Abstract

FBF, a PUF RNA-binding protein, is a key regulator of the mitosis/meiosis decision in the Caenorhabditis elegans germline. Genetically, FBF has a dual role in this decision: it maintains germ cells in mitosis, but it also facilitates entry into meiosis. In this article, we explore the molecular basis of that dual role. Previous work showed that FBF downregulates gld-1 expression to promote mitosis and that the GLD-2 poly(A) polymerase upregulates gld-1 expression to reinforce the decision to enter meiosis. Here we ask whether FBF can act as both a negative regulator and a positive regulator of gld-1 expression and also investigate its molecular mechanisms of control. We first show that FBF co-immunoprecipitates with gld-1 mRNA, a result that complements previous evidence that FBF directly controls gld-1 mRNA. Then we show that FBF represses gld-1 expression, that FBF physically interacts with the CCF-1/Pop2p deadenylase and can stimulate deadenylation in vitro, and that CCF-1 is partially responsible for maintaining low GLD-1 in the mitotic region. Finally, we show that FBF can elevate gld-1 expression, that FBF physically interacts with the GLD-2 poly(A) polymerase, and that FBF can enhance GLD-2 poly(A) polymerase activity in vitro. We propose that FBF can affect polyadenylation either negatively by its CCF-1 interaction or positively by its GLD-2 interaction.

PUF (Pumilio and FBF) RNA-binding proteins influence many aspects of development and physiology. In Drosophila and Caenorhabditis elegans, both Pumilio and FBF are required for maintenance of germline stem cells (Lin and Spradling 1997; Forbes and Lehmann 1998; Crittenden et al. 2002), and PUF proteins have been implicated in stem cell controls in other organisms, including humans (Wickens et al. 2002; Salvetti et al. 2005; Xu et al. 2007). In addition, PUF proteins influence embryonic patterning (Barker et al. 1992), germline sex determination (Zhang et al. 1997), and memory formation (Dubnau et al. 2003). A molecular understanding of PUF regulation will therefore affect a broad spectrum of critical biological processes.

This work focuses on C. elegans FBF (fem-3 binding factor), a collective term for the nearly identical and largely redundant FBF-1 and FBF-2 proteins (Zhang et al. 1997). Biochemically, FBF-1 and FBF-2 bind the same RNA sequence, the FBF binding element (FBE) (Zhang et al. 1997; Bernstein et al. 2005), and also bind the same proteins, including GLD-3 (Eckmann et al. 2002). Genetically, fbf-1 and fbf-2 single mutants are virtually wild-type and fertile, but fbf-1 fbf-2 double mutants fail to maintain germline stem cells, fail to embark on oogenesis, and are sterile (Zhang et al. 1997; Crittenden et al. 2002; Lamont et al. 2004). Thus, FBF-1 and FBF-2 have similar biochemical activities in vitro and similar effects on the mitosis/meiosis decision. Work on PUF proteins in other organisms demonstrated that they repress mRNA activity, at least in part, by recruiting the deadenylation machinery (Goldstrohm et al. 2006, 2007), but the mechanism of FBF action has not yet been examined.

FBF promotes germline self-renewal by repressing regulators of meiotic entry (Figure 1A). Indeed, two regulatory branches control meiotic entry (Kadyk and Kimble 1998) and FBF represses an mRNA in each branch (Crittenden et al. 2002; Eckmann et al. 2004). One branch includes GLD-1, a translational repressor (Jan et al. 1999; Lee and Schedl 2001; Marin and Evans 2003), and the other branch consists of GLD-2/GLD-3, a translational activator and poly(A) polymerase (Wang et al. 2002; Suh et al. 2006). Meiotic entry is dramatically curtailed in double mutants that delete key components of both branches, but not in the single mutants (Kadyk and Kimble 1998; Eckmann et al. 2004; Hansen et al. 2004b). Of most relevance to this article, FBF directly represses gld-1 mRNA (Crittenden et al. 2002; Merritt et al. 2008), and GLD-2 directly activates gld-1 mRNA, a positive regulatory step that reinforces the decision to enter meiosis (Figure 1B) (Suh et al. 2006). GLD-3 has not yet been confirmed molecularly as a direct regulator of gld-1 mRNA, but it seems likely and therefore is shown in Figure 1B.

The gld-1 mRNA switches from FBF repression to GLD-2 activation in the “mitotic region” of the distal gonad (Figure 1B) (reviewed in Kimble and Crittenden 2007). FBF extends throughout the mitotic region, and decreases more proximally, in the transition zone where germ cells have entered meiotic prophase I (Crittenden et al. 2002; Lamont et al. 2004). By contrast, GLD-1 protein first appears in the proximal mitotic region, where germ cells are beginning to switch from the mitotic cell cycle into meiosis (Jones et al. 1996; Hansen et al. 2004b). GLD-3 appears in the proximal mitotic region as well, and has been proposed to act together with GLD-2 to promote meiotic entry (Eckmann et al. 2004).

In addition to its essential role in promoting germline self-renewal, FBF has a nonessential role in promoting meiotic entry. Meiotic entry is dramatically curtailed in gld-1; fbf-1 fbf-2 triple mutants, much as it is in gld-1 gld-2 or gld-1; gld-3 double mutants (Crittenden et al. 2002; Hansen and Schedl 2006; Kimble and Crittenden 2007). Thus, FBF acts genetically as part of the GLD-2/GLD-3 regulatory branch which promotes meiotic entry (Figure 1A). The molecular mechanism by which FBF promotes meiotic entry is not known, but we envision two simple possibilities, which are not mutually exclusive. FBF might act directly with GLD-2 and GLD-3 to activate mRNAs that promote meiotic entry (Figure 1B) or FBF might repress a repressor of meiotic entry. Because gld-1 mRNA is a known target of FBF (Crittenden et al. 2002) and can be activated by GLD-2 (Suh et al. 2006), we investigated whether FBF can positively affect gld-1 expression.

FBF and the GLD proteins control meiotic entry in both XX hermaphrodites and XO males and in both spermatogenic and oogenic germlines (Kadyk and Kimble 1998; Crittenden et al. 2002; Eckmann et al. 2004; Hansen et al. 2004a,b; S. Crittenden, unpublished results). Wild-type XX germlines make sperm during larval development and switch to oogenesis late in larval development so that adult germlines are oogenic; XX germlines can be feminized using one of several mutations (e.g., fem-3) so that only oocytes are made. Wild-type XO male germlines make sperm continuously. Our analyses were complicated by the facts that germline sex is dynamic during wild-type hermaphrodite larval development, that FBF affects germline sex (Zhang et al. 1997), and that GLD-1 abundance is controlled in a sex-specific manner (Jones et al. 1996). Our analyses therefore manipulate germline sex as necessary to facilitate our ability to quantitate FBF effects on GLD-1 accumulation.

