Developmental regulation of a cyclin-dependent kinase inhibitor controls postembryonic cell cycle progression in Caenorhabditis elegans

The proper development of a multicellular organism requires the precise orchestration of cell proliferation and differentiation. Despite considerable progress toward understanding the mechanisms of cell cycle progression in single cells, less is known about the coordination of the cell cycle with development in a multicellular context. Developmental control of cell cycle is likely implemented by signals acting on limiting components of the cell cycle machinery (Edgar and Lehner, 1996). For example, in Drosophila, the timing and number of embryonic cell cycles is controlled by the developmental regulation of Cyclin E, String (cdc25), and E2F, which are limiting cell cycle activators (Duronio and O’Farrell, 1994, 1995; Knoblich et al., 1994; Duronio et al., 1995; Richardson et al., 1995; Sauer et al., 1995; Lehner and Lane, 1997), and by the activity of Dacapo, the Drosophila member of the CIP/KIP family of cyclin-dependent kinase inhibitors (CKIs) (de Nooij et al., 1996; Lane et al., 1996). Cells at the anterior side of the morphogenetic furrow in the fly eye imaginal disc are synchronized in G₁ of their penultimate cell cycle by roughex and rca1, which regulate cyclin A (Thomas et al., 1994; Dong et al., 1997). In C. elegans, the number of cell divisions in diverse cell lineages is limited by the activity of cul-1, which likely acts by regulating the degradation of G₁ cyclins (Kipreos et al., 1996; Mathias et al., 1996).

CKIs block cell division by inhibiting cyclin-dependent kinase activity necessary for the G₁/S transition (Harper and Elledge, 1996) and hence these proteins are potential mediators of the developmental control of G₁ progression. Two groups of CKIs have been described (Sherr and Roberts, 1995). These include the Ink4 family, members of which specifically inhibit Cyclin D-Cdk4/6 activity, and the CIP/KIP family which includes p21CIP1, p27KIP1 and p57KIP2 that exhibit promiscuous CDK-inhibitory activity. The CIP/KIP family of CKIs are highly expressed during terminal differentiation in mammalian cells (Parker et al., 1995; Macleod et al., 1995; Lee et al., 1995; Matsuoka et al., 1995), and cell culture experiments suggest a clear role for CKIs in promoting cell cycle exit for certain terminally differentiating cells (Casaccia-Bonnefil et al., 1997). In the Drosophila eye imaginal disc, Dacapo is expressed in terminally differentiating cells, and not in the cells anterior to the morphogenetic furrow that are temporarily held in G₁, suggesting a major role for Dacapo at the termini of cell lineages, rather than in precursor cells. Although there is evidence implicating CKIs in the developmental regulation of G₁ in precursor cells, it is relatively circumstantial. Genetic disruption of p27KIP1 or p57KIP2 in mice leads to organ hyperplasia and endocrine tumors, apparently reflecting enhanced proliferation (Fero et al., 1996; Kiyokawa et al., 1996; Nakayama et al., 1996; Zhang et al., 1997). However, it was not possible to determine directly whether the extra cells in these mouse mutants resulted from shortened G₁ phases in their progenitor cell cycles.

The nematode C. elegans is an excellent model for the study of developmental cell cycle control, since this animal develops according to a nearly invariant cell lineage (Sulston and Horvitz, 1977) that is conveniently accessible to observation and experimentation. Many of the postembryonic cell cycles of C. elegans, particularly those of extended duration, are regulated at the level of the G₁/S transition (Hedgecock and White, 1985; Euling and Ambros, 1996). Genetic pathways have been identified that developmentally regulate postembryonic cell cycles in C. elegans. These pathways include the daf genes that control entry into and exit from the developmentally arrested dauer larva stage (Riddle et al., 1997), the gon genes that control cell cycle progression in somatic gonad lineages (Sun and Lambie, 1997) and the heterochronic genes, that control the length of G₁ in the vulva precursor cell (VPC) cell cycle (Ambros and Horvitz, 1985; Euling and Ambros, 1996). Although it is not known how the heterochronic pathway affects G₁ progression selectively in the VPCs, it likely regulates components of the general cell cycle machinery critical for G₁ progress.

In this study we show that C. elegans developmental regulators may control the timing of the G₁/S transition by temporally regulating the activity of cki-1. The cki-1 promoter confers a temporal and cell type-specific pattern of reporter gene expression that coincides with developmental cell cycle progression, consistent with a role for cki-1 in the developmental control of G₁/S in diverse cell types. Using misexpression and RNA-mediated interference (RNAi) to perturb cki-1 activity we show that a major role for cki-1 is in the developmental regulation of G₁/S progression throughout postembryonic development. For the vulva cell lineage, we demonstrate that cki-1 may act as a downstream effector of heterochronic genes to control the timing of VPC cell division.

