Sex determination in the germ line^*

Two facts imply that fog-1 and fog-3 control sexual fate in germ cells and probably act at the end of the sex-determination pathway. First, mutations in fog-1 or fog-3 cause germ cells to differentiate as oocytes rather than sperm, but have no effect on other parts of the body. Second, genetic analyses indicate that mutations in fog-1 and fog-3 act downstream of mutations in all other genes in the pathway (Figure 2). Thus, these two genes might be directly responsible for initiating spermatogenesis.

The fog-1 gene produces two major transcripts (Jin et al., 2001; Luitjens et al., 2000). The large transcript is necessary for fog-1 function and is sufficient to rescue fog-1(q253) mutants, but the small one has no detectable activity (Jin et al., 2001a; Jin et al., 2001b). Other sex-determination genes, like her-1, tra-1 and tra-2, also make minor transcripts of unknown function. Here we discuss only the essential products of each gene.

Sequence analysis reveals that FOG-1 is a Cytoplasmic Polyadenylation Element Binding (CPEB) protein. All CPEB proteins contain two RNA Recognition Motifs (RRMs) and a C-H domain that binds zinc and interacts with RNA (reviewed by Mendez and Richter, 2001). Because other CPEB proteins regulate the translation of specific messages, FOG-1 probably regulates the translation of mRNAs that control sexual fate in the germ line. Interestingly, FOG-1 can bind the fog-1 message itself (Jin et al., 2001a), so it might be auto-regulatory. Other targets of FOG-1 have not yet been identified.

Although three other CPEB proteins are also expressed in the germ line (Luitjens et al., 2000), only FOG-1 regulates sexual fate. How is this specificity achieved? Analyses of FOG-1 homologs from other Caenorhabditis species show that FOG-1 has a conserved insertion in the C-H domain that is not found in other CPEB proteins (Cho et al., 2004); this domain might specify FOG-1’s unique function. One possibility is that FOG-1 binds specific elements found only in mRNAs that act in sex determination. Alternatively, FOG-1 might interact with other proteins that confer the ability to regulate sexual fate.

FOG-3 is a member of the Tob family of proteins (Chen et al., 2000). These proteins are found in all animals, and share a common amino-terminal domain of 115 amino acids. This family has been implicated in the control of cell division and differentiation, but so far little is know about their biochemical functions. Tob itself has both nuclear import and export signals in the conserved amino terminus (Kawamura-Tsuzuku et al., 2004; Maekawa et al., 2004), and associates with several Smad transcription factors during osteogenesis (Yoshida et al., 2000), so perhaps FOG-3 also regulates transcription.

TRA-1 appears to control germ cell fate by regulating the transcription of fog-1 and fog-3. Each gene has three to six potential TRA-1 binding sites in its promoter; more than are found in any other gene in the genome (Chen and Ellis, 2000; Jin et al., 2001). Furthermore, the fog-3 promoter can bind TRA-1 in gel shift assays, and this binding requires intact TRA-1 sites (Chen and Ellis, 2000). However, analyses of promoter mutations show that some of the TRA-1 binding sites mediate activation of fog-3 transgenes, whereas studies of fog-3 expression imply that at least one site mediates repression. Furthermore, tra-1(null) mutants usually make sperm early in life and oocytes later on (Hodgkin, 1987; Schedl et al., 1989). One explanation for these complex observations is that TRA-1 acts on the fog-3 promoter to repress spermatogenesis in some circumstances, and to activate it in others.

Just as in the soma, the fem genes regulate the activity of tra-1 in germ cells. For example, quantitative RT-PCR studies of fem; tra-1 double mutants show that the FEM proteins act through TRA-1 to control the level of fog-3 transcripts (Chen and Ellis, 2000). However, these double mutants make oocytes instead of sperm, which shows that the FEM proteins also act downstream of TRA-1 (Figure 2, Hodgkin, 1986). Perhaps the FEM proteins directly activate FOG-1 or FOG-3, or regulate an unknown target.

Gain-of-function mutations in tra-2 cause XX animals to make only oocytes, but have little effect on males (Doniach, 1986). Sequence analysis showed that these tra-2(gf) mutations disrupt sequences in the 3’ UTR known as TGEs (Goodwin et al., 1993). In wild-type animals, these elements bind a protein complex that includes GLD-1 (Goodwin et al., 1993; Jan et al., 1999). Three observations imply that GLD-1 represses translation of tra-2 (Figure 3). First, GLD-1 is required for spermatogenesis in hermaphrodites but not in males (Francis et al., 1995; Francis et al.,1995). Second, somatically expressed GLD-1 can repress translation of somatic transgene RNAs that contain TGEs (Jan et al., 1999). Third, TRA-2 levels are increased in gld-1 null mutant germ lines, based on in situ antibody staining (Jan et al., 1999).

GLD-1 is a member of the STAR family of proteins, which contain a maxi-KH RNA binding domain and two other conserved domains, called QUA1 and QUA2 (Jones et al., 1996). In vitro studies indicate that GLD-1 acts as a dimer (Ryder et al., 2004). Because GLD-1 has pleiotropic roles in germline development (Francis et al., 1995; Francis et al., 1995) it must regulate other messages in addition to tra-2. Some of these mRNA targets have been identified (Lee and Schedl, 2001; 2004; Marin and Evans, 2003; Mootz et al., 2004). Recently, in vitro biochemical studies implicated a hexanucleotide sequence as part of the GLD-1 binding site, at least for some RNA targets (Ryder et al., 2004).

fog-2 is required specifically for hermaphrodite spermatogenesis (Schedl and Kimble, 1988). FOG-2 binds GLD-1 in the yeast two-hybrid assay and in vitro, which implies that they interact to regulate tra-2 (Figure 3; Clifford et al., 2000). Genetic data support this model, since epistasis analyses place fog-2 upstream of tra-2, and loss of fog-2 activity causes the same phenotype as tra-2(gf) mutations (Schedl and Kimble, 1988). The C-terminus of FOG-2, which mediates its interaction with GLD-1, has been evolving rapidly (Nayak et al., 2004). The N-terminus of FOG-2 contains an F-box (Clifford et al., 2000) and binds the Skp1-related protein SKR-1 (Nayak et al., 2004); in other proteins, this would imply an interaction with the SCF ubiquitin ligase machinery. However, FOG-2 cannot target GLD-1 for degradation, since both promote spermatogenesis. Perhaps it recruits other proteins to the tra-2 3’UTR to block translation.

Some data suggest that laf-1 also regulates the translation of tra-2 messages (Goodwin et al., 1997). However, definitive tests of epistasis have not been possible, because laf-1 alleles cause recessive lethality as well as dominant feminization.

Several missense mutations in the intracellular domain of TRA-2 prevent spermatogenesis in hermaphrodites but not in males (Doniach, 1986; Kuwabara et al., 1998). Although this feminizing effect is dominant, these mutations also cause recessive masculinization in the soma, so they were named mixomorphic (mx) alleles. Surprisingly, these alleles alter a region of TRA-2 that normally binds TRA-1, and disrupt this interaction (Lum et al., 2000; Wang and Kimble, 2001). What remains unclear is the purpose of this TRA-2/TRA-1 interaction. One possibility is that the TRA-2/TRA-1 complex directly promotes oogenesis and other female fates; this model is consistent with the fact that tra-2(mx) alleles behave as if they cause increased wild-type activity (Doniach, 1986; Schedl and Kimble, 1988). Alternatively, the TRA-2/TRA-1 complex might promote spermatogenesis, as implied by the fact that blocking this interaction leads to oogenesis.

Since TRA-2 is a membrane protein, and TRA-1 is a transcription factor, how do they find each other? Lum et al. (2000) showed that the intracellular portion of TRA-2 is imported to the nucleus. Since tra-2 requires tra-3 to promote female development, and tra-3 encodes a calpain protease (Barnes and Hodgkin, 1996), TRA-3 might cleave TRA-2 to release this intracellular fragment. This interaction has been confirmed in vitro (Sokol and Kuwabara, 2000). Although germ cells produce a small transcript, tra-2B, that encodes a similar intracellular fragment, this transcript is associated with oogenesis rather than spermatogenesis (Kuwabara et al., 1998), and is unlikely to act in this regulatory process.

Since TRA-1 is a transcription factor, it needs to be in the nucleus to control cell fates. Surprisingly, TRA-1 binds sequences in the 3’-UTR of tra-2 messages (Graves et al., 1999). Furthermore, a mutation in tra-2 that blocks this interaction increases the levels of nuclear TRA-1 (Segal et al., 2001). This mutation also promotes oogenesis, though this effect is not unexpected since it deletes both TGEs, which are needed for translational repression of tra-2. Taken together, these results suggest that tra-2 messages might negatively regulate TRA-1 activity. The export of tra-2 messages requires CRM-1, as well as a group of genes that prevent export by the alternative NXF-1 pathway (Kuersten et al., 2004). Perhaps these complex interactions between tra-2 and TRA-1 help modulate TRA-1 activity during hermaphrodite development.

Hermaphrodites cannot switch from the production of sperm to that of oocytes unless fem-3 activity is repressed by a “Point Mutation Element,” or PME, located in the 3’UTR (Ahringer and Kimble, 1991; Ahringer et al., 1992; Rosenquist and Kimble, 1988). Gain-of-function mutations in fem-3 alter this PME, resulting in increased fem-3 activity and constitutive spermatogenesis. Since these mutations don’t change the level of fem-3 transcripts, but appear to increase the length of the poly(A) tail, they probably increase fem-3 translation.

Using the yeast three-hybrid assay, Zhang et al. (1997) identified two proteins, called FBF-1 and FBF-2, that bind the fem-3 PME. These proteins are remarkably similar in amino acid sequence, and have overlapping functions. The gain-of-function mutations in the PME prevent the FBF proteins from binding the fem-3 3’-UTR. Moreover, inactivation of the fbf genes by RNA-mediated interference recapitulates the phenotype of these fem-3(gf) mutants — the XX animals produce sperm instead of oocytes. Thus, FBF-1 and FBF-2 repress the translation of fem-3 messages by binding the PME (Figure 5). Although fbf-1 and fbf-2 single mutants are usually wild type, recent results suggest that they are not completely redundant, but have slightly different roles in the control of sexual fate and cell proliferation (Lamont et al., 2004).

The FBF proteins are homologs of the Drosophila translational regulator Pumilio, which interacts with Nanos to inhibit translation of hunchback messages. Similarly, the C. elegans FBF proteins bind NOS-3 in the yeast two-hybrid system and in vitro (Kraemer et al., 1999). Furthermore, inactivation of NOS-3 and the two other NOS homologs by RNAi causes many XX animals to produce sperm instead of oocytes (Kraemer et al., 1999). Thus, NOS-3 is likely to act with the FBF proteins to repress translation of fem-3.

The same yeast two-hybrid screen that identified NOS-3 also showed that the GLD-3 protein interacts with FBF-1 and FBF-2 (Eckmann et al., 2002). GLD-3 is a homolog of Bicaudal-C from Drosophila. Mutant analysis showed that GLD-3 plays several roles in germline development, one of which is the promotion of spermatogenesis — in both gld-3(null) males and hermaphrodites, some cells that should have become sperm instead develop as oocytes. This gld-3 phenotype is suppressed by mutations in the fbf genes, which implies that GLD-3 negatively regulates FBF activity (Figure 5).

The activity of fem-3 mRNA is also controlled by a second group of genes. Recessive mutations in mog-1, mog-2, mog-3, mog-4, mog-5 or mog-6 cause hermaphrodites to make sperm instead of oocytes, and also affect the size of the germ line and the viability of embryos (Graham and Kimble, 1993; Graham et al., 1993). By using reporter assays, Gallegos et al. (1998) showed that mog gene activity can also be detected in the soma, where they act through the fem-3 3’UTR to repress reporter constructs. However, mutations in the mog genes don’t cause masculinization of the soma in hermaphrodites. One possible explanation is that maternal mog activity is sufficient for normal development. Alternatively, the level of fem-3 transcripts in the soma of hermaphrodites might be outside the range that is responsive to repression by the mog genes.

Four of the mog genes have been cloned; mog-1, mog-4, and mog-5 encode DEAH-box proteins that are similar in sequence to the yeast proteins PRP16, PRP2, and PRP22, respectively (Puoti and Kimble, 1999; Puoti and Kimble, 2000), and mog-6 encodes an unusual nuclear cyclophilin (Belfiore et al., 2004). All four proteins bind the MEP-1 zinc-finger protein (Belfiore et al., 2002; Belfiore et al., 2004), which is also located in the nucleus, but has essential functions not shared by the mog genes. At this time, the manner in which the MOG complex controls target messages remains unknown.

Two other genes have similar phenotypes. First, mag-1 is the homolog of the Drosophila mago nashi gene, which regulates several aspects of germline development. In C. elegans, mag-1(RNAi) causes hermaphrodites to produce sperm instead of oocytes, and affects embryonic viability (Li et al., 2000). Although these phenotypes suggest that mag-1 functions in the same process as the mog genes, the mag-1(RNAi) phenotype is suppressed by mutations in fog-2 (Li et al., 2000), whereas the mog-1 phenotype is not (Graham and Kimble, 1993). Second, atx-2(RNAi) causes hermaphrodites to make sperm instead of oocytes, in addition to numerous other defects in germline development (Ciosk et al., 2004; Maine et al., 2004).

We suggested above that the crucial factor controlling whether germ cells become sperm or oocytes is the ratio of TRA-2 activity to FEM-3 activity, and described several regulatory pathways that influence these activities. However, we do not yet know why hermaphrodites make sperm early in life, and oocytes later on, since we don’t know which regulatory activities change at the appropriate time. For example, GLD-1 and FOG-2 are expressed in both larval and adult hermaphrodites, so it seems unlikely that repression of tra-2 ends when animals reach sexual maturity, thus allowing the switch to oogenesis, although it remains possible that the activity of a third factor, or a modification to either FOG-2 or GLD-1, might be responsible for the switch.

As proposed by Goodwin and Ellis (2002), three explanations might account for this puzzle. First, one of these regulatory proteins might be modified to change its activity in a manner that could not be detected by RNA or protein cytological analysis. Second, the switch might be caused by a change in the activity of a gene that has not yet been discovered. Third, the increasing size and cell number in the developing germ line might slowly change the relative activities of the tra-2 and fem-3 regulators, so that the ratio of tra-2 to fem-3 activity automatically flips during late larval development.

In theory, these two decisions might be completely independent, so that germ cells entering meiosis merely check the status of the sex-determination pathway to learn if they should begin spermatogenesis or oogenesis. However, many of the genes that act in one pathway also regulate the other. For example, the FBF proteins and GLD-1 control both processes. Furthermore, genes that control progression through the pachytene stage of meiosis also promote spermatogenesis. Are these pleiotropic effects accidental, resulting only because many genes have pleiotropic roles in development? Or do these dual activities help coordinate the overall development of the germ line? Furthermore, do these data shed any light on the developmental point at which germ cells commit to one sexual fate or the other?

The ablation of a single sheath/spermathecal precursor cell in the somatic gonad can completely feminize the germ line in XX animals (McCarter et al., 1997). This result suggests that a second somatic signal might influence germ cell fates, since her-1 is not expressed in XX animals. However, no ligand or receptor has been identified, and it is possible that these ablations exert their influence indirectly, by reducing the size of the germ line. Indeed, direct ablation of a subset of germ cells at the L2/L3 molt also completely feminizes some XX animals (McCarter et al., 1997).

There is also a potential signal transduction pathway that promotes spermatogenesis, but whose mechanism of operation remains unknown. Genes of the ERK MAP Kinase pathway are required for several different germ cell fates, including the progression of cells through the pachytene stage of meiosis I (Church et al., 1995; Hsu et al., 2002). Partial loss of function of lin-45 Raf, mek-2 MEK and mpk-1 ERK can causes hermaphrodites to make only oocytes instead of sperm. Furthermore, complete inactivation of these genes causes germ cells to arrest at the pachytene stage of meiosis and produce oocyte-specific transcripts, even in males (M. Ohmachi, M. H. Lee, E. Lambie and T. Schedl, personal communication). Thus, this pathway normally promotes spermatogenesis. However, the source of the signal, or even whether these genes act in their familiar mode as transducers of a signal, remains unknown.

Genes that regulate sexual fate in germ cells also determine if XX animals develop as females or hermaphrodites. Since new phylogenies show that hermaphroditism likely has evolved independently in C. elegans and C. briggsae (Cho et al., 2004 Kiontke et al., 2004), these nematodes provide an ideal subject for comparative evolutionary studies. In particular, we need to know how the sex-determination pathway is modified over time to create new mating systems (see The evolution of nematode sex determination).

Although we have a solid working model of sex-determination in the C. elegans germ line, the picture from Drosophila and mammals remains confusing. However, homologs of many genes that act in the C. elegans germ line are expressed in the germ cells of other animals. Do any of these genes play conserved roles in germline development, like mab-3 and doublesex do in the soma?

A major problem facing research on the C. elegans germ line is the difficulty in making reliable reporter constructs. Often, transgenic reporters are silenced in the germ line, while expression in the soma is not disrupted. This problem is at least partially caused by the ability of the germ line to shut down transcription of repetitive regions of the genome, such as are often created by transgenes. Despite the use of high-complexity DNA (Kelly et al., 1997) and particle bombardment (Praitis et al., 2001) to counter this problem, these transgenes generally take much more work to produce than similar ones in somatic tissues. New methods to simplify this process, or the identification of related nematodes with less stringent controls on expression in the germ line, could revolutionize the field.