Germ cells can divide mitotically to replenish germline tissue or meiotically to produce gametes. In this article, we report that GLD-3, a Caenorhabditis elegans Bicaudal-C homolog, promotes the transition from mitosis to meiosis together with the GLD-2 poly(A) polymerase. GLD-3 binds GLD-2 via a small N-terminal region present in both GLD-3S and GLD-3L isoforms, and GLD-2 and GLD-3 can be co-immunoprecipitated from worm extracts. The FBF repressor binds specifically to elements in the gld-3S 3′-UTR, and FBF regulates gld-3 expression. Furthermore, FBF acts largely upstream of gld-3 in the mitosis/meiosis decision. By contrast, GLD-3 acts upstream of FBF in the sperm/oocyte decision, and GLD-3 protein can antagonize FBF binding to RNA regulatory elements. To address the relative importance of these two regulatory mechanisms in the mitosis/meiosis and sperm/oocyte decisions, we isolated a deletion mutant, gld-3(q741), that removes the FBF-binding site from GLD-3L, but leaves the GLD-2-binding site intact. Animals homozygous for gld-3(q741) enter meiosis, but are feminized. Therefore, GLD-3 promotes meiosis primarily via its interaction with GLD-2, and it promotes spermatogenesis primarily via its interaction with FBF.
GERM cells can divide by either the mitotic or the meiotic cell cycle and, in animals, they can differentiate as either sperm or oocytes. These fundamental decisions are crucial for animal development and their misregulation can lead to clinical problems. For example, germline tumors are a frequent type of cancer (e.g., Talerman 1985; Ulbright 2004), and defects in germline stem cells may underlie infertility (Fox and Reijo Pera 2001). However, the regulation of mitotic proliferation in the human germline remains poorly understood. Germline stem cells are regulated by signaling from a somatic niche in Caenorhabditis elegans, Drosophila, and probably vertebrates (Kimble and White 1981; Meng et al. 2000; Spradling et al. 2001). Within the germline, stem cells are maintained by RNA-binding proteins of the PUF [for Pumilio and FBF (fem-3 binding factor)] family in both C. elegans and Drosophila (Lin and Spradling 1997; Forbes and Lehmann 1998; Crittenden et al. 2002). In the human testis, Pumilio-2 is expressed in germline stem cells (Jaruzelska et al. 2003; Moore et al. 2003). Therefore, the molecular regulation of germline stem cells may be conserved throughout the animal kingdom.
We have focused on regulation of the mitosis/meiosis decision in C. elegans. Organization of the germline is essentially the same in the two sexes (XX self-fertilizing hermaphrodite and XO male; for review see Hubbard and Greenstein 2000). The mitotic region is located at the distal end; more proximally, germ cells transition into early meiotic prophase; and yet more proximally, germ cells enter the pachytene stage of meiotic prophase. Therefore, progression from mitosis into meiotic prophase is spatially organized in a linear fashion extending from the distal end. We note that virtually all germline “cells” are connected to a common core of cytoplasm, but we call them “cells” for several reasons: they do not divide synchronously, they adopt distinct fates within the syncytium (e.g., mitosis vs. meiosis, sperm vs. oocyte), and each “cell” possesses its own cytoplasmic territory that is largely enclosed by a plasma membrane (reviewed in Hubbard and Greenstein 2000). Therefore, germline mitosis/meiosis and sperm/oocyte controls probably act at the level of individual “cells.”
The mitotic region of the C. elegans germline includes germline stem cells (Crittenden et al. 2003). Unlike Drosophila, where a few well-defined stem cells are located adjacent to the niche (Xie and Spradling 2000; Yamashita et al. 2003), the C. elegans mitotic region comprises about 225 germ cells, most organized in a cylinder that extends ∼20 cell diameters along the gonadal axis from the niche (Figure 2A; Table 2; Crittenden et al. 1994); specific stem cells within the mitotic region have not been identified. It seems most likely that C. elegans germline stem cells are not controlled by the lineage strategy typical of Drosophila germline stem cells, but instead by a regulative strategy, typical of many vertebrate stem cells (Watt and Hogan 2000). The regulative strategy generates distinct populations of cells—some with stem cell potential and others that have begun to differentiate. In the C. elegans germline, a molecular understanding of how those populations are established is emerging (see discussion).
This article focuses on the role of GLD-3 (germline development) in controlling the mitosis/meiosis and sperm/oocyte decisions in the C. elegans germline. To place our GLD-3 studies in context, we first introduce other RNA regulators critical to these decisions. Table 1 summarizes their molecular identities and biological functions. Briefly, FBF-1 and FBF-2 are nearly identical and collectively are called FBF; they belong to the PUF family of RNA-binding proteins and control both mitosis/meiosis and sperm/oocyte decisions (Zhang et al. 1997; Crittenden et al. 2002). PUF proteins bind specifically to 3′-UTR regulatory elements and repress their target mRNAs (Wickens et al. 2001). FBF controls germline fates, at least in part by repressing gld-1 (Crittenden et al. 2002). FBF achieves the switch from spermatogenesis to oogenesis together with NOS-3, one of three nematode Nanos homologs, by repressing fem-3 (Zhang et al. 1997; Kraemer et al. 1999).
The decision to leave the mitotic cell cycle and enter meiosis is promoted by two GLD proteins (Kadyk and Kimble 1998) and by NOS-3 (Hansen et al. 2004). GLD-1 is a STAR/GSG/quaking RNA-binding protein and translational repressor (Jones and Schedl 1995; Jan et al. 1999). GLD-2 is the catalytic subunit of an atypical poly(A) polymerase (Wang et al. 2002). GLD-1 and GLD-2 act redundantly to promote the switch from the mitotic to the meiotic cell cycle (Kadyk and Kimble 1998). Thus, in gld-1 and gld-2 single mutants, germ cells make the transition from mitosis into meiosis, but in gld-1 gld-2 double mutants, the germline is largely mitotic with little or no entry into meiosis (Kadyk and Kimble 1998). This synergistic effect suggested that gld-1 and gld-2 activities act in parallel to control entry into meiosis. The nos-3 gene promotes GLD-1 accumulation and appears to act in the gld-1 branch of regulation (Hansen et al. 2004).