Nematode strains were grown and maintained as described by Wood et al. (1988). All animals were grown at 20°C unless otherwise indicated. The following strains were used: VT284 lin-14(ma135)/szT1, VT573 lin-4(e912);lin-14(n179ts), VT815 unc-36(e251);maIs109, VT825 dpy-20(e1282);maIs113, VT826 gon-2(q388ts);maIs109, VT827 maEx154, VT828 maEx155, VT829 maIs103;maEx156, VT774 unc-36(e251);maIs103, VT843 lin-4(e912);maIs113, VT844 lin-14(ma135);maIs113, VT845 lin-29(n1440);maIs113, EJ186 gon-2(q388ts), CB1282 dpy-20(e1282ts), NH2466 dpy-20(e1282ts); ayIs4, CB1372 daf-7(e1372ts), CB1370 daf-2(e1370ts).

Worms were transformed by microinjection as previously described (Epstein and Shakes, 1995). maEx154 was generated by co-injecting hsp16-41::cki-1 (pVT365; 20 μg/ml), hsp16-41::cki-1::GFP (pVT365G; 20 μg/ml), and rol-6(su1006) (pRF4; 100 μg/ml). maEx155 was generated by co-injecting egl-17::CKI-1::GFP (pVT363G; 20 μg/ml) and rol-6(su1006) (pRF4; 125 μg/ml). maEx156 was generated by co-injecting egl-17::CKI-1 (pVT363; 60 μg/ml) and rol-6(su1006) (pRF4; 120 μg/ml). maIs109 was generated by co-injecting cki-1(P2kb)::GFP (pVT353; 20 μg/ml) and unc-36 (Rlp16; 50 μg/ml). maIs113 was generated by co-injecting cki-1::GFP (pVT371; 20 μg/ml) and dpy-20 (pMH86; 100 μg/ml). maIs103 was generated by co-injecting rnr::GFP (50 μg/ml), unc-36 (Rlp16; 50 μg/ml) and pBluescript (100 μg/ml). Extrachromosomal arrays were integrated by gamma irradiation from a ¹³⁷Cs source (Epstein and Shakes, 1995).

pVT371(cki-1::GFP) is a transcriptional fusion between 8 kb of cki-1 upstream sequence from cosmid T05A6 and the NLS::GFP sequence from pPD95.02, and includes 1.5 kb of 3’ sequence from cki-1. pVT353 (cki-1(P2kb)::GFP) is identical to pVT371 except that it has only 2 kb of cki-1 upstream sequence. pVT352 and pVT352G are identical to pVT353 except that the NLS::GFP sequence has been replaced by the coding sequence for CKI-1 or CKI-1::GFP, respectively. pVT363 (egl-17::CKI-1) and pVT363G (egl-17::CKI-1::GFP) were made by replacing the 2kb promoter sequence in pVT352 and pVT352G with 4.5 kb of the egl-17 promoter from NH354 (Burdine et al. 1998). This promoter fragment also contains the EGL-17 signal peptide coding sequence (MLKVLLTLMLS) which was fused in frame with CKI-1. pVT365 (hsp16-41::CKI-1) and pVT365G (hsp16-41::CKI-1::GFP) were made by inserting the CKI-1 (from pVT352) or CKI-1::GFP (from pVT352G) coding sequence with 1.5kb 3’ cki-1 sequence into pPD79.83.

The large subunit of the ribonucleotide reductase gene (rnr-1) was identified by the C. elegans sequencing consortium on chromosome I. 1.0 kb of promoter sequence upstream of the predicted translational start was amplified by PCR. This fragment, which contains putative E2F binding sites, was inserted into pPD95.67 to yield pVT501 (rnr::GFP).

The cki-1 cDNA was amplified from an oligo dT-primed C. elegans cDNA library (kindly provided by Robert Barstead). The XmaI fragment was inserted into pBluescript and the resulting construct was linearized with BamHI and HindIII to synthesize antisense and sense cki-1 RNA, respectively. 1 μg of the gel-purified template was used for in vitro transcription reactions and the RNA was then phenol/chloroform extracted, ethanol precipitated and annealed (Fire et al., 1998). Double stranded RNA was injected at concentrations of 1-2 mg/ml and injected animals were recovered, transferred daily to new plates, and the resulting F₁ progeny were monitored for phenotypic abnormalities.

Images of live animals anesthetized with 1 mM levamisole were captured with an Optronics DEI750 integrating color CCD video camera and a Scion CG-7 RGB video capture board on a Power Macintosh 8500AV. GFP fluorescence images were taken with exposures ranging from 1/8 to 2 seconds. The digital images were processed using Adobe Photoshop. Images used to display relative levels of GFP expression were all processed identically.

The C. elegans genome project identified two loci on chromosome II (cki-1 and cki-2) that are predicted to encode proteins of the p21/p27 class of CKI (de Nooij et al., 1996; Lane et al., 1996). cki-1 was chosen for this investigation due to its greater sequence similarity to p21/p27. The genomic structure of cki-1 predicted using GeneFinder (Epstein and Shakes, 1995) was confirmed by cDNA sequencing (Kipreos, personal communication; data not shown).

