Cyclin C, a Cyclin-Related CLN Complementing Gene

The ability of the unique non-Dmcdc2 plasmid, CL24, to rescue the Cln-deficient strain was confirmed by retransformation of YPL1. The CL24 insert was subcloned in a Bluescript vector and sequenced. The 1086 bp sequence showed a single open reading frame of 267 amino acids (Figure 2). Comparison with the size of the RNA transcript detected on Northern blots (see below) suggests that about 150 bp of noncoding sequences are not represented in the cDNA. The absence of a substantial poly(A) sequence in the 3′ sequence of the cDNA suggests that priming of cDNA synthesis by oligo(dT) may have initiated at an internal A-rich sequence. While this implies loss of 3′ sequences, some 5′ sequences might also be missing.

Open in a separate window

Figure 2

Nucleotide Sequence and Predicted Amino Acid Sequence of a Cyclin C cDNA

The nucleotide sequence (1086 bp) has an open reading frame encoding a 267 amino acid protein. The putative initiating ATG is underlined along with flanking bases matching the consensus sequences for translational start sites in Drosophila (Cavener, 1987).

Computer searches of data bases with the CL24 aminoacid sequence detected homology with cyclins. This homology was explored in detail by manual comparisons. The many well conserved G2 cyclin sequences served as a guide to align more distantly related cyclin sequences. Alignment of 15 G2 cyclin sequences identified conserved positions, permissible gaps, and allowable substitutions at modestly conserved positions (Figure 3). The CL24 sequence, the CLN sequences, and a distant cyclin homolog (prad1; Motakura et al., 1991) were then aligned by inspection, emphasizing conserved residues and optimizing the matches to alternative residues at other positions. Three regions gave unambiguous and uninterrupted alignments for all of the divergent cyclin sequences (Figure 3). Homologies that are interrupted by frequent gaps or that did not extend to all of the divergent cyclin sequences are considered tenuous and are not discussed here.

Open in a separate window

Figure 3

Cyclin C Sequence Alignment with Conserved Cyclin Sequences

The bars in the upper panel are scaled representations of five divergent cyclins. The location of the three regions identified as having high conservation are indicated as shaded segments. The scale is indicated on the right. For each of the conserved regions, a consensus of G2 cyclin sequences is shown. This was derived as follows: 15 G2 cyclin sequences were compared (clam cyclins A [Swenson et al., 1986] and B [Westendorf et al., 1989], human cyclins A [Wang et al., 1990] and B [Pines and Hunter, 1989], Drosophila cyclins A [Lehner and O’Farrell, 1989] and B [Lehner and O’Farrell, 1990], frog cyclins A [Minshull et al., 1990], B1 [Minshull et al., 1989], and B2 [Minshull et al., 1989], sea urchin cyclin [Pines and Hunt, 1987], S. pombe cdc13 [Booher and Beach, 1988], and S. cerevisiae Clb1, Clb2, Clb3, and Clb4 [Ghiara et al., 1991]). At each position, the most commonly occurring amino acid was selected for the consensus. In the few cases of a numerical tie, we selected the residue that was the most widely distributed among the more evolutionarily diverged cyclins. Below the consensus, all of the different residues that occur at a position are listed as alternatives. The cyclin C sequence and the other diverged cyclin sequences were aligned with the G2 consensus sequence. The number to the left of each aligned sequence is the position of the first residue of the segment. As a reference and an example of a G2 cyclin, the clam cyclin A sequence is included in this alignment. Residues matching the G2 consensus are boxed and shaded. Residues considered homologous to the residues of the consensus are boxed, according to the following groupings: A, L, V, I, M; D, E; K, R; N, Q; F, Y; S, T. The scores on the right of each alignment were calculated as follows. The 15 G2 sequences were used as a reference. The number of times each amino acid occurred at a given position in this reference set was determined. For an aligned sequence, each position was given a score corresponding to the number of times the corresponding amino acid occurred in the reference set. The total scores for an alignment were then normalized by dividing by the maximum score (number of positions scored × number of sequences in the reference set). For the ras homology within region III, homology scores were similarly calculated based only on the 21 amino acid sequence indicated for human H-ras.

Pairwise sequence comparisons were made within the stringently aligned regions (Table 1). Two divergent B-type cyclins, human and S. pombe, are 52% identical, while the CL24 sequence has 23% identity with the human cyclin B sequence. The CL24 sequence did not show especially high homology to any of the cyclin subtypes, A, B, or Cln. In particular, CL24 and Cln sequences are both more similar to G2 cyclins than to each other (Table 1). Thus, CL24 represents a distant homolog of both the G2 cyclins and the CLN genes.

