Cell-cycle regulation^*

The C. elegans genome encodes multiple members of the cyclin-dependent kinase (CDK) family. At least two CDKs, CDK-1 and CDK-4, are essential for cell-cycle progression (Boxem et al., 1999; Boxem and van den Heuvel, 2001; Park and Krause, 1999). These CDKs act at distinct times in the cell cycle and use specific cyclin partners, similar to their mammalian orthologs (Table 1).

CDK-1, previously known as NCC-1 for nematode cell cycle, was identified based on its close similarity to the prototypical yeast CDK (Mori et al., 1994). In contrast to yeast, but similar to mammalian Cdk1, cdk-1/(ncc-1) is specifically required for G₂/M progression and not for G₁ or S phase (Boxem et al., 1999). Maternal cdk-1 product suffices for embryogenesis, and candidate null mutant animals arrest cell division during L1 development. Several observations indicate that the post-embryonic precursor cells in these mutants arrest in G₂ phase: such cells show normal expression of the rnr::GFP reporter and BrdU incorporation during S phase, but fail to proceed into mitosis (as indicated by absence of chromosome condensation and nuclear envelope breakdown). Moreover, endoreduplication cycles, which skip M phase, continue in cdk-1 mutants and intestinal nuclei accumulate the normal 32n DNA content. Following RNAi of cdk-1 in adult hermaphrodites, oocytes show delayed meiotic maturation, form an eggshell upon fertilization, but neither align nor segregate homologous chromosomes. Thus, cdk-1 is required for meiotic as well as mitotic M phase.

CDK-1 likely acts with mitotic cyclins of the A and B subfamilies (Kreutzer et al., 1995). A single full-length cyclin A gene (cya-1) is expressed in C. elegans, as well as three typical B-type cyclins (cyb-1, cyb-2.1 and cyb-2.2), and a distinct member of the cyclin B3 subfamily (cyb-3). While cyb-1 and cyb-3 each are individually required during embryonic development, simultaneous inactivation of these mitotic cyclins causes a more severe phenotype: cyb-1;cyb-3(RNAi) embryos arrest at the one-cell stage and resemble cdk-1(RNAi) embryos (our unpublished results). These data support the notion that different mitotic cyclins have functions that are partly distinct and partly overlapping.

The cdk-4 Cdk4/6 kinase and cyd-1 D-type cyclin genes are required for progression through G₁ phase during larval development (Boxem and van den Heuvel, 2001; Park and Krause, 1999). CYD-1 and CDK-4 likely act in complex, as indicated by their direct interaction in vitro and close similarity in null phenotypes (Park and Krause, 1999). In contrast to larval divisions, only a few very late embryonic divisions depend on cyd-1/cdk-4 activity (Boxem and van den Heuvel, 2001; Yanowitz and Fire, 2005). Possible explanations for this difference include that the early embryonic divisions lack a G₁ phase and therefore will not need a G₁ cyclin or CDK. Also, as one of its most important functions, cyd-1 and cdk-4 act to antagonize the transcriptional repressor lin-35 Rb (see below; (Boxem and van den Heuvel, 2001). In the absence of cyd-1/cdk-4 function, lin-35 Rb may inappropriately repress cell-cycle genes, but this cannot prevent divisions that are driven by maternal products. Also, cyd-1 and cdk-4 could primarily promote growth, as in Drosophila (Datar et al., 2000; Meyer et al., 2000), which is not incorporated in the embryonic divisions. However, larval divisions arrest in G₁ while cells continue to grow in the mutants, and growth retardation occurs later (Boxem and van den Heuvel, 2001). Therefore, absence of G₁ phases and maternal contribution of DNA replication components likely explain the limited requirement for cyd-1/cdk-4 during embryogenesis

Taken together, CDK-1 and CDK-4 act in G₂/M and G₁, respectively, like their mammalian orthologs Cdk-1 and Cdk4/6. It is currently not clear whether C. elegans also uses a Cdk2 ortholog, which acts subsequent to Cdk4/6 in mammals to promote G₁/S and S phase progression. The best candidate is K03E5.3, which shares 43% amino-acid identity with human Cdk2 (Boxem et al., 1999). Inhibition of this gene by RNAi resulted in a variable phenotype, with animals arresting during embryogenesis, during early or late larval development, and as sterile adults.