To test whether cki-1 encodes a cell cycle inhibitor, we examined the effects of ectopic expression of cki-1 on cell cycle progression. First, we used a heat shock promoter to express CKI-1 and CKI-1::GFP fusion proteins, resulting in embryonic and postembryonic cell cycle arrest (Fig. 1A; Table 1). Second, we employed the egl-17 promoter (Burdine et al., 1998) to express CKI-1 or CKI-1::GFP specifically in a single vulva precursor cell, P6.p. The egl-17 promoter is activated specifically in P6.p during the late G₁ of the cell cycle (V. Ambros, unpublished). Driven by the egl-17 promoter, the CKI-1::GFP fusion protein was expressed specifically in P6.p and blocked P6.p from dividing (Fig. 1B; Table 1). The egl-17::CKI-1 transgene was transformed into animals carrying an S phase-specific reporter (Duronio and O’Farrell, 1994) consisting of the worm ribonucleotide reductase (rnr) promoter fused to GFP. In these doubly transgenic animals, no rnr::GFP expression was detectable in the non-dividing P6.p, whereas P5.p and P7.p divided in the L3 and their daughter cells showed strong rnr::GFP expression (Fig. 1C; Table 1). We conclude that CKI-1 expression cell-autonomously blocks G₁/S.

There are no predicted coding sequences within 20 kb upstream of cki-1. To begin an analysis of cki-1 promoter activity, the CKI-1 protein coding sequence was excised from a clone containing 8 kb of cki-1 upstream and 1.5 kb of cki-1 3’ downstream sequence and replaced with GFP coding sequence. The resulting transcriptional fusion, pVT371, was transformed into worms and attached to a chromosome to generate maIs113 (see Materials and Methods). Unless otherwise noted, the term cki-1::GFP will be used in reference to maIs113.

Embryonic expression of cki-1::GFP begins at the comma stage in the pharyngeal primordium as pharyngeal muscles begin terminal differentiation (Wood et al., 1988). cki-1::GFP is highly expressed in most cells during late embryogenesis, when cells are either differentiating or undergoing cell cycle arrest prior to hatching. At hatching, and in L1 animals maintained in the absence of food, cki-1::GFP is detected in Q, M, Z1, Z4 and the V cells. The expression in these cells fades after feeding, when cell division resumes. Throughout postembryonic development, cki-1::GFP is expressed with a temporal and cell type-specific pattern that corresponds to developmental patterns of cell cycle progression (see below and Table 2). Expression is also strong in many postmitotic neurons and muscle cells. cki-1::GFP expression tends to be stronger in newly differentiated cells and then gradually decreases. cki-1::GFP is also expressed in dauer larvae (see below).

V cells express cki-1::GFP strongly (Fig. 4D) until they divide in the mid-L1, when fluorescence decreases significantly. Seam cells express at a moderately high level during resting phases between molts, and at relatively reduced levels during division. Expression increases substantially at the L4 stage coincident with seam cell terminal differentiation (Fig. 4H).

The two sex myoblasts (SMs) are born in the late L1, migrate to the lateral midline (Sulston and Horvitz, 1977) and, in the mid-L3, divide to produce sex muscles. cki-1::GFP expression is high during SM migration, reduced during SM division and high again as the sex muscles differentiate (Fig. 2A-D).

cki-1::GFP expression is undetected in P cells, but at the L1 molt, the progeny of the Pn.a neuroblasts express high levels of cki-1::GFP. cki-1::GFP expression in P3.p-P8.p (the VPCs) is described below.

cki-1::GFP expression in Z1 and Z4 (Fig. 4B) diminishes prior to cell division in the mid-L1. cki-1::GFP is strongly expressed in Z1.aa and Z4.pp, the distal tip cells (DTCs), beginning in the L2 (Fig. 4F), but is undetectable in the rest of the Z1/Z4 lineage until the late L3 and early L4, when cki-1::GFP expression in the somatic gonad increases dramatically at the onset of terminal differentiation (Fig. 3E).

After the L1 molt, intestine cells undergo cycles of endoreplication (Hedgecock and White, 1985). cki-1::GFP is expressed in the intestine after the L1 molt and throughout the larval stages.

To determine if cki-1::GFP expression correlates with developmental G₁ arrest, we compared the timing of expression of cki-1::GFP to that of the S phase reporter, rnr::GFP. The expression pattern of rnr::GFP is consistent with the previously reported timing of S phase in the lateral and ventral hypodermal lineages (Hedgecock and White, 1985; Euling and Ambros, 1996; R. Roy, unpublished data). For example, VPC G₁ extends from the mid-L1 until shortly after the L2 molt when the DNA content of VPCs doubles (Euling and Ambros, 1996) and rnr::GFP expression appears in VPCs (Fig. 3). cki-1::GFP expression is first detected in late L1 or early L2 VPCs, and peaks at the L2 molt. cki-1::GFP expression is almost undetectable in dividing vulva cells, while rnr::GFP expression is strong. After the final vulva cell division, cki-1::GFP fluorescence increases markedly and rnr::GFP fades (Fig. 3A-F).