The above alignments define cyclin consensus sequences, as has been done for other multigene families, such as those encoding homeodomain proteins or zinc finger–containing proteins. Once such precise consensus sequences are defined, computer searches can identify even very distant homologs. Three largely independent cyclin consensus sequences identified the ras oncogene as a homolog (Figure 3 and Experimental Procedures). The match across a 21 amino acid ras region is excellent (it would occur randomly with a probability of about 10⁻⁹ for a single comparison) and extends to the diverged cyclin sequences (Cln1, Cln2, Cln3, prad1, and Drosophila cyclin C).

Homology between Human and Drosophila Cyclin C

Using a selection similar to ours, D. Lew and colleagues (see Lew et al., 1991, this issue) isolated human clones that rescued G1 cyclin–deficient S. cerevisiae. In addition to G2 cyclin genes, they isolated three rescuing sequences that have weak cyclin homology. Comparison of the CL24 sequence with these sequences uncovered homology far more extensive than that found in comparisons with any previously known cyclin. The CL24 sequence is 72% identical, with only a single gap of one residue, to one of their human sequences (Figure 4). The human sequence is somewhat longer, extending 20 residues beyond the C-terminus and 17 residues beyond the N-terminus of the Drosophila sequence. Given that CL24 is homologous to cyclins and does not belong to existing classes of cyclins, we will refer to this sequence as Drosophila cyclin C (see Lew et al., 1991, this issue).

Open in a separate window

Figure 4

Amino Acid Sequence Comparison of Drosophila and Human Cyclin C

Total Drosophila and human sequences are aligned. Shaded boxes correspond to identical residues. The dash indicates a 1 amino acid gap introduced into the human sequence.

The Drosophila Cyclin C Gene

Cyclin C hybridization probes were prepared and used for Southern blots of Drosophila genomic DNA and for in situ hybridization to polytene chromosomes (Figure 5). The results show that Drosophila cyclin C is encoded by a single copy gene (DmcycC) at 88E on the polytene map.

Open in a separate window

Figure 5

Characterization of the DmcycC Gene

(A) Total genomic DNA from Drosophila was digested with EcoRI (lane 1) and Pstl (lane 2) and probed on Southern blots with a DmcycC probe. The scale on the left is in kilobases.

(B) A DmcycC probe was used to localize the gene on polytene chromosomes from salivary glands of third instar larvae. The cyclin C cDNA clone was labeled with digoxigenin and hybridized to salivary gland squashes. The position of the phosphatase signal following immunological detection was identified in several squashes. It consistently mapped to 88E.

Northern blots of RNA from staged embryos revealed a single low abundance 1.25 kb transcript (Figure 6). The presence of this RNA in newly laid eggs indicates a maternal contribution, and its persistence suggests either that this maternal RNA is stable for about 5 hr or that the RNA levels are maintained by balanced degradation and zygotic synthesis. Importantly, there is no correlation of RNA levels and cell proliferation.

Open in a separate window

Figure 6

Cyclin mRNA Levels during Embryonic Cell Cycle Progression

Total RNA samples (corresponding to: lane 1, rapid cleavage stage embryos; lane 2, embryos in interphase of cycle 14; lane 3, embryos engaged in divisions 15 and 16; lane 4, older embryos, whose cell divisions are restricted to the nervous system) were probed with the cyclin C cDNA clone on a Northern blot. The lower panel shows a control hybridization of the same filter with an actin 5C probe. Note that the second lane is shifted down owing to loading of an unusually large volume of sample and that cyclin C and actin bands are affected similarly. Given the specific activities of the cyclin C and actin probes and the exposure times of the two hybridizations, we estimate that the intensity of the cyclin C signal is about 200 times less than that of actin. Signals were detected and analyzed by a Molecular Dynamics Phospholmager, which produces a relatively grainy image. The estimated size of the cyclin C mRNA is indicated in kilobases. Embryos were collected and aged at 25°C for the times indicated.

In situ hybridization of embryos with a digoxigenin-labeled DmcycC cDNA probe gave a faint, dispersed signal. This signal was detected prior to onset of significant transcription in the embryo. Like many maternal transcripts, this RNA is concentrated in the islands of yolk-free cytoplasm associated with nuclei, and it relocates to the cortical region of the syncytial egg upon nuclear migration to the surface. The signal persists through germ band extension, and no evidence of localized accumulation is seen (data not shown). This pattern is most simply consistent with a maternal origin of the embryonic DmcycC RNA with little or no zygotic expression. We have yet to study expression at later stages.