GLD-3 belongs to the Bicaudal-C family of RNA-binding proteins, and it has multiple roles in germline development (Eckmann et al. 2002). In germline sex determination, GLD-3 controls continued spermatogenesis in males and the number of sperm produced in self-fertilizing hermaphrodites; in addition, GLD-3 is required for germline survival, progression through meiosis, and early embryogenesis (Eckmann et al. 2002). GLD-3 interacts with GLD-2 as a subunit of an atypical poly(A) polymerase activity (Wang et al. 2002). Whereas the canonical poly(A) polymerase is monomeric and enriched in the nucleus (Raabe et al. 1994; Martin et al. 1999), the GLD-2/GLD-3 enzyme is heterodimeric and resides predominantly in the cytoplasm (Eckmann et al. 2002; Wang et al. 2002).
In this article, we investigate the role of GLD-3 in the mitosis/meiosis decision. We show that GLD-3 works together with GLD-2 to control entry into meiosis and that it works in parallel to GLD-1 and NOS-3. In addition, we demonstrate that FBF acts upstream of GLD-3 in the mitosis/meiosis decision, binds the gld-3S 3′-UTR and represses GLD-3 protein expression. By contrast, GLD-3 acts upstream of FBF in the sperm/oocyte decision by antagonizing the binding of FBF to its 3′-UTR regulatory element (Eckmann et al. 2002). To explore the relationships among FBF, GLD-3, and GLD-2, we first show that FBF and GLD-2 bind to distinct regions of GLD-3. We then generate a GLD-3 deletion mutant, gld-3(q741), that removes the FBF-binding region, but leaves the GLD-2-binding region and KH domains intact. This gld-3 deletion mutant is defective in the sperm/oocyte decision, but the mitosis/meiosis decision is virtually wild type. We conclude that GLD-3 promotes meiosis primarily via its interaction with GLD-2 and that it promotes spermatogenesis primarily via its interaction with FBF.
MATERIALS AND METHODS
Nematode strains and RNAi:
Worm strains are derivatives of the wild-type Bristol strain N2 and were maintained at 20° as described by Brenner (1974). All mutations used were either internal deletions or nonsense mutations, and all are putative nulls: LGI—gld-1(q485) (Jones and Schedl 1995), gld-2(q497) (Kadyk and Kimble 1998; Wang et al. 2002); LGII—fbf-1(ok91), fbf-2(q704) (Crittenden et al. 2002), gld-3(q730) (Eckmann et al. 2002). The gld-3(q741) allele was isolated in a PCR-based screen for deletion mutants induced by ethyl methane sulfonate [see Kraemer et al. (1999) for method], outcrossed against wild-type (Bristol, N2) eight times, and balanced with mIn1. Primers used for screening were as described (Eckmann et al. 2002). The deletion removes nt 1912–4076 in the genomic sequence, where the A in ATG is the first nucleotide. The gld-3(q741) mutation is recessive with respect to fertility and makes a few Fog hermaphrodites. To balance mutations linked to LGI, we used unc-15(e73) with a closely linked GFP transgene (ccIs4251), and to balance mutations on chromosome II, we used mIn1[mIs14 dpy-10(e128)]. Double, triple, and quadruple mutants on LGII were generated by recombination, balanced by mIn1[mIs14 dpy-10(e128)], and validated by PCR for deletions or by sequencing for q704. Adult germline phenotypes were scored 24 hr past L4 stage at 20°.
By standard procedures (Kamath and Ahringer 2003), HT115 cells were transformed with a construct to generate double-stranded RNA corresponding to gld-3 (nt 1–1719), a region common to both gld-3S and gld-3L. When animals were fed gld-3(RNAi) bacteria, defects were similar to those observed after injection of a similar gld-3 region (Eckmann et al. 2002). Zygotic defects were visible in the first window (16–36 hr) after feeding. The success of gld-3(RNAi) for any given experiment was scored by continued feeding to ensure that later progeny displayed the maternal germline survival phenotype (Eckmann et al. 2002). In addition, nos-3(q650) gld-3(RNAi) animals had the same phenotype as nos-3(q650) gld-3(q730) double mutants, and GLD-3 protein was either significantly decreased or undetectable when animals fed gld-3(RNAi) bacteria were immunostained for GLD-3 (data not shown).
DAPI staining and immunocytochemistry:
Germlines were extruded from the animal and stained with 4,6-diamidino-2-phenylindole (DAPI) staining or immunocytochemistry by standard methods (Crittenden et al. 1994). The extent of each region was assayed by nuclear morphology using DAPI staining and confirmed using affinity-purified antibodies: anti-GLP-1 for the mitotic region (Crittenden et al. 1994), anti-HIM-3 (kindly provided by M. Zetka) for meiotic cells (Zetka et al. 1999), and anti-phosphohistone H3 (PH3) to mark cells in mitosis (Hendzel et al. 1997). Antibodies used for examining RNA regulators included: affinity-purified anti-GLD-1 (kindly provided by T. Schedl), anti-GLD-2 (Wang et al. 2002), anti-GLD-3 (Eckmann et al. 2002), anti-FBF-1 (Crittenden et al. 2002), and anti-NOS-3 (Kraemer et al. 1999). Anti-RME-2 antibodies (kindly provided by B. Grant) were used to demonstrate the production of oocytes in males. Positive controls for antibodies were included in all experiments. Epifluorescent images were captured with a Zeiss Axioskop equipped with a Hamamatsu digital CCD camera (Hamamatsu Photonics) and processed with Openlab 3.1.5 (Improvision).