A reciprocal pattern of cki-1::GFP and rnr::GFP expression also occurs in animals that are subject to conditional cell cycle arrest; in the absence of food, newly hatched L1 larvae do not begin larval development, and cell division is arrested in G₁ as evidenced by a lack of rnr::GFP expression. cki-1::GFP expression is very high in these arrested larvae, and then fades by about 8-12 hours postfeeding, as cells re-enter the cell cycle and activate rnr::GFP expression (Fig. 3G-J). Similarly, under crowded or starved conditions, L2 larvae differentiate into developmentally arrested dauer larvae (Riddle et al., 1997) in which rnr::GFP expression is reduced while cki-1::GFP expression is elevated (Table 2).

To test whether 5’ upstream sequences of cki-1 affect the developmental expression pattern of a cki-1::GFP reporter, we removed the distal 6 kb of 5’ upstream sequence from the cki-1::GFP construct to generate cki-1(P2kb)::GFP. maIS109, the integrated transgene containing cki-1(P2kb)::GFP, is expressed in some of the cells that express the 8 kb construct, and not in others (Table 2; Fig. 4). Specifically, maIS109 is not expressed in resting V cells, the VPCs, SM or in terminally differentiating seam cells. This result suggests that the distal 6 kb of cki-1 5’ upstream sequences contains enhancer elements required for cki-1 expression in these cells. Replacing the cki-1 3’ untranslated region (UTR) in the 8 kb cki-1::GFP construct with the unc-54 3’UTR did not alter the pattern of GFP expression (data not shown).

During normal gonadogenesis, Z1 and Z4 divide in the mid-L1, while in gon-2 animals Z1 and Z4 are severely delayed or totally blocked from dividing (Sun and Lambie, 1997). This early requirement for gon-2 activity by Z1/Z4, together with the temporal pattern of expression of cki-1::GFP in these cells, suggests that gon-2 may be required for down-regulating cki-1 in Z1 and Z4 in the L1. To test whether this decrease in cki-1::GFP expression depends on gon-2 activity, we examined the expression of cki-1::GFP in Z1/Z4 of a gon-2(q388ts) mutant. Approximately 50% (n=11) of undivided Z1/Z4 cells strongly express cki-1::GFP in gon-2(q388ts);cki-1::GFP animals as late as the L3 or L4 (Fig. 5A) suggesting that gon-2 activity is directly or indirectly involved in the down-regulation of cki-1 in Z1 and Z4.

lin-29 encodes a zinc-finger transcription factor which is required for the terminal differentiation of hypodermal seam cells at the L4 molt (Rougvie and Ambros, 1995). In lin-29(lf) animals, seam cells fail to withdraw from the cell cycle or to differentiate at the L4 molt, and instead they reiterate larval fates including cell divisions. In the wild type, cki-1::GFP is expressed strongly in terminally differentiated seam cells (Fig. 4H). In contrast, lin-29(n1440);cki-1::GFP animals display an intermediate level of cki-1::GFP expression similar to that of resting hypodermal cells between each molt (Fig. 5B). This result suggests that lin-29 is required to activate the high level of cki-1::GFP expression associated with seam cell terminal differentiation.

The heterochronic gene lin-14 controls the length of VPC G₁ (Euling and Ambros, 1996). To test whether cki-1::GFP expression in VPCs (Fig. 3A and Table 2) depends on lin-14 activity, lin-14(ma135); cki-1::GFP larvae (n=11) were examined for cki-1::GFP expression throughout the L1 and L2. With the exception of one VPC (P3.p) with weak cki-1::GFP expression in one animal, VPCs displayed no detectable cki-1::GFP expression (Fig. 5C). Expression in other cell types appeared essentially normal. Thus, removal of lin-14 activity reduces cki-1 expression to below detectable levels specifically in VPCs.

The normal timing of VPC cell division requires the down-regulation of lin-14 activity after the L1 by lin-4 (Lee et al., 1993). In a lin-4(lf) mutant, where lin-14 activity remains high throughout development, VPCs display complex defects, including delayed or blocked cell divisions, or irregular cell division patterns (Euling and Ambros, 1996). Nevertheless vulval morphogenesis never occurs in lin-4(lf), indicating that high lin-14 activity prevents vulval differentiation (Euling and Ambros, 1996). Consistent with blocks in both VPC cell cycle progression and vulval differentiation in lin-4(e912);cki-1::GFP animals, GFP was detected at high levels in VPCs exhibiting cell cycle delay, and was low or undetectable in the progeny of divided VPCs (Fig. 5D).

