greatly reduce the levels of TRA‐2A (L.Graves and E.B.Goodwin, unpublished results). One would predict that if GLD‐1 is controlling the translation of tra‐2 there would be an increase in TRA‐2A protein levels in gld‐1(null) animals as compared with wild‐type. To examine this, we used immunocytochemistry to compare the levels of TRA‐2A in the gonads of third larval stage (L3) worms in wild‐type and gld‐1(null) backgrounds. L3 gonads were used since at this time tra‐2 must be repressed to allow for spermatogenesis (Hodgkin, 1986). As predicted, there were higher levels of TRA‐2A in the gld‐1(null) animals as compared with wild‐type (Figure 7A). The increase in TRA‐2A levels is not due to the presence of abnormal germline, since the germline is normal at this time of development (Francis et al., 1995a). We also measured the amount of tra‐2 mRNA to ensure that the increased TRA‐2A was not due to an increase in mRNA levels. We found that tra‐2 mRNA levels are lower in the gonads of gld‐1. lf) animals as compared with wild‐type gonads (Figure 7B). The fact that lower tra‐2 mRNA levels are observed yet there is an increase in the amounts of TRA‐2A in the mutant backgrounds suggests that tra‐2 is translated more actively in gld‐1(null) backgrounds as compared with wild‐type, and that GLD‐1 binding is required to repress tra‐2 translation. The lower levels of tra‐2 mRNA may indicate a role for TGEs and GLD‐1 in stabilizing the mRNA, as deletions that remove the TGEs also result in decreased levels of tra‐2 mRNA (Goodwin et al., 1993).
GLD‐1 directly represses translation via the TGEs in vitro
We directly asked whether GLD‐1 represses translation via the TGEs by assaying whether purified GLD‐1 can inhibit translation in an in vitro translation assay. Using yeast extracts, we assayed the translation of reporter RNAs containing different 3′‐UTRs. Yeast extracts were used since translation in these extracts is sensitive to poly(A) tail length (Iizuka et al., 1994), and previously we found that TGEs may control translation by regulating the length of the poly(A) tail (Jan et al., 1997). The reporter RNAs encoded lacZ and carried wild‐type tra‐2 3′‐UTR [tra‐2(+)] or mutant tra‐2 3′‐UTRs in which one [tra‐2(−32)] or both TGEs plus some flanking sequences [tra‐2(−108)] were deleted. The RNAs were capped and contained a poly(A) tail of 30 A residues. The reporter RNAs were added to the yeast extract with or without a 70‐fold molar excess of purified GST–GLD‐1 protein and the expression of lacZ was quantitated at specific times. The addition of GST–GLD‐1 protein to reactions containing RNAs carrying tra‐2(+) or tra‐2(−32) 3′‐UTRs resulted in a decrease in lacZ expression (Figure 8A, squares, and data not shown). However, incubation of GLD‐1 protein with RNAs carrying tra‐2(−108) 3′‐UTR did not affect lacZ expression (Figure 8B, circles). Northern analysis showed that the addition of GLD‐1 did not significantly alter RNA levels during the course of the experiment (data not shown), indicating that differences in lacZ expression are due to differences in translation. The repression is specific to wild‐type GLD‐1 since addition of mutant GLD‐1(q361) protein did not repress translation (Figure 8A, triangles). Our data demonstrate that GLD‐1 represses translation via the TGEs in vitro.
Sexual development in C.elegans is controlled in part by the translational regulation of the sex determination gene tra‐2. The TGE control regulates the translation of multiple mRNAs and is present in both invertebrates and vertebrates (Jan et al., 1997). Here, we demonstrate that the STAR protein GLD‐1 controls sexual fate by repressing the translation of tra‐2 via the TGEs.
GLD‐1 and translational control of tra‐2
GLD‐1 plays multiple roles in germline development. It is essential in oogenesis, and is also necessary for spermatogenesis and inhibition of germline proliferation (Francis et al., 1995a). We propose that the GLD‐1 is necessary for repressing tra‐2 translation in the germline to allow spermatogenesis to occur. The oogenic and proliferation phenotypes of GLD‐1 may result from mis‐regulation of additional mRNAs that contain TGEs.
