Cyclin E and Cdk2 Control GLD-1, the Mitosis/Meiosis Decision, and Germline Stem Cells in Caenorhabditis elegans

Coordination of the cell cycle with developmental events is crucial for generation of tissues during development and their maintenance in adults. Defects in that coordination can shift the balance of cell fates with devastating clinical effects. Yet our understanding of the molecular mechanisms integrating core cell cycle regulators with developmental regulators remains in its infancy. This work focuses on the interplay between cell cycle and developmental regulators in the Caenorhabditis elegans germline. Key developmental regulators control germline stem cells (GSCs) to self-renew or begin differentiation: FBF RNA–binding proteins promote self-renewal, while GLD RNA regulatory proteins promote meiotic entry. We first discovered that many but not all germ cells switch from the mitotic into the meiotic cell cycle after RNAi depletion of CYE-1 (C. elegans cyclin E) or CDK-2 (C. elegans Cdk2) in wild-type adults. Therefore, CYE-1/CDK-2 influences the mitosis/meiosis balance. We next found that GLD-1 is expressed ectopically in GSCs after CYE-1 or CDK-2 depletion and that GLD-1 removal can rescue cye-1/cdk-2 defects. Therefore, GLD-1 is crucial for the CYE-1/CDK-2 mitosis/meiosis control. Indeed, GLD-1 appears to be a direct substrate of CYE-1/CDK-2: GLD-1 is a phosphoprotein; CYE-1/CDK-2 regulates its phosphorylation in vivo; and human cyclin E/Cdk2 phosphorylates GLD-1 in vitro. Transgenic GLD-1(AAA) harbors alanine substitutions at three consensus CDK phosphorylation sites. GLD-1(AAA) is expressed ectopically in GSCs, and GLD-1(AAA) transgenic germlines have a smaller than normal mitotic zone. Together these findings forge a regulatory link between CYE-1/CDK-2 and GLD-1. Finally, we find that CYE-1/CDK-2 works with FBF-1 to maintain GSCs and prevent their meiotic entry, at least in part, by lowering GLD-1 abundance. Therefore, CYE-1/CDK-2 emerges as a critical regulator of stem cell maintenance. We suggest that cyclin E and Cdk-2 may be used broadly to control developmental regulators.


Introduction
Metazoan development coordinates progression through the cell cycle with key developmental events, such as stem cell maintenance, cell fate specification and patterning. An emerging theme is that core cell cycle regulators can influence development in addition to their more traditional role. For example, cyclin E affects the Drosophila endocycle [1,2], CDK1 influences Drosophila neuroblast asymmetric divisions [3] and cyclin D controls the Caenorhabditis elegans asymmetric division of somatic gonadal precursors [4]. Yet our understanding of the molecular interplay between cell cycle and developmental regulators remains in its infancy with only a few exceptions (e.g. MyoD and CKI [5][6][7]).
The C. elegans adult germline provides a superb model to investigate cell cycle controls in a developmental context [8]. Germ cells progress from mitotic divisions at the distal end of the adult gonad through meiotic prophase I in the middle to overt differentiation as sperm or oocyte at the proximal end. Here we focus on the distal germline, including the ''mitotic zone'' and adjacent ''transition zone'' ( Figure 1A). The mitotic zone houses a pool of .200 mitotically dividing germ cells, whereas the transition zone contains germ cells that have entered the meiotic cell cycle. The transition zone is recognized by the presence of crescent-shaped DAPI-staining of nuclei, typical of the first step in pairing homologous chromosomes [9]. As germ cells move proximally through the mitotic zone, most enter meiotic S-phase prior to their entry into the transition zone [10][11][12]. The distal germline therefore presents an orderly maturation from stem cell through transit-amplifying cell to meiotic entry and differentiation.
Major developmental regulators controlling the choice between germline self-renewal and differentiation are well established ( Figure 1B) [8]. A single somatic cell, the distal tip cell (DTC), provides the stem cell niche, and Notch signaling from the DTC drives self-renewal at the expense of meiotic entry and differentiation. Germ cells within the niche express FBF-1 and FBF-2, two nearly identical and largely redundant PUF (Pumilio and FBF) RNA-binding proteins [13,14]; the FBF proteins promote germline self-renewal and repress differentiation [15][16][17]. When the DTC is removed by laser ablation or when GLP-1/Notch signaling or both FBFs are removed by either mutation or RNAi, germ cells leave the mitotic cell cycle, enter the meiotic cell cycle, and progress through meiotic prophase and gametogenesis [13,15,18].