If cki-1 is critical for maintaining cells in G₁ during development, then reduced cki-1 activity should disrupt the normal pattern of cell division, particularly in those cell types where cki-1 temporal expression is coupled to development. To test this we used RNA-mediated interference (RNAi) (Rocheleau et al., 1997; Fire et al., 1998) to compromise cki-1 gene activity. The RNAi method involves injecting adult hermaphrodites with double-stranded RNA corresponding to exon sequences of a target gene. The progeny of injected animals display a phenotype corresponding to loss of function of the target gene (Fire et al., 1998). cki-1(RNAi) animals were visibly uncoordinated and showed morphological abnormalities accompanied by misregulated cell division in numerous postembryonic cell lineages. cki-1(RNAi) animals show extra lateral hypodermal cell divisions in the early larval stages, accompanied by alterations in the timing of the L1 molt (data not shown). The apparent absence of strong embryonic effects of cki-1(RNAi) was somewhat unexpected since cki-1::GFP is expressed in mid- to late embryos. Perhaps other gene(s) function redundantly with cki-1 during embryogenesis, or the RNAi method may not entirely eliminate cki-1 gene activity (Fire et al., 1998). In the analysis described here, we did not exhaustively characterize cki-1(RNAi) phenotypes in all cell types (such as postmitotic neurons and muscles), but rather we concentrated on cases of developmental G₁ progression.

VPCs were monitored from the L1 through adult stages in cki-1(RNAi) animals, and up to six extra ventral hypodermal cells were observed to arise in the L2 from precocious division of P3.p-P8.p (Fig. 6A). In cki-1(RNAi) animals, VPCs undergo only one round of precocious division. These precocious divisions produce an apparent expanded set of VPCs that remain quiescent until the L3 when vulval development ensues as usual. In some cases the subsequent vulval morphogenesis is essentially normal.

To confirm that the precocious VPC cell divisions observed in cki-1(RNAi) animals produce cells with full VPC potential, we performed cki-1(RNAi) in a lin-12(gf) background. lin-12 mediates lateral signaling among VPCs to control the pattern of vulval fates (Ferguson et al., 1987). In lin-12(gf) animals, the six VPCs execute secondary vulva fates independently of gonadal signals, resulting in six ventral protrusions, or pseudovulvae. If the precocious VPC daughters in cki-1(RNAi) animals possess full VPC potential, then they would each have the capacity to adopt a secondary fate and generate a pseudovulva in a lin-12(gf) background (Fig. 6B). Indeed, cki-1(RNAi);lin-12(gf) animals displayed as many as 12 pseudovulvae (Fig. 6C) indicating that the precocious VPC daughters in cki-1(RNAi) animals each have the potential to execute vulval fates.

RNAi was used to test whether cki-1 is required for cell cycle arrest in dauer larvae and starved L1 animals. rnr::GFP(maIs103) larvae were hatched in the absence of food and these starved L1s were examined for evidence of cell division by monitoring expression of rnr::GFP. Control rnr::GFP(maIs103) larvae (from uninjected parents) showed no evidence of rnr::GFP expression. cki-1(RNAi);rnr::GFP(maIs103) animals (from injected parents) showed strong rnr::GFP expression in hypodermal cells and the M cell, and some hypodermal cells were observed to divide (Fig. 7A). These results indicate that cki-1 is required to maintain cells in a cytostatic state at the end of embryogenesis until the hatched L1 begins to feed.

daf-7(e1372ts) animals constitutively form dauer larvae at nonpermissive temperature, and no cell divisions occur after developmental arrest. However, in daf-7(e1372ts); cki-1(RNAi) dauer larvae, cell division, accompanied by rnr::GFP expression, was observed in the hypodermal and gonadal cell lineages (Fig. 7D,E). In these cki-1(RNAi) animals, formation of the lateral alae characteristic of dauer larva cuticle was blocked or severely impaired (Fig. 7B,C). Similar effects on dauer cuticle formation were observed in daf-2(e1370ts); cki-1(RNAi) animals (cell division and rnr::GFP were not monitored in this case). These results suggest that cki-1 acts in the dauer larva to block cell division and permit dauer-specific hypodermal differentiation. Not all aspects of dauer morphology are sensitive to cki-1(RNAi), as the affected animals exhibit some characteristics of normal dauer larvae, including intestinal lipid droplets, SDS-resistance and radial shrinkage.

cki-1(RNAi) animals show a dramatic expansion of the gonad after the L3 compared to the wild type. The affected gonads often are morphologically abnormal and consist of one to six extra distal tip cells (DTCs) and a large excess of germ cells. DTCs signal to the germ cells via a lag-2/glp-1 signaling pathway to promote germline proliferation (Lambie and Kimble, 1991). The cki-1(RNAi)-induced germline hyperplasia is blocked in a glp-1 background (data not shown), suggesting that the extra germ cells result from ectopic lag-2/glp-1 signaling, likely from the extra DTCs.