Normal sexual development of both XX and XO animals requires that tra‐2 is translationally repressed in both the germline and the soma (Doniach, 1986). However, GLD‐1 is germline specific (Jones et al., 1996). These findings suggest that other gene product(s) function in the soma to repress tra‐2 translation, and that the TGE control is regulated by tissue‐specific factors. For example, there may be a somatic factor that is a homolog of GLD‐1. Indeed, several GLD‐1‐like sequences have been found in the C.elegans database that may be the somatic factor for tra‐2 translational control. Alternatively, translational control may require multiple factors, and some of these factors may be tissue specific and others not. Two results possibly support this hypothesis. First, GLD‐1 has a moderate affinity for the TGEs (500 nM). This may be due to the fusion protein, as the first 80 amino acids are missing. Alternatively, it may indicate that high affinity binding of GLD‐1 to TGEs requires another factor(s). Secondly, the yeast in vitro assay showed that purified GLD‐1 is sufficient to inhibit translation 2.5‐ to 3‐fold. Although tra‐2 is dosage sensitive and small changes in activity have significant effects on phenotype (Doniach, 1986), it is possible that in vivo GLD‐1 interacts with other factors to increase its ability to repress translation.
The sex determination gene, laf‐1, is an excellent candidate for a factor that interacts with GLD‐1 (Goodwin et al., 1997). laf‐1 is required for the translational repression of tra‐2, and affects the TGE control in the germline (Goodwin et al., 1997). However, laf‐1 may also interact with a somatic factor, perhaps a GLD‐1‐like factor, since laf‐1 affects tra‐2 translation in the soma (Goodwin et al., 1997). Presently, it is unclear whether laf‐1 is a component of DRF or promotes DRF binding. It is possible that a homolog of laf‐1 in yeast is participating with GLD‐1 to repress translation in the in vitro translation assay. Spermatogenesis also requires that the germline‐specific gene fog‐2 inhibits tra‐2 activity (Schedl and Kimble, 1988; Figure 1). It is possible that fog‐2 may act with GLD‐1 in the germline to repress tra‐2 translation. Alternatively, fog‐2 may act at another level to regulate TRA‐2A.
gld‐1 mutant phenotypes may be due to loss of TGE control
Previous genetic studies have suggested that mutations in the KH domain of GLD‐1 result in a dramatic loss of GLD‐1 activity. Three missense mutations (q361, oz89 and q93oz55) that change the absolutely conserved Gly227 to Asp in the GLD‐1 KH motif result in a loss‐of‐function phenotype that in homozygous XX animals abolishes sperm production (Jones and Schedl, 1995). Here, we show that the q361 mutation causes a loss of GLD‐1 binding to TGEs. This result agrees with previous data that show that a similar mutation in the related STAR protein, SAM68, eliminates specific RNA binding (Lin et al., 1997). Moreover, we show that the mutant GLD‐1(q361) protein is not able to repress the activity of a reporter RNA in vivo or in vitro. These results suggest that the loss of sperm in gld‐1. q361) mutants is due to a reduction in GLD‐1 repression of tra‐2 translation.
These same three alleles also have a semi‐dominant Fog phenotype (feminization of germline) (Francis et al., 1995a); a percentage of heterozygous animals produce only oocytes. There are several possibilities for this phenotype. The semi‐dominant phenotype may result from the sensitivity of sexual development to levels of tra‐2 activity (Doniach, 1986). Alternatively, Francis et al. (1995a) proposed that the semi‐dominant Fog phenotype may result from a stable mutant protein either titrating out a limited supply of factor(s) or poisoning a GLD‐1 multimer. Interestingly, other STAR family proteins bind to one another (Chen et al., 1997). It is possible that GLD‐1 likewise self‐associates and that this self‐association is necessary for proper GLD‐1 function.