Worm extracts, in vitro translation, immunoprecipitation, and yeast two hybrid:
Worm extracts were generated by boiling worms in SDS sample buffer for 5 min. Protein separation was achieved on 4–12% gradient gels and subsequent immunoblots onto PVDF membranes were carried out according to standard procedures. Secondary antibodies were used as described (Eckmann et al. 2002).
In vitro translation in TNT-T7-coupled reticulocyte lysate was carried out according to the manufacturer's protocol (Promega). A full-length cDNA clone for gld-3LΔ was generated by RT-PCR from gld-3(q741) worms with primers including convenient restriction sites (EcoRI and SalI). The open reading frame and the entire 3′-UTR was cloned into pTNT (Promega). A construct containing the entire gld-3S cDNA in pTNT was generated with similar primers.
For immunoprecipitation, N2 worms were grown on standard NGM agar plates, collected by centrifugation, and then washed multiple times in M9 buffer, with a final wash in NP40 extraction buffer (150 mm NaCl, 1.0% NP-40, 50 mm Tris, pH 8.0). The equivalent of 1 ml packed worms was resuspended in 5 ml total extraction buffer, supplemented with an EDTA-free protease inhibitor cocktail (Roche). Worms were frozen immediately in liquid nitrogen and homogenized with a cooled mortar and pestle. The frozen extract was thawed on ice, cleared by centrifugation at full speed for 20 min in a cooled Eppendorf tabletop centrifuge and diluted to 1 mg protein/1 ml buffer. For each immunoprecipitation, 500 μg extract was incubated with 10 μl preimmune serum or anti-GLD-2 antibody (Wang et al. 2002) coupled protein G beads (Pharmacia) for 2 hr at 4° on a rotating platform. Protein G beads were subsequently washed five times with 1 ml NP40 buffer. Bound proteins were eluted in SDS sample buffer, boiled, and separated on 4–12% gradient gels, which were then used to prepare Western blots. GLD-3L antibodies were used as a primary antibody as in Eckmann et al. (2002).
Yeast two-hybrid experiments were done as described by Eckmann et al. (2002). Full-length gld-2 ORF in pACT is described in Wang et al. (2002). Fragments of gld-3 ORF were inserted into pBTM116. Full-length gld-3 ORF and N-terminal deletion is described in Eckmann et al. (2002).
Yeast three-hybrid and gel retardation assays:
Three-hybrid experiments were performed as described by Bernstein et al. (2002). gld-3S RNA sequences were cloned into the XmaI and SphI sites of the vector pIIIA/MS2-2, using either PCR-amplified fragments (an RNA segment containing FBE-a + b and RNA segment FBE-c) or annealed synthetic oligonucleotides (individual FBF-binding sites, FBE-a and FBF-b). For binding specificity, a mutation (UG to AC) was engineered by site-directed mutagenesis using Quickchange (Stratagene).
Fusion proteins were made as described by Crittenden et al. (2002). Purification of GST-FBF-1 protein was essentially carried out according to Crittenden et al. (2002) and dialyzed against 10 mm Hepes buffer, pH 7.9. In vitro binding reactions were performed with purified GST-FBF-1 protein incubated in the presence of buffer D (Ohno et al. 2000) supplemented with indicated radiolabeled and gel-purified gld-3S RNA fragments. The reactions were resolved on a 6% native polyacrylamide gel under standard conditions.
gld-1 and gld-3 activities are redundant for control of entry into meiosis:
To determine if gld-3 acts in parallel to gld-1, we generated animals deficient for both gld-1 and gld-3 activities. We first found gld-1; gld-3 double mutants to be inviable; characterization of that lethality is beyond the scope of this work. We therefore used RNAi to deplete gld-3 activity in gld-1 mutants (see materials and methods). The gld-1; gld-3(RNAi) germlines were tumorous: no nuclei typical of early meiosis were identified by DAPI staining (Figure 1A), the mitotic marker anti-PH3 stained many nuclei throughout the germline (Figure 1C), and the meiotic marker anti-HIM-3 was not detectable (Figure 1D). GLP-1, a marker that correlates with mitosis, was also evenly distributed throughout the gld-1; gld-3(RNAi) tumorous germlines (data not shown). The germlines of gld-1 gld-2; gld-3(RNAi) animals were also exclusively tumorous (Figure 1, E and F). Enlarged defective nuclei were frequent in gld-1; gld-3(RNAi) germlines (60%, n = 25; Figure 1B), whereas these unusual nuclei were rare in gld-1 gld-2 tumors (6%, n = 16) and gld-1 gld-2; gld-3(RNAi) tumors (25%, n = 18). Therefore, gld-1 gld-2 and gld-1; gld-3 are similar, but not identical, germline tumors. We conclude that wild-type gld-3 promotes meiosis in the absence of gld-1, and vice versa; therefore, these two genes appear to act in parallel to control entry into meiosis.
Entry into meiosis is delayed in gld-2 and gld-3 single mutants:
Animals homozygous for either gld-2 or gld-3 possess a mitotic region (MR), transition zone (TZ), and pachytene region (PR) arranged normally (Kadyk and Kimble 1998; Eckmann et al. 2002). To ask if gld-2 and gld-3 single mutants have some minor defect in the transition from mitosis to meiosis, we examined the MR/TZ and TZ/PR borders. To measure the position of these borders, we counted number of nuclei along the distal-proximal axis from the distal end to the MR/TZ boundary and from the distal end to the TZ/PR boundary; in addition, we counted total number of nuclei in the MR; all measurements were made in DAPI-stained adult hermaphrodite germlines that had been extruded from the animals at 24 hr past L4. Some animals were also stained with the mitotic marker anti-phosphohistone (Hendzel et al. 1997) and the meiotic marker anti-HIM-3 (Zetka et al. 1999) to confirm the DAPI staining results (data not shown).
In wild-type animals, the MR extends ∼20 nuclei along the distal-proximal axis and includes a total of ∼225 nuclei (Crittenden et al. 1994; Figure 2A; Table 2). By contrast, the MRs of gld-2 and gld-3 single mutants were longer and had more nuclei than normal (Figure 2, B and C; Table 2). The MR/TZ boundary was shifted to an average of ∼27 or 28 nuclei in gld-2 and gld-3 mutant germlines (Table 2). Furthermore, the boundaries between MR and TZ and TZ and PR were not as sharp in gld-2 and gld-3 mutants as they were in wild-type germlines. In gld-2 and gld-3 single-mutant L4s, we observed a similar, although less pronounced, extension of the MR. By comparison, the MRs of gld-1 and nos-3 mutants were not longer than normal (Table 2). Indeed, the gld-1 mitotic region was shorter (Table 2), suggesting that GLD-1 plays some minor role in maintaining mitosis in addition to its major role in directing entry into meiosis. We conclude that both gld-2 and gld-3 are required for determining the position at which germ cells enter meiosis and that removal of either activity extends the mitotic region.
We next examined animals lacking both gld-2 and gld-3. If gld-2 and gld-3 act together to control entry into meiosis, then the double mutant should be no more defective than either single mutant. We first attempted to make a gld-2; gld-3 double mutant, but it was inviable; therefore, these genes act synergistically in embryogenesis, but characterization of that synergy is beyond the scope of this work. We therefore depleted gld-3 from gld-2 mutants using RNAi. The transition from mitosis to meiosis in gld-2; gld-3(RNAi) germlines was virtually identical to that of gld-2 and gld-3 single mutants (Figure 2D; Table 2). Surprisingly, most gld-2; gld-3(RNAi) germ cells arrested in pachytene, a synthetic phenotype suggesting redundancy for progression into diplotene (data not shown). However, the focus of this article is entry into meiosis, and we emphasize the lack of a synergistic effect for gld-2 and gld-3 in the mitosis/meiosis decision.
GLD-2/GLD-3 molecular interaction:
The genetic analyses described above indicate that GLD-2 and GLD-3 work together to control entry into meiosis. Consistent with this hypothesis, GLD-2 and GLD-3 function biochemically as subunits of an atypical poly(A) polymerase (Wang et al. 2002). The gld-3 gene encodes two major proteins, GLD-3L and GLD-3S; in the N-terminal half, these two proteins share five KH motifs as well as parts of a serine-rich region (Eckmann et al. 2002; Figure 3A). The larger protein, GLD-3L, possesses a C-terminal extension that binds FBF (Eckmann et al. 2002). To learn what part of GLD-3 is responsible for its specific binding to GLD-2, we generated GLD-3 deletion mutants and assayed them for interactions with GLD-2 by yeast two-hybrid assay. Full-length GLD-3L interacts with GLD-2, as do two N-terminal GLD-3 fragments containing amino acids 1–81 or 34–81 (Figure 3B). These N-terminal fragments did not bind FBF (data not shown). An N-terminal fragment containing amino acids 55–81 failed to interact with GLD-2, suggesting that amino acids 34–55 are required for GLD-2 interaction. Furthermore, a large C-terminal fragment carrying amino acids 80–969 failed to interact with GLD-2 (Figure 3B), but did retain its ability to bind FBF (Eckmann et al. 2002). In addition, an N-terminal fragment of GLD-3 (amino acids 1–102) binds GLD-2 in vitro (data not shown). Interestingly, aa 34–81 are strongly conserved in GLD-3 homologs from each of three Caenorhabditis species (Figure 3A). We conclude that specific binding to GLD-2 is conferred by a small region that is present in both GLD-3L and GLD-3S.
We also tested for protein complexes containing both GLD-2 and GLD-3 proteins in worm extracts. To this end, we immunoprecipitated GLD-2 using anti-GLD-2 antiserum bound to protein G beads. Specifically, we incubated extracts prepared from mixed stage worms with beads carrying either anti-GLD-2 antibodies or preimmune serum. GLD-3L antibodies recognized a band of the correct size in the GLD-2 immunoprecipitate, but did not recognize any band in the precipitate using preimmune serum (Figure 3C). We conclude that GLD-2 and GLD-3 are associated with each other in nematodes, consistent with the idea that they work together to control the mitosis/meiosis decision.
Do gld-2 and gld-3 promote meiosis in fbf-1 fbf-2 double mutants?
Whereas wild-type germlines make >1000 descendants/gonadal arm, the fbf-1 fbf-2 double mutant generates only ∼70 germ cells/arm, and all of these enter meiosis and make sperm (Crittenden et al. 2002; Table 3; Figure 5A). By contrast, gld-2 and gld-3 single mutants have a longer than normal mitotic region. To determine if gld-2 or gld-3 is required for abnormal entry into meiosis in fbf-1 fbf-2 double mutants, we examined gld-2; fbf-1 fbf-2 and fbf-1 fbf-2 gld-3 triple-mutant germlines. On average, these triple mutants [and the quadruple gld-2; fbf-1 fbf-2 gld-3(RNAi)] made about twice as many germ cells as the fbf-1 fbf-2 double mutant (Table 3). We conclude that the mitosis/meiosis defect typical of gld-2 and gld-3 single mutants, which is an increased capacity for mitosis, was also observed in the triple and quadruple mutants that lack both fbf and gld activities. Therefore, fbf appears to act upstream of gld-2 and gld-3 (also see below).