In other metazoans, Cdk2 acts with Cyclin E to promote S phase entry. C. elegans cye-1 Cyclin E deletion animals show surprisingly normal development until the L3 stage, at which time the VPC divisions proceed slowly and incompletely (Fay and Han, 2000). This modest phenotype apparently depends on long lasting maternal function, as RNAi results in embryonic lethality at approximately the hundred-cell stage (Brodigan et al., 2003; Fay and Han, 2000).

Several other members of the Cdk superfamily are present in C. elegans, including a Cdk7/Mo15 ortholog (Boxem et al., 1999; Liu and Kipreos, 2000). Cdk7 was identified as CDK-activating kinase (CAK) in mammalian cells, but also as part of the TFIIH transcription factor, responsible for phosphorylating the C-terminal domain (CTD) of RNA polymerase II (Fisher and Morgan, 1994; Shiekhattar et al., 1995). The combination of a temperature-sensitive mutation and RNAi of cdk-7 resulted in one-cell arrest similar to cdk-1(RNAi) embryos (Wallenfang and Seydoux, 2002). In addition, partial inactivation of cdk-7 interfered with transcription and phosphorylation of the RNA polymerase CTD. These data support dual activities of CDK-7 as both CDK-activating kinase (CAK) and CTD kinase in vivo.

Several other CDKs are likely to act independent of the cell cycle. CDK-5 is remarkably close to human Cdk5, sharing 74% identity at the amino-acid level, which has neuronal functions (Cruz and Tsai, 2004). Two other CDKs, CDK-8 and CDK-9, likely are involved specifically in regulating transcription (Liu and Kipreos, 2000; Shim et al., 2002).

Association with small inhibitory proteins is a universal mechanism of CDK regulation (Sherr and Roberts, 1999), though the CKIs (Cyclin-dependent Kinase Inhibitors) involved are highly divergent between yeasts and metazoans. Three different proteins, p21^Cip1, p27^Kip1 and p57^Kip2, form the "CDK inhibitory Protein/Kinase Inhibitor protein" (Cip/Kip) family in mammals. The C. elegans genome encodes two members of this family: CKI-1 and CKI-2 (Feng et al., 1999; Hong et al., 1998). Although both predicted proteins are similarly close in amino-acid sequence to p21^Cip1 and p27^Kip1, only CKI-1 appears to act generally in cell-cycle control (Boxem and van den Heuvel, 2001; Feng et al., 1999; Fukuyama et al., 2003; Hong et al., 1998).

Several results have shown that CKI-1 acts to promote cell-cycle arrest throughout development, analogous to p27^Kip1 in mammals and Dacapo in flies. Absence of cki-1 as a result of the mnDf100 deletion results in embryonic arrest with hyperplasia in multiple lineages, including the intestinal and hypodermal lineages, as well as increased apoptosis and defects in morphogenesis (Fukuyama et al., 2003). mnDf100 also eliminates cki-2 and other genes, but cki-1 genomic sequences partly suppress the phenotype. Moreover, similar defects were observed in a fraction of cki-1(RNAi) embryos. However, RNAi usually causes incomplete inactivation of cki-1 and predominantly gives rise to sterile adults with extra divisions in the postembryonic lineages and extra gonad arms (Shiva phenotype; Hong et al., 1998). Postembryonic precursor cells in such cki-1(RNAi) animals fail to arrest in G₁ and ectopically express the S phase marker rnr::gfp. Thus, cki-1 Kip1 function is rate limiting for S phase entry, particularly in cells that enter a prolonged quiescent state before re-entering the cell cycle.

Interestingly, loss of cki-1 Kip1 also affects aspects of cell-fate determination, as the extra distal tip cells (DTCs) in cki-1(RNAi) animals are derived from a different cell type and not from DTC duplication (Kostic et al., 2003).

The tumor suppressor pRb is a well-known transcriptional repressor of, among others, genes involved in S phase progression such as cyclin E. Proteins of the pRb family exert this role through association with transcription factors, primarily with E2F/DP heterodimers (together referred to as "E2F"; Stevaux and Dyson, 2002). This association prevents "activating E2Fs" from promoting transcription. In addition, to inhibit transcription, "repressive E2Fs" recruit pRb family members and associated chromatin remodeling complexes such as the Nucleosome Remodeling and Deacetylase (NuRD) complex.

Several members of the Rb/E2F pathway in C. elegans have been identified as class B synthetic Multivulva (synMuv) genes (see Vulval development). As such, lin-35 Rb, efl-1 E2F4/5, dpl-1 DP and several putative NuRD-complex components all act in a redundant pathway to antagonize Ras-mediated induction of the vulval cell fate. The fact that these genes have similar, rather than opposite, loss-of-function phenotypes provided strong in vivo evidence that E2F/DP can act as a transcriptional repressor, in concert with pRb and NuRD. The contribution of lin-35 Rb in vulval precursor cell (VPC) determination is non cell-autonomous (Myers and Greenwald, 2005), and cannot be explained by lack of cell-cycle control in these cells.

Homozygous lin-35 Rb mutants do not display a prominent increase in cell division, although a fraction of such animals form extra intestinal nuclei (Saito et al., 2004). However, the contribution of lin-35 Rb to negative regulation of G₁ progression became apparent in double mutant combinations. Inactivation of lin-35 substantially rescues the larval arrest of cell division in mutants that lack the positive G₁ regulators cyd-1 and cdk-4 (Boxem and van den Heuvel, 2001). In addition, lin-35 inactivation synergistically increases the number of extra cell divisions when combined with negative G₁ regulators, such as cki-1 Cip/Kip, cdc-14 Cdc14 and fzr-1 Cdh1/FZR (Boxem and van den Heuvel, 2001; Fay et al., 2002; Saito et al., 2004). Together, such results indicate that lin-35 acts redundantly to inhibit G₁ progression, likely downstream of cyd-1 and cdk-1 and parallel of cki-1, fzr-1 and cdc-14 (Figure 3).

Examination of additional double mutant combinations revealed that some of the other synMuv class B genes also contribute to G₁ control (Boxem and van den Heuvel, 2002; Fay et al., 2002). Specifically, efl-1 E2F negatively regulates cell-cycle entry, while dpl-1 DP appears to act both as a positive and negative regulator. In addition, a negative G₁ regulatory function was identified for lin-9 Mip130/TWIT, as well as lin-15B and lin-36, which encode novel proteins. Class A synMuv genes and class B genes that encode NURD components have not been observed to affect the cell cycle.

Candidate targets of Rb/E2F regulation in C. elegans include cye-1 Cyclin E and rnr-1, which encodes the ribonucleotide reductase large subunit. These genes have multiple E2F-binding sites within their promoter regions (Brodigan et al., 2003; Hong et al., 1998) and are regulated by E2F in other species. Several genetic observations are also consistent with lin-35 acting upstream of cye-1 to repress its transcription (Boxem and van den Heuvel, 2001; our unpublished observations).

In all eukaryotes studied, CDKs are regulated by phosphorylation and dephosphorylation of critical residues. Important targets of this regulation are the neighboring threonine and tyrosine in the ATP-binding loop (GXGTYG, corresponding to Thr14 and Tyr15 in fission yeast p34^cdc2). Phosphorylation of these residues by the related Wee1 and Myt1 kinases prevents activation of the CDK/cyclin complex. Members of the Cdc25 family of dual-specificity phosphatases counteract Wee1/Myt1 phosphorylation and control the appropriate timing of CDK activation (Figure 2).

Inhibitory phosphorylation allows for developmental regulation of CDK activity, as demonstrated for the String and Twine phosphatases in Drosophila. In C. elegans, two different Wee1/Myt1 kinases are expressed (wee-1.1 and wee-1.3), as are four phosphatases of the Cdc25 family (cdc-25.1 to cdc-25.4). Despite the multiplicity, each family contains at least one essential member: homozygous wee-1.3 null mutants are embryonic lethal and cdc-25.1 null mutants develop into sterile adults (Ashcroft and Golden, 2002; Lamitina and L'Hernault, 2002). In the latter mutants, maternal product likely masks the embryonic function of cdc-25.1, as RNAi studies have indicated roles for cdc-25.1 during female meiosis and embryonic divisions (Ashcroft et al., 1999; Clucas et al., 2002).

For both cdc-25.1 and wee-1.3, gain-of-function mutants have been identified with surprising lineage-specific defects. Two independent dominant mutations in cdc-25.1 induce extra divisions in the intestine (Clucas et al., 2002; Kostic and Roy, 2002). Gain-of-function alleles of spe-37/wee-1.3 Myt1 cause spermatogenesis defects, resulting from G₂/M arrest during male meiosis (Lamitina and L'Hernault, 2002). The molecular consequences of these mutations are not precisely known, but the phenotypes indicate either a different CDK-activity threshold for specific developmental decisions, or differential regulation of CDKs within certain lineages.

Which CDKs are regulated by inhibitory phosphorylation in C. elegans? The ATP-binding sites of CDK-1 and K03E5.3 each contain adjoining threonine and tyrosine residues, while CDK-4 contains alanine and tyrosine at the corresponding positions, similar to vertebrate Cdk4/Cdk6. It is likely that the dominant wee-1.3 mutations affect CDK-1, while cdc-25.1 gain of function may cause premature activation of a G₁ CDK/Cyclin complex.

In addition to inhibition, CDKs are also positively regulated by phosphorylation. Activation of the CDK/Cyclin complex requires phosphorylation of a specific threonine residue in the activation loop. Based on results in other species, CDK-7 was examined as the candidate positive regulator (Wallenfang and Seydoux, 2002). Indeed, combined inactivation of cdk-7 through a temperature-sensitive mutation and RNAi resulted in a one-cell arrest, similar to cdk-1(RNAi) embryos. These results indicate strongly that CDK-7 kinase activity is required for CDK-1 activation (see also: CDKs and cyclins).

Protein degradation through ubiquitin-mediated proteolysis plays an important role in cell-cycle regulation. In C. elegans, this was first established through the characterization of lin-19/cul-1 and lin-23, which were originally defined in screens for mutants with abnormal postembryonic cell lineages. Blast cells undergo excessive divisions in all post-embryonic lineages in lin-19/cul-1 and lin-23 loss-of-function mutants (Kipreos et al., 2000; Kipreos et al., 1996). Cells divide at the appropriate times in these mutants, but additional cycles continue when cells should terminally differentiate. Thus, cul-1 and lin-23 act to promote cell-cycle exit.

lin-19 is the founding member of the conserved cullin gene family and was renamed "cul-1" (Kipreos et al., 1996). lin-23 encodes an F-box/WD-repeat protein, most similar to MET30 in yeast, human β-TRCP and Slmb in Drosophila (Kipreos et al., 2000). Based on the similar loss-of-function phenotype and molecular identity, CUL-1 and LIN-23 probably encode components of a Skp1-Cul1-F box (SCF) protein complex that acts as an E3 ubiquitin ligase and targets proteins for degradation (see Ubiquitin-mediated pathways in C. elegans ). This complex likely promotes cell-cycle arrest during larval development by promoting destruction of positive G₁/S regulators. CYE-1 Cyclin E is a candidate critical target of CUL-1/LIN-23 SCF, analogous to the role of vertebrate Cul1 SCF.

The cullin family is comprised of six different genes in C. elegans and cell-cycle related functions have also been reported for three additional members. Deletion of cul-2 results in G₁ arrest of germ-precursor cells, while cul-2(RNAi) embryos show defects in chromosome condensation and segregation (Feng et al., 1999) and are defective in progression through meiotic anaphase II (Liu et al., 2004; Sonneville and Gonczy, 2004). CUL-3 is specifically required for degradation of MEI-1, a subunit of a katanin complex that severs microtubules and is essential for meiotic chromosome segregation (Furukawa et al., 2003; Pintard et al., 2003; van den Heuvel, 2004; Xu et al., 2003). Finally, cul-4 is essential to prevent re-replication of DNA, likely by promoting degradation of the replication licencing factor CDT-1 (Zhong et al., 2003). These results demonstrate the importance of various cullin-based ubiquitin ligases in cell-cycle progression.