Jones and Schedl identified two classes of gain‐of‐function GLD‐1 mutations, one class which result in a Fog phenotype and the other in a Mog phenotype (Jones and Schedl, 1995). It is possible that Mog mutations cause an increase in the ability of GLD‐1 to bind TGEs. Alternatively, GLD‐1 (Mog) protein may alter the interaction of GLD‐1 with a regulator and, consequently, there is an increase in GLD‐1 activity. GLD‐1 (Fog) protein may reduce GLD‐1 binding to the tra‐2 TGEs or affect the interaction of GLD‐1 with factors that regulate its activity.
How is GLD‐1 regulating the translation of tra‐2?
There are several possibilities of how GLD‐1 represses tra‐2 translation. One model is that GLD‐1 may bind directly or indirectly to translational initiation factors and consequently inhibit their activities. Alternatively, GLD‐1 may sequester or mask the tra‐2 transcript from the translational machinery. Finally, GLD‐1 may influence the lengths of the poly(A) tail. Support for this last model comes from our studies that show that the presence of the TGEs correlates with a short poly(A) tail and the absence of the elements with a long poly(A) tail (Jan et al., 1997).
A number of factors have been identified that bind 3′‐UTR elements and are implicated in repressing translation. Two related proteins work via elements in the 3′‐UTR to inhibit mRNA activity in C.elegans and Drosophila. In C.elegans, FBF‐1 inhibits the activity of the fem‐3 mRNA (Zhang et al., 1997), and in Drosophila Pumilio inhibits the translation of hunchback mRNA (Murata and Wharton, 1995; Wharton et al., 1998). Interestingly, database analysis suggests that FBF‐1 and Pumilio are part of a larger family of proteins that may be translational regulators (Zhang et al., 1997). Also in Drosophila, Bruno and a 55 kDa protein repress oskar translation (Kim‐Ha et al., 1995; Gunkel et al., 1998), Bicoid inhibits caudal translation (Dubnau and Struhl, 1996; Rivera‐Pomar et al., 1996), and Smaug inhibits the translation of nanos (Smilbert et al., 1996). In C.elegans, a non‐coding RNA, called lin‐4, is a translational repressor of the heterochronic genes lin‐28 and lin‐14 (Wightman et al., 1993; Moss et al., 1997).
Although a number of factors have been identified that bind to 3′‐UTR elements and are required for translational control in vivo, only two factors have been shown to inhibit translation in vitro. Ostareck et al. (1997) demonstrated that purified hnRNP K and E1 can bind to LOX 3′‐UTR elements and repress translation in a reticulocyte cell‐free translation assay. Here, we demonstrate that purified GLD‐1 represses translation in a TGE‐dependent manner in a yeast cell‐free translation system. Interestingly, although hnRNP K, E1 and GLD‐1 repress translation via elements in the 3′‐UTR, LOX 3′‐UTR translational control occurs independently of changes in poly(A) tail length (Ostareck‐Lederer et al., 1994), whereas the tra‐2 3′‐UTR translational control correlates with changes in poly(A) tail lengths (Jan et al., 1997). These differences indicate that hnRNP K and E1 and GLD‐1 may act by different mechanisms to inhibit translation. This is interesting since regulation of a number of 3′‐UTR translational control elements, such as fem‐3 (Ahringer and Kimble, 1991), bicoid (Salles et al., 1994) and hunchback (Wreden et al., 1997), correlate with changes in poly(A) tail lengths, and others such as nanos (Salles et al., 1994) and oskar (Webster et al., 1997) do not.