In contrast to the effect of gld-2 and gld-3, removal of nos-3 had a more dramatic effect. Whereas nos-3 germlines have no detectable effect on the mitosis/meiosis decision (Table 2), fbf-1 fbf-2 nos-3 germlines made ∼500 descendants/gonadal arm (Table 3; Hansen et al. 2004). This effect is more dramatic than that of gld-2 or gld-3 and is reminiscent of gld-1; fbf-1 fbf-2 germlines, which are fully tumorous (Crittenden et al. 2002). We conclude that nos-3 is critical for promoting meiosis in the absence of FBF.
The nos-3 gene acts in the gld-1 branch and parallel to gld-2/gld-3 to promote meiosis:
To determine how nos-3 fits into the regulatory pathway controlling mitosis/meiosis, we examined the transition from mitosis to meiosis in double and triple mutants (Figure 4; Table 4). No synergistic effect was seen between the gld-1 and nos-3 genes: gld-1; nos-3 double-mutant germlines were similar to those of gld-1 single mutants (Figure 4A; Tables 2 and 4; Hansen et al. 2004). By contrast, we did see a synergistic effect between nos-3 and both gld-2 and gld-3. In gld-2; nos-3 double mutants, aberrant mitoses were observed in the proximal germline (Figure 4B; Hansen et al. 2004). Such proximal germline mitoses are not typical of either gld-2 or nos-3 single mutants, although they are seen in 2% of strong loss-of-function gld-2(q497) germlines (Kadyk and Kimble 1998; Kraemer et al. 1999). More dramatically, gld-3 nos-3 germlines were uniformly tumorous, as assayed by DAPI staining and antibodies to PH3 (Figure 4, C and D) and antibodies to GLP-1 (not shown). Indeed, we observed no sign of early meiosis in gld-3 nos-3 germlines. Therefore, both gld-2 and gld-3 exhibit synergistic defects with nos-3, but the effect is more severe with gld-3.
The difference between the gld-2; nos-3 and gld-3 nos-3 double-mutant germlines underscores the idea that gld-2 and gld-3 do not have identical roles in maintaining the mitosis/meiosis decision. Because nos-3 is not synergistic with gld-1, but is synergistic with both gld-2 and gld-3, albeit at different levels, we suggest that nos-3 acts in the same regulatory branch as gld-1 to promote meiosis (Figure 9A; see also Hansen et al. 2004). Furthermore, we suggest that nos-3 acts in parallel with gld-3 to control entry into meiosis and in parallel with gld-2 to maintain commitment to meiosis. This difference between the effects of gld-2 and gld-3 may reflect the reciprocal repression between FBF and GLD-3 (see discussion).
Regulatory relationship between fbf and gld/nos genes:
We next investigated epistasis in quadruple mutants, which included the fbf-1 fbf-2 double mutant in combination with various synthetic tumorous double mutants. All germlines were scored in young adults, when fbf-1 fbf-2 germlines normally have no mitotic cells, but instead consist solely of mature sperm (Figure 5A). By contrast, the gld-1 gld-2; fbf-1 fbf-2 and gld-1; fbf-1 fbf-2 gld-3(RNAi) germlines were fully tumorous (Table 4). This tumorous state was similar to that of the gld-1; fbf-1 fbf-2 triple mutant reported previously (Crittenden et al. 2002).
The fbf-1 fbf-2 gld-3 nos-3 and gld-2; fbf-1 fbf-2 nos-3 quadruple-mutant germlines were also largely tumorous: they proliferated more than any of the triple mutants (fbf-1 fbf-2 gld-3, gld-2; fbf-1 fbf-2, or fbf-1 fbf-2 nos-3; Figure 5B; not shown). In both quadruple mutants, the distal mitotic region was even longer than that in gld-2 or gld-3 single mutants and proximal mitoses were consistently observed (Figure 5, C and D). However, the germlines did enter meiosis. Therefore, the gld-3 nos-3 tumor is not completely epistatic to fbf-1 fbf-2. We suggest that this partial epistasis may reflect reciprocal repression between FBF and GLD-3 (see discussion), although alternative explanations are plausible.
FBF binds elements in the gld-3S 3′-UTR:
To ask whether gld-3 might be a direct target of FBF, we examined the sequence of its 3′-UTRs for potential FBF-binding sites. All known PUF-binding sites contain a core UGU(G/A) tetranucleotide, but FBF does not bind specifically to all such sequences (Wickens et al. 2000). We therefore scanned 3′-UTRs for this core sequence in a context similar to that found for established FBF-binding sites in fem-3 and gld-1 mRNAs (Zhang et al. 1997; Crittenden et al. 2002). The gld-3 gene expresses two major transcripts, gld-3L and gld-3S (Figure 8A). Two potential sites in the gld-3L 3′-UTR were tested by three-hybrid assay, but they did not bind FBF-1 (data not shown). However, FBF-1 and FBF-2 did interact specifically with two of the three potential sites identified within the 191-nt 3′-UTR of the gld-3S mRNA (Figure 6). By three-hybrid assay, a region covering two closely linked core elements (FBE-a and FBE-b) in the gld-3S 3′-UTR (Figure 6A) interacted strongly and reproducibly with FBF (Figure 6, B and C). Furthermore, individual sequence elements FBE-a and FBE-b were sufficient to bind FBF in yeast three-hybrid assays (Figure 6, B and C). Indeed, both of these sites are strongly conserved in gld-3S 3′-UTRs from each of the three Caenorhabditis family members (Figure 6A, bottom). Another putative site (FBE-c, Figure 6A) did not bind FBF, although it expressed well in yeast (Figure 6, B and C; data not shown). In gel shifts, a fragment carrying both FBE-a and FBE-b sites was bound in vitro by purified FBF protein (Figure 6D, left), as was a fragment carrying only the FBE-b site (Figure 6D, right). As controls for specificity, we mutated the first two core nucleotides, which abolished binding for the individual sequence elements FBE-a and FBF-b. We conclude that FBF-1 and FBF-2 bind specifically to two sites within the gld-3S 3′-UTR.
