FOG-2, a novel F-box containing protein, associates with the GLD-1 RNA binding protein and directs male sex determination in the C. elegans hermaphrodite germline

Translational regulation is an important mechanism for temporal and spatial restriction of gene expression. Biological processes utilizing translational repression/activation include anterior-posterior axis formation in Drosophila, sex determination in C. elegans, and iron homeostasis in mammalian cells (reviewed by Anderson and Kimble, 1997; Hentze and Kuhn, 1996; van Eeden and St Johnston, 1999). Translational repression, in many cases, is controlled by sequences in the 3′UTR of the mRNA. While sequences in 3′UTRs that are important for translational repression and proteins that bind these sequences have been identified, the mechanism(s) by which repression occurs is largely unknown (Gray and Wickens, 1998).

C. elegans GLD-1, a germline-specific cytoplasmic protein, is necessary for three aspects of germline development (Francis et al., 1995a,b; Jones et al., 1996). First, gld-1 has an essential function in meiotic prophase progression and oocyte development. Second, it is necessary for spermatogenesis in the hermaphrodite. Third, it directs the initiation of meiotic development in a pathway that is redundant with the activity of the gld-2 gene (Kadyk and Kimble, 1998). Recent data indicates that GLD-1 carries out at least some of these functions by acting as a translational repressor (see below: Jan et al., 1999; M.-H. Lee, B. Grant, D. Hirsh, and T. Schedl, unpublished data). GLD-1 is a member of a family of proteins, including mouse quaking and Drosophila How, that contain an approx. 200 amino acid region of similarity called the GSG or STAR domain (Di Fruscio et al., 1998; Jones and Schedl, 1995; Vernet and Artzt, 1997), which consists of a centrally located approx. 100 amino acid KH RNA binding region flanked by an approx. 75 amino acid conserved N-terminal region and an approx. 25 amino acid conserved C-terminal region. GSG domain proteins differ from other KH RNA binding proteins (e.g. fragile X protein FMR1) in three respects: (1) the KH domain contains an additional approx. 26 amino acids present in two conserved loops (Gibson et al., 1993; Musco et al., 1996; Siomi et al., 1993); (2) there is conservation outside the KH region; and (3) there is only a single KH domain.

To begin to understand how GLD-1 functions as a translational repressor for diverse aspects of germline development, we performed a yeast two-hybrid screen to identify proteins that associate with it. This screen identified the fog-2 gene product that, like GLD-1, acts in the specification of the male sexual fate in the hermaphrodite germline (Schedl and Kimble, 1988). Below we briefly review hermaphrodite germline sex determination.

In C. elegans there are two sexes, hermaphrodites and males. The hermaphrodite has a female soma but a germline that makes sperm then oocytes. The first approx. 40 germ cells that differentiate in each gonad arm develop as sperm in the fourth larval stage, then a cell fate switch occurs so that all remaining germ cells develop as oocytes in the adult. Sexual fate in the germline is determined by a negative regulatory cascade (Fig. 1; reviewed by Ellis, 1999; Hansen and Pilgrim, 1999; Kuwabara et al., 1998). The genes fem-1, fem-2 and fem-3 and fog-1 and fog-3 direct spermatogenesis (Barton and Kimble, 1990; Ellis and Kimble, 1995; Hodgkin, 1986). The activities of these genes are negatively regulated by tra-2 and tra-3 to allow oogenesis. tra-2 encodes a transmembrane receptor protein (Kuwabara et al., 1992). The TRA-2 cytoplasmic domain binds FEM-3; this interaction appears to lead to FEM-3 sequestration or inactivation (Mehra et al., 1999). In males, secreted HER-1 product presumably binds to the extracellular domain of TRA-2, freeing FEM-3 from the TRA-2 intracellular domain, resulting in continuous spermatogenesis (Kuwabara et al., 1992; Perry et al., 1993). Control in the hermaphrodite germline is more complex as there is a brief period of spermatogenesis during larval development before the switch to oogenesis in the adult. Furthermore, her-1 is not responsible for inhibiting tra-2 to allow hermaphrodite spermatogenesis (Hodgkin, 1980; Perry et al., 1993; Trent et al., 1991). Instead, the gene fog-2 acts at an equivalent position in the cascade, being necessary for spermatogenesis in the hermaphrodite and functioning upstream of tra-2 (Schedl and Kimble, 1988). The function of fog-2, unlike the fem genes and fog-1 and fog-3, is restricted to the hermaphrodite.

Posttranscriptional control of tra-2 and fem-3 is important for the hermaphrodite pattern of spermatogenesis then oogenesis. tra-2 mRNA is translationally repressed to allow hermaphrodite spermatogenesis. Gain-of-function (gf) mutations that disrupt one or both of the 28 base Direct Repeat Elements (DRE) in the tra-2 3′UTR cause feminization of the hermaphrodite germline; a role of the DREs in translational repression is inferred from the shift of tra-2 mRNA containing gf mutant 3′UTR into larger polysome fractions compared to mRNAs containing a wild-type 3′UTR and on derepression of a lacZ reporter when the tra-2 3′UTR contains gf mutations (Doniach, 1986; Goodwin et al., 1993; Schedl and Kimble, 1988). Goodwin and coworkers identified GLD-1 as a protein that binds the DREs and used reporter constructs and antibody staining to identify that its function is in the translational repression of tra-2 mRNA (Jan et al., 1999). This is consistent with the loss-of-function (lf) phenotype of gld-1 (feminization of the hermaphrodite, and not the male, germline; Francis et al., 1995a) where the absence of tra-2 translational repression would result in high levels of TRA-2 protein that could bind and inactivate FEM-3.

fem-3 mRNA is posttranscriptionally repressed to allow oogenesis. Mutations that alter the fem-3 3′UTR result in a gf masculinization where the switch to oogenesis fails to occur (Ahringer and Kimble, 1991; Barton et al., 1987). The fbf-1 and fbf-2 Pumilio homologs and the genes mog-1 to mog-6 are important for this negative regulation (Gallegos et al., 1998; Zhang et al., 1997).

Genetic studies of tra-2 gf and fem-3 gf mutants suggested that a balancing of TRA-2 feminizing and FEM-3 masculinizing activities controls the hermaphrodite pattern of spermatogenesis, then oogenesis (Barton et al., 1987; Schedl and Kimble, 1988). This can now be understood in terms of posttranscriptional regulation of tra-2 and fem-3 mRNAs as well as inactivation of FEM-3 by binding to the intracellular domain of TRA-2. In larvae, translational repression of tra-2 mRNA would lead to a lower TRA-2/FEM-3 ratio resulting in spermatogenesis, while in adults, posttranscriptional repression of fem-3 mRNA would lead to a higher TRA-2/FEM-3 ratio resulting in oogenesis. However, it is not known how the switch from spermatogenesis to oogenesis is achieved (Gallegos et al., 1998). Is translational repression of tra-2 mRNA constant over time or is it relieved in the adult? Similarly, is the posttranscriptional repression of fem-3 implemented only in the adult? One route to understanding this switch is to characterize the mechanism of tra-2 mRNA translational repression. We have identified FOG-2 as a cofactor that physically associates with GLD-1 in the translational repression of tra-2 mRNA. FOG-2 contains an F-box motif, which is usually associated with proteins involved in ubiquitin mediated degradation of substrates. We propose that in C. elegans sex determination a FOG-2/GLD-1/tra-2 3′UTR complex functions in translational control, representing a novel role for an F-box protein.

Standard methods were used for culturing and handling C. elegans (Brenner, 1974) at 20°C. fog-2 oz alleles were isolated in a screen for failure to complement fog-2(q71) feminization of the hermaphrodite germline phenotype following EMS mutagenesis (oz123, oz168, oz169, oz170 and oz184), psoralen mutagenesis (oz40) or gamma-ray mutagenesis (ozDf1 and ozDf2).

fog-2 corresponds to the predicted gene Y113G7B.5. We determined the intron and exon boundaries of fog-2 by sequencing cDNA clones cm16g5 (ACeDb) and CD13.1 (this study). fog-2 contains four exons that correspond to nucleotides 11,630-11,702, 11,834-11,911, 11,959-12,598 and 12,649-13,008 of Y113G7B (gi:5824688;

EMBL:AL110477.1). The start codon is at nucleotides 11662-11664; the termination codon is at nucleotides 12871-12873.

Our mRNA sequence includes nucleotides 12,806-12,850 and thus differs from EST y335g6 and the predicted Y113G7B.5 sequence. We believe y335g6 represents a mis-spliced or rare form of the fog-2 message because the putative intron is present in cm16g5, CD13.1 and each of the four 3′ cDNA clones we isolated. In addition, we cannot detect the y335g6 mRNA species by RT-PCR using a primer specific to the isoform or primers flanking it.

To identify fog-2 lesions, we isolated genomic DNA from two mutant worms to serve as template for a PCR reaction (Williams et al., 1992). PCR products were purified and then used for sequencing. Sequence alignments were performed with the Lasergene (DNAStar) and ClustalW 1.7 (Thompson et al., 1994). Phylogenetic analysis was performed using PHYLIP (Lim and Zhang, 1999; version 3.57C) and homology searches were performed using the NCBI BLAST (Version 2.0) server (http://www.ncbi.nlm.nih.gov:80/BLAST/).

GLD-1 constructs were cloned into the pAS1 vector (Fields and Sternglanz, 1994) to generate in-frame fusions to the DNA binding domain of Gal4p at the C terminus. The amino acids included in each ‘bait’ construct are indicated in parentheses; pAS1-GLD-1(84-457), pAS1-GLD-1(84-227), pAS1-GLD-1(84-341), pAS1-GLD-1(245-457), pAS1-GLD-1(245-341), pAS1-GLD-1(273-457), and pAS1-GLD-1(1-270). A set of mutant gld-1 cDNAs kindly provided by A. Jones was used to construct two-hybrid vectors containing the missense mutations oz10, oz17, oz47, q126, q266 and q361 in pAS1-GLD-1(84-341).

The complete fog-2 coding region was PCR amplified from the cDNA cm16g5 using Vent polymerase (NEB). The following pAS-1 FOG-2 constructs were used for two-hybrid analysis; pAS1-FOG-2(1-327) and pAS1-FOG-2(79-327). ftr-1 was obtained by RT-PCR and used to construct pAS1-FTR-1(72-314).

Yeast strain Y190 was transformed as described by Yamada et al. (1998) using R. Barstead’s oligo(dT)-primed and random-primed C. elegans cDNA two-hybrid libraries (Kraemer et al., 1999). Two-hybrid screens were performed with the bait constructs pAS1-GLD-1(1-270), pAS1-GLD-1(84-227), pAS1-GLD-1(84-341) and pAS1-GLD-1(373-457). Transformants were selected on His^–/Leu^–/Trp^–medium with 25 mM 3-amino-triazole (Sigma). His⁺Leu⁺Trp⁺ yeast colonies were tested for expression of the β-galactosidase reporter gene by a nitrocellulose filter lift assay (Durfee et al., 1993). The library plasmid was isolated from β-galactosidase-positive yeast grown in His^–/Leu^–/Trp^– liquid medium using the glass bead miniprep and then rescued in E. coli by electroporation. Minipreps were prepared from two bacterial colonies from each transformation and re-tested for specificity (Bartel et al., 1993) using GLD-1, Tat and p53. Clones showing specific interaction with GLD-1 were sequenced for further analysis.

The C-terminal FOG-2 peptide (Fig. 7A, CFDSALKIEKVSITE-DDLALL) was coupled to KLH using the Imject Activated Immunogen Conjugation kit (Pierce) and used to immunize a rabbit. A peptide affinity column was prepared using the Sulfolink kit (Pierce) and antibodies specific to the peptide were purified essentially as described by Harlow and Lane (1988). Gonad dissections (Francis et al., 1995b), whole-mount staining of L1 to L3 larvae, antibody incubations and washes, and western blots (Jones et al., 1996) were performed as described. Affinity purified anti-FOG-2 antibody was used at a 1:100 dilution. Epifluorescent images were captured with a Zeiss Axioskop equipped with a Hamamatsu digital CCD camera (Hamamatsu Photonics), and processed with Photoshop 5.5 (Adobe).

³²P-labeled RNAs were synthesized from PCR products that have a T3 primer site with [³²P]UTP (ICN) and T3 RNA polymerase at 37°C for 2 hours. Template DNAs were then removed by incubating with 10 Units RNase-free DNase (Roche) at 37°C for 15 minutes and ³²P-labeled RNAs were purified through Sephadex G-50 quick spin column (Roche). A volume of RNA giving 10⁵ cts/minute was incubated with cytosol extracts in 5 mM Hepes (pH 7.6), 1 mM MgCl₂, 75 mM KCl, 1 mM DTT, 1% glycerol, 300 μg/ml tRNA, 3 mg/ml Heparin in a final volume of 30 μl for 20 minutes at room temperature and UV-crosslinked in a Stratalinker UV Crosslinker 2400 (Strategene) at maximum output for 10 minutes on ice. 80 μg RNase A was then added and incubated for 20 minutes at room temperature to digest any uncrosslinked RNA. After digestion, 30 μl of homogenization buffer (HB, 15 mM Hepes pH 7.6, 10 mM KCl, 1.5 mM MgCl₂, 0.1 mM EDTA, 0.5 mM EGTA, 44 mM sucrose; Lichtsteiner and Tjian, 1995) containing 100 mM NaCl was added and immunoprecipitation was performed with anti-GLD-1 antibody, anti-FOG-2 antibody, or control IgG (SC94; Santa Cruz Biotech.) bound to protein G-Sepharose (Sigma) for 1 hour at 4°C. The beads were then isolated, washed 4 times with HB containing 100 mM NaCl, and boiled in SDS sample buffer. The supernatant was resolved in 10% SDS-PAGE and exposed to X-OMAT film (Kodak).

Primers used for sequencing and RT-PCR, details of the various pAS1-GLD-1 and pAS1-FOG-2 constructs, results and methods for the GST-pull downs, and details of the fog-2/ftr-1 gene family can be accessed from our web site (http://www.genetics.wustl.edu/tslab/SN_FOG-2_Page.html).

We used the yeast Gal4p two-hybrid system (Fields and Sternglanz, 1994) to identify proteins that physically interact with GLD-1. We recovered two identical cDNAs in two-hybrid screens (Fig. 2B, row 7). One (OG2.3) using GLD-1 residues 84-341 and the other (CD13.1) using residues 273-457, both fused to the Gal4p DNA binding domain (Fig. 2A, rows 1 and 2). The two cDNAs are products of the gene Y113G7B.5 (Fig. 3A; Consortium, 1998), located on chromosome V in the vicinity of the genetically mapped germline sex determination gene fog-2 (Schedl and Kimble, 1988). Since the lf phenotype of fog-2 is essentially identical to the lf germline sex determination phenotype of gld-1 (Francis et al., 1995a,b), the fog-2 product is a candidate GLD-1 binding protein.

Three lines of evidence indicate that Y113G7B.5 is the fog-2 gene. First, RNA mediated interference (RNAi; Fire et al., 1998) reveals that the lf phenotype of Y113G7B.5 is identical to that of fog-2. Introduction of dsRNA synthesized from cDNA clone CD13.1 produces a hermaphrodite-specific feminization of the germline: 91% of gonad arms (n=292) in somatically female cross-progeny have feminized germlines (contain oocytes where sperm would normally be produced) while none of their male siblings (n=100) show any germline or somatic sex determination defect. Second, 18 fog-2 mutations are in the Y113G7B.5 transcription unit. These fog-2 alleles include six different nonsense mutations, three unique missense mutations, three distinct splice site mutations and one complex lesion consisting of a trans-splice acceptor mutation and a missense mutation (Table 1). Finally, the Y113G7B.5 protein is not detected in animals homozygous for the fog-2 nonsense alleles oz40 or q71 (Figs 4 and 6).

We further characterized the GLD-1-FOG-2 interaction using the yeast two-hybrid system (Fig. 2A). GLD-1 amino acids 84-341, 273-457 or 273-341 fused to the Gal4p DNA binding domain interact with FOG-2 residues 79-327, while GLD-1 amino acids 1-270 failed to interact. Fusion proteins containing the Gal4p DNA binding domain and either residues 1-327 or residues 79-327 of FOG-2 interact with amino acids 283-463 of GLD-1. These results indicate that amino acids 283-341 of GLD-1 bind FOG-2 and demonstrate that the N-terminal part of FOG-2 (F-box) is not required for this interaction. The 58 amino acids of GLD-1 (283-341) that interact with FOG-2 contain the third alpha helical region of the KH domain (Lewis et al., 2000; Musco et al., 1996) as well as the C-terminal part of the redefined GSG domain (Di Fruscio et al., 1998).

A number of gld-1 mutations alter residues in the portion of the protein that interacts with FOG-2 (Jones and Schedl, 1995). These include the missense mutations q126 G308E, oz47 D310N, and oz10 P335S and the q266 mutation, an in-frame deletion of residues 322-331. Both the q126 and q266 mutants show strong feminization of the germline phenotypes (Francis et al., 1995a; Francis et al., 1995b), which might be due to a failure of GLD-1 to bind FOG-2. To test if any of these mutations, as well as q361 G227D and oz17 G248R, disrupt the interaction between GLD-1 and FOG-2, each was separately introduced into the Gal4 DNA binding domain construct containing residues 84-341 of GLD-1. Surprisingly, all the mutant fusion proteins retain the ability to interact with FOG-2 (Fig. 2, row 6), although we cannot rule out a small quantitative decrease in binding. The feminization phenotype of gld-1 q126 and q266 mutants thus appears not to be caused by a defect in the GLD-1-FOG-2 interaction.

To corroborate the GLD-1-FOG-2 interaction, we used an in vitro system. GLD-1 protein synthesized in the reticulocyte lysate system was tested for binding to GST-FOG-2 fusion protein linked to beads. Consistent with the two-hybrid assay results, we found that amino acids 84-457 of GLD-1, which do not bind GST, do bind residues 79-327 of FOG-2 fused to GST (data not shown).

If FOG-2 and GLD-1 interact in vivo they must be co-expressed at the appropriate time and reside in the same subcellular compartment. We used an affinity purified polyclonal antibody raised against a unique 20 amino acid sequence at the C terminus of FOG-2 to examine its accumulation during nematode development. The anti-C-terminal antibodies are highly specific because they detect a single polypeptide with a mobility similar to the predicted molecular mass of FOG-2 (37 kDa) in wild-type adult hermaphrodites and males but not in the fog-2 nonsense mutants oz40 and q71 (Fig. 4), and they fail to stain oz40 or q71 larvae or dissected adult gonads (Fig. 5, data not shown). FOG-2 is detected uniformly throughout the germline of L1, L2 and L3 larvae (Fig. 5). In dissected gonads from L4 (not shown) and young adult hermaphrodites and males, FOG-2 accumulates in proliferating and meiotic prophase germ cells with the amount of protein decreasing in late gametogenesis such that it is not observed in sperm or in full grown oocytes. FOG-2 accumulates exclusively in the cytoplasm, as is the case for GLD-1 (Jones et al., 1996). FOG-2 was not detected in somatic cells of larvae or somatic tissue liberated in gonad dissections (Fig. 5) and was not observed in western blots of glp-1 mutants that lack a germline (Fig. 4). Therefore, FOG-2 appears to be limited to the germline. The presence of FOG-2 in proliferating and early meiotic prophase germ cells is consistent with its genetically defined role in germline sex determination. Importantly, this pattern overlaps, although is not identical to, that of GLD-1 (Jones et al., 1996). The significance of FOG-2 accumulation in the male germline or during gametogenesis, both of which are unaffected in fog-2 mutants, is unknown.

Since GLD-1 binds to the tra-2 3′UTR to mediate translational repression (Jan et al., 1999) and FOG-2 binds to GLD-1, perhaps FOG-2, GLD-1 and the tra-2 3′UTR are present in a complex. To examine whether FOG-2 might bind directly to the tra-2 3′UTR, or associate indirectly via GLD-1, we used UV-induced RNA/protein cross-linking followed by immunoprecipitation (IP) with either anti-GLD-1 or anti-FOG-2 antibodies (Materials and Methods). 32P-labeled wild-type or e2020 mutant tra-2 3′UTRs were incubated with adult hermaphrodite cytoplasmic extracts from either wild-type, gld-1(q361) or fog-2(oz40) mutants, the RNA cross-linked to protein by UV treatment, and the unprotected RNA removed by RNaseA treatment. The ³²P-labeled RNA-protein complexes were either untreated (no IP), IP with anti-GLD-1 antibodies, or IP with anti-FOG-2 antibodies and the products resolved by SDS-PAGE (Fig. 6). The tra-2 e2020 gain-of-function mutant 3′UTR is a negative control as it fails to bind GLD-1 (Jan et al., 1999) and is not translationally repressed. Both the anti-GLD-1 IP and the anti-FOG-2 IP identify a doublet protein band labeled with the wild-type tra-2 3′UTR but not e2020 mutant 3′UTR. The bands co-migrate with the two isoforms of GLD-1 on western blots (data not shown; Jones et al., 1996). The gld-1(q361) extract (Fig. 6B) demonstrates that the presence of labeled GLD-1 in both IPs depends on GLD-1 and its ability to bind the tra-2 3′UTR; the q361 missense mutation is in a GLY residue that is absolutely conserved in all KH-domain containing proteins and GST-GLD-1(q361) does not bind the tra-2 3′UTR in vitro (Jones and Schedl, 1995; Jan et al., 1999). We did not detect FOG-2 labeled with wild-type or e2020 mutant tra-2 3′UTR in the anti-FOG-2 IP (Fig. 6; data not shown) indicating that FOG-2 does not directly bind the tra-2 3′UTR. The IP of GLD-1 by anti-FOG-2 from worm cytoplasmic extracts therefore provides further evidence of a physical interaction between the two molecules. GLD-1 was found to bind to the tra-2 3′UTR even in the absence of FOG-2 (Fig. 6B, fog-2(oz40)). These results suggest that translational repression of tra-2 mRNA is mediated by a FOG-2/GLD-1/tra-2 3′UTR ternary complex, where FOG-2 associates with the tra-2 3′UTR indirectly via GLD-1.

The fog-2 gene encodes a predicted protein of 327 amino acids (see Materials and Methods). FOG-2 is acidic and lacks an obvious signal sequence or nuclear localization sequence. However, it possesses two conserved motifs as determined by sequence alignment and by searching against the Pfam protein pattern database (Bateman et al., 2000).

Amino acids 16-65 of FOG-2 constitute an F-box motif (Fig. 7B; Bateman et al., 2000), which is a protein-protein interaction domain conserved in plants, fungi and animals (Bai et al., 1996). Residues 102 to 317 form the second conserved motif in FOG-2 (Fig. 7C), which we term the FOG-2 Homology domain (FTH). This motif was previously described as Pfam entry Duf38 (Domain of Unknown Function 38) by Bateman et al. (2000): http://pfam.wustl.edu/hmmsearch.shtml. Unlike the F-box, which has been conserved throughout eukaryotic evolution, we have not detected the FTH domain outside of C. elegans and C. briggsae. The FTH region is critical for FOG-2 function, as two residues within this domain are altered by three missense mutations: G219R (oz170, q113 and q170), G271R (q123) and G271E (q70) (Fig. 3a; Table 1).

The best characterized F-box-containing proteins are components of the SCF complex, which directs ubiquitin-mediated degradation of specific substrates (Deshaies, 1999). However, since both gld-1 and fog-2 promote spermatogenesis, FOG-2 cannot function simply by targeting GLD-1 for ubiquitin-mediated degradation (see Discussion). F-box containing proteins (e.g. Cdc4) act as bridges; the N-terminal F-box recruits the ubiquitin ligase machinery while a C-terminal protein-protein interaction domain (e.g. WD40 for Cdc4) binds specific substrates. FOG-2 containing the FTH domain, but lacking the F-box, can physically associate with GLD-1 (Fig. 2, row 7). We thus propose that the C-terminal FTH motif represents a novel protein-protein interaction domain, analogous to the WD40 domain of Cdc4. The three fog-2 missense mutations in the FTH domain may be defective in GLD-1 binding as they produce stable protein at essentially wild-type levels (Fig. 4, data not shown), yet have a strong lf mutant phenotype. The, oz170, q123 and q70 mutations were separately introduced into Gal4 DNA binding domain fusion constructs containing residues 81-317 of FOG-2 and tested for interaction with GLD-1. The three mutant constructs were able to interact with GLD-1 equivalent to wild-type FOG-2 (Fig. 2, row 9). Thus, the two FOG-2 residues affected by these mutations are not required for interaction with GLD-1 in yeast, but may instead be necessary for interaction with another factor.

There are approx. 102 predicted C. elegans proteins related to FOG-2, each containing an N-terminal F-box and a C-terminal FTH (Duf38) domain. We have called these ftr genes (for fog-2related). The most similar gene, ftr-1, lies just 5′ to fog-2 and is 63% identical at the amino acid level (Figs 3A and 7A). Eight of the ftr genes (including ftr-1) may be functional, as ESTs have been identified (Y. Kohara, personal communication). There is a further set of approx. 75 predicted genes that contain only the FTH domain, some of which have EST hits. From the small amount of C. briggsae sequenced to date (approx. 10%) two predicted genes apparently containing only the FTH domain have been identified. (For additional information see http://www.genetics.wustl.edu/ tslab/SN_FOG-2_Page.html).

Many of the ftr genes are found in clusters. fog-2 itself lies within a group of six tandemly arranged ftr genes (Fig. 3A). Based on phylogenetic analysis, the fog-2 cluster has expanded from 3 to 6 members by local gene duplication such that fog-2 is most similar to ftr-1, Y113G7B.7 to Y113G7B.6 and Y113G7B.1 to Y113G7B.3 (Fig. 3B). For the most part, ftr genes from different clusters are more distantly related while genes within a cluster are more closely related. ftr genes, whether in clusters or not, tend to be found in the chromosomal arms, which in C. elegans are thought to be somewhat heterochromatic.

The high level of sequence identity between FOG-2 and FTR-1 suggests that they might function redundantly in sex determination or some other process. To investigate this possibility, we first used the yeast two-hybrid system to ask if FTR-1 could interact with GLD-1. We found that FTR-1 does not interact with GLD-1 using an FTR-1 fragment that was essentially identical to the FOG-2 fragment identified in the original screen (Fig. 2, line 10). We next examined the phenotypic consequences of loss of ftr-1 function using RNAi. Using a 3′ region of ftr-1 that has reduced identity with fog-2, we found that ftr-1(RNAi) essentially produces no phenotype in either wild-type hermaphrodites or males, in fog-2 mutant hermaphrodites (females) or males or in a fem-3(gf); fog-2 (Schedl and Kimble, 1988) background that is sensitized to additive feminization (data not shown). These results suggest that ftr-1 does not work redundantly with fog-2 in sex determination and that FTR-1 does not bind to GLD-1, analogous to FOG-2, to mediate translational repression in a separate context (e.g. RNAs involved in oogenesis).

GLD-1, a cytoplasmic RNA binding protein, regulates multiple aspects of C. elegans germ cell development. Among these are meiotic prophase progression and oogenesis, male sex determination and initiation of meiotic development (Francis et al., 1995a,b; Kadyk and Kimble, 1998). To better understand how GLD-1 performs such diverse functions we used the yeast two-hybrid system to identify proteins that physically interact with GLD-1. One protein recovered in our two-hybrid screen was the product of the fog-2 germline sex determination gene. Like gld-1, fog-2 plays an important role in the adoption of the male sexual fate by germ cells developing in the hermaphrodite (Schedl and Kimble, 1988). GLD-1 appears to act as a translational repressor of tra-2 mRNA by binding to the DRE sequences in its 3′UTR (Jan et al., 1999). A function in translational repression is supported by the observation that deletions affecting the DREs result in a gf feminization of the hermaphrodite germline, similar to the lf feminization phenotype of gld-1 or fog-2, and causes missexpression of tra-2 3′UTR-containing reporter constructs (Doniach, 1986; Goodwin et al., 1993; Schedl and Kimble, 1988). Here we show that in worm cytoplasmic extracts FOG-2 and GLD-1 can form a ternary complex with the tra-2 3′UTR (Fig. 6). FOG-2 appears to associate indirectly with the tra-2 3′UTR by binding to GLD-1. Although GLD-1 does not require FOG-2 for binding to the tra-2 3′UTR (Fig. 6B), the completely penetrant fog-2 lf feminization of the germline argues that FOG-2 function is important for tra-2 translational repression in vivo and that GLD-1 is not sufficient. FOG-2 thus appears to act as a cofactor with GLD-1 in tra-2 translational repression (Fig. 1).

FOG-2 contains an N-terminal F-box motif and a C-terminal FTH (Duf38) domain (Fig. 7). The F-box motif is distributed throughout eukaryotes, and within species multiple F-box proteins are found. The best characterized F-box proteins are components of the SCF complex that functions as an E3 ubiquitin protein ligase. F-box proteins, like Cdc4, act as bridges between the core ubiquitination complex and specific substrates (Fig. 8A). This occurs by the N-terminal F-box motif binding to Skp1, which recruits the core ubiquitination complex (CDC34, Rbx and CDC53; Deshaies, 1999). The C-terminal protein-protein interaction domain of the F-box protein (e.g. WD40 for Cdc4) binds specific substrates to be ubiquitinated and degraded (Winston et al., 1999). The C-terminal FTH domain of FOG-2 is likely a protein-protein interaction region as GLD-1 binds the C-terminal 246 amino acids of FOG-2, of which 215 belong to the FTH domain. Thus, FOG-2 has an overall arrangement that is typical of F-box proteins.