Possible relationship of TGE and the STAR family of proteins
The TGE control regulates the translation of several mRNAs in both invertebrates and vertebrates (Jan et al., 1997). The STAR proteins are also present in a number of organisms and are important in many developmental decisions, including embryogenesis and myelination in mice, notochord differentiation in Xenopus embryos and muscle development in Drosophila (Vernet and Artzt, 1997). However, the RNA targets and mechanisms by which they control RNA activity are poorly understood. The STAR proteins can be separated into several subfamilies that may indicate distinct functions. For example, the mammalian STAR family member, SF1, and its yeast homolog, BBP, are splicing factors and bind to or very near the conserved branchpoint sequence (Abovich and Rosbash, 1997; Berglund et al., 1997). SF1/BBP are distant from other STAR members (Vernet and Artzt, 1997). They contain only half of the QUA1 domain and it is not well conserved (Vernet and Artzt, 1997). Moreover, they have a zinc‐knuckle RNA‐binding motif (Berglund et al., 1997). GLD‐1 is most similar to the subfamily that includes the Drosophila HOW/WHO (Baehrecke, 1997; Zaffran et al., 1997) and vertebrate QUAKING proteins (Ebersole et al., 1996; Vernet and Artzt, 1997; Zorn and Krieg, 1997). The fact that the TGE control is conserved raises the possibility that other STAR family members, perhaps some of the family members most similar to GLD‐1, act via TGEs to control translation in different organisms.
Materials and methods
General procedures and strains
Routine maintenance was as described by Brenner (1974). All strains were raised at 20°C unless otherwise indicated.
The following mutant C.elegans alleles were used in this study: (i) LGI, unc‐13. e51), gld‐1(q485lf) (Francis et al., 1995a), gld‐1. q93Mog) (Francis et al., 1995a); (ii) LGII: tra‐2. e2020gf) (Doniach, 1986); (iii) LGIV: unc‐24(e138)fem‐3(e1996lf)/DnT1 (Hodgkin, 1986), him‐8(e1488); and (iv) LGV: her‐1(e1518lf) (Hodgkin, 1980).
Yeast three‐hybrid system
The RNA hybrid contains the tra‐2 TGEs arranged in tandem upstream of the MS2 stem–loops. The sequence of the tra‐2 TGEs produced by the RNA hybrid: UAUUUAAUUUCUUAUCUACUCAUAUCUA and CUCAUAUUUAAUUUCUUAUCUACUCAUAUCUA. To construct the hybrid RNA plasmid, the tra‐2 TGEs were amplified from cDNA using PCR primers EBG‐74 and EBG‐75 that contained a SmaI site (see below for primer sequences). The resulting PCR products were subcloned into the unique SmaI site of pIII/MS2‐2 (a gift of Dr M.Wickens) that contains the URA3 gene. This chimeric RNA plasmid was called pIII/DRE‐MS2. pIII/DRE‐MS2 was transformed into the yeast strain, L40‐coat (SenGupta et al., 1996; gift of Dr M.Wickens). In this strain, the HIS3 and lacZ genes are under the control of the LexA‐binding sites. A strain carrying the hybrid RNA plasmid subsequently was transformed with a mixed stage C.elegans cDNA library, PRB‐2 (kindly provided by Dr R.Barstead). Transformants were selected for growth on plates lacking histidine and uracil plus 5 mM 3‐aminotriazole. After several days of growth, the colonies were lifted onto nitrocellulose and stained for β‐gal activity. Only colonies that expressed β‐gal were analyzed further.
The binding specificity of a clone was tested by transforming the yeast strain L4‐coat with the test cDNA plasmid and plasmids that encoded different hybrid RNAs. These plasmids coded for the MS2‐binding site or the MS2‐binding site fused with either the C.elegans tra‐2 TGEs, an IRE or a poly(A) tract containing 30 A residues. Yeast triple transformants were assayed for β‐gal activity and for the ability to grow on plates lacking histidine. Only one clone was found to bind specifically to the TGEs and it was given the name pBG515. The insert of pBG515 was sequenced and found to encode the full‐length GLD‐1 coding region.