Most relevant here is the FBF repression of gld-1 mRNA, which is exerted directly through a cis-acting regulatory element in the gld-1 39 untranslated region and which is responsible for lowering gld-1 expression in the distal mitotic zone [15,27]. In addition, the roles of FBF-1 and FBF-2 are subtly different even though they are redundant for maintenance of germline self-renewal [15,30]. For example, in fbf-1 single mutants (but not in fbf-2 mutants), GLD-1 is expressed ectopically in the distal-most germ cells within the niche, even though those distal germ cells remain in the mitotic cell cycle and germline self-renewal is maintained in both single mutants. Therefore, simply the presence of GLD-1 protein does not drive germ cells from the mitotic into the meiotic cell cycle [15,30].
Here we focus on cyclin E, which is best known as a regulator of the G1/S transition [39,40]. In C. elegans, the cyclin E homolog, known as CYE-1, acts with CDK-2, the Cdk2 homolog, to facilitate the G1/S transition in somatic cells and to promote proliferation of germ cells [36,[41][42][43][44][45]. In the adult germline, CYE-1 is easily detectable throughout the mitotic zone but becomes undetectable in the transition zone and pachytene region [43]. Its presence in germ cells throughout the mitotic zone suggests that CYE-1 may have a broader role than control of the G1/S transition, because half of the mitotic zone germ cells are in Sphase [11]. In the transition zone, the GLD-1 translational repressor acts directly on cye-1 mRNA to reduce its activity [23]. This GLD-1 control of cye-1 mRNA translation provided the first link between developmental and cell cycle regulators in this system.
In addition to its role in cell cycle regulation, cyclin E can affect developmental events. In C. elegans, non-vulval cells are transformed into vulval cells upon CYE-1 removal [42], the asymmetric division of somatic gonadal precursor cells is rendered symmetrical [4] and quiescent somatic precursor cells are spurred to differentiate [45]. In Drosophila, cyclin E affects neuroblasts [46,47] and ovarian follicle stem cells [48]. Most recently, cyclin E has been implicated in the control of human embryonic stem cells [49]. Therefore, the coordination of cell cycle with development appears to be achieved, at least in part, by a dual role of key cell cycle regulators [39,50].
Here we report that CYE-1/CDK-2 controls the balance between proliferation and differentiation in the C. elegans germline, at least in part, by lowering GLD-1 abundance in dividing cells. We provide multiple lines of evidence to suggest that GLD-1 is a direct substrate of CYE-1/CDK-2 phosphorylation. These results forge a molecular and functional link between the CYE-1/ CDK-2 cell cycle regulator and the GLD-1 developmental regulator. Because GLD-1 represses cye-1 mRNA translation [23], these two regulators engage in a double negative feedback loop. Finally, we demonstrate that CYE-1/CDK-2 and FBF-1 normally function together to repress GLD-1 and prevent meiotic entry of germ cells in the stem cell pool. Therefore, CYE-1/ CDK-2 emerges as a critical regulator of germline stem cell maintenance. Schematic of adult distal germline. The distal tip cell (DTC) niche (red) resides at the distal end. The mitotic zone contains mostly germ cells in the mitotic cell cycle (yellow) but also includes some germ cells in meiotic S-phase (green) at its proximal edge; the germline stem cell (GSC) pool is located at the distal end and consists of 30-70 germ cells. The transition zone contains germ cells in meiotic S-phase (green circles) and meiotic prophase (green crescents). The dashed line marks the boundary between mitotic and transition zones. (B) Skeleton of regulatory network controlling the decision between germline selfrenewal and differentiation, a decision that also controls the choice between mitotic and meiotic cell cycles. See Introduction for details. doi:10.1371/journal.pgen.1001348.g001

Author Summary
How are cell cycle regulators coordinated with cell fate and patterning regulators during development? Several studies suggest that core cell cycle regulators can influence development, but molecular mechanisms remain unknown for the most part. We have tackled this question in the nematode Caenorhabditis elegans. Specifically, we have investigated how cell cycle regulators affect germline stem cells. Previous work had identified conserved developmental regulators that control the choice between self-renewal and differentiation in this tissue. In this work, we focus on cyclin E/Cdk-2, which is a core cell cycle kinase, and GLD-1, a key regulator of stem cell differentiation. Our work shows that cyclin E/Cdk-2 phosphorylates GLD-1 and lowers its abundance in stem cells via a posttranslational mechanism. We also find that a posttranscriptional GLD-1 regulator, called FBF-1, works synergistically with cyclin E/Cdk-2 to ensure that GLD-1 is off in germline stem cells. When both FBF-1 and cyclin E/Cdk-2 are removed, the stem cells are no longer maintained and instead differentiate. Our findings reveal that cyclin E/Cdk-2 kinase is a critical stem cell regulator and provide a paradigm for how cell cycle regulators interface with developmental regulators.