Drosophila p34^cdc2 Has CLN-lndependent Function in Yeast

The rescue of cln deficiency by the Dmcdc2 gene was unanticipated but might be explained if the Drosophila p34^cdc2 functioned with residual basal level of Cln expression in YPL1 grown on glucose. To test whether leaky expression of GAL^p–CLN2 is required for Dmcdc2 rescue, we removed the galactose-regulated CLN2 gene using a plasmid shuffle. A triple deletion strain (cln1⁻, cln2⁻, cln3⁻, leu⁻) bearing a GAL^p–CLN2, URA3 plasmid was transformed with an ADH^p–Dmcdc2 plasmid marked with LEU2⁺. The LEU⁺ transformants were picked and streaked on plates that selected against URA3 (plates containing fluoroorotic acid) and for LEU2⁺ and Dmcdc2 (glucose media lacking leucine). After 4 days of growth, colonies appeared. Thus, when overproduced in S. cerevisiae, Dmcdc2 can bypass the requirement for the known CLN genes. However, these cells grew poorly (see below), and the presence of the GAL1^p–CLN2, URA3 plasmid improved growth even when glucose repressed. Thus, leaky expression of GAL1^p–CLN2 appears to be significant, though not essential for growth of the Dmcdc2 transformants.

The Dmcdc2 gene might bypass the need for CLN functions either by virtue of properties distinct from the endogenous cdc2 homolog, the CDC28 gene, or by virtue of over-expression. We tested whether high copy overexpression of CDC28 bypasses the CLN requirement. The YPL1 strain was transformed with pRD46a, a multicopy plasmid carrying the CDC28 gene with its own promoter. For comparison we also transformed this strain with ADH^p-Dmcdc2-and ADH^p-CLN2-expressing plasmids. On glucose plates (minus tryptophan and uracil), CLN2-rescued colonies appeared in 2 days and Dmcdc2-rescued colonies in 3 days, but no growth of CDC28-transformed cells was detected within 5 days. Thus, overproduction of CDC28 does not appear to result in bypass of the CLN requirement.

Growth of Cyclin C- and Dmcdc2-Rescued Cells

Cross (1988) reported that alterations of cln activity change cell size but not the doubling time. Apparently, increased G1 cyclin activity shortens the G1 phase, but some other part of the cell cycle compensates, so that total doubling time is unchanged (Cross, 1988; Hadwiger et al., 1989). Accordingly, the level of G1 cyclin activity should be reflected in the size at which the rescued cells initiate S phase rather than in the growth rate.

We compared the cell size of CLN-deficient cells rescued by CLN2, DmcycC, or Dmcdc2. The DmcycC-rescued cells are smaller than the CLN2-rescued cells, whether assessed by microscopy (Figure 7) or by fluorescence-activated cell sorting analysis (data not shown). This difference suggests that DmcycC encodes high or unregulated G1 cyclin activity. In contrast, the Dmcdc2-rescued cells are larger than the CLN2-rescued cells. This suggests that the basis of the rescue by Dmcdc2 is different from that of DmcycC.

Open in a separate window

Figure 7

Photomicrographs of cln⁻ Rescued Mutants

(A) Haploid YPL1 strain grown either in galactose (GAL1^p–CLN2) or in glucose after transformation with plasmids containing ADH1 promoter–driven DmcycC, Dmcdc2, or CLN2 (respectively, ADH1^p–DmcycC, ADH1^p–Dmcdc2, and ADH1^p–CLN2).

(B) Enlargements of representative YPL1 transformants with ADH1^p–CLN2, ADHI^p–DmcycC, and ADH1^p–Dmcdc2.

(C) CLN⁺(bar1-1) cells either transformed with ADH1^p–DmcycC or non-transformed (CONTROL), grown in glucose selective media minus uracil, or supplemented with uracil in the case of the nontransformed control.

We expected that DmcycC would drive premature division in CLN⁺ strains as well as in the CLN-deficient strain. However, when we transformed a CLN⁺ strain we found that the DmcycC cells were not reduced in size. Thus, the wild-type CLN functions or possibly some other aspect of the strain background blocks DmcycC promotion of the small cell phenotype.

The viability of cells rescued by the Drosophila genes was less than that of the same cells rescued by ADH^p–CLN2. The cell number in actively growing cultures was determined by direct counts, and a constant number of cells from each culture were plated. If normalized to the number of colonies obtained from YPL1 grown on galactose media, the plating efficiency was 85% for CLN2-rescued YPL1, 50% for DmcycC-rescued YPL1, 50% for Dmcdc2-rescued YPL1, and 15% for Dmcdc2-rescued cells that lack a GAL^p–CLN2 gene. Thus, the rescued strains have reduced viability, and a constant attrition of the growing population of cells might be responsible for the slow rate of colony formation.

α factor arrest of the cell cycle appears to be mediated by inhibition of CLN expression or function and can be bypassed by constitutive production of an active CLN product (Cross, 1988; Nash et al., 1988; Hadwiger et al., 1989; Chang and Herskowitz, 1990). Thus, we expected constitutive expression of DmcycC or Dmcdc2 to bypass the α factor block. However, a CLN⁺ α factor–sensitive strain (CLN⁺ bar1-1) remained α factor sensitive when transformed with ADH^p–Dmcdc2 or ADH^p–DmcycC. This sensitivity was evident as an inhibition of budding assessed 4 hr after α factor addition and as growth inhibition on a plate assay (data not shown).