Unlike the cki-1(RNAi)-induced extra VPCs, ectopic distal tip cells in cki-1(RNAi) animals do not necessarily arise from duplications of differentiated distal tip cells, as might have been expected from the expression of cki-1::GFP in the DTCs. When we removed all the DTCs from cki-1(RNAi) L2 larvae by laser ablation, new DTCs were observed by the L4. Thus, extra DTCs can arise from as yet unidentified somatic gonadal precursors -suggesting that cki-1 normally acts to restrict the potential of those other cells, perhaps by inhibiting cell division during a critical developmental period.

cki-1 shares sequence identity with mammalian and Drosophila members of the CIP/KIP family of cyclin-dependent kinase inhibitors. Using an S phase-specific reporter, rnr::GFP, we have shown that expression of CKI-1 in worms results in a cell-autonomous G₁ arrest, consistent with its role as a G₁/S regulator. Like other members of this family of CKIs, CKI-1 may function by inhibiting cyclin/cdk complexes responsible for the G₁/S transition (Sherr and Roberts, 1995).

cki-1::GFP is expressed strongly in differentiating cells, a pattern shared by mammalian p21CIP, p27KIP1 and p57KIP2. cki-1::GFP is also expressed dynamically in non-differentiating progenitor cells in the larva, suggesting a role for cki-1 in coupling cell cycle progression to developmental signals. DNA quantitation has been previously used to determine the timing of S phase in certain lineages (Hedgecock and White, 1985; Euling and Ambros, 1996), wherein cki-1::GFP expression occurs clearly during a predicted G₁ phase while rnr::GFP is expressed coincident with the predicted S phase. For other cells, cell cycle phases had not been previously determined, but the reciprocal pattern of cki-1::GFP and rnr::GFP expression suggests that G₁/S is also developmentally regulated in these cells. Overall, the postembryonic pattern of cki-1::GFP expression correlates with developmentally regulated cell cycle progression in diverse cell lineages. This result implicates cki-1 as a downstream effector of developmental pathways controlling postembryonic larval arrest in the absence of food, dauer larva arrest, and developmental timing during continuous development.

The continuous cki-1::GFP expression in intestinal cells was somewhat unexpected, since these cells undergo endoreduplicative cell cycles. A similar expression in certain endoreduplicating cells was reported for Drosophila Dacapo (de Nooij et al., 1996). This observation suggests that CKI-1 may play a role during endoreduplicating cell cycles, although that role is likely to be distinct from its activity in mitotic cell cycles.

Since no cki-1 mutations were available, we used RNA-mediated interference (RNAi) to analyze cki-1 function. The most striking cki-1(RNAi) phenotype that we observed was the inappropriate cell cycle progression by progenitor cells in the developing larva. This RNAi phenotype is consistent with the expression of cki-1::GFP in numerous postembryonic lineages at times when cells are temporarily or conditionally quiescent. For example, cki-1::GFP expression in starved L1s and dauer larvae is consistent with the abnormal cell cycle entry at these stages in cki-1(RNAi) animals; similarly, cki-1::GFP expression in lateral hypodermal cells between molts, and in VPCs, is consistent with the extra cell divisions observed in these lineages in cki-1(RNAi) larvae. These results suggest a major developmental role for cki-1 in holding progenitor cells in G₁ prior to later proliferation.

cki-1::GFP was strongly expressed in certain neurons, muscles and hypodermal cells at the terminus of cell lineages, and in dauer larva hypodermal cells, which produce a dauer-specific cuticular structure called lateral alae. We did not exhaustively monitor the effects of cki-1(RNAi) in all these differentiated cell types, but for specific cases, namely the differentiation of hypodermal cells in dauer larvae and at the L4 molt, cki-1(RNAi) had a clear effect. In cki-1(RNAi) animals, some hypodermal cells fail to form adult-specific alae (data not shown), and hypodermal cells in cki-1(RNAi) dauer larvae fail to withdraw from the cell cycle and differentiate.

The inability of cells to exit the cell cycle during differentiation likely contributes to the extensive hyperplasia in some organs in p27 or p57 deficient mice (Fero et al., 1996; Kiyokawa et al., 1996; Zhang et al., 1997), supernumerary rounds of epidermal cell division in Dacapo mutant embryos, and the inability of p27−/− oligodendrocytes to efficiently differentiate into dendrocytes (Cassaccia-Bonnefil et al., 1997).

We are confident that the cki-1(RNAi) phenotypes described here reflect cki-1 loss of function. RNAi is a convenient method for reducing the level of activity of a gene for which exon sequence is available (Rocheleau et at, 1997; Fire et al., 1998). The method appears to be specific to the targeted gene, likely reflecting a mechanism involving base-pairing. In cases where genomic mutations were available for test genes targeted by RNAi, the RNAi phenotype accurately reflected the known loss-of-function phenotype, with the possible exception of members of gene families that are highly conserved in DNA sequence (Fire et al., 1998). However, cki-1 and cki-2 are not well conserved at the DNA sequence level, and we observed that the cki-1(RNAi) phenotype was the same whether the injected RNA corresponded to the entire cki-1 coding sequence, or only a carboxy-terminal segment unrelated to cki-2 (data not shown). It should be noted that cki-1(RNAi) did not lead to uniformly expressed or completely penetrant defects. For example, some cells that expressed cki-1::GFP either were not affected, or the effects were weak or variable. We cannot say whether this incomplete expressivity and penetrance of cki-1(RNAi) reflects functional redundancy with other CKIs such as cki-2, epistatic effects of positive effectors of cell cycle, or simply an incomplete loss of cki-1 function. A critical test of the null phenotype of cki-1 will require isolation of the appropriate mutant alleles.