The anaphase-promoting complex (APC) is another multi-subunit E3 ubiquitin ligase that regulates cell-cycle transitions. As in other eukaryotes, C. elegans APC promotes sister chromosome separation in M phase. emb-30 is required for the metaphase-to-anaphase transition in meiosis and mitosis and encodes the APC subunit APC4 (Furuta et al., 2000). Moreover, a screen for temperature-sensitive embryonic lethal mutations revealed a large number of metaphase-to-anaphase transition (mat) mutants (Golden et al., 2000). These mat mutants arrest in metaphase of meiosis I and helped define five different genes that all encode components of the APC (Table 1).

In other species, the adaptor Cdc20/Fizzy recruits the APC to the substrate securin. Degradation of securin releases separase, which cleaves cohesins and triggers sister chromosome separation. Similar mechanisms are likely used in C. elegans, involving FZY-1 as the substrate specificity factor homologous to Cdc20/Fizzy, the interactor of FZY-1, as candidate securin and SEP-1 as the C. elegans separase (Kitagawa et al., 2002). The APC uses another substrate specificity factor, known as Cdh1p in yeast and Fizzy-related in Drosophila, in the destruction of mitotic cyclins. Loss of function of fzr-1, the C. elegans Cdh1/FZR ortholog, promotes hyperproliferation when combined with lin-35 Rb loss of function (Fay et al., 2002). This observation may indicate that FZR-1 also contributes to G₁ cyclin degradation, or, alternatively, it may counteract CKI-1 Kip1 degradation. The latter mechanism has been observed in human cells, in which Cdh1/FZR promotes accumulation of p27^Kip1 through degradation of the F box factor that targets p27^Kip1 for destruction (Bashir et al., 2004; Wei et al., 2004).

As metazoans go through development, their cells progress through various types of cell cycles. These include the embryonic cell cycle, somatic cell cycle, endoreduplication cycle, and meiotic cell cycle. The cell-cycle machinery used in each case is tailored towards the individual cycle and shows different requirements for critical regulators. For instance, C. elegans cyd-1 cyclin D and cdk-4 Cdk4/6 are not required for most of embryogenesis and cdk-1 is not needed for the endoreduplication cycles (Boxem et al., 1999; Boxem and van den Heuvel, 2001; Park and Krause, 1999). Switching from one division cycle to another involves important developmental decisions that remain poorly understood.

Checkpoint controls prevent progression through cell-cycle transitions prior to the completion of critical earlier events. For instance, the cell cycle is arrested and cell death may be triggered in response to DNA damage, progress into mitosis is halted when DNA replication is ongoing, and sister-chromatid separation is delayed until all kinetochores are attached to the spindle. The DNA damage checkpoint, replication checkpoint and spindle-assembly checkpoint, respectively, impose these "brakes" on the cell cycle and couple the initiation of later events to completion of earlier events. Recent studies in C. elegans have demonstrated conservation of these checkpoint pathways, identified novel components and revealed developmental functions.

At present, there is only limited understanding of developmental regulation of cell cycle entry and arrest in C. elegans. Several environmental conditions and mutations cause a global interruption of the reproducible pattern of cell division. For instance, when eggs hatch in the absence of food, larvae arrest early in L1 and do not initiate development and cell division until food is restored. Similarly, entry into and exit from the dauer state coincide with arrest and re-initiation of cell division. In addition, heterochronic mutations affect the timing of cell division simultaneously in multiple postembryonic lineages.

It is unclear how such developmental programs control the cell cycle and whether systemic signals are involved. Most likely however, regulation occurs via the cyclin/Cdk complexes. Inactivation of cyd-1 and cdk-4 leads to G₁ arrest (see above, CDKs and cyclins; Figure 3), and heat-shock induced expression of cyd-1 plus cdk-4 is sufficient to induce expression of the S phase marker rnr::GFP in L1 animals arrested through starvation (Park and Krause, 1999). Conversely, ectopic expression of the CDK inhibitor CKI-1 induces cell-cycle arrest and prevents rnr::gfp expression, while cki-1 loss-of-function results in extra cell division (see: CDK inhibitory proteins). RNAi of cki-1 induces rnr::gfp expression and limited cell division even in starved L1 animals and arrested dauer larvae (Hong et al., 1998). Mutations in heterochronic genes that cause precocious or retarded patterns of cell division affect the time cells spend in G₀/G₁, probably also acting upstream of the basic cell cycle machinery (Euling and Ambros, 1996b). Thus, both the normal developmental timing of cell division and environmental, possibly systemic, regulation is likely accomplished through regulators acting upstream of G₁ cyclin/CDK complexes.