The Family of Cyclin Homologs

The cell cycle p34^cdc2 kinase is thought to require association with a cyclin for activity. What of the other members of the family of cdc2-related kinases (Garrett and Broach, 1989; Courchesne et al., 1989; Lehner and O’Farrell, 1990a; Elion et al., 1990; Paris et al., 1991; Johnson and Smith, 1991)? We propose that kinase activation by cyclin–p34^cdc2 association is a more widely used regulatory paradigm and that there is a family of cyclin-like molecules that function to activate cdc2-like kinases.

Overexpression or transfer of a gene to a different species might erode specificity and allow a cyclin-like molecule to activate kinases other than its normal target. How ever, independent isolation of the DmcycC gene (E. Lahue and T. Orr-Weaver, personal communication) and isolation of the homologous human cyclin C gene (Lew et al., 1991, this issue) argue that the screen for G1 cyclin activity in yeast is relatively selective. This selectivity could represent a specificity for genes with a G1 cyclin function. Alternatively, it could represent a broader specificity for genes capable of activating the S. cerevisiae p34^cdc2 encoded by the CDC28 gene. Or perhaps, the selectivity is for a still more fundamental ability to interact with and activate cdc2-like kinases.

How extensive might the family of cyclin homologs be? A search for sequences including key amino acid residues found in all cyclin subtypes identified the ras oncogene as a homolog. In a 21 amino acid region, ras and G2 cyclins share the same level of homology as the more distantly related cyclins (Figure 3). Further comparisons of cyclin and ras sequences have identified a second region of extensive ras–cyclin sequence homology (O’Farrell and Léopold, unpublished data). The pattern of sequence homology and alignment with the ras crystal structure suggest that ras and cyclins share a common core structure as well as conserved surface features. The conservation of exposed residues leads us to suggest that these proteins may have common targets of interaction.

The Cyclin C Gene

The distant relationship of cyclin C to the other members of the cyclin family does not appear to be the result of rapid evolutionary divergence of this cyclin. Cyclin C conservation from Drosophila to human (72% identity) is much higher than conservation of other cyclins (e.g., 31% identity between Drosophila and human cyclin A). This suggests that cyclin C has important functions that severely constrain evolutionary divergence. Indeed, genes showing such levels of sequence conservation between species as distant as human and Drosophila tend to be well conserved among all eukaryotes (Wilson et al., 1977).

Despite an expectation that DmcycC should be particularly closely related to the CLN genes because it was selected by its ability to complement defects in these genes, a variety of sequence comparisons failed to detect any particular affinity between these sequences. Indeed, cyclin C appears to be a diverged cyclin homolog and does not appear to belong to any of the cyclin subfamilies.

Based on cyclin homology and independent evolutionary conservation, the name cyclin C seems appropriate. However, it should be noted that there is as yet no indication of whether cyclin C protein levels show the periodic accumulation that originally drew attention to the cyclins and inspired the name. In this context, it should be noted that cyclin C lacks the sequences that are thought to target other cyclins for rapid degradation, the degradation box of the G2 cyclins (Glotzer et al., 1991) and the PEST sequences of the Cln proteins (Rogers et al., 1986).

Cyclin C and Regulation of Cln Function in Yeast

Dominant alleles of CLN1 and CLN3 genes and high level constitutive expression of CLN1 are known to give a small cell phenotype (Cross, 1988; Nash et al., 1988; Hadwiger et al., 1989). On this basis it has been suggested that limitation of CLN function is involved in normal cell size control (Reed, 1991). DmcycC-transformed Cln-deficient yeast are very small, as would be expected for constitutive expression of G1 cyclin activity. In contrast, rescue by CLN2 did not produce a small cell phenotype. This difference leads us to suggest that a component of normal size control still operates in the CLN2-transformed cells but not in the DmcycC-transformed cells. Since both of these introduced genes are expressed from a constitutive ADH promoter, it appears that control at a posttranscriptional level affects Cln2 activity.

We expected constitutive expression of cyclin C to drive premature START in CLN⁺ cells as well as in cln⁻ cells. However, the size of CLN⁺ cells was little influenced by introduction of the ADH^p–DmcycC gene. Thus, one or more of the CLN genes suppresses the G1 cyclin activity of cyclin C. Such suppression can be accounted for by a competition model in which a yeast Cln forms a complex to the exclusion of cyclin C. This preferential interaction could restore size control, if the Cln complex is regulated while the cyclin C complex is not. Control of a cyclin–p34^cdc2 complex has a precedent in the control of cyclin B–p34^cdc2 in G2 by phosphorylation and dephosphorylation.

The mating pheromone α factor arrests the yeast cell cycle prior to START. