Although the decision between stem cell self‐renewal and differentiation has been linked to cell‐cycle modifications, our understanding of cell‐cycle regulation in stem cells is very limited. Here, we report that FBF/Pumilio, a conserved RNA‐binding protein, promotes self‐renewal of germline stem cells by repressing CKI‐2Cip/Kip, a Cyclin E/Cdk2 inhibitor. We have previously shown that repression of CYE‐1 (Cyclin E) by another RNA‐binding protein, GLD‐1/Quaking, promotes germ cell differentiation. Together, these findings suggest that a post‐transcriptional regulatory circuit involving FBF and GLD‐1 controls the self‐renewal versus differentiation decision in the germline by promoting high CYE‐1/CDK‐2 activity in stem cells, and inhibiting CYE‐1/CDK‐2 activity in differentiating cells.
Several lines of evidence suggest that the decision to self‐renew or differentiate is closely connected to the cell‐cycle status. While pluripotent cells typically display rapid G1‐to‐S phase transitions in the cell cycle, differentiation correlates with slower divisions, longer gap phases and introduction of cell‐cycle checkpoints (Orford and Scadden, 2008; Lange and Calegari, 2010). Also, factors promoting the G1‐to‐S phase transition impede differentiation (Kato and Sherr, 1993; Skapek et al, 1995; Richard‐Parpaillon et al, 2004; Lange et al, 2009). Despite growing evidence for the importance of cell‐cycle regulation in self‐renewal and differentiation, the connection between the cell cycle and critical stem cell factors is yet to be understood.
A conserved feature of the animal germline is a key role of RNA‐binding proteins in controlling the self‐renewal versus meiosis/differentiation decision. For example, members of the PUF (Pumilio and FBF) family of RNA‐binding proteins maintain germline stem cells (GSCs) in diverse organisms (Forbes and Lehmann, 1998; Crittenden et al, 2002; Salvetti et al, 2005). In C. elegans, two redundant PUF proteins, FBF‐1 and ‐2 (collectively referred to as FBF), are required for stem cell maintenance, as GSCs are lost from fbf‐1 fbf‐2 mutant gonads (Crittenden et al, 2002). Previously, FBF was shown to repress expression of the differentiation‐promoting protein GLD‐1 (Suh et al, 2009). This finding provided a paradigm for compartmentalization of factors regulating self‐renewal versus differentiation, but did not explain how these factors affect the cell cycle.
Recently, we have shown that the differentiation‐promoting role of GLD‐1 involves translational repression of cye‐1 (Cyclin E) mRNA, which prevents ectopic activation of CYE‐1/CDK‐2 in germ cells undergoing meiosis and their reversal to self‐renewal (Biedermann et al, 2009). Here, we show that repression of CKI‐2, a member of the Cip/Kip family of cyclin‐dependent kinase inhibitors (CKIs; Buck et al, 2009), is important for the maintenance of GSCs. We demonstrate that CKI‐2 is repressed in GSCs and that this repression is mediated by conserved elements in the cki‐2 mRNA 3′UTR that recruit FBF. Importantly, while GSCs are lost from fbf‐1 fbf‐2 gonads (Crittenden et al, 2002), GSCs are restored in fbf‐1 fbf‐2 gonads upon depletion of CKI‐2, suggesting that FBF‐mediated repression of CKI‐2 is critical for maintenance of GSCs. To our knowledge, these findings establish the first direct link between a conserved stem cell factor and the cell cycle in adult stem cells. Because Cip/Kip CKIs in worms and other animals inhibit Cdk2 activity (Besson et al, 2008), we propose that FBF and GLD‐1 regulate the self‐renewal versus differentiation decision, at least in part, by patterning CYE‐1/CDK‐2 activity; ensuring high levels of CYE‐1/CDK‐2 in GSCs to promote self‐renewal, and low levels in cells undergoing meiosis to promote differentiation.
The Cip/Kip protein CKI‐2 is repressed in GSCs
The C. elegans genome encodes two members of the Cip/Kip family: CKI‐1 and CKI‐2 (Figure 1A). While CKI‐1 is required in somatic blast cells for the proper timing of cell‐cycle withdrawal (Hong et al, 1998), CKI‐2 is not essential, having a minor role during vulval development (Buck et al, 2009). By semiquantitative RT–PCR and immunofluorescent detection, we found that of the two Cip/Kip proteins, CKI‐2 is the predominant CKI in the adult germline (Figure 1C and D; Supplementary Figure S1A and B). Immunodetection studies revealed that CKI‐2 protein was absent from GSCs and became expressed in cells entering meiosis (Figure 1D). In both cki‐2(ok2105) mutants and cki‐2(RNAi) animals, CKI‐2 protein is essentially absent (Figure 1B and C; unpublished observation). Yet, the animals are viable and fertile (Buck et al, 2009), suggesting that CKI‐2 is not essential for the entry into and progression through meiosis. However, since the Cip/Kip proteins tend to be absent from pluripotent cells (Ramalho‐Santos et al, 2002; Ginis et al, 2004), we wondered whether repression of CKI‐2 in GSCs is important for their maintenance.
CKI‐2 repression is mediated through the 3′UTR of its mRNA
The self‐renewal of C. elegans GSCs depends on GLP‐1/Notch signalling, but its relevant targets remain unclear (Kimble and Crittenden, 2007). To determine if Notch signalling regulates CKI‐2 expression, we measured cki‐2 mRNA levels by quantitative RT–PCR in gonads in which GLP‐1 was either active or inactive. Specifically, we dissected gonads from gld‐1 gld‐2; glp‐1(gf) (gf: gain‐of‐function) animals (GLP‐1 ON) and from gld‐1 gld‐2; glp‐1(lf) (lf: loss‐of‐function) animals (GLP‐1 OFF) (Priess et al, 1987; Kodoyianni et al, 1992; Kadyk and Kimble, 1998; Pepper et al, 2003; Supplementary Figure S1C). We found that cki‐2 mRNA abundance increased moderately in the absence of GLP‐1/Notch signalling (Supplementary Figure S1D). Surprisingly, CKI‐2 protein was absent from both GLP‐1 ON and GLP‐1 OFF gonads (unpublished observation). Thus, though GLP‐1 signalling affects cki‐2 mRNA abundance (directly or indirectly), an additional regulatory mechanism preventing CKI‐2 protein expression must exist.
Post‐transcriptional regulation is common in germ cells and is often mediated by the 3′UTR of mRNAs (Merritt et al, 2008). To test if the cki‐2 3′UTR mediates repression in GSCs, we produced single‐copy integrated transgenic lines expressing a GFP–H2B reporter from a constitutive (mex‐5) germline promoter. When this reporter was fused to the housekeeping tbb‐2 (tubulin) 3′UTR, it was expressed throughout the germline. In contrast, a reporter fused to the cki‐2 3′UTR was repressed in stem cells (Figure 2A). Thus, a 3′UTR‐based mechanism may be sufficient to repress CKI‐2 in GSCs.
Putative FBF‐binding elements in cki‐2 3′UTR mediate reporter repression in GSCs
To identify RNA elements regulating cki‐2 mRNA expression, we aligned cki‐2 3′UTRs from various nematode species. We observed several blocks of homology. One contains a putative MEX‐3 recognition element (MRE; Pagano et al, 2009); an additional MRE was found in a non‐conserved region. Another conserved region contains several potential FBF‐binding elements (FBEs) that are recognized by PUF proteins (Wickens et al, 2002; Supplementary Figure S2). Because PUF and MEX‐3 proteins have been implicated in germ cell proliferation (Crittenden et al, 2002; Ariz et al, 2009), we tested whether their putative binding sites were required for cki‐2 repression.
We produced a set of reporter constructs fused to the cki‐2 3′UTR containing overlapping deletions (Figure 2B); these reporters were each integrated as a single copy into a defined genomic locus using MosSCI (Frøkjaer‐Jensen et al, 2008). Removing the conserved MEX‐3 binding site (ΔM2) was not sufficient to cause reporter derepression in wild‐type GSCs. Consistently, the wild‐type 3′UTR reporter remained repressed in GSCs in the mex‐3 mutant (unpublished observation). Nevertheless, since MEX‐3 can promote proliferation in sensitized backgrounds (Supplementary Figure S3C and D; Ariz et al, 2009), we initially combined the MEX‐3 binding site mutation with larger deletions (Δ1–5). We found that deleting fragments 3 and M2 caused reporter derepression (Figure 2B). Importantly, also a smaller deletion removing only the potential FBEs (ΔFBE1–4) but leaving MRE2 intact was equally sufficient to derepress the reporter (Figure 2B). Removing only the highly conserved motif FBE1 resulted in barely detectable derepression (unpublished observation), indicating that multiple FBEs within the cluster contribute to repression of cki‐2 mRNA. Thus, though MRE2 is conserved, associates with MEX‐3, and may contribute to cki‐2 regulation under special circumstances (Supplementary Figure S3), it is FBEs that are normally critical for cki‐2 mRNA repression in GSCs.
FBF represses the cki‐2 3′UTR reporter in GSCs
To complement the above analysis, we tested if a reporter containing the wild‐type cki‐2 3′UTR was derepressed in fbf‐1 fbf‐2 gonads. Since FBF maintains self‐renewal in the adult germline, adults have no GSCs in which to assay reporter repression. For this reason, and as has been done previously (Merritt and Seydoux, 2010), we assayed reporter expression in fbf‐1 fbf‐2 mutant L4 larvae, at which stage FBF is not yet essential. In contrast to wild‐type animals that repressed the reporter in GSCs, we observed its partial derepression in fbf‐1 fbf‐2 animals (Figure 3A). To determine if the residual repression depends on the region containing putative FBEs, we analysed expression of the ΔFBE1–4 reporter at the same larval stage and found that it was fully derepressed (Figure 3B). Because another PUF protein, PUF‐8, has been implicated in germ cell proliferation (Ariz et al, 2009), we tested whether PUF‐8 contributes to cki‐2 reporter repression. We found that RNAi‐mediated depletion of PUF‐8 neither derepressed the reporter (Supplementary Figure S4A), nor enhanced reporter derepression in fbf mutants (unpublished observation). These results demonstrate that FBEs mediate cki‐2 repression in larval and adult GSCs, and suggest that an additional factor (apparently distinct from FBF, PUF‐8 and MEX‐3) contributes to repression of cki‐2 in GSCs during larval development when FBF is not critical.
FBF binds the FBEs mediating cki‐2 repression
To test if FBF associates with its putative binding elements in the cki‐2 3′UTR, we used fluorescence electrophoretic mobility shift (F‐EMSA) and fluorescence polarization (FP) assays (Pagano et al, 2009, 2011). Since FBF‐1 and FBF‐2 act redundantly (Lamont et al, 2004) and have similar RNA binding characteristics (Bernstein et al, 2005), these assays were performed only with FBF‐2. Within the conserved region of the cki‐2 3′UTR that mediates regulation of the reporter (Figure 2B), five UGU motifs are found in the context of four possible FBEs, FBE1 through 4 (Supplementary Figure S2). FBF‐2 was found to bind with high affinity to FBE1; this binding is comparable to the previously described FBF affinity to FBEa, a binding element in the gld‐1 3′UTR (Bernstein et al, 2005; Wang et al, 2009; Figure 4; Table 1). We also found that FBF‐2 bound with modest affinity to FBE3 and FBE4 elements, and weakly to FBE2, which is the least conserved element (Supplementary Figure S2; Table 1). Mutating the UGU sequence to ACA eliminated FBF binding to these elements (Table 1). These experiments demonstrate that FBF directly binds FBEs that mediate repression of cki‐2 in GSCs.
Removing CKI‐2 rescues the stem cell loss observed in fbf‐1 fbf‐2 mutant gonads
Expectedly, we found that CKI‐2 was derepressed in the distal‐most cells of fbf‐1 fbf‐2 mutant gonads (Figure 5A). To test if this ectopic expression of CKI‐2 in fbf‐1 fbf‐2 gonads is responsible for the stem cell loss observed in these animals, we constructed fbf‐1 fbf‐1 cki‐2 triple mutants. As previously observed, fbf‐1 fbf‐2 mutant animals rapidly lost GSCs after the adult moult (Crittenden et al, 2002; Figure 5B and C). In contrast, fbf‐1 fbf‐2 cki‐2 adult animals retained GSCs beyond day 2 of adulthood (Figure 5B and C). This interaction appeared to be specific to FBF and CKI‐2, as the cki‐2 mutation did not suppress the reduced proliferation phenotype seen in puf‐8(RNAi) (Supplementary Figure S4B) or puf‐8(RNAi); mex‐3 animals (Ariz et al, 2009; unpublished observation). Together, these results demonstrate that CKI‐2 is repressed in GSCs by FBF, which binds a conserved regulatory region in cki‐2 mRNA 3′UTR, and suggest that FBF‐mediated repression of CKI‐2 is critical for the maintenance of GSCs.
Several members of the PUF protein family have been associated with stem cell self‐renewal in diverse organisms (Wickens et al, 2002). Like many RNA‐binding proteins, FBF associates with, and presumably regulates, hundreds of mRNAs (Kershner and Kimble, 2010). Apart from cki‐2, its targets include mRNAs encoding structural components of meiotic chromosomes such as HIM‐3, HTP‐1 and ‐2, SYP‐1 and SYP‐2, which are ectopically expressed in FBF(−) gonads (Kershner and Kimble, 2010; Merritt and Seydoux, 2010). However, in contrast to CKI‐2, depletion of those meiotic proteins does not restore stem cells in fbf‐1 fbf‐2 mutants (Merritt and Seydoux, 2010). This suggests that cki‐2 mRNA is the key target of FBF in stem cell maintenance.
Though cki‐2 deletion has no obvious consequences for germ cell development, our findings suggest that CKI‐2 repression in GSCs is critical for stem cell maintenance. This agrees with observations that cell‐cycle inhibitors, including Cip/Kip family members, are usually depleted from stem cells (Stead et al, 2002; Ginis et al, 2004). The regulation of Cip/Kip mRNAs by PUF proteins appears to be evolutionarily conserved (Morris et al, 2008). A vertebrate PUF protein, PUM1, binds to mRNA encoding p27, allowing miRNA‐mediated repression (Kedde et al, 2010). However, cki‐2 is a weak candidate for miRNA‐mediated regulation based on the available prediction tools (MirWip, PicTar, TargetScan and MirZ; our observation). Only the TargetScan software identified the (non‐conserved) seed region of miRNAs 44, 45, 247 and 61 (CUAGUCA) in the cki‐2 3′UTR. However, the Δ1 deletion removes this site without affecting distal repression, indicating that these miRNAs are not essential for CKI‐2 regulation in the germline. On the other hand, PUF proteins have been shown to repress mRNAs by recruiting deadenylases (Goldstrohm and Wickens, 2008). In C. elegans, at least one mRNA, gld‐1, appears to be repressed by FBF in a similar manner (Suh et al, 2009), suggesting a possible mechanism for cki‐2 regulation. Somewhat consistently with this possibility, the cki‐2 message is less abundant in GSCs (this study and Kim and Roy, 2006).
Cip/Kip CKIs are well‐established inhibitors of Cyclin E/Cdk2 activity. Therefore, repression of CKI‐2 in GSCs (this study), and of CYE‐1/Cyclin E in differentiating cells (Biedermann et al, 2009), is expected to produce a high‐low gradient of CYE‐1/CDK‐2 activity in the germline, which could be important for the self‐renewal versus differentiation decision (Figure 6). Consistently with this model, proliferation of GSCs is sensitive to CYE‐1 and CDK‐2 levels (Seydoux et al, 1993; Fox et al, 2011; Jeong et al, 2011). Although this work is focused on regulation by RNA‐binding proteins, other mechanisms may contribute. For example, CKIs are subject to proteasomal degradation (Lu and Hunter, 2010), and depletion of the C. elegans E3 ligase CUL‐2 (Feng et al, 1999) or core proteasome components (our unpublished observation) causes proliferation defects that may involve stabilization of CKIs.
Interestingly, murine Cyclin E and Cdk2, which are mostly dispensable for development, have critical functions in the mouse germline (Geng et al, 2003; Ortega et al, 2003). Conversely, mitotic arrest and differentiation of germ cells in the fetal testes correlate with suppression of Cyclin E and activation of several CKIs, including p27Kip1 (Western et al, 2008). Also, accumulation of p27, achieved by protection of p27 mRNA from miRNA‐mediated repression through the RNA‐binding protein DND1 (Kedde et al, 2007), is thought to facilitate cell‐cycle arrest and differentiation of spermatogenic cells (Cook et al, 2011). Thus, Cyclin E/Cdk2 emerges as an important player in germ cell development. Although the direct role of Cyclin E/Cdk2 in the self‐renewal versus differentiation decision in the mammalian germline remains to be tested, it is intriguing that Cdk2 has been implicated in the self‐renewal of embryonic stem (ES) cells (Neganova et al, 2009). This might reflect an intimate connection between Cyclin E/Cdk2 and self‐renewal in general, or a shared mechanism operating in germ cells and early embryonic cells, from which ES cells are derived.
Materials and methods
Nematode culture, mutants, constructs and transgenic lines
Standard procedures were used to maintain animals and perform RNAi by feeding. Worms were grown at 25°C unless stated otherwise. Alleles used were mex‐3(or20), gld‐2(q492), gld‐1(q485), glp‐4(bn2)ts (I); fbf‐1(ok91), fbf‐2(q704), cki‐2(ok2105) (II), glp‐1(q175), glp‐1(ar202)ts, glp‐1(e2141)ts (III). Unless stated otherwise, all mutations are, or phenocopy, null alleles.
The cki‐2 cDNA was determined by 3′RACE using SL1‐complementary forward primers. Transgenic lines were generated using the Gateway System (Merritt et al, 2008) and the MosSCI direct insertion protocol (Frøkjaer‐Jensen et al, 2008). For coordinates of deletions, see Supplementary Table S2.
Immunofluorescence and microscopy
Immunofluorescence was performed as described previously (Lin et al, 1998). Antibody dilutions were 1:15 for CKI‐1 and 1:20 for CKI‐2 (Feng et al, 1999), 1:2000 for H3Ser10p (Upstate 06‐570; Hendzel et al, 1997) and 1:200 for HTP‐3 (Goodyer et al, 2008). Secondary antibodies: goat anti‐mouse alexa‐568, anti‐rabbit alexa‐568 and anti‐rabbit alexa‐488 were from Molecular Probes. A Zeiss AxioImager Z1 microscope equipped with AxioncamMRm REV 2 CCD camera was used to capture images. Quantification of reporter derepression in fbf‐1 fbf‐2 mutants: Fluorescence was quantified using ImageJ and compared with fluorescence in distal‐most nuclei of wild‐type animals carrying the same reporter. The P‐value was calculated by t‐test in R. All images subject to direct comparison were taken at identical exposure times and were processed with Adobe Photoshop CS2 in an identical manner.