cki-1(RNAi) caused one round of precocious vulva precursor cell (VPC) divisions in the L2. The precocious VPC daughters in cki-1(RNAi) animals exhibited characteristics of VPCs with full competence for vulval development, suggesting that loss of cki-1 causes VPCs to enter the cell cycle, undergo one round of division, but retain VPC developmental potential. We did not observe more than one round of precocious VPC cell division, suggesting that cki-1 may act in parallel with other factors that control G₁ progression in VPCs. Cullins such as CUL-1 might play roles in VPC G₁ progression by regulating the degradation of G₁ cyclins (Mathias et al., 1996; King et al., 1996; Kipreos et al., 1996). However, although cul-1 mutants display supernumerary cell divisions in the vulval cell lineages (Kipreos et al., 1996), extra VPCs were not observed (E. Kipreos, personal communication), suggesting that cki-1 and cul-1 are not functioning equivalently in VPCs during the L2.

The precocious VPC divisions in cki-1(RNAi) animals are reminiscent of the precocious vulval development of lin-14(lf) or lin-28(lf) mutants, which also commences in the L2, but with a distinct difference: in lin-14(lf) or lin-28(lf) mutants, multiple rounds of precocious divisions produce precocious differentiated vulval cells. The fact that vulval fates are not expressed until the normal time in the L3 of cki-1(RNAi) animals indicates that the VPCs are still subject to the normal temporal control on VPC competence and vulval cell fate determination, presumably via lin-14 activity (Euling and Ambros, 1996).

Our analysis of the effects of cki-(RNAi) on VPCs clarifies the interpretation of defects observed in p27−/− mice. In the mouse mutants, organs such as the spleen and thymus were observed to be abnormally large without any obvious defects in differentiation or morphogenesis. One inference from those results was that loss of p27 caused over-proliferation of early progenitor cells in certain organs (Fero et al., 1996). The VPC duplication observed in cki-1(RNAi) animals validates this particular interpretation of the p27 phenotype, and suggests a general role for cki-1/p27 in the control of organ size by regulating G₁ progression in progenitor cells. Defects in progenitor cell G₁ progression were not observed in the postembryonic development of Dacapo mutants, consistent with the observation that Dacapo expression was limited to differentiating cells (de Nooij et al., 1996; Lane et al., 1996).

In larvae that hatch in the absence of food, cells are arrested in G₁, and elevated levels of cki-1::GFP expression persist until animals begin feeding and cells reenter the cell cycle. In starved cki-1(RNAi) L1 larvae, certain postembryonic blast cells escape the starvation-induced G₁ arrest and enter the cell cycle prematurely. Perhaps the transition between embryonic and larval development coincides with a reduction in maternally supplied growth factors, causing an increase in cki-1 expression and consequent cell cycle arrest. Development in the absence of food may be detrimental to the larva. It should be noted that cells in cki-1(RNAi) starved L1 larvae undergo only limited rounds of division, suggesting additional constraints, besides the CKI-1 level, on cell cycle progression in the absence of food.

Our observations implicating cki-1 in a larval cell division checkpoint are analogous to the stimulation of p27KIP levels observed after serum withdrawal in fibroblasts (Coats et al., 1996), and the reduced dependence on growth factors observed in p27−/− mammalian cells (Fero et al., 1996; Coats et al., 1996). These results suggest a general role for CKIs in mediating growth factor responses.

The developmental patterns of mRNA accumulation for Drosophila Dacapo and mouse p21 and p57 have suggested that CIP/KIP CKIs are transcriptionally controlled (Parker et al., 1995; de Nooij et al., 1996; Lane et al., 1996), but did not provide direct evidence of promoter-dependent expression. Our experiments demonstrate that the pattern of cki-1::GFP expression in C. elegans larvae is governed by 5’ upstream sequences, and hence suggest that cki-1 transcription is developmentally regulated.

The spatial pattern of expression of cki-1::GFP depended on the length of the cki-1 5’ upstream sequence included in the fusion gene. The 2 kb of upstream sequence contained in maIs109 was sufficient for expression in certain differentiated cells, while expression in specific lateral and ventral hypodermal progenitor cells required additional upstream sequences contained in maIs113. It should be noted that even maIs113, with 8 kb of upstream sequence, was not expressed in all resting cells, particularly the P cells and some somatic gonadal progenitors. If endogenous cki-1 is expressed in these cells, then additional sequence may be required for the full range of cki-1 developmental expression.