What then controls the proper timing of cell-cycle entry and arrest? Promoter studies have shown that cyd-1, cdk-4 and cye-1 all are transcriptionally induced in dividing cells (Brodigan et al., 2003; Park and Krause, 1999). However, homozygous cye-1 mutants show normal temporal control of cell division during embryonic and early larval development, illustrating that zygotic cye-1 transcription does not drive lineage-specific timing of cell-cycle entry (Brodigan et al., 2003; Fay and Han, 2000). In contrast to cyd-1 and cye-1, the cki-1 promoter region is large and complex, and includes separate control elements for transcription in different lineages (Hong et al., 1998). Thus, transcription of cki-1 likely contributes to timing of cell cycle arrest.

Another level of CKI-1 regulation may be mediated by phosphorylation. Just like CKI-1, the CDC-14 phosphatase is needed to maintain the vulval precursor cells (VPCs) and other postembryonic precursor cells in a resting state (Saito et al., 2004). In genetic studies, cdc-14 was found to act positively in a cki-1 pathway. Moreover, a CKI-1::GFP fusion protein, expressed from a heterologous promoter, failed to accumulate in the absence of cdc-14 function. Based on analogy with other systems, CKI-1 turn-over is likely promoted by phosphorylation and ubiquitin-dependent proteolysis. CDC-14 phosphatase activity could contribute to cell-cycle arrest by counteracting CKI-1 phosphorylation, thereby preventing CKI-1 degradation.

Most cells exit the cell cycle and differentiate appropriately in cki-1 mutants, and even in animals that lack both lin-35 and cki-1 function. Cell-cycle exit is delayed in lin-23 and cul-1 mutants, but the smaller extra cells still differentiate (Kipreos et al., 1996; Kipreos and Pagano, 2000). Thus, cells can transit from G₁ to G₀ even when these negative regulators of G₁ progression are missing, possibly because redundant mechanisms are in place.

Differentiation is often surprisingly independent from cell division. Even in the absence of larval divisions, cyd-1 mutants and cdk-1 mutant animals continue molting cycles and form adult alae. These observations confirm earlier conclusions that aspects of development and differentiation can continue in the absence of cell division (Albertson et al., 1978). Vulval cells in cye-1 mutants undergo normal differentiation and morphogenesis of the vulva, despite delayed and fewer division cycles (Fay and Han, 2000). Extra divisions of the VPCs, in cki-1(RNAi) and cdc-14(0) larvae, or extra vulval cells, in cul-1 and lin-23 mutants, also does not prevent acquisition of a normal vulval fate or vulval morphogenesis (Hong et al., 1998; Kipreos et al., 2000; Kipreos et al., 1996; Saito et al., 2004).

Despite this independence, receptiveness and response to cell-cell signaling does include a cell-cycle component. The number of VPCs adopting a vulval fate is somewhat increased in cye-1 mutants, possibly because prolonged time in G₁ provides greater opportunity for induction (Fay and Han, 2000). Different VPC fates may use a different "window of opportunity" for lin-12-mediated lateral inhibition, with the 1° versus 2° decision requiring signaling earlier than the 2° versus 3° decision (Ambros, 1999). In precocious lin-28 mutants, VPCs adopt vulval fates and initiate fate-specific division patterns in L2. Dauer induction interrupts this pattern and arrests the VPC daughter cells (Euling and Ambros, 1996a). Interestingly, these arrested cells lose their lineage commitment and respond as multipotent VPCs to vulval induction during post-dauer development. In all three examples, the time spent in G₁ affects the VPC fate in response to signals.

Developmental decisions and cell-cycle regulation interconnect at many levels and much remains to be discovered. Germline proliferation and its control describes proliferation in the germline, which involves some of the best characterized connections between developmental signals and cell-division control.