RT quantitative PCR on dissected gonads and semiquantitative PCR
gld‐2(q492) gld‐1 (q485); glp‐1(ar202) and gld‐2(q492) gld‐1(q485); glp‐1(e2141) animals were grown at the restrictive temperature (25°C). Fifty gonads per strain were dissected for each of three replicates in M9 buffer+1 mM Levamisole and transferred to Trizol. 1.5 μg of mouse RNA was added to the sample to facilitate precipitation. RNA was extracted according to the manufacturer's instructions. cDNA was synthesized with random hexamers using the ImProm II Reverse transcription system from Promega according to the manufacturer's instructions from 1 μg RNA. Real‐time PCR was performed in duplicate on 1/20 of obtained cDNA as previously described (Biedermann et al, 2009).
For semiquantification of cki‐1 mRNA, RNA was extracted with Trizol from young non‐gravid adults and from embryos. cDNA was produced as above from 1 μg RNA. PCR was performed on 1/200 of the cDNA reaction using FastTaq Polymerase from Roche for 30 cycles. Images were acquired with a Syngene Transilluminator and GeneSnap software and processed with Adobe Photoshop CS2.
RNA was extracted with Trizol from young non‐gravid adults, from glp‐4 animals grown at restrictive temperature, and from cki‐2 animals. 40 μg RNA was loaded per lane. The probe corresponds to SL1, 5′UTR and CDS until nucleotide 462 (exons 1–3). The DIG Northern Starter Kit from Roche was used to prepare the probe and perform hybridization according to the manufacturer's instruction, except that the blocking solution was used at 10 × concentration.
The PUF domain (amino acids 121–632) of FBF‐2 was cloned into pGEX‐6P‐1 (GE Healthcare) and transformed into BL21(DE3) codon plus cells. Cultures were grown at 37°C to an OD600 of 0.4, shifted to 4°C for 15 min, induced with 0.1 mM IPTG, then incubated at 16°C overnight. Cell pellets were resuspended in lysis buffer (50 mM Tris pH 7.5, 500 mM KCl and 2 mM DTT), supplemented with one Complete Mini EDTA‐free protease tablet (Roche) per 50 ml of buffer, and disrupted using a microfluidizer. Following clarification, the protein was purified by affinity chromatography using a glutathione sepharose 4B (GE Healthcare) column. To remove co‐purifying nucleic acids, the column was washed with lysis buffer supplemented with 1 M KCl then exchanged into 50 mM Tris pH 7.5, 150 mM KCl, 1 mM EDTA and 2 mM DTT. The protein was eluted by incubation with Precission protease (GE Healthcare) at 4°C overnight which cleaves in between FBF and the N‐terminal GST tag. Additional contaminants were removed using a HiTrap Q HP (GE Healthcare) column where purified FBF was recovered in the flow‐through. MEX‐3 was purified as previously described (Pagano et al, 2009).
RNA oligonucleotides (IDT) were labelled as described by Pagano et al (2011) and MEX‐3 binding assays were performed as described in Pagano et al (2009). FBF‐binding reactions were performed as described in Koh et al (2011) and Wang et al (2009), with the exception that F‐EMSA and FP data were collected as described in Pagano et al (2011). See Supplementary Methods for oligonucleotide sequences.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We thank Asja Moerkamp and Mathias Senften for experimental assistance; Hans R Hotz for bioinformatic support; Ed Kipreos for CKI antibodies; Christian Lehner, Iskra Katic and Susanne Finger for comments on the manuscript. Some strains were provided by the Caenorhabditis Genetics Center funded by the NIH. IK was supported by a Boehringer Ingelheim Fonds PhD fellowship; RC by a Leukemia & Lymphoma Society Special Fellow Award; SPR by NIH grant GM081422. The Friedrich Miescher Institute is supported by the Novartis Research Foundation.
Author contributions: IK, RC, BF and SR designed experiments. BF performed in vitro association assays. IK performed all other experiments. SP provided critical technical assistance. IK and RC wrote the paper.
- Copyright © 2011 European Molecular Biology Organization