RNA gel mobility shifts
The vectors, pGEX3‐14N (gift of Dr T.Schedl) and pGEXq361, were used to produce wild‐type and mutant GLD‐1 proteins fused with GST (Jones et al., 1996). pGEX3‐14N and pGEXq361 encode amino acids 84–457 which is 82% of the coding region (nucleotides 465–1370) (Jones et al., 1996). Fusion proteins were isolated as described (Ausubel et al., 1989). The mutant GLD‐1 protein results from a Gly227 to Asp change (Jones and Schedl, 1995). pGEXq361 was made by cloning an NdeI and MfeI fragment from pMALq361 (gift of Dr T.Schedl) into the same restriction sites of pGEX3‐14N. Wild‐type and mutant GLD‐1 fusion proteins were expressed and purified identically.
32P‐labeled and unlabeled RNA probes containing different 3′‐UTRs were produced by standard methods. The different 3′‐UTRs were subcloned in KSII (+) pBluescript vector (Stratagene). The 3′‐UTR‐containing pBluescript vectors were linearized and the sense 3′‐UTR RNAs were transcribed in vitro by either T3 or T7 RNA polymerase. Other 32P‐labeled and unlabeled RNA probes [EBG‐9, EJ‐19, EJ‐32, EJ‐35, EJ‐38 and EBG‐11; see Jan et al. (1997) for description and below for sequences] were produced using the method of Milligan and Uhlenbeck (1989). Cold RNA probes were produced by the RiboMAX kit (Promega). Quantitation of the cold RNA probes was measured by spectrophotometry at OD260.
The sequences of the wild‐type (EJ‐19) and mutant (EJ‐32 and EJ‐35) C.briggsae tra‐2 TGEs are:
EJ‐32, CAGATCTGTGTTTCCTACTTTCCTGCCTAG( )GAACACA;
EJ‐35, CAGATCTCACTTTCCTACTTTCCTGCCTAG( )GAACACA.
The conserved nucleotides are in bold, underlined nucleotides represent mutations and parentheses indicate deletions.
Apparent binding constants were determined by titrating increasing amounts of purified GST–GLD‐1 with a constant amount of radiolabeled RNAs. Reactions were loaded on a 3.75% native acrylamide gel. The gels were dried and the radioactive bands were quantified by a phosphoimager (FUJIX BAS 2000). For each lane, radioactivity migrating above free probe was indicative of binding and was quantified. The apparent binding constants were defined as the protein concentration at which half of the total amount of RNA was bound.
Analysis of endogenous GLD‐1 binding to tra‐2 3′‐UTR
Purified bacterially expressed GST–GLD‐1 was used to immunize rabbits (Cocalico Inc., Reamstown, PA). GLD‐1 antibodies were purified using a GST–GLD‐1 affinity column (Bio‐Rad, Affigel). To test whether GLD‐1 is a component of DRF, RNA gel shifts were performed as described (Goodwin et al., 1993). GLD‐1 antibodies were incubated with C.elegans extract for 5 min at room temperature before adding to the reaction mix containing the radiolabeled RNAs. Reactions were loaded on a 6% polyacrylamide gel and autoradiographed. GLD‐1 antibodies were pre‐absorbed with GST–GLD‐1 protein prior to incubation with C.elegans extract. Depletion of the GLD‐1 antibodies was confirmed by Western blot analysis. Western blot analysis was performed as described (Harlow and Lane, 1988).
All β‐gal reporter transgenes were derived from the same parent vector, pPC16.41 (gift of Dr Peter Candido). This vector contains the C.elegans inducible promoter, hsp16‐41, the lacZ coding sequence and a polylinker. The construction of pBG2 [lacZ::tra‐2(+)3′UTR], pBG3 [lacZ::tra‐2(−32)3′UTR] and pBG4 [lacZ::tra‐2(−60)3′UTR] are described in Goodwin et al. (1997). hsp::GLD‐1 contains the heat shock‐inducible promoter, 16‐41, the entire full‐length gld‐1 protein‐coding region and the unc‐54 3′‐UTR. The hsp16‐41 promoter plus the HSP initiating methionine was amplified from genomic DNA by PCR using oligos CKM‐5 and CKM‐6 that contain ClaI sites. The subsequent product was cloned into the ClaI site of KSII (+) Bluescript, resulting in the plasmid pBG516. The unc‐54 3′‐UTR was amplified from genomic DNA with primers CKM‐7 and EBG‐45, which contained translational stop codons in three frames, and both primers contained ApaI sites. The PCR product was cloned into the ApaI site of pBG516, resulting in pBG517. To insert the full‐length gld‐1 coding region, pBG515 (see above) was cleaved with XhoI and the resulting fragment was cloned into the XhoI site of pBG517, resulting in the plasmid pBG518, which is also referred to as hsp::GLD‐1. hsp::GLD‐1(q361) was constructed by cloning an NdeI and MfeI fragment from pMALq361 into the same restriction sites of hsp::GLD‐1.