The temporal dynamics of cki-1 expression within a cell lineage may also depend on promoter sequences. For example, in the M lineage, both maIs113 and maIs109 were expressed in the M cell and its descendant vulval muscles, while only maIs113 was expressed in SM, the immediate precursor to the vulval muscles. Thus, the observed dynamics of cki-1::GFP expression in other lineages, such as the vulva, could also be transcriptionally programmed. Substituting the unc-54 3’UTR for the cki-1 3’UTR had no discernible effect on cki-1::GFP expression, so mRNA translation or stability does not contribute significantly to the observed temporal patterns of cki-1::GFP expression. However, these results with a GFP reporter do not necessarily reflect the endogenous levels of cki-1 mRNA, nor do they rule out the possibility that posttranscriptional regulation may be significant to the expression of endogenous CKI-1 protein. Furthermore, cki-1 protein sequences are not included in the maIs113 or maIs109 transgenes, so our results do not rule out the possibility that regulated protein degradation contributes to the cell cycle dynamics of endogenous CKI-1 expression, as is the case for p27KIP1 (Hengst and Reed, 1996; Sheaff et al., 1997; reviewed by King et al., 1996).

Although developmental control of the cell cycle is universal, few genetic pathways have been directly implicated in controlling the developmental signals that influence cell cycle progression. These pathways include the anachronism and roughex genes of Drosophila which regulate cell cycle entry in certain neuroblasts, and in the eye, respectively (Ebens et al., 1993; Thomas et al., 1994). In Drosophila embryogenesis, certain blastoderm cell cycles occur in spatial patterns owing to the regulation of string transcription by spatial patterning genes (Edgar et al., 1994).

In C. elegans, a genetic pathway that includes gon-2 (Sun and Lambie, 1997) mediates feeding-dependent postembryonic cell cycling in somatic gonadal lineages. gon-2 encodes a TRP channel protein that may mediate the transduction of signals for gonadal growth (R. Biron, Y. Sun, D. Church and E. Lambie, personal communication), resulting directly or indirectly in cki-1 down-regulation (Fig. 8A). cki-1 appears to be activated in dauer larvae to affect cell cycle arrest, as one consequence of a complex signaling cascade that includes daf-7 (TGF-β; Ren et al., 1996), daf-2 (insulin-related receptor; Kimura et al., 1997), and daf-12 (nuclear hormone receptor; Riddle et al., 1997). Our experiments do not distinguish whether cki-1 is regulated directly or indirectly by specific components of the dauer regulatory pathway (Fig. 8B). Interestingly, the possible role of cki-1 as a downstream effector of the TGF-β (daf-7) signaling is analogous to the role of p27KIP as a mediator of TGF-β-induced epithelial cell division arrest (Polyak et al., 1994). However, daf-7 does not activate, but rather antagonizes, dauer arrest suggesting that the effect of TGF-β/daf-7 on downstream CKIs may be cell-type specific (Ravitz et al., 1996).

A pathway of heterochronic genes, including lin-14 and lin-28 and their repressor, lin-4 (Lee et al., 1993; Moss et al., 1997), controls the timing of G₁ progression and developmental competence in VPCs (Euling and Ambros, 1996). Our observation that cki-1::GFP expression is reduced specifically in the VPCs of lin-14(0) animals strongly suggests that cki-1 is activated in VPCs by lin-14. lin-14 and lin-28 expression are mutually interdependent (Moss et al., 1997), so we cannot say whether cki-1 is a direct target of either (or both) of lin-14 or lin-28. Since the cki-1(RNAi) phenotype results in precocious VPC division, but is not identical to the lin-14(0) VPC defect, we conclude that lin-14 acts via cki-1 to keep VPCs in G₁ during the first half of the L2 stage, but acts independently of cki-1 to maintain VPC G₁ through the end of L2 (Fig. 8C). Further molecular and biochemical analysis are required to determine precisely how cki-1 is targeted for regulation by the heterochronic pathway in VPCs. lin-31 and lin-25 mutants also display precocious VPC divisions, although the lin-25 defect is of low penetrance (Ferguson et al., 1987), and in the case of lin-31, the precocious divisions can sometimes produce neuroblasts (Miller et al., 1993). Despite these differences between lin-25, lin-31 and cki-1(RNAi) VPC phenotypes, these observations suggest that lin-31 and lin-25 may function with lin-14 to potentiate cki-1 activity in VPCs.

We are grateful to Rebecca Burdine and Michael Stern for egl-17 plasmids and worm strains, the Sanger Center for providing cosmid T05A6, Andy Fire for plasmids, Eric Lambie for worm strains, and to the Stern, Kipreos and Rothman labs for sharing information prior to publication. We also thank Eric Moss, Philip Olsen, Edward Kipreos and Eric Lambie for discussion, and Rosalind Lee for assistance. R. R. was supported by a Human Frontiers Science Program fellowship. This work was supported by a U. S. Public Health Service research grant GM34208 (V. A.). Some nematode strains used in this work were provided by the Caenorhabditis Genetics Center, which is funded by NIH National Center for Research Resources (NCRR).