In this article, we investigate FBF controls of gld-1 expression. We first extend evidence that FBF represses gld-1 expression in oogenic germlines and then explore the molecular mechanism of that repression. We next report that FBF elevates gld-1 expression in spermatogenic germlines and explore one possible molecular mechanism for that activation. Our results support the idea that FBF is capable of controlling gld-1 expression either negatively or positively. Our results are also consistent with the idea, put forward previously (Suh et al. 2006), that FBF may mark its repressed target mRNAs in preparation for the molecular transition of those same mRNAs into a GLD-2-activated complex.

MATERIALS AND METHODS

Nematode strains and methods:

All strains were maintained at 20° (Brenner 1974). Mutations and balancers are as follows: LGI, gld-1(q485) (Jones and Schedl 1995); balancer, hT2[qIs48]. LGII, fbf-1(ok91) fbf-2(q704) hermaphrodites and spontaneous males; balancer, mnIn1[mIs14 dpy-10(e128)]. LGIII, ccf-1(gk40) (Molin and Puisieux 2005); balancer, hT2[qIs48]. LGIV, fem-3(e1996) (Hodgkin 1986); balancer, nT1[qIs51]. Strains were as follows: fbf-1(ok91) fbf-2(q704)/mnIn1[mIs14 dpy-10(e128)]; fem-3(e1996)/qIs24 and ccf-1(gk40); qEx655. qIs24 contains Plag-2∷GFP integrated on LGIV. qEx655 contains Psur-5∷GFP along with genomic ccf-1 DNA, including 5′ and 3′ flanking sequences. To rescue the larval phenotype of ccf-1(gk40), 1 ng/μl ccf-1 genomic PCR product and 100 ng/μl Psur-5∷GFP were injected into ccf-1(gk40)/hT2[qIs48] animals.

Immunohistochemistry:

Two fixation methods were used with similar results. For one method, germlines were extruded into 7 μl M9 containing 0.25 mm levamisole. Next, 2 μl 2% paraformaldehyde were added and coverslip set gently on the slide. After freezing on dry ice, the coverslip was popped off, 100 μl 2% paraformaldehyde were added for 10 min at room temperature followed by incubation with 0.2% Triton X-100 for 5 min at room temperature. Alternatively, germlines were extruded and fixed with 1.6% formaldehyde/0.1 m K2HPO4 (pH 7.2) for 2 hr at room temperature followed by 10 min incubation with 100% methanol at −20°. For both methods, samples were blocked in phosphate-buffered saline containing 0.5% BSA, then fixed germlines were incubated with primary antibodies overnight at 4°. Rabbit α-GLD-1 (gift of Goodwin lab), mouse α-SP56 (gift of Ward and Strome labs), and mouse α-PH3 (Cell Signaling Technology) were all diluted 1:100. Slides were washed and incubated with secondary antibodies, washed again, and mounted for microscopy. Images were obtained on a Zeiss LSM 510 confocal microscope.

Quantitation of GLD-1 protein:

Samples to be compared were processed for immunohistochemistry in parallel, using the same reagents, and images were taken with the same settings. The gain was set for little or no saturation. Raw images were opened in ImageJ and staining quantitated by measuring pixel intensity in a 5.22 × 5.22 μm square at sites described in the text. Data were copied into Excel and averaged.

Cell number counts:

Dissected adult germlines (∼24 hr past L4) were DAPI (4′,6′-diamidino-2-phenylindole) stained (Crittenden and Kimble 2006) to examine nuclear morphology. The mitotic region (MR)/transition zone (TZ) boundary was defined as the distalmost row of cells containing multiple nuclei with crescent-shaped DAPI-stained morphology, typical of leptotene/zygotene of meiotic prophase I (Francis et al. 1995; Dernburg et al. 1998; Eckmann et al. 2004; Crittenden et al. 2006). Cell numbers were obtained by counting nuclei focal plane by focal plane through the width of the germline.

Deadenylation in vitro:

The yeast Pop2p-TAP complex was purified as described (Goldstrohm et al. 2006). GST-FBF-2 (121–632) was purified from Escherichia coli (Bernstein et al. 2005). Deadenylation reactions were carried out in 20 μl of 50 mm Tris-HCl (pH 8.0), 20 mm NaCl, 0.1 mm MgCl2, 10% (v/v) glycerol. FBE-a substrate RNAs (IDT) (sequences in Figure 3E) were radioactively labeled with T4 polynucleotide kinase (Promega) at the 5′ end and added to reactions at a final concentration of 10 nm. GST-FBF-2 (121–632) (100 nm) and purified Pop2p-TAP complex (10 ng) were added to reactions as indicated.

GST pull downs:

Purified GST, GST-FBF-2 (121–632), GST-FBF-1 (121–614), GST-PUF-8 (127–519), GST-Mpt5p (126–626), and GST-GLD-2 (482–1113) were immobilized on glutathione agarose beads (Amersham) at 1 μg/10 μl bed volume and used for pull-down assays. CCF-1 or Pop2p fused to a T7 tag (Figure 3) or 35S-radiolabeled GLD-2 full length (1–1113) and GLD-2 C terminus (482–1113) (Figure 6) were translated using rabbit reticulocyte TNT translation system (Promega). In Figure 6I, GST-GLD-2 pull downs were performed with recombinant purified MBP-FBF-2. Pull downs were then processed as described in Goldstrohm et al. (2006).

Immunoprecipitations:

For protein–mRNA interactions, immunoprecipitation experiments were performed as described (Suh et al. 2006) with minor modifications: fbf-2(q738) mutants were suspended in PBS containing EDTA-free protease inhibitor cocktail (Roche). Worms were homogenized using a mortar and pestle and then dounce homogenizer. Immunoprecipitations were performed with rabbit α-FBF-1 antiserum or preimmune serum, bound to protein-A beads (Pierce). Rabbit anti-FBF-1 (animal no. 695) was produced by injecting rabbits with purified glutathione-S-transferase (GST)-tagged FBF-1 protein (containing amino acids 121–614). Anti-FBF-1 antibodies predominantly recognize FBF-1 on Western blots of FBF-1 and FBF-2 translated in vitro.

RT–PCR analysis of eft-3 and gld-1 were performed using gene-specific primers. RT–PCR products (40%) were analyzed with 2% agarose gel.

For protein–protein interactions, extracts from fbf-2(q738) mutants or fbf-2(q738) gld-3(q730) mutants were made as described above. After washing, bound proteins were eluted in SDS sample buffer, boiled, and separated on 7.5% gels, which were then used to prepare Western blots with α-GLD-2, α-FBF-1, and α-actin, respectively.

Bridge yeast two-hybrid assay:

Bridge yeast two-hybrid assays were performed as described (Bernstein et al. 2002; Eckmann et al. 2002) with minor modification. Plasmids encoding chimeric proteins were cotransformed into strain L40-ura3 (Invitrogen). Levels of 3-aminotriazole (3-AT) resistance were determined by assaying multiple independent transformants as described (Kidd et al. 2005). Reporter (lacZ) expression was assayed using the Beta-Glo system (Promega).

Polyadenylation in vitro:

GST-GLD-2 (482–1113) and GST-FBF-1 (121–614) were purified from E. coli (Bernstein et al. 2005) and eluted with 20 mm Tris-HCl (pH 8.0), 50 mm KCl, 1 mm MgCl2, 0.2 mm EDTA, 30% glycerol, and 50 mm glutathione. The GST was cleaved with PreScission Protease (GE Healthcare) overnight at 4°. Polyadenylation reactions were done in 25 μl of 20 mm Tris-HCl (pH 8.0), 50 mm KCl, 1 mm MgCl2, 0.2 mm EDTA, 10% (v/v) glycerol, and 0.1 mg/ml BSA. gld-1 mRNAs (IDT) (sequences in Figure 3) were radioactively labeled with T4 polynucleotide kinase (NEB) at the 5′ end and added to reactions at a final concentration of 10 nm. GLD-2 (1, 5, 10, 50, and 100 ng) was added to the reactions shown in Figure 7B, lanes 2–6, and 10 ng of GLD-2 was added to other reactions. GST-FBF-1 (170 nm) (corresponding approximately to the measured Kd to the RNA) was added to reactions.

RESULTS

FBF-1 co-immunoprecipitates with gld-1 mRNA:

Previous work showed that FBF binds tightly and specifically to FBEs in the gld-1 3′-UTR (Figure 1C) (Crittenden et al. 2002). Here we asked if gld-1 mRNA co-immunoprecipitates with FBF-1, an experiment that complements the in vitro binding studies in support of a direct interaction. To ask whether gld-1 mRNA is physically associated with FBF, we used FBF-1-specific antibodies for immunoprecipitation from worm extracts (see materials and methods). The gld-1 mRNA was indeed enriched in FBF-1 immunoprecipitates compared to those with preimmune serum (Figure 1D, top), but eft-3, a control germline mRNA, was not enriched (Figure 1D, bottom). Together with previous experiments (Crittenden et al. 2002; Merritt et al. 2008), we conclude that FBF directly represses gld-1 expression.

Figure 1.—

The mitosis/meiosis decision and its control. (A) The core regulatory circuit controlling the mitosis/meiosis decision. FBF acts genetically in two positions: (1) upstream of gld mRNAs to promote mitosis and (2) together with GLD-2 and GLD-3 to promote meiotic entry. Colored thick arrows depict positive regulation; colored thick bars depict negative regulation; black thin arrows depict translation of mRNA into protein. See text for further explanation and references. (B) Diagrams of the distal germline (top) and its control by FBF and GLD-2/GLD-3 (bottom). (Top) The somatic distal tip cell (red) promotes germline self-renewal by Notch signaling; all other cells are germ cells. Yellow, germ cells in mitotic cell cycle; green, cells in meiotic cell cycle; gradient from yellow to green, germ cells switching from mitosis to meiosis. (Bottom) molecular controls. (Left) FBF represses gld-1 mRNA to prevent meiotic entry; (right) GLD-2 and GLD-3 activate gld-1 mRNA to reinforce the decision to enter meiosis. (C) The gld-1 3′-UTR harbors two FBF binding elements (FBEs) (Crittenden et al. 2002). (D) Co-immunoprecipitation of FBF-1 and gld-1 mRNA. Pre-Imm, preimmune serum. eft-3, control mRNA. PCRs were continued similarly for both gld-1 and eft-3, with sufficient cycles to detect nonspecific association of eft-3 mRNA after incubation with both preimmune and α-FBF-1 antibodies.

FBF-1 and FBF-2 repress gld-1 expression in oogenic germlines:

Previous work showed that GLD-1 protein is elevated in the distal mitotic region of most fbf-1 single mutants (Crittenden et al. 2002). Here we assayed GLD-1 abundance in fbf-1 fbf-2 double mutants, a stronger experiment than that in fbf-1 single mutants. To compare GLD-1 levels in mitotic regions with and without FBF-1 and FBF-2, both germlines had to have a mitotic region and to be in the same sexual state. We focused on L4 germlines, because at this stage fbf-1 fbf-2 germlines still have a mitotic region. In addition, because fbf-1 fbf-2 germlines are masculinized throughout development and wild-type hermaphrodite germlines are in the process of switching from spermatogenesis to oogenesis during the fourth larval stage (L4), we feminized both fbf-1 fbf-2 and wild-type germlines so that the two germlines were in a comparable state. To this end, we used a fem-3 mutant and examined XX fem-3 m+z− and XX fbf-1 fbf-2; fem-3 m+z− germlines at L4; both have mitotic regions and both are oogenic (Figure 2A) (Thompson et al. 2005). The fem-3 and fbf-1 fbf-2; fem-3 germlines were stained simultaneously with DAPI to assess nuclear morphology and meiotic entry, with α-phosphorylated histone 3 (anti-PH3) antibodies to visualize mitotic M phase and with α-GLD-1 antibodies to measure GLD-1 abundance. We first confirmed that fem-3 and fbf-1 fbf-2; fem-3 L4 germlines both contained mitotically dividing cells (Figure 2, B and E), and then assayed GLD-1. In fem-3 germlines, GLD-1 abundance was low in the distal mitotic region, increased in the proximal mitotic region, and was abundant in the transition zone, where germ cells had entered early meiotic prophase (Figure 2C). In fbf-1 fbf-2; fem-3 triple mutants, by contrast, GLD-1 was abundant throughout both the mitotic region and transition zone (Figure 2F). Using ImageJ software, GLD-1 levels were quantitated in both the distalmost mitotic region and transition zone (Figure 2, D and G). Distally, GLD-1 was approximately sixfold more abundant on average in fbf-1 fbf-2; fem-3 than in the same region of fem-3 germlines. Indeed, the high distal GLD-1 was fully penetrant in fbf-1 fbf-2; fem-3 germlines whereas it had been only partially penetrant in fbf-1 single mutants (Crittenden et al. 2002). In the transition zone, by contrast, GLD-1 levels were comparable in the two germlines (Figure 2H). We conclude that FBF-1 and FBF-2 normally reduce GLD-1 abundance in distal oogenic germlines, but that they have no apparent effect in the transition zone.