CYE-1/CDK-2 controls the mitosis/meiosis balance in adult germlines
To assess CYE-1 function in the adult hermaphrodite germline, we placed mid to late L4 larvae on bacteria expressing doublestranded RNA (dsRNA) complementary to cye-1 mRNA and scored germlines after 48 hours, well into adulthood. This regimen effectively reduced cye-1 expression (Figure 2A), and is referred to henceforth as adult RNAi (aRNAi).
cye-1 aRNAi had a dramatic effect on the mitotic zone, which extends from the distal end of the adult germline along the gonadal axis to the first set of crescent-shaped nuclei in early meiotic prophase. Normally, the mitotic zone harbors actively cycling germ cells, it extends 18-20 germ cell diameters (gcd) along the gonadal axis from the distal end, and its total germ cell number exceeds 200 [51]. However, after cye-1 aRNAi, this region was abnormal: germ cell nuclei were grossly enlarged (Figure 2A-2C), the mitotic index plummeted from 3% to 0.5% ( Figure 2E), the length of the region, measured in gcd from the distal end, was reduced from 19 gcd in controls to 12 gcd (Figure 2A, 2D), and total number of germ cells was reduced from 219 to 77 ( Figure 2F). After 72 or 94 hrs of cye-1 aRNAi treatment, these abnormal distal germ cells began to deteriorate ( Figure S1A). Similar results were obtained after cdk-2 aRNAi ( Figure 2D-2F). Therefore, CYE-1 and CDK-2 appear to act together to keep about one-third of the germ cells in the mitotic zone from enlarging abnormally and to prevent the remaining two-thirds from entering meiotic prophase.
We wondered if the enlarged distal germline nuclei might reflect partial rather than full depletion of CYE-1/CDK-2 activity. A cye-1 null mutant or cye-1 RNAi treatment at an early stage in larval development affects both somatic and germ cell divisions [42]; moreover, somatic cells control germ cell proliferation [52,53]. We therefore turned to other methods to address this issue. First, we enhanced aRNAi with eri-1 and rrf-3 mutants [54,55]; however, the mitotic zone length was still shorter after cye-1 depletion ( Figure S1B) and enlarged distal germ cells were still observed. Second, we restricted RNAi to the germline using rrf-1 mutants, which are defective for somatic but not germline RNAi; this mutant therefore permitted CYE-1 depletion in young larvae without affecting somatic divisions [55]. However, cye-1 RNAi treatment initiated in rrf-1 first stage larvae (L1s) again resulted in typical distal germline defects ( Figure S1B). Third, we generated a genomic cye-1 transgene, qIs134 ( Figure S2A), using a method predicted to allow expression in somatic cells but not germ cells [56] (see Experimental Procedures). Consistent with that prediction, CYE-1 protein was detected in somatic but not germ cells of cye-1(0); qIs134 embryos ( Figure S2B) and qIs134 rescued cye-1(0) somatic defects ( Figure S2C). However, germlines in these somatically rescued cye-1(0); qIs134 animals were smaller than normal, with abnormally enlarged distal germ cells ( Figure S2D, bottom), as seen previously in cye-1 null mutants [41][42][43]. Therefore, the enlarged germline nuclei observed after cye-1 and cdk-2 aRNAi are likely to represent the effect of eliminating CYE-1/CDK-2 activity. In addition, these experiments show that CYE-1 functions autonomously within germ cells.
We next attempted to ask if the enlarged distal germ cells were in the mitotic or meiotic cell cycle. To this end, we first stained for cytoplasmic REC-8, a marker of mitotic germ cells [10], and chromosomal HIM-3, a marker of meiotic germ cells [28]. In most cye-1(aRNAi) germlines (97.1%, n = 35), REC-8 was lower in the mitotic zone relative to controls ( Figure 2B) and HIM-3 expanded distally (82.5%, n = 40; Figure 2C). This shift in marker distribution suggested that the enlarged distal germ cells may have entered the meiotic cell cycle. However, the finite mitotic index, albeit low, suggested that at least some germ cells remained in the mitotic cell cycle. To explore this question further, we examined DNA content of DAPI-stained nuclei, comparing values for the abnormally enlarged germline nuclei to those of nuclei in the somatic DTC (2n) and pachytene germ cells (4n). This analysis revealed a range of DNA content from 2n to 4n in the abnormally enlarged germline nuclei that was essentially the same as seen in wild-type mitotic zone nuclei ( Figure 2G). DNA content therefore does not distinguish between germ cells being in mitotic or meiotic S-phase. Figure 2H depicts one interpretation of the phenotype typical of cye-1(aRNAi) distal germlines. The wild-type mitotic zone contains .200 germ cells in the mitotic cell cycle ( Figure 2H, above). However, the cye-1(aRNAi) ''mitotic zone'' contains only ,75 enlarged germ cells whose state is questionable ( Figure 2H, below). Figure 2H suggests that the distal most ,75 germ cells enlarge after cye-1 depletion while the more proximal .100 germ cells enter meiotic prophase prematurely ( Figure 2H, dashed lines between wild-type and cye-1(aRNAi) distal germlines). An alternative explanation might have been that .100 germ cells die, but no germ cell corpses or degenerating germ cells were seen during the transition from the wild-type .200 germ cells mitotic zone to the cye-1(aRNAi) ,75 germ cells mitotic zone. A third possibility is that the 75 enlarged germ cells come from some other part or parts of the mitotic zone and that the remaining germ cells are squeezed proximally as the ,75 enlarge. Regardless, we conclude that cye-1 depletion causes ,75 of the .200 germ cells in the mitotic zone to enlarge abnormally and that it also causes the remaining .100 germ cells in the mitotic zone to enter meiotic prophase. Therefore, CYE-1/CDK-2 is required to achieve and maintain the normal balance between numbers of germ cells in the mitotic and meiotic cell cycles.

CYE-1/CDK-2 acts through GLD-1 to control distal germ cells
We postulated that CYE-1/CDK-2 might exert its influence in the mitotic zone through GLD-1, a key repressor of germline proliferation (see Introduction). We first tested for a change in GLD-1 abundance after CYE-1/CDK-2 depletion. In wild-type and control germlines, GLD-1 was barely detectable in the distal mitotic zone, increased in the proximal mitotic zone and became abundant in the transition zone ( Figure 3A, top), as seen previously [26,57]. By contrast, GLD-1 was abundant throughout the mitotic zone in both cye-1(aRNAi) and cdk-2(aRNAi) germlines ( Figure 3A, middle and bottom). A similar GLD-1 increase was also observed in gld-1(q361) homozygotes ( Figure S3A, S3B), which harbor a G227D point mutation in the KH domain resulting in loss of GLD-1 function [58]. Therefore, the GLD-1 increase upon CYE-1/CDK-2 depletion is not dependent on GLD-1 function. We conclude that CYE-1 and CDK-2 lower the abundance of GLD-1 in the wild-type mitotic zone.

CYE-1/CDK-2 likely phosphorylates GLD-1
We next considered the idea that GLD-1 protein might be a direct target of the CYE-1/CDK-2 kinase. This prospect seemed plausible because the GLD-1 amino acid sequence possesses three consensus CDK phosphorylation sites (X-X-X-S/T-P-X-K/R) [59,60] ( Figure 4A and 4B, red) and four predicted cyclin binding sites (R/K-X-L-(X)-F/Y/L/I/V/M/P) [61,62] ( Figure 4A and 4B, blue). To ask if GLD-1 is a phosphoprotein, we examined GLD-1 on Western blots, with or without phosphatase treatment ( Figure 4C). The GLD-1 antibody recognized two bands in proteins prepared from wild-type extracts ( Figure 4C, left panel). Although the resolution of these two bands was not optimal and the lower band was much more abundant, phosphatase treatment reduced the upper band whereas phosphatase inhibitor restored it ( Figure 4C, left panel).
We reasoned that the relative amount of phosphorylated GLD-1 might be low in wild-type extracts, because CYE-1 and GLD-1 only overlap in a small region within the proximal part of the mitotic zone ( Figure S5A), whereas abundant GLD-1 extends over a much larger region [23,43,57]. In an attempt to increase the ratio of phosphorylated to unphosphorylated GLD-1, we used gld-3(0) nos-3(0) double mutants, which have a germline tumor [20] with broadly distributed CYE-1 and less GLD-1 overall ( Figure  S5B). Before assaying phosphorylation, we confirmed that CYE-1/ CDK-2 influences GLD-1 abundance in the gld-3 nos-3 tumorous germline ( Figure 4E). We then prepared extracts from the double mutants and assayed GLD-1 phosphorylation. The two GLD-1 bands were better resolved in the gld-3 nos-3 extracts ( Figure 4C, right), and again phosphatase treatment reduced the upper band while phosphatase inhibitor restored it ( Figure 4C, right panel). Finally, we assayed proteins from gld-3 nos-3 mutants after treatment with cye-1 or cdk-2 aRNAi and found that phosphorylated GLD-1 was indeed reduced ( Figure 4D). We conclude that GLD-1 is a phosphoprotein and that its phosphorylation depends on CYE-1 and CDK-2 in vivo.

GLD-1 CDK phosphorylation sites affect its abundance and mitotic zone size
To ask if CYE-1/CDK-2 controls GLD-1 abundance via its predicted phosphorylation sites, we generated transgenes encoding a GFP-and FLAG-tagged GLD-1 that was either wild-type or mutated at its putative CDK phosphorylation sites [dubbed transgenic (tr) GLD-1(wt) and trGLD-1(AAA), respectively] ( Figure 5A). Both trGLD-1(wt) and trGLD-1(AAA) made functional protein, as assayed by gld-1(0) mutant rescue to fertility. The distribution of the transgenic proteins was analyzed in five independent lines for each protein with equivalent results. Expression of the trGLD-1(wt) protein was very low or undetectable in the distal mitotic zone ( Figure 5B, top), but trGLD-1(AAA) was easily seen ( Figure 5B, bottom). Indeed, the abundance of trGLD-1(AAA) in the distal germline was greater than that of trGLD-1(wt), when quantified using ImageJ ( Figure 5C). The mitotic zone length in trGLD-1(AAA) germlines was also shorter: 13.9 gcd on average in trGLD-1(AAA) germlines (n = 40) compared to 17.8 gcd on average in trGLD-1(wt) control germlines (n = 36) ( Figure 5B, 5D). A similar effect was observed in a gld-1(0 mutant background, both with respect to higher protein abundance at the distal end and shortening of the mitotic zone ( Figure S8). We note however that cye-1 depletion increased trGLD-1(AAA) in the distal-most germ cells even more ( Figure  S9), suggesting that CYE-1/CDK-2 represses GLD-1 abundance both directly and indirectly. We also measured total transgenic GLD-1 protein with anti-GLD-1 antibodies in Western blots, and total transgenic gld-1 mRNA with primers to the gfp portion of the mRNA. The total trGLD-1(AAA) protein was twice as abundant as total trGLD-1(wt) protein ( Figure 5E), but mRNA levels were comparable ( Figure 5F). We conclude that the CDK phosphorylation sites in GLD-1 contribute to the control of its abundance in the distal germline and also contribute to control of mitotic zone length.

Discussion
This work explores the role of CYE-1 and CDK-2 in the C. elegans germline and comes to two major conclusions. First, CYE-1 and CDK-2 regulate the mitosis/meiosis balance, and do so, at least in part, by downregulating abundance of the GLD-1 translational repressor. Second, CYE-1 and CDK-2 work together with FBF-1 to prevent cells in the germline stem cell pool from precocious meiotic entry and differentiation. Our discussion places these findings in context with previously known controls of the mitosis/meiosis decision and stem cell maintenance.
The GLD-1 developmental regulator is likely a direct CYE-1/CDK-2 substrate GLD-1 normally facilitates the switch of germ cells from the mitotic cell cycle into the meiotic cell cycle (see Introduction). In wild type, the GLD-1 protein is either extremely low or absent in the germline stem cell pool, but after depletion of either CYE-1 or CDK-2, GLD-1 becomes abundant in the distal-most germ cells. That GLD-1 increase is not simply a reflection of a mitotic cell cycle defect since depletion of other cell cycle regulators (e.g. CDK-1, CDK-4, CYA-2, EMB-30) does not lead to a similar GLD-1 increase in the distal germline (this work and [63]). Therefore, CYE-1/CDK-2 appears to govern GLD-1 independently of its canonical role in the cell cycle.
Several lines of evidence suggest that GLD-1 is a direct substrate of CYE-1/CDK-2. First, endogenous GLD-1 is a phosphoprotein, as shown here and corroborated by mass spectrometry [64]. Indeed, CDK-2 was predicted to be the best candidate kinase for the GLD-1 peptides identified in the C. elegans phosphoproteome [64]. Second, GLD-1 phosphorylation depends on CYE-1 and CDK-2 in vivo, and human cyclin E/CDK2 phosphorylates GLD-1 in vitro. Third, transgenic GLD-1 with serine/threonine to alanine substitutions at predicted CDK phosphorylation sites is ectopically expressed in the distal germline. Yet no obvious cell cycle perturbation was observed in the transgenic GLD-1(AAA) germlines, supporting the suggestion that GLD-1 phosphorylation represents a cell cycle independent role of CYE-1/CDK-2. Intriguingly, cyclin B/Cdk1 phosphorylates a mammalian GLD- Distal end of dissected gonads with DAPI staining. Left panels, GFP fluorescence reports transgenic GLD-1; right panels, DAPI staining reveals boundary between mitotic and transition zones. Solid triangle marks distal end; v and V mark steps of comparable GLD-1 abundance; dashed line is boundary between mitotic and transition zones. Germlines were treated identically and confocal fluorescent images taken at the same settings and then stacked through Z-axis. Scale bar, 20 mm. (C) ImageJ quantitation of GFP in the distal-most germline. Units on y-axis refer to pixel intensity; background was subtracted from both measurements. Error bars indicate 95% confidence limits; asterisk denotes a statistically significant difference (p,4.46E 26 using Student's t-test). (D) Mitotic zone length is reduced in trGLD-1(AAA) compared to trGLD-1(wt). Error bars indicate 95% confidence limits; asterisk denotes a statistically significant difference (p,7.36E 210 using Student's t-test). (E) Total trGLD-1(AAA) is higher than total trGLD-1(wt). Western blots of proteins prepared at 48 hrs after L4. Top panel, GLD-1 antibodies detect both transgenic GLD-1 (upper band) and endogenous GLD-1 (lower band); middle panel, actin; bottom panel, quantitation of band intensities, measured with ImageJ and analyzed with Microsoft Excel. (F) Transgenic gld-1(wt) and gld-1(AAA) mRNAs were of comparable abundance. mRNAs were extracted at 48 hrs after L4 stage, cDNA synthesized by RT-PCR, and DNA fragments were amplified using primers against a GFP-encoding part of transgenic mRNA and against unc-54 for a control. Quantitation of PCR products was done using KODAK MI program. doi:10.1371/journal.pgen.1001348.g005 1 homolog, Sam68 (Src-associated in mitosis), during mitosis [65], implying a conserved link between proteins in the GLD-1 family and regulators of the cell cycle.

Mutual repression between GLD-1 and CYE-1/CDK-2
This work together with a previous study [23] reveals a doublenegative feedback loop between CYE-1/CDK-2 and GLD-1 (Figure 7, blue shading). Mutual repression is a classical network motif and has been suggested to serve as a 'toggle switch' for controlling decisions between two states [reviewed in 68]. Here the two relevant states are the mitotic and meiotic cell cycles. CYE-1/ CDK-2 represses GLD-1 to promote mitotic divisions [this work], and GLD-1 represses cye-1 mRNA to promote meiotic progression [23]. However, loss of CYE-1 or CDK-2 does not flip germ cells from the mitotic to meiotic cell cycle; instead it shifts the balance between the two cycles in favor of meiotic entry. Therefore, the mutual repression between GLD-1 and CYE-1/CDK-2 does not provide a toggle switch on its own. We suggest instead that this CYE-1/CDK-2 and GLD-1 double-negative feedback loop contributes to the bi-stable switch between the mitotic and meiotic cell cycles and helps to maintain the normal balance between them.
The C. elegans mitosis/meiosis decision is influenced not only by CYE-1/CDK-2 and GLD-1 but also by a distinct double-negative feedback loop between Notch signaling and GLD-1 (Figure 7, yellow shading). Notch signaling promotes mitotic divisions through GLD-1 repression, albeit indirectly via the FBFs [15,30], while GLD-1 directly represses translation of the glp-1/ Notch receptor mRNA [69]. Therefore, two circuits of mutual repression converge on GLD-1, one involving core cell cycle regulators and the other involving differentiation regulators. We suggest that this two-fisted mutual repression renders the switch more robust and helps to ensure that the two cell cycle programs are incompatible.
Mutual repression between key regulators of the cell cycle and cell differentiation is emerging as a theme of developmental control. In addition to the example reported here, other examples exist. One example is in developing skeletal muscle of mammals.  Figure 2). (C-E) GLD-1 quantitation in confocal images measured using ImageJ, averaged, and plotted. Error bars indicate 95% confidential limits; asterisks denote a statistically significant difference with p,0.002 using Student's t-test. (C) Quantitation was done at two sites: the distal-most germ cells (Distal) and more proximal germ cells (Proximal). The proximal site was located just beyond the MZ/TZ boundary in those germlines with a boundary; however in fbf-1; cye-1(aRNAi) and fbf-1; cdk-2(aRNAi), no MZ/TZ boundary exists so this proximal site was defined arbitrarily as 12 germ cell diameters from the distal end. (D) GLD-1 abundance at the distal-most site, quantitated in strains as noted. (E) GLD-1 abundance at the more proximal site, quantitated in strains as noted. doi:10.1371/journal.pgen.1001348.g006 MyoD activates expression of CDK inhibitors and thereby inhibits cell cycle progression; conversely, CDK1 or CDK2 phosphorylates MyoD and promotes its turnover to prevent cell differentiation [reviewed in 6]. A second example occurs in Drosophila neuroblasts. Prospero represses transcription of cell cycle regulators, including cyclin E, to block cell cycle progression [70,71], and conversely cyclin E inhibits Prospero by promoting its cortical localization [47]. Those two examples focus on mutual repression between cell cycle regulators and transcription factors. With this work, the mutual repression between GLD-1 and CYE-1/CDK-2 extends this important theme to translational regulators.

CYE-1/CDK-2 is a critical regulator of germline stem cell maintenance
A stem cell pool resides at the distal end of the adult germline [8,63]. The FBF-1 and FBF-2 proteins are well-established and largely redundant regulators of germline self-renewal: fbf-1 and fbf-2 single mutants maintain germline stem cells, but fbf-1 fbf-2 double mutants do not -instead, their germline stem cells enter meiotic prophase and differentiate [15,30]. We have found that CYE-1/CDK-2 depletion in fbf-1 single mutants similarly causes germline stem cells to enter meiotic prophase and differentiate. Therefore, CYE-1/CDK-2 emerges as a critical regulator of germline self-renewal.
The cye-1/cdk-2 effect on germline self-renewal occurs in fbf-1, but not in fbf-2 mutants. Why might this be true? One interpretation is that FBF-1 and CYE-1/CDK-2 function in parallel pathways, while FBF-2 and CYE-1/CDK-2 function together in the same pathway. We cannot exclude this idea, but suggest that an alternate explanation, which is based on FBFspecific effects of FBF cross-regulation, may be simpler. In fbf-1 and fbf-2 single mutants, the remaining FBF protein increases in abundance, but effects on distribution are different. The increased FBF-1 expands proximally in fbf-2 mutants with the result that more germ cells than normal possess FBF; by contrast, the increased FBF-2 remains limited to a small region at the distalmost end in fbf-1 mutants with the result that fewer germ cells than normal have FBF [30]. We suspect that this difference may result from a differential stability of FBF-1 and FBF-2 proteins, although this idea has not been tested. Regardless of the underlying mechanism, the spatially expanded FBF-1 may compensate for the loss of FBF-2 better than the spatially restricted FBF-2 can compensate for loss of FBF-1. This spatial difference could explain why GLD-1 expands distally in fbf-1 but not fbf-2 mutants [15] and why cye-1(aRNAi) has a synergistic effect with fbf-1 but not fbf-2 mutants (this work). Based on this reasoning, Figure 7 depicts FBF-1 and FBF-2 in one pathway and CYE-1/CDK-2 in a parallel pathway. However, we cannot exclude regulatory interactions between the FBF and CDK pathways.
Both FBF-1 and CYE-1/CDK-2 down-regulate GLD-1 in the germline stem cell pool (Figure 7) ( [15,27] and this work). FBF-1 keeps GLD-1 low by its direct post-transcriptional repression of gld-1 mRNA, while CYE-1/CDK-2 keeps GLD-1 low posttranslationally. One simple idea is that GLD-1 must be increased to a critical threshold to switch from the mitotic cell cycle into meiotic prophase. Consistent with that idea, reduction of gld-1 by one copy prevents entry into meiotic prophase in the distal germ cells of either an fbf-1 fbf-2 double mutant or an fbf-1; cye-1(aRNAi) animal ( [15] and this work). It is important to emphasize that other factors (e.g. GLD-2/GLD-3) also affect the decision [19,20]. Moreover, FBF-1 is a broad-spectrum regulator of .1000 mRNAs [17], with multiple targets that promote differentiation (e.g ERK/ MAP kinase [16]). We suspect that CYE-1/CDK-2 may similarly have many substrates. Therefore the regulatory network depicted in Figure 7 is highly simplified: it highlights the two doublenegative feedback loops that converge on GLD-1 but does not illustrate the complete network.
A central role for cyclin E and CDK2 in stem cell maintenance may be conserved. In human embryonic stem cells (hESCs), CDK2 promotes the G1/S transition and prevents differentiation, perhaps by maintaining a short G1 that is not subject to differentiation signals [49]. In addition, Drosophila CycE maintains neuronal stem cells by inhibiting the Prospero transcription factor, either directly or indirectly [47]. Our results support and extend this idea in several ways. Most importantly, our analysis places the C. elegans cyclin E and CDK2 homologs into the developmental pathway controlling self-renewal or differentiation. We identify a double-negative feedback loop between CYE-1/CDK-2 and GLD-1 that operates in parallel to a second double-negative feedback loop between Notch signaling and GLD-1. Although details will surely be different among the various stem cell systems, this regulatory logic provides a paradigm for the control of stem cell identity.