FBF controls GLD-3 expression:
To further test the idea that FBF represses gld-3 expression, we compared the level of GLD-3 protein in wild-type and fbf-1 fbf-2 germlines by immunocytochemistry (Figure 7). Specifically, we examined GLD-3 in late L3 or early L4 males using affinity-purified antibodies that recognize both GLD-3S and GLD-3L (Eckmann et al. 2002). We used immunocytochemistry rather than Western blots in this experiment because GLD-3 levels vary with position and we wanted to monitor expression at specific locations in the germline (Eckmann et al. 2002; Figure 7, C and D). We used males in this experiment to eliminate sexual differences between the two germlines (both wild-type males and fbf-1 fbf-2 mutants make sperm only), and we used late L3 or early L4 animals to ensure the presence of mitotic, transition, and pachytene regions (fbf-1 fbf-2 mutants lose their mitotic region in late L4 or adulthood). Wild-type and mutant germlines were stained in parallel and images were captured and processed using identical settings.
In wild-type males, the midlarval germline has a distal mitotic region, a transition zone, and some pachytene cells (Figure 7A). The same regions are present in fbf-1 fbf-2 males in larval stage four (Figure 7B). In the wild type, GLD-3 was detectable, but only at a low level in the mitotic region, barely detectable in the transition zone, and higher in the pachytene region (Figure 7C). In contrast, the majority (65%, n = 20) of fbf-1 fbf-2 germlines displayed higher levels of GLD-3 protein than did wild type in all three regions (Figure 7D), but some (25%, n = 20) showed higher GLD-3 levels restricted to the TZ and the pachytene region. The remaining 10% (n = 20) showed no difference in GLD-3 levels. The anti-GLD-3 antibodies recognize both GLD-3L and GLD-3S, so we cannot exclude the possibility that FBF regulates gld-3L expression as well. We conclude that FBF negatively regulates GLD-3 expression and suggest that FBF represses gld-3 directly by binding an element in the gld-3S mRNA.
GLD-3 protein contains two distinct regulatory domains:
Using a standard PCR-based method to screen for deletion mutants, we isolated gld-3(q741), which removes most of the C-terminal region from GLD-3, including the FBF-binding domain (Eckmann et al. 2002; Figure 8A). Wild-type animals make two major GLD-3 proteins, GLD-3L and GLD-3S (Figure 8B, lane 1), and the strong loss-of-function gld-3(q730) mutant makes no detectable GLD-3 protein (Figure 8B, lane 2; Eckmann et al. 2002). In gld-3(q741) homozygotes, we observed two major proteins (Figure 8B, lane 3): the larger protein in lane 3 corresponds in size to in vitro translated protein from a gld-3L(Δ) cDNA carrying the same deletion as q741 (Figure 8B, lane 4), and the smaller protein in lane 3 corresponds in size to in vitro translated GLD-3S (Figure 8B, lane 5). Therefore, gld-3(q741) homozygotes express full-length GLD-3S protein and a stable truncated version of GLD-3L.
We next examined the mitosis/meiosis and sperm/oocyte decisions in gld-3(q741) homozygotes. The mitotic region of gld-3(q741) germlines was similar to that of wild type, albeit slightly longer (Table 2). Therefore, GLD-3 protein that retains its GLD-2-binding region, but lacks an FBF-binding region, does not have the dramatic mitosis/meiosis delay seen in gld-3 null mutants. By contrast, sex determination was not normal; gld-3(q741) males produced oocytes, as assayed by both Nomarski appearance (Figure 8C) and anti-RME-2 staining (100%, n = 45). We suggest that the primary role of GLD-3 protein in the mitosis/meiosis decision is carried out by its interaction with GLD-2 and that its primary role in the sperm/oocyte decision relies on its interaction with FBF. Consistent with this interpretation, the gld-3(q741) nos-3(q650) double-mutant germ cells enter meiosis (data not shown).
This article focuses on GLD-3 and its control of the mitosis/meiosis decision in the C. elegans germline. Our results support four major conclusions. First, GLD-3 works together with GLD-2 to control the mitosis/meiosis decision. Second, GLD-3 acts in parallel to GLD-1 and NOS-3 in the regulatory network controlling mitosis or meiosis. Third, FBF represses gld-3 expression. And fourth, the molecular mechanisms by which GLD-3 controls the mitosis/meiosis and sperm/oocyte decisions are largely distinct. Our discussion places these results in the context of a regulatory circuit that controls the mitosis/meiosis decision and contrasts that circuit and its molecular underpinnings to one controlling the sperm/oocyte decision (Figure 9). Collectively, our results demonstrate how a molecular switch can be modified and refined to control distinct decisions, e.g., mitosis/meiosis and sperm/oocyte, in a developing tissue.
GLD-3 acts with GLD-2 to promote entry into meiosis:
The gld-2 gene controls entry into meiosis (Kadyk and Kimble 1998), but the role of gld-3 in this process had not been explored. Instead, gld-3 was known to function in the sperm/oocyte decision, germline survival, progression through meiosis, and early embryogenesis (Eckmann et al. 2002). The discovery that GLD-3 partners with GLD-2 to form a poly(A) polymerase (Wang et al. 2002) raised the question of whether GLD-3 might work together with GLD-2 to control entry into meiosis. In this article, we provide several lines of evidence that this is the case. First, both gld-2 and gld-3 function in parallel with gld-1 to direct meiosis (Kadyk and Kimble 1998; this article). Second, both gld-2 and gld-3 single mutants fail to enter meiosis at the normal position, but instead both enter meiosis more proximally than normal. And third, the gld-2; gld-3 double mutant is indistinguishable from either single mutant with respect to entry into meiosis. The idea that GLD-2 and GLD-3 act together as a protein complex is supported by identification of a specific region within GLD-3 that is responsible for binding to GLD-2 as well as by their co-immunoprecipitation from worm extracts. Therefore, both genetic and molecular lines of evidence support the idea that GLD-2 and GLD-3 act together to control germline fates and, in particular, that they control entry into meiosis.