We propose that FOG-2 acts as a bridge or scaffold to recruit tra-2 mRNA bound to GLD-1 and other proteins into a multiprotein complex (Fig. 8B). The FOG-2/GLD-1 multiprotein complex would then mediate translational repression of tra-2 mRNA. This represents a novel function for an F-box containing protein. Typically, F-box proteins are part of the SCF complex that targets specific substrates for polyubiquitination and degradation by the 26S proteosome. However, both FOG-2 and GLD-1 promote male sex determination by acting to translationally repress tra-2 mRNA; therefore, FOG-2 cannot target GLD-1 for ubiquitin mediated degradation.

How might FOG-2 function in the translational repression of tra-2 mRNA? Here we consider three models for FOG-2 function. In the first, FOG-2 acts like a typical F-box protein of the SCF complex. In this case, the FOG-2 FTH domain binds an as yet unidentified positive regulator of tra-2 translation (or a negative regulator of translational repression) while the FOG-2 F-box binds a Skp1 homolog to recruit the core ubitquitination machinery. Translational repression of tra-2 mRNA occurs when the positive regulator is polyubiquitinated and degraded by the proteosome. Ubiquitin-mediated proteolysis appears to be important in the decay of cytokine mRNA that contain AU-rich elements in their 3′UTR (Laroia et al., 1999). For the second model, FOG-2 directs mono-ubiquitination of GLD-1 or an as yet unidentified co-factor that promotes translational repression of tra-2 mRNA. Mono-ubiquitination might change the activity of the target protein or complex, but not lead to degradation. Examples of mono-ubiquitination include the alpha factor receptor, which results in ligand stimulated internalization (Terrell et al., 1998), and histones, which are associated with chromatin remodeling (Baarends et al., 1999). For the third model, FOG-2 assembles a novel multiprotein translational repression complex that does not involve covalent attachment of ubiquitin or ubiquitin related proteins. S. cerevisiae p58/Ctf13 is an example of an F-box protein that assembles a novel multiprotein complex, the Cbf3 centromere binding complex (Kaplan et al., 1997). Distinguishing between these and other models will require identification of other proteins that bind FOG-2.

GLD-1 appears to translationally repress a number of mRNAs. FOG-2 is likely specific for GLD-1-mediated translational repression of tra-2 mRNA as fog-2 mutants do not display other phenotypes observed in gld-1 mutants and do not affect translational regulation of another GLD-1 mRNA target that encodes the RME-2 yolk receptor (M.-H. Lee, B. Grant, D. Hirsh, and T. Schedl, unpublished data). Specificity may be achieved by FOG-2 interacting, directly or indirectly, with another RNA binding protein that specifically associates with tra-2 mRNA.

The hermaphrodite makes sperm in the L4 stage and then switches to oocytes in adulthood. Translational repression of tra-2 is important to allow the fem genes and fog-1 and fog-3 to direct spermatogenesis while posttranscriptional repression of fem-3 is important for oogenesis (Ahringer and Kimble, 1991; Barton et al., 1987; Doniach, 1986; Gallegos et al., 1998; Goodwin et al., 1993; Schedl and Kimble, 1988; Zhang et al., 1997). However, it is not known whether translational repression of tra-2 (and/or control of fem-3) is modulated in larvae or the adult. tra-2 translational repression occurs in males and the soma (Goodwin et al., 1993; Jan et al., 1999). An activity defined by RNA gel shift that binds the tra-2 DRE (called DRF for Direct Repeat Factor) is found in wild-type hermaphrodites and in mutants that lack a germline (Goodwin et al., 1993). However, GLD-1 and FOG-2 are not detected in the soma and do not have essential functions in the specification of spermatogenesis in the male. Therefore, tra-2 translational repression, and DRF components, differ between the hermaphrodite germline and the soma and males.

It is possible that tra-2 is translationally regulated in two ways. In the hermaphrodite soma and in males, there may be only a basal or constitutive level of translational repression as there is no need for modulation. In the male, this regulation may help to adjust the concentration of TRA-2 to be in a range where regulation by HER-1 can occur. In the hermaphrodite germline, modulation of translational repression may be an important part of the sperm/oocyte switch: strong repression in larvae to allow for spermatogenesis and weak repression in the adult to assist in oogenesis. How might FOG-2 and GLD-1 participate in this modulation? FOG-2 and GLD-1 protein accumulation is unlikely to be a major factor as both are present in the germline throughout larval and adult life. Instead, protein modification mediated by FOG-2 is a possibility. For SCF-mediated degradation, often modification (phosphorylation) of specific substrates is the signal for destruction (Deshaies, 1999). If FOG-2 acts like a typical F-box protein (first model above), then modification of a positive regulator of tra-2 translation followed by degradation could lead to temporal modulation of repression. Alternatively, for the assembly of a multiprotein translational repression complex (third model above) modification of FOG-2 (like phosphorylation of the F-box protein p58/Ctf13; Kaplan et al., 1997) or another protein may control assembly and thus modulate translational repression.