RNA interference (RNAi) was performed essentially as described [76]. For adult RNAi, synchronized L4s were placed on NGM plates seeded with dsRNA-expressing or empty vector control bacteria, and incubated at 20uC.
trGLD-1(wt): pMM016 DNA (gift from T. Schedl) harbors (1) a wild-type genomic gld-1 fragment that includes 1.1 kb 59 and 1.5 kp 39 flanking sequences and that was altered to include GFPand FLAG-encoding sequences fused to the GLD-1 C-terminus and (2) unc-119 to select for transformants. Microparticle bombardment was done using established protocols ( [79] and T. Schedl lab protocol), with modifications. Briefly, 4 mg of linear plasmids were mixed with 40 ml of 0.1 M spermidine-treated 1 mm gold-beads (Bio-Rad) and 100 ml of 2.5 M CaCl 2 ; particles were transferred to the macrocarrier after ethanol washes and then bombarded as normal. The bombarded worms were collected with M9 buffer after 1 hr incubation at room temperature, placed on rich media plates, and grown at 20uC. Bombardment of pMM016 generated five independent transgenic lines, qIs168 -qIs172. All five had similar effects on GLD-1 abundance and distribution; JK4626 (qIs170) was selected for further analyses, including confirmation of gld-1(q485) rescue.

Measuring mitotic zone size and mitotic index
Mitotic zone size was scored in DAPI-stained dissected gonads. We first identified the boundary between mitotic and transition zones, as described [11], and then counted germ cell diameters from the distal end to the boundary, or total number of germ cells within the mitotic zone. We measured mitotic index in extruded gonads stained with DAPI and antibodies to phosphohistone H3 (PH3), dividing the number of PH3-positive germ cells by the total number of DAPI-stained nuclei within the relevant region [11].
Immunostaining and quantification of GLD-1 Antibody staining was performed essentially as described [57], except we fixed tissues in 3% paraformaldehyde for 30 min. We used affinity purified antibodies against CYE-1 [43], REC-8 [80], HIM-3 [28], phosphohistone H3 (PH3; Cell Signaling), and GLD-1 [21], captured images on confocal (Carl Zeiss LSM 510) and compound (Carl Zeiss Axio Imager D1) microscopes, and processed them using ImageJ and Adobe Photoshop. To quantitate GLD-1, samples to be compared were processed in parallel, and confocal images were captured using the same settings. Raw images were opened and stacked in ImageJ and quantitated by measuring pixel intensity in a 5.1265.12 mm square encompassing germ cells in the relevant location. Data were copied into Excel, averaged, and graphed.
In vivo GLD-1 phosphorylation Synchronized wild-type or gld-3 nos-3 animals were harvested at 24 hrs past the L4 stage and then washed 36 with M9 buffer and 16 with 50 mM Tris-Cl, 5 mM MgCl 2 . Worm extracts were prepared with freeze-and-thaw cycles in liquid nitrogen; extracts were treated with Lambda Protein Phosphatase (l PPtase, NEB), with or without phosphatase inhibitor cocktail (Thermo Scientific) at 37uC for 30 min. Phosphorylated and dephosphorylated GLD-1 were separated on 7.5% SDS-PAGE gels and visualized by Western blot using 1:1000 affinity-purified anti-GLD-1 (Rabbit) and 1:10,000 HRP-conjugated anti-rabbit antibodies.

Preparation of recombinant GLD-1
A full-length wild-type gld-1 cDNA was cloned into pET-21b using Sal I and Xho I restriction sites to make pJK1341. To create pJK1347 (GLD-1 S22A, S39A, T348A), we changed the three consensus CDK phosphorylation sites in GLD-1 by site-directed mutagenesis (Stratagene). Primer sequences are available upon request. Recombinant proteins were expressed in BL21(DE3) (Novagen) cells by adding 0.1 mM IPTG and incubating 4 hrs at 30uC. Proteins were purified using Ni-NTA columns (Novagen) under native conditions using the manufacturer's protocol.

Supporting Information
(A) Distal end of dissected gonads. Left panels, stacked confocal images of GFP fluorescence to show abundance of transgenic GLD-1; right panels, DAPI staining to show distal end and mitosis/meiosis boundary. Solid triangle marks distal end; small v and larger V mark steps of comparable GLD-1 abundance; dashed line indicates boundary between mitotic and transition zones. Germlines were treated identically and confocal fluorescent images taken at the same settings and then stacked through Z-axis. Scale bar, 20 mm. (B) Quantitation of GFP intensity in the distal-most germline. Pixel intensity was measured using ImageJ and plotted using Microsoft Excel. Error bars indicate 95% confidence limits; asterisk denotes a statistically significant difference (p,1.06E 25 using Student's t-test). (C) Mitotic zone length is reduced in gld-1(0); trGLD-1(AAA) compared to gld-1(0); trGLD-1(wt). Error bars indicate 95% confidence limits; asterisk denotes a statistically significant difference (p,3.856E 211 using Student's t-test). Error bars indicate 95% confidence limits; asterisk denotes a statistically significant difference (p,0.0003 using Student's t-test). (C) Mitotic zone length, measured in gcd from the distal end, is reduced after cye-1 aRNAi. Error bars indicate 95% confidence limits; asterisk denotes a statistically significant difference (p,2.196E 28 using Student's t-test).