Figure 2.—

FBF represses gld-1 expression. All samples were processed in parallel. (A) Summary of the fbf and fem mutants used in this study, focusing on the presence of a mitotic region (MR) and the sexual state. (+), presence of mitotic region (MR); (−) absence of mitotic region; (sp) sperm; (oo) oocyte; see text for further explanation and references. (B–G) Arrowhead, distal end of gonad; dashed line, boundary between mitotic region and transition zone. (B–D) Same L4 XX fem-3 m+z− germline. (E–G) Same L4 XX fbf-1 fbf-2; fem-3 m+z− germline. (B and E) DAPI (blue, all nuclei) and α-PH3 (yellow, mitotic nuclei) staining. (C and F) α-GLD-1 staining (green). (D and G) Areas of GLD-1 quantitation (squares) in distalmost gonad (yellow) or transition zone (green). (H) GLD-1 quantitation. Herm, hermaphrodite. P-value = 0.0008 and 0.38 for distal MR and transition zone, respectively.

FBF binds CCF-1 and stimulates deadenylation in vitro:

The Pop2p class of deadenylases interacts with PUF proteins in several species (see discussion). We therefore asked if FBF binds CCF-1, the C. elegans ortholog of Saccharomyces cerevisiae Pop2p (Molin and Puisieux 2005). CCF-1 possesses a DEDD-type, magnesium-dependent exoribonuclease domain and shares significant sequence identity with other Pop2p orthologs (Figure 3, A and B). Mammalian Pop2 is an enzymatically active deadenylase (Thore et al. 2003; Viswanathan et al. 2004; Bianchin et al. 2005). Since CCF-1 contains all active site residues (Figure 3A), it too is likely an active deadenylase. To examine FBF interactions with CCF-1, we incubated CCF-1 protein fused to a T7 tag, which was translated in vitro, with each of four purified, recombinant PUF proteins fused to GST and immobilized on glutathione agarose beads. The four PUF proteins were FBF-1 and FBF-2 plus two positive controls, C. elegans PUF-8 and yeast Mpt5p, both of which were already known to bind CCF-1 (Goldstrohm et al. 2006); GST alone served as a negative control. FBF-1 and FBF-2 both retained CCF-1, as did PUF-8 and Mpt5p; GST did not (Figure 3C). We conclude that FBF-1 and FBF-2 bind CCF-1.

Figure 3.—

FBF directly binds Pop2p/CCF-1 and enhances deadenylation. (A) Sequence alignment of nuclease domain in Homo sapiens (Hs) CNOT8 (NP_004770.4), C. elegans (Ce) CCF-1 (NP_499553.1), and S. cerevisiae (Sc) Pop2p (NP_014450.1) using ClustalV algorithm. Arrowheads, active site residues. Solid boxes, identical amino acids; shaded boxes, similar amino acids. (B) Phenogram of Pop2p orthologs, including that of Drosophila melanogaster (Dm), was created using MegaAlign (DNASTAR). (C) FBF-1 and FBF-2 bind CCF-1. Western blot of GST pull downs with beads bound to GST fusions indicated at the top using T7-tagged, in vitro translated CCF-1. (D) FBF-2 binds CCF-1 and Pop2p. Western blot of GST pull downs with control GST or GST-FBF-2 beads using T7-tagged, in vitro translated yeast Pop2p and C. elegans CCF-1. (E) FBF-2 enhances deadenylation in vitro. (Bottom) Substrate RNA from gld-1 3′-UTR, 5′ end labeled, and polyadenylated, containing wild-type (WT) or mutant FBF binding elements (FBE). (Top) Reactions contained 200 fmol of substrate RNA and, where indicated at the top, 10 ng of purified Pop2p-TAP complex and 100 nm of recombinant GST-FBF-2. Reactions were initiated by addition of substrate and incubated for the times indicated at the top. As markers, substrate RNAs without (A0) or with (A14) poly(A) tails were included in lanes 1 and 2, respectively.

We next tested whether FBF can stimulate deadenylation. To this end, we asked if FBF could stimulate deadenylation with active yeast Pop2p complex. We first showed that FBF-2 binds yeast Pop2p (Figure 3D) and then assayed deadenylation in vitro using a substrate RNA derived from the gld-1 3′-UTR that carried 14 adenosine residues at the 3′ end (Figure 3E, bottom). When the wild-type FBE-carrying RNA was incubated with recombinant FBF alone, no degradation was observed, even after 60 min, a result that excludes the possibility of contaminating exonucleases (Figure 3E, lanes 3–5). When a limiting amount of the Pop2p complex was incubated with either the same wild-type RNA or an FBE mutant RNA that does not bind FBF, only a few adenosines were removed (Figure 3E, lanes 6–8 and lanes 12–14). However, when FBF was combined with the Pop2p complex, deadenylation was efficient (Figure 3E, lanes 9–11). This FBF enhancement was dependent on FBF recognition of the substrate mRNA, because the FBE mutant RNA was not similarly deadenylated (Figure 3E, lanes 15–17). We conclude that FBF can activate deadenylation by recruiting the Ccr4p-Pop2p-Not deadenylase complex.