Transgenic C.elegans animals were generated using standard methods (Mello and Fire, 1995). hsp::GLD‐1 or hsp::GLD‐1(q361) (100 ng/μl) were injected into wild‐type C.elegans with 100 ng/μl of pRF4 and 50 ng/μl emb‐9::GFP (a kind gift from C.Witkowski and J.Kramer). emb‐9::GFP contains the coding region of the green fluorescent protein (GFP) directed by the emb‐9 promoter. This construct will be expressed predominantly in the muscle of the adult animal. Transgenic animals carrying extra‐chromosomal arrays containing hsp::GLD‐1 and emb‐9::GFP were crossed to transgenic animals carrying lacZ reporter transgenes that carry different tra‐2 3′‐UTRs. For studies on lacZ::tra‐2(−60)3′UTR, 100 ng/μl of hs::GLD‐1 with 50 ng/μl emb‐9::GFP were injected into transgenic animals carrying lacZ::tra‐2(−60)3′UTR. Only progeny expressing GFP in the adult muscle were analyzed for β‐gal activity.
Cytochemistry and RNA in situ analyses
Immunofluorescence and in situ hybridization experiments were performed on the extruded gonads of L3 animals. The gonads were permeabilized by freeze‐cracking in liquid nitrogen. For immunofluorescence, animals were then fixed in cold methanol for 10 min, followed by 10 min in cold acetone. The polyclonal antibodies used in the experiment were generated by immunization of rabbits with a GST–TRA‐2 fusion consisting of the N‐terminal portion of the TRA‐2A protein. The resulting antibodies were affinity purified as described for GLD‐1. Characterization of TRA‐2 antibodies will be described elsewhere (L.E.Graves and E.B.Goodwin, in preparation). For in situ hybridizations, animals were treated as described (Seydoux and Fire, 1994). A single‐stranded, digoxigenin‐labeled DNA probe was synthesized using PCR run‐off from a linearized plasmid containing a 774 bp fragment of the tra‐2 cDNA. This probe is specific to the 4.7 kb tra‐2 transcript. The probe was detected using a fluorescein‐conjugated anti‐digoxigenin antibody (Boehringer Mannheim). An anti‐mouse actin monoclonal antibody was used (ICN Biomedicals).
Yeast in vitro translation assay
Yeast lysates were produced as described previously (Iizuka et al., 1994) and modified as described (Preiss and Hentze, 1998), where extracts were not treated with micrococcal nuclease. The in vitro translation assay was performed as described (Iizuka et al., 1994). Capped RNAs containing the lacZ gene and different 3′‐UTRs were produced by a standard in vitro transcription reaction (Ausubel et al., 1989). The ratio of cap analog to GTP was 5:1. pBG51, pBG52 and pBG53 correspond to constructs containing the lacZ gene and wild‐type tra‐2(+) 3′‐UTR, a mutant tra‐2(−32) 3′‐UTR in which one TGE was deleted or a mutant tra‐2(−108) 3′‐UTR in which both TGEs plus some flanking sequences were deleted, respectively. All three plasmids contain a poly(A)30 tract downstream of the 3′‐UTRs. An NsiI site is immediately 3′ to the poly(A)30 tract. The constructs were linearized using NsiI and transcribed using a Promega SP6 Ribomax kit. RNAs were quantitated by spectrophotometry and checked for purity on an agarose gel. RNA was added to the yeast lysates to a final concentration of 4 nM. GST–GLD‐1 was added to reaction to a final concentration of 280 nM. At specific time points, aliquots were taken from each reaction and frozen by liquid nitrogen to stop the reaction or added to TRIZOL (Gibco‐BRL). β‐gal activity was assayed using a colorimetric assay (Tropix) and a luminometer (Monolight 2010 Luminometer). RNA levels were quantitated by Northern blot analysis. All presented data are representative of at least five independent experiments.
EBG‐9, 5′‐GGACGATTAGATATGAGATGATAAG AAATTAAATATGAGTAGATATGAGTAGATAAGAAA TTAAATAATGAAATGGAAATTGTCG CCCTATAGTGAGTCGTATTA‐3′; EBG‐11, 5′‐TGGACGATTATGAAATGGAAATTGTACAAATAATAGAAA CGAAAATGAGTAAGAAATGAAATTTT GGAACCAAATTCTCG CCCTATAGTGAGTCGTATTA‐3′; EJ‐19, 5′‐GTGTTCAGAAAACTAGGCAGGAAAGTAGGAAAGTGAGA TCTGTTAATCGCCCTATAGTGAGTCGTATTA‐3′ (the underlined nucleotides represent the nucleotides deleted and/or mutated in EJ‐32 and EJ‐35); EJ‐24, 5′‐GGAAGGATAGAAACCCCTTA GGAAATGCGATCTGTGATGGATGAGATT CCCTCGCCCTATAGTGAGTCGTATTA‐3′; EJ‐32, 5′‐TACAAGATCTGTGTT CCTAGGCAGGAAAGTAGGAAA CACAGATCTGTTCGCCC TATAGTGAGTCGTATTA‐3′; EJ‐35, 5′‐GAATTCTCGAGTACAAGATCTGTGTTCCTAGGCAG GAAAGTAGGAAAGTGAGATCTGTTCG CCCTATAGTGAGTCGTATTA‐3′; EJ‐38, 5′‐TGCAGCTCCCCCAATTTTT CTGGAAGGATAGAAACCCCTTAGGAAA TGCGATCTGTGATGGATGAGATT CCCTCGCCCTATAGTGAGTCGTATTA‐3′; EBG‐45, 5′‐CTAGGGGCCCTGCGGTTTTTTCTATGAT‐3′; EBG‐74, 5′‐TCCCCCGGGGGATTTGTACAATTTCCATTTCAT‐3′; EBG‐75, 5′‐TCCCCCGGGGGAAGTTGAGGTCGAGTGGACGAT‐3′; CKM‐5, 5′‐CCATCGATGCATTTTCGAAGTTTTTTAGAT‐3′; CKM‐6, 5′‐CCATCGATGGCGAGCTGCTTGTTGCAAAAGG‐3′; CKM‐7, 5′‐GCTAGGGCCCTAACTAAGTAATAGGGGCCGCTGTCATCA‐3′.
We are very grateful to Beilin Zhang and Marvin Wickens for the yeast three‐hybrid constructs, reagents and valuable assistance. We especially thank Tim Schedl for the GLD‐1 expression constructs and worm strains. We are grateful to Aurelia Haller and Peter Sarnow for assistance in the yeast in vitro assay, and Laimonis Laimins for use of the luminometer. We thank Brian Ackley, Bob Barstead, Peter Candido, Stella Doktor, Jeff Johnson and Colette Witkowski for equipment, constructs and technical assistance. We thank Steve Adam, Pete Okkema and Hank Seifert for critical reading of the manuscript, and Karen Hofstra, Jim Kramer, Eric Punkay, Tim Schedl and Sejal Shah for valuable discussions. We acknowledge the Caenorhabditis Genetics Center for worm strains. This work was supported by NIH grant GM 51836‐01 to E.B.G., by the NIH Carcinogenesis Training Grant to E.J. and by the Chicago Baseball Cancer Charities grant to L.E.G.
- Copyright © 1999 European Molecular Biology Organization