Genetic studies suggest that tra-2 translational repression may also occur in adults. fem-3(gf) mutants undergo spermatogenesis throughout adulthood (Barton et al., 1987). The adult spermatogenesis phenotype of fem-3(gf) is suppressed by fog-2(lf) and by tra-2(gf) suggesting that disruption of tra-2 translational repression during adulthood will produce sufficient additional TRA-2 to counteract the excess FEM-3 produced by fem-3(gf). If tra-2 translational repression is temporally modulated, then modulation may be quantitative rather than qualitative.

fog-2 is a member of a rapidly evolving gene family that may be nematode specific

In C. elegans there are more than 100 genes related to fog-2 (ftr genes) that contain an N-terminal F-box and a C-terminal FTH domain. There is an additional set of approx. 75 predicted genes that have only the FTH domain. The FTH domain (and thus FOG-2) does not appear to exist outside the nematodes. By contrast, the worm genome does have orthologs of conserved F-box-containing proteins (e.g. sel-10 corresponds to Cdc4; Hubbard et al., 1997). The large number of predicted F-box proteins in C. elegans is therefore primarily due to the large number of ftr genes. Many of the ftr genes are found in clusters, with expansion within a cluster due to local gene duplications. The ftr genes thus appear to be rapidly evolving, and at least in C. elegans are greatly expanded. The expansion of worm-specific gene families such as ftr may account, in part, for the greater number of genes in C. elegans compared to flies (Consortium, 1998; Rubin et al., 2000).

fog-2 and ftr-1 represent an example of a recent gene duplication followed by divergence: FOG-2 binds GLD-1 and is important for translational repression of tra-2 mRNA while FTR-1 appears not to bind GLD-1 or have a role in germline sex determination. fog-2 is likely under selection for function in sex determination while ftr-1 is not. However, the fog-2–ftr-1divergence is non-random. First, exons 1, 2 and 3 are >80% identical at the nucleotide level while exon 4 is only 41% identical. Second, the introns are conserved, with intron 3 being 82% identical. The bulk of the amino acid changes are in exon 4 (Fig. 7A); surprisingly most of these are due to single base insertion/deletions that shift the ftr-1 reading frame relative to fog-2. Thus, ftr-1 remains under selection, although its function is not currently known.

Within the Caenorhabditis genus, there are species like C. elegans that have self-fertile hermaphrodites and males (androdioecious) and species like C. remanei that have females and males (gonochoristic). Fitch (1997) proposed that within the Rhabditidae family female/male reproduction is the ancestral state while self-fertile hermaphroditism has evolved independently multiple times. The important feature in the evolution of self-fertility is the ability of the germline to briefly undergo spermatogenesis, as C. elegans hermaphrodites and C. remanei females are otherwise essentially identical at the morphological level. fog-2, gld-1 and tra-2(gf) mutants can be propagated as obligate XX female/XO male reproducing strains suggesting that implementation of germline translational control of tra-2 mRNA may be significant in the evolution of self-fertile hermaphroditism (Doniach, 1986; Francis et al., 1995b; Haag and Kimble, 2000; Schedl and Kimble, 1988). C. remanei has a tra-2 ortholog that functions in female sex determination, analogous to C. elegans (Haag and Kimble, 2000). C. remanei also has a homolog of gld-1 (Jones and Schedl, 1995). A DRF-like activity is found in C. remanei extracts that can gel shift the 3′UTRs of both C. remanei and C. elegans tra-2, leading Haag & Kimble (2000) to propose that tra-2 is also translationally regulated in C. remanei. However, the tissues or sex specificity of the C. remanei DRF-like activity is not known.

Since the fog-2/ftr gene family appears to be rapidly evolving, there is a distinct possibility that C. remanei may not have a fog-2 ortholog, although it likely has ftr genes. In the absence of a fog-2 ortholog, C. remanei would be unable to translationally repress tra-2 in the germline (or modulate translational repression), although it may retain a basal, fog-2 independent translational repression mechanism in the female soma and/or the male, like C. elegans. With this scenario, fog-2 would be a recently evolved gene that facilitates self-fertility in C. elegans. However, it is unlikely that the presence or absence of fog-2 is the only difference between C. elegans and the ancestral female/male species given the temporal control of the sperm/oocyte switch, the regulation of fem-3, and the importance of the number of sperm produced for reproductive success (Hodgkin and Barnes, 1991).

We thank Bob Barstead, Frank Li and Valerie Lantz for two-hybrid reagents and technical assistance, Yuji Kohara for cDNA clones, and Lutfiyya and Jim Dover for sharing their yeast knowledge. We are grateful to Dave Hansen, Patty Kuwabara, Eric Haag and Mark Johnston for comments on the manuscript and to Betsy Goodwin, Eric Haag and Patty Kuwabara for communicating information prior to publication. This work was supported by NIH grant HD25614 to the Schedl lab and GM15622 to B. Clifford.