CCF-1 affects both mitotic region size and gld-1 expression:

ccf-1 null mutants normally arrest during larval development (Molin and Puisieux 2005). To examine the role of CCF-1 in the adult germline, we generated transgenic animals predicted to be mosaic for CCF-1. Specifically, we rescued ccf-1 larval lethality with a “simple” extrachromosomal array, which is typically active in somatic tissues but silenced in the germline (Kelly et al. 1997). Our array, qEx655, contained genomic ccf-1 DNA plus Psur-5∷GFP, a reporter that makes nuclearly localized GFP in essentially all cells (see materials and methods). As predicted, qEx655 expressed GFP in somatic but not germline tissues (Figure 4, A and B). Moreover, the array efficiently rescued ccf-1 larval lethality, but adults were sterile, as previously reported for a different array (Molin and Puisieux 2005). Therefore, expression from the array is likely to be silenced in the germline. Our CCF-1 antibodies do not work for immunocytochemistry, and therefore we have not been able to confirm the CCF-1 mosaicism. However, given the larval rescue and germline sterility, we suggest that CCF-1 is likely to be at least reduced in the germline tissue. Hence, we refer to the ccf-1(0); qEx655 strain as ccf-1(GLrf), for ccf-1 germline reduction of function.

Figure 4.—

CCF-1 affects the size of the mitotic region and gld-1 expression. (A) Diagram of the ccf-1 null mutant carrying rescuing somatic array, qEx655, which includes genomic ccf-1 DNA and Psur-5∷GFP. GFP is expressed only in somatic tissues. (B) Germline extruded from ccf-1(GLrf). Small arrowheads indicate somatic gonadal nuclei that have strong GFP signal. No GFP was detected in germline (dashed outline). (C–H) Germlines extruded from either wild-type (C and D) or ccf-1(GLrf) (E–H) adult hermaphrodites. (C, E, and G) DAPI (blue) staining. (D, F, and H) α-GLD-1 (red) staining. Arrowhead, distal end of gonad; dashed lines, mitotic region/transition zone boundary. In ccf-1(GLrf) germlines, GLD-1 levels are higher than wild type. (I) GLD-1 quantitation. P-value = 0.02, 0.002, and 0.009 for distal MR, proximal MR, and transition zone, respectively.

To ask if CCF-1 affects mitotic region size, we examined DAPI-stained wild-type and ccf-1(GLrf) adult XX germlines. Control wild-type mitotic regions contained ∼230 germ cells on average and extended about 20 cell diameters from the distal end (Figure 4C). By contrast, the ccf-1(GLrf) mitotic region possessed only ∼170 (170 ± 21, n = 7) germ cells and extended ∼14 germ cell diameters (14 ± 3, n = 28) on average from the distal end (Figure 4, E and G), although its size ranged considerably. The simplest explanation is that CCF-1 normally acts in the germline to control the proper mitotic region size.

To ask if CCF-1 affects gld-1 expression, we compared GLD-1 protein in wild-type and ccf-1(GLrf) adult XX germlines. In this experiment, GLD-1 was quantitated in three regions, the distal mitotic region, the proximal mitotic region, and the transition zone. In wild-type germlines, GLD-1 was difficult to detect in the distal mitotic region, became easily detectable in the proximal mitotic region, and increased dramatically in the transition zone (Figure 4D), as previously reported (Jones et al. 1996). In ccf-1(GLrf) germlines, by contrast, GLD-1 was easily detectable in the distal mitotic region and became abundant in the proximal mitotic region and transition zone of most ccf-1(GLrf) germlines (80%, n = 15) (Figure 4, F and H), visual results that were borne out by quantitation (Figure 4I). Therefore, CCF-1 normally reduces GLD-1 throughout the mitotic region and transition zone. The increased GLD-1 in the ccf-1(GLrf) germlines together with biochemical data for FBF/CCF-1 binding and FBF-dependent deadenylation (Figure 3) suggest that FBF and CCF-1 repress gld-1 expression together. However, derepression is not as dramatic in ccf-1(GLrf) germlines as in fbf-1 fbf-2 germlines (compare Figures 2 and 4), suggesting that FBF represses gld-1 by both CCF-1-dependent and CCF-1-independent mechanisms. However, we cannot exclude the possibility that this milder derepression is due to low levels of expression from the transgene in ccf-1(GLrf) germlines.

FBF-1 and FBF-2 elevate gld-1 expression in spermatogenic germlines:

In addition to inhibiting entry into meiosis, FBF behaves genetically together with GLD-2 and GLD-3 to promote meiotic entry (see Introduction). However, in oogenic germlines, no FBF activation of gld-1 expression was seen as germ cells entered meiosis (Figure 2). We reasoned that a positive effect of FBF on GLD-1 accumulation might be overwhelmed by other regulators that drive an oocyte-specific GLD-1 amplification. We therefore examined masculinized germlines in an attempt to see FBF positively regulate gld-1 expression. Specifically, we compared GLD-1 in L4 XO wild-type, L4 XO fbf-1 fbf-2, and L4 XX fbf-1 fbf-2 germlines, all of which are spermatogenic. Both XO wild-type and XX fbf-1 fbf-2 germlines contained mitotic germ cells at the distal end, and XO fbf-1 fbf-2 germlines contained germ cells that had not entered meiotic prophase in a similar position (Figure 5, A, C, and E). In wild-type male germlines, GLD-1 was low in the distalmost mitotic region, but increased visibly in the proximal mitotic region and transition zone (Figure 5B), as previously reported (Jones et al. 1996). However, in fbf-1 fbf-2 germlines (both XX and XO), GLD-1 was less abundant than in wild type (Figure 5, D and F). Upon quantitation, we found that GLD-1 quantity was not affected dramatically in the distal mitotic region, but that it was approximately fourfold lower in the transition zone of both XX and XO fbf-1 fbf-2 germlines than in wild-type males (Figure 5G). As one control, we costained with antibodies to FOG-1, a male germline protein (Thompson et al. 2005), and found both germlines equally permeable (not shown). As a second control, we stained gld-1 null mutant germlines to quantitate background (Figure 5G). We conclude that FBF can promote GLD-1 accumulation, consistent with its role in promoting meiotic entry. The failure of GLD-1 to increase in distal fbf-1 fbf-2 male germlines indicates the existence of other gld-1 regulators.

Figure 5.—

FBF can promote increased gld-1 expression. (A–F) Arrowhead, distal end of gonad; dashed line, boundary between mitotic region and transition zone. (A and B) Dissected germline from L4 XO wild type. (C and D) Dissected germline from L4 XX fbf-1 fbf-2 mutant. (E and F) Dissected germline from L4 XO fbf-1 fbf-2 mutant. (A, C, and E) Germline nuclei stained with DAPI (blue); mitotic germ cells stained with α-PH3 (yellow). (B, D, and F) GLD-1 detected with α-GLD-1 antibodies. In fbf-1 fbf-2 germlines, GLD-1 levels are low compared to wild type. (G) GLD-1 quantitation. GLD-1 levels are similar in the mitotic regions of L4 XO wild-type and XX and XO fbf-1 fbf-2 mutant germlines; P-value = 0.03 and 0.06 for XX and XO, respectively. GLD-1 levels are reduced in the transition zones of L4 XX and XO fbf-1 fbf-2 mutants compared to XO wild-type germlines; P-value = 2 × 10−7 and 2 × 10−5 for XX and XO, respectively.

FBF can recruit GLD-2 indirectly using a GLD-3 bridge:

A positive effect of FBF on gld-1 expression can be explained by either direct FBF activation of gld-1 mRNA or indirect activation (e.g., repression of a gld-1 repressor other than FBF itself). These two ideas are not mutually exclusive. In this work, we tested the idea that FBF might be a direct activator of gld-1 expression, because FBF binds the gld-1 3′-UTR (Crittenden et al. 2002) (Figure 1, C and D) and because GLD-2 directly activates gld-1 mRNA (Suh et al. 2006). We therefore asked if FBF might be capable of recruiting GLD-2. To this end, we first tested whether FBF might interact with GLD-2 via a common binding partner, GLD-3, a C. elegans homolog of Bicaudal-C that contains five KH domains and is predicted to bind RNA (Eckmann et al. 2002). GLD-3 is expressed in the proximal mitotic region where germ cells are switching into meiosis, and GLD-3 binds both FBF and GLD-2 (Eckmann et al. 2002, 2004). Therefore, GLD-3 has the qualities expected for a bridge between FBF and GLD-2.

To test the idea that GLD-3 serves as a bridge, we used a variant yeast two-hybrid assay (Figure 6A). Specifically, we coexpressed three proteins: FBF-1 fused to the LexA DNA-binding domain, GLD-2 fused to the GAL4 activation domain (AD), and a test protein (e.g., GLD-3) on its own. To monitor interactions, we assayed lacZ and HIS3 reporter genes (Figure 6B, top panels) and used Western blots to ensure that proteins were expressed comparably (Figure 6B, bottom panels). A weak interaction was observed when LexA-FBF-1 and AD-GLD-2 were coexpressed [Figure 6, B (lane 1) and C]. That interaction was dramatically enhanced upon coexpression of FBF-1 and GLD-2 together with GLD-3 (Figure 6B, lane 2). The GLD-3 enhancement was specific since two other test proteins, FOG-1, which is a CPEB-related protein (Luitjens et al. 2000; Jin et al. 2001), and PK (chicken pyruvate kinase), had no effect on the weak binding between FBF-1 and GLD-2 (Figure 6B, lanes 3 and 4). We conclude that GLD-3 is capable of bridging FBF-1 and GLD-2. To ask whether the minor interaction between LexA-FBF-1 and AD-GLD-2 was above background, we tested the two both alone and together. The interaction was indeed above background (Figure 6C). Therefore, FBF and GLD-2 interact with each other weakly in addition to the strong interaction with GLD-3 as a bridge.

Figure 6.—

FBF physically associates with GLD-2 both in vitro and in vivo. (A) Schematic of modified yeast two-hybrid assay used to test GLD-3 bridge mechanism. (B) GLD-2 can interact with FBF indirectly, using GLD-3 as a bridge. Top, interaction assays; bottom, Western blots of yeast cell lysates probed with α-LexA to detect FBF-1, α-Gal4AD to detect GLD-2, α-HA to detect test proteins, and α-actin to monitor loading. The results in each panel are representative of at least two independent assays and the β-gal graph compiles three independent assays. (C) Yeast two-hybrid assay with GLD-2 alone, FBF-1 alone, or the two together. (D) Immunoprecipitation (IP) from worm extracts using either α-GLD-3 antibody or preimmune serum. Both FBF-1 and GLD-2 are detected in the GLD-3 IP. (E–F) IP from worm extracts with α-FBF-1 antibody or control preimmune serum. Western blots were probed for α-GLD-2, α-FBF-1, and α-actin. fbf-2 null mutants have an essentially wild-type phenotype, but make excess FBF-1. (G) FBF–GLD-2 interaction is RNA independent. Worm extracts were treated with or without RNase A before immunoprecipitation with α-FBF-1 antibody. (H) FBF-1 associates with GLD-2 in reticulocyte lysates. GST-FBF-1 or GST alone were immobilized on sepharose-conjugated glutathione beads and incubated with radiolabeled in vitro translated GLD-2 fragments and a control protein. GLD-2 was specifically retained by GST-FBF-1 beads. (I) FBF-2 binds GLD-2 in vitro. Western blots of GST and GST-GLD-2 pull downs with increasing MBP-FBF-2 (100, 200, and 400 nm). (J) Diagram of all pairwise combinations of FBF, GLD-2, and GLD-3 interactions.

We next asked if FBF, GLD-2, and GLD-3 associate with each other in C. elegans extracts. Previous work showed that GLD-2 and GLD-3 co-immunoprecipitate (Eckmann et al. 2004). Here, we show that FBF-1 and GLD-2 are brought down with α-GLD-3 antibodies (Figure 6D, lane 6) and that GLD-2 is brought down with α-FBF-1 antibodies (Figure 6E, lane 6). Thus, FBF, GLD-2, and GLD-3 can be co-immunoprecipitated in all pairwise combinations (Figure 6J). We conclude that GLD-2, GLD-3, and FBF are likely to be components of a complex in the intact germline.

FBF can interact with GLD-2 directly and is RNA independent:

To learn whether GLD-3 is an essential bridge for the FBF–GLD-2 interaction, we used α-FBF-1 antibodies to immunoprecipitate proteins from a gld-3 null mutant. To our surprise, GLD-2 came down in roughly the same amount from gld-3(+) and gld-3(0) extracts (Figure 6F, lane 6). Therefore, GLD-3 is not essential for the FBF–GLD-2 association. One explanation is that the FBF–GLD-2 interaction is sufficient for physical interaction, though it may be reinforced by either GLD-3 or other factors (e.g., posttranslational modifications, other bridging proteins). To test the possibility that RNA might mediate binding between GLD-2 and FBF-1, we treated worm extracts with RNase A before immunoprecipitation with α-FBF-1 antibody. Although most RNA was degraded (data not shown), the FBF–GLD-2 interaction was still seen (Figure 6G, lanes 2 and 3). Therefore, the FBF–GLD-2 interaction is RNA independent.