In addition to their ability to function together, GLD-2 and GLD-3 also have distinct roles in germline development. For example, GLD-3 is required for maintenance of spermatogenesis in male germlines (Eckmann et al. 2002), while GLD-2 is apparently not involved in this process (Kadyk and Kimble 1998). Similarly, GLD-3 is required for germline survival (Eckmann et al. 2002), while GLD-2 has not been implicated in this process. Finally, most germ cells in gld-2; gld-3 double mutants arrest in pachytene (this work), whereas those in single mutants progress into diplotene (Kadyk and Kimble 1998; Eckmann et al. 2002), suggesting that GLD-2 and GLD-3 act independently to control the transition from pachytene to diplotene. We envision that GLD-2 and GLD-3 interact with distinct proteins to accomplish their individual roles in germline development. For example, GLD-2 may interact with other RNA-binding proteins in addition to GLD-3 to polyadenylate mRNAs that are different from those targeted by the GLD-2/GLD-3 heterodimer. And GLD-3 may interact with other proteins to accomplish its unique roles. Consistent with this idea, GLD-3 promotes continued spermatogenesis by antagonizing FBF (Eckmann et al. 2002).
Two regulatory branches promote entry into the meiotic cell cycle:
Entry into meiosis is controlled by two major regulatory branches (Kadyk and Kimble 1998; Hansen et al. 2004; this work; Figure 9A). The GLD-1 and NOS-3 proteins reside in one branch, and GLD-2 and GLD-3 reside in the other. The GLD-1/NOS-3 branch is defined by two translational repressors (Jan et al. 1999; Kraemer et al. 1999; Lee and Schedl 2001; Marin and Evans 2003; Ryder et al. 2004). One simple model is that the GLD-1/NOS-3 branch represses mRNAs that promote mitosis (Figure 9A). Of particular note in this context, GLD-1 represses glp-1 mRNA (Marin and Evans 2003), which encodes the GLP-1/Notch receptor essential for germline mitoses (Crittenden et al. 2003). Therefore, GLD-1 ensures entry into meiosis, at least in part, by acting in a negative feedback loop to inhibit a mitosis-promoting regulator.
The GLD-2/GLD-3 regulatory branch is defined by two subunits of an atypical poly(A) polymerase (Wang et al. 2002). We suggest that GLD-2 and GLD-3 may polyadenylate and activate meiosis-promoting mRNAs (Figure 9A); in other systems, polyadenylation usually stimulates mRNA expression (Groisman et al. 2002). One possible target of the GLD-2/GLD-3 poly(A) polymerase is gld-1 mRNA; preliminary results indicate that the gld-1 poly(A) tail is shorter in gld-2 mutants than in wild type (N. Suh and J. Kimble, unpublished observations). In addition, Hansen et al. (2004) found that gld-2 and nos-3 act redundantly to control GLD-1 protein accumulation, which led them to suggest that gld-1 might be a GLD-2 target (Hansen et al. 2004). Although gld-1 mRNA may indeed be one direct target, GLD-2/GLD-3 poly(A)polymerase must target additional mRNAs since GLD-2 can promote meiosis in the absence of GLD-1 activity.
The GLD-1/NOS-3 and GLD-2/GLD-3 regulatory branches are redundant for entry into meiosis, but they most likely exert their effects by opposite molecular mechanisms—RNA repression for GLD-1/NOS-3 and activation for GLD-2/GLD-3. Such a dual mechanism drives the switch from mitosis to meiosis by simultaneously turning on meiosis-specific activities and turning off mitosis-specific activities. That dual mechanism also makes the circuit remarkably robust, a feature that would be selected to control fundamental decisions central to the survival of the species.
A similar two-pronged mechanism controls mating type in the yeast Saccharomyces cerevisiae, but does so at the level of transcription. In this case, the MATα mating-type allele encodes α1, a positive regulator of α-specific gene transcription as well as α2, a negative regulator of a-specific gene transcription; as a result, yeast cells carrying MATα adopt an α-specific fate, turning on α-specific genes and turning off a-specific genes (Herskowitz 1989). The same basic logic regulates entry into meiosis in C. elegans, but here the circuitry relies on RNA controls instead of DNA controls.
FBF inhibits both major meiosis-promoting branches:
FBF promotes mitosis by repressing mRNAs in both of the major meiosis-promoting regulatory branches (Figure 9A). Specifically, it represses both gld-1 expression (Crittenden et al. 2002) and gld-3 expression (this work). Thus, FBF binds specifically to sites within the gld-1 and gld-3S 3′-UTRs, and the abundance of GLD-1 and GLD-3 proteins is increased in fbf mutants. Mutant analysis is consistent with the idea that FBF represses gld-3: when gld-3 is removed from an fbf-1 fbf-2 double mutant, an extra cell division occurs. However, a complexity in the interpretation of these results arises from a comparison of the gld-3 nos-3 and fbf-1 fbf-2 gld-3 nos-3 germline tumors. Although both strains display vastly overproliferating germlines, the former is fully tumorous, while the latter possesses some cells that have entered meiosis. Therefore, loss of fbf does have a minor effect on the gld-3 nos-3 tumorous phenotype. This lack of complete epistasis may be explained by reciprocal repression between FBF and GLD-3 (Figure 9A; see below), although other explanations are possible.
FBF/GLD balance controls size of the mitotic region:
A balance between FBF and GLD-2/GLD-3 activities appears to control the number of cells committed to mitosis or meiosis. fbf-1 single mutants possess a smaller mitotic region than normal, suggesting that a full complement of both FBF-1 and FBF-2 activities is required to achieve a normally sized mitotic region (Crittenden et al. 2002). By contrast, gld-2 and gld-3 mutants have a larger mitotic region than normal (this work). Therefore, GLD-2 and GLD-3 are also required for a normally sized mitotic region, but their effect is opposite to that of FBF. In the case of GLD-1, the mitotic region becomes smaller, which we do not yet understand, but Hansen et al. (2004) suggest that GLD-1 affects the size of the mitotic region. Taking these results together, we suggest that a balance between FBF and GLD-2/GLD-3 controls cell number in the mitotic region. In an fbf-1 mutant, which has less FBF than normal, the balance is tipped toward GLD activity and premature meiosis; in gld-2 and gld-3 mutants, the balance is tipped in favor of FBF and extended mitosis.
Distinct roles of GLD-3 in the mitosis/meiosis and sperm/oocyte decisions:
GLD-3 has distinct roles in the mitosis/meiosis and sperm/oocyte decisions (Figure 9, A and B). In the sperm/oocyte decision, gld-3 acts upstream of fbf (Eckmann et al. 2002), but in the mitosis/meiosis decision, gld-3 acts largely downstream of fbf (this work). At a molecular level, GLD-3L protein can bind FBF and antagonize FBF RNA-binding activity (Eckmann et al. 2002) and, conversely, FBF can bind the gld-3S 3′-UTR and repress gld-3 expression (this work). Therefore, the fbf and gld-3 genes are capable of reciprocal repression.
We find that the two opposing FBF/GLD-3 regulatory mechanisms are differentially important for the mitosis/meiosis and sperm/oocyte decisions. Essential to this argument is gld-3(q741), a deletion mutant that removes the FBF-binding region from GLD-3L, but leaves its GLD-2-binding region and KH motifs intact. The germlines of gld-3(q741) homozygotes enter meiosis at a position close to that of the wild type, but fail to make sperm continuously in males. Therefore, the FBF-binding region of GLD-3 is crucial for the sperm/oocyte decision, but it plays a minor role, if any, in the mitosis/meiosis decision. The presence of distinct regulatory domains raises the possibility that GLD-3 was co-opted into control of sex determination by incorporation of an FBF-binding region into one of its isoforms.
The mitosis/meiosis and sperm/oocyte regulatory circuits and their evolution:
Many mitosis/meiosis regulators also control the sperm/oocyte decision (Table 1). And both circuits control cell number: fbf-1 mutants have fewer mitotic cells and more sperm, while gld-3 mutants have more mitotic cells and fewer sperm (Crittenden et al. 2002; Eckmann et al. 2002; this work). Nonetheless, only one regulatory relationship is fully conserved in the two circuits. FBF represses gld-1 mRNA in both circuits, but all the other regulators appear to have different roles (Figure 9, A and B). GLD-2 is a key regulator of meiosis, but has no known role in sex determination (Kadyk and Kimble 1998); GLD-3 works downstream of FBF to promote meiosis, but acts primarily upstream of FBF in sex determination (Eckmann et al. 2002; this work); and NOS-3 promotes meiosis in the GLD-1 regulatory branch (Hansen et al. 2004; this work), but works with FBF to promote oogenesis (Kraemer et al. 1999). Therefore, although composed of largely the same regulators, the two pathways have diverged considerably. The mechanism by which two distinct pathways operate in the same tissue to control two distinct sets of fates is not known, but it is likely to involve separation of the regulators in time or space.
We suggest that the mitosis/meiosis and sperm/oocyte circuits evolved from an ancient pathway designed to control opposing fates (e.g., mitosis vs. meiosis or sperm vs. oocyte). The circuit controlling the mitosis/meiosis decision in the C. elegans germline can be thought of more broadly as a circuit that controls the balance between growth and differentiation. Analogous circuits are likely to exist in other stem cell systems. Indeed, homologous circuits may control stem cells in other tissues, given the conserved role of PUF proteins in controlling stem cells in C. elegans and Drosophila (Lin and Spradling 1997; Forbes and Lehmann 1998; Crittenden et al. 2002). In C. elegans, we suggest that the mitosis/meiosis and sperm/oocyte circuits have evolved together since they both control early decisions in germline development that must be integrated to achieve the mature tissue. The future challenge is to understand how these overlapping pathways work in both sexes to maintain a stem cell population and yet generate germlines with different gametes.
We thank Tim Schedl for anti-GLD-1, Monique Zetka for anti-HIM-3, Sam Ward for the sperm-specific antibody SP56, Barth Grant for anti-RME-2, and the Caenorhabditis Genetics Center (CGC) for providing strains. We thank Dave Hansen, Tim Schedl, and members of the Kimble laboratory for helpful discussions and Beth Thompson, Liana Lamont, Jen Bachorik, Dave Rudel, and the reviewers for comments on the manuscript. We are also grateful to Peggy Kroll-Conner for assistance with generating the gld-3 deletion mutant and some double and triple mutants on chromosome II and Anne Helsley-Marchbanks and Laura Vanderploeg for help with preparing the manuscript and figures. C.R.E. was supported by a Human Frontiers Science Program Long-Term Fellowship. J.K. is an investigator with the Howard Hughes Medical Institute (HHMI).
↵ 1 Present address: Max Planck Institute of Molecular Cell Biology and Genetics, 01307 Dresden, Germany.
Communicating editor: A. P. Mitchell
- Received March 26, 2004.
- Accepted May 27, 2004.
- Genetics Society of America