The evidence presented above suggested that FBF is capable of directly binding GLD-2. To test this idea, we first produced two different GLD-2 proteins (full length and a C-terminal fragment) and a control protein, all translated in vitro using reticulocyte lysates (Figure 6H, lanes 1–3). Lysates were then incubated with purified recombinant FBF-1 fused to glutathione S-transferase (GST) and attached to glutathione beads. Both full-length GLD-2 and the C-terminal GLD-2 fragment bound FBF (Figure 6H, lanes 7 and 8) but not the control protein (Figure 6H, lane 9). None of the proteins bound beads carrying GST alone (Figure 6H, lanes 4–6). To confirm that FBF and GLD-2 bind one another in vitro using purified recombinant proteins, we purified FBF-2 fused to maltose-binding protein (MBP-FBF-2) from bacteria. Immobilized GST-GLD-2 bound MBP-FBF-2 (Figure 6I, lanes 5–7). We conclude that FBF directly binds GLD-2. Therefore, FBF can associate with GLD-2 either indirectly (via GLD-3) or directly and that association is not dependent on RNA.

FBF stimulates GLD-2 PAP activity in vitro:

We next asked if FBF can stimulate GLD-2 PAP activity in vitro. To this end, we reconstituted sequence-specific polyadenylation in vitro using purified GLD-2 and FBF-1 recombinant proteins (Figure 7A) and a substrate RNA derived from the gld-1 3′-UTR. To assay GLD-2 activity, we first incubated a 5′ end-labeled RNA substrate with increasing amounts of GLD-2 and analyzed polyadenylation by gel electrophoresis. GLD-2 was capable of polyadenylation on its own in a concentration-dependent manner (Figure 7B). At the highest concentration tested, GLD-2 polyadenylated the substrate RNA by hundreds of nucleotides (Figure 7B, lanes 2–6; data not shown). We conclude that GLD-2 is capable of polyadenylation without accessory RNA-binding proteins, even though accessory proteins can enhance its PAP activity (K. W. Kim, K. Nykamp, N. Suh, J. L. Bachorik, L. Wang and J. Kimble, unpublished results; Wang et al. 2002).

Figure 7.—

FBF enhances GLD-2 activity in vitro. (A) Coomassie-stained SDS–PAGE gel of purified GLD-2 (482–1113) (left) and GST-FBF-1 (121–614) (right). (B) In vitro poly(A) polymerase assay. Increasing GLD-2 amounts (lanes 2–6; 1, 5, 10, 50, 100 ng, respectively) were incubated with 5′ end-labeled substrate RNA (WT A14) (see Figure 3E, bottom). GLD-2 added the poly(A) tail in a concentration-dependent manner. As a marker, substrate RNA alone is shown in lane 1. (C) In vitro poly(A) polymerase reactions containing GLD-2 and GST-FBF-1 proteins, as marked above gel, were incubated with either WT A14 (lanes 2–4, 6, and 7) or mutant A14 (lanes 8 and 9) (see Figure 3E, bottom). FBF enhanced GLD-2 activity (lanes 3, 4, 6, and7) and this effect was FBE dependent (8 and 9). This gel is representative of four independent assays. Asterisks, end positions of detectable signal at the top of background: *GLD-2 alone; **FBF + GLD-2.

To ask whether FBF could stimulate GLD-2 PAP activity, we incubated recombinant FBF-1 with a limiting concentration of GLD-2 and substrate RNA. The optimal in vitro conditions for FBF binding to RNA and for GLD-2 enzymatic activity differed significantly, forcing us to compromise. Under these suboptimal conditions and with a limiting GLD-2 concentration, GLD-2 alone was capable of adding only a few adenosines (Figure 7C, lanes 3, 6, and 8). FBF-1 alone had no effect on the RNA (Figure 7C, lane 2). However, FBF and GLD-2 together extended the poly(A) tail (Figure 7C, compare lanes 3 and 4 as well as lanes 6 and 7). This increase was observed in four independent assays and is therefore highly reproducible. Importantly, the increase was sequence specific: substrates with mutations in the critical UGU of FBF binding sites failed to show any activation (Figure 7, lanes 8 and 9). The lack of a more robust enhancement by FBF may be due to the weak interaction between FBF and GLD-2 or the suboptimal reaction conditions used. GLD-3 may further enhance FBF-stimulated GLD-2 polyadenylation; however, we have not been able to purify recombinant GLD-3. We conclude that FBF can enhance GLD-2 polyadenylation activity.

DISCUSSION

FBF controls gld-1 expression both negatively and positively:

The key conclusion of this work is that FBF has dual and opposite roles in regulating gld-1 expression. Its role in repression is not surprising, since PUF proteins are established repressors of translation or mRNA stability (e.g., Wickens et al. 2002). Moreover, previous work had already provided evidence that FBF represses gld-1 expression (Crittenden et al. 2002; Merritt et al. 2008). However, the role of FBF in activation is surprising. Only recently has a role for PUF proteins in activation been reported. First, Xenopus Pumilio enhances translation of RNAs containing a weak CPEB binding site (Pique et al. 2008). Second, C. elegans FBF-1 activates RNA in olfactory sensory neurons (Kaye et al. 2009). Thus precedents exist for dual and opposite roles for PUF proteins in general and FBF in particular.

The notion that FBF controls gld-1 expression by a direct interaction with gld-1 mRNA is based on two lines of evidence: the presence of high-affinity FBF binding sites in the gld-1 3′-UTR (Crittenden et al. 2002) and co-immunoprecipitation from worm extracts of FBF protein and gld-1 mRNA (this work). The idea that FBF represses gld-1 expression is supported by several lines of evidence. We demonstrate in this article that GLD-1 abundance increases dramatically in the distal mitotic region of fbf-1 fbf-2; fem-3 triple-mutant germlines compared to fem-3 single-mutant germlines. The only difference between these two germlines is the lack of FBF. The GLD-1 increase observed was fully penetrant and therefore more severe than the partially penetrant increase seen in fbf-1 single mutants (Crittenden et al. 2002). In addition, GFP expression from a reporter transgene linked to the gld-1 3′-UTR was low in wild type but elevated upon depletion of FBF using RNAi (Merritt et al. 2008). In sum, our data together with previous work (Crittenden et al. 2002; Merritt et al. 2008) provide strong evidence that FBF directly represses gld-1 expression.

We also show that FBF can positively regulate gld-1 expression: GLD-1 abundance decreases dramatically after removal of FBF activity in spermatogenic germlines (e.g., fbf-1 fbf-2 XX or XO). We interpret that decrease to mean that FBF normally promotes gld-1 expression, at least in spermatogenic germlines. We were unable to see a similar effect in oogenic germlines. That failure may represent a detection problem, caused by extremely abundant GLD-1 typical of oogenic germlines, or it may represent a difference in FBF activities in oogenic and spermatogenic germlines. Regardless, the important point is that FBF is capable of both negative and positive regulation of gld-1 expression.

FBF controls polyadenylation:

Regulated polyadenylation was implicated as a mechanism for FBE-dependent mRNA repression even before discovery of FBF itself: When the wild-type fem-3 mRNA is repressed, it has a short poly(A) tail, whereas mutant fem-3 mRNAs carrying either of two FBE point mutations are derepressed and have a longer poly(A) tail (Ahringer and Kimble 1991). This work extends that finding to report that FBF can interact with yeast Pop2p, a subunit of the Ccr4p-Pop2p-Not deadenylase complex, as well as with CCF-1, the C. elegans homolog of Pop2p. We find that FBF promotes deadenylation together with an active yeast Pop2p complex in an FBE-dependent manner. Moreover, germlines defective for CCF-1 have a smaller than wild-type mitotic region and higher than wild-type GLD-1 levels. Together our results suggest that FBF repression relies, at least in part, on the deadenylation of target mRNAs. In yeast, PUF-mediated repression is similarly dependent on Pop2p to promote deadenylation, degradation, and translational repression of target mRNA (Goldstrohm et al. 2006, 2007; Hook et al. 2007). Furthermore, both human and Drosophila PUF proteins bind directly to the relevant Pop2p/Caf1 homolog (Goldstrohm et al. 2006; Kadyrova et al. 2007). Therefore, PUF proteins are likely to repress mRNAs generally by recruiting deadenylation machinery.

The finding that FBF can promote gld-1 expression led us to test the idea that FBF might interact directly with the GLD-2 poly(A) polymerase. Using a combination of yeast two-hybrid assays and in vitro biochemistry, we found that FBF can indeed interact with GLD-2 either indirectly, using GLD-3 as a bridge, or directly. Furthermore, FBF, GLD-2, and GLD-3 are physically associated within the C. elegans germline, since the three proteins co-immunoprecipitate in all pairwise combinations (Eckmann et al. 2004; this work). Given the evidence for an FBF–GLD-2 interaction (Figure 6J), we tested the idea that FBF might stimulate GLD-2 poly(A) polymerase activity. Using purified proteins and an FBE-bearing substrate RNA, we found a reproducible FBE-dependent stimulation. The in vitro FBF–GLD-2 binding and polyadenylation enhancement were both weak, and therefore unlikely to be sufficient for mRNA activation on their own. However, in vivo, FBF and GLD-2 are likely to be associated in a larger complex that includes GLD-3 (and perhaps other proteins since the FBF–GLD-2 interaction persists in the absence of GLD-3) (Figure 6J).

We envision two simple models by which FBF might directly activate mRNAs. The simplest idea is that FBF functions in an FBF/GLD-2/GLD-3 complex to promote polyadenylation and activate or stabilize the mRNA. This model is consistent with genetic results that FBF plays a role in meiotic entry that is equivalent to that of GLD-2 and GLD-3 (Crittenden et al. 2002) (see Introduction). This model is also consistent with our molecular results that FBF, GLD-2, and GLD-3 co-immunoprecipitate and that FBF can enhance GLD-2 PAP activity. Alternatively, FBF might function simply to recruit GLD-2 and GLD-3 to its target mRNAs, but FBF itself may leave the complex before or as active polyadenylation begins. By this model, the FBF repressor would mark its target mRNAs in preparation for the transition to an activated state, an idea put forward previously (Suh et al. 2006). Other models remain possible, including the possibility that FBF activates gld-1 expression indirectly (by repressing a repressor). Nonetheless, this work provides both genetic and molecular evidence to support the idea that FBF can both repress and activate its target mRNAs.

Might FBF be part of a molecular switch?

The identification of both FBF repressive and activating activities suggests that FBF may function as part of a molecular switch. FBF might first recruit CCF-1 and the deadenylation machinery to repress its target mRNAs and then help recruit the GLD-2 poly(A) polymerase and its partners to activate those same target mRNAs. The work presented in this article suggests that this idea is biochemically feasible. The major caveat is that FBF repression and activation have not been observed in the same germline. That failure may mean that FBF does not act like a switch or it may mean that redundancy masks the FBF effects. Given the rampant redundancy and complexity typical of the network (Hansen and Schedl 2006; Kimble and Crittenden 2007), the latter explanation is certainly plausible. If true, we postulate that FBF facilitates the transition toward an activated complex but is not an essential component.

The idea that the FBF RNA-binding protein may be part of a molecular switch is reminiscent of other molecular switches that control gene expression. Several DNA-binding proteins can recruit chromatin-modifying enzymes to either repress or activate transcription, and the CPEB RNA-binding protein can interact with the PARN deadenylase (Kim and Richter 2006) or GLD-2 poly(A) polymerase to repress or activate translation (Barnard et al. 2004; Rouhana et al. 2005; Kim and Richter 2006). This work suggests that PUF proteins may also act as a molecular switch, a notion that awaits further genetic analysis and in vitro reconstruction.

Acknowledgments

We are grateful to members of the Kimble and Wickens laboratories for helpful discussions during the course of this work. We thank Anne Helsley-Marchbanks and Laura Vanderploeg for help with preparing the manuscript and figures. J.K. and M.W. are funded by the National Institutes of Health (GM31892 and GM50942 to M.W. and GM69454 to J.K.); J.K. is an investigator with the Howard Hughes Medical Institute.

Footnotes

  • 1 These authors contributed equally to this work.

  • 2 Present address: Department of Biological Chemistry, University of Michigan, 1150 W. Medical Center Dr., 3220A MSRB3, Ann Arbor, MI 48109-5606.

  • Communicating editor: A. Villenueve

  • Received December 5, 2008.
  • Accepted February 7, 2009.

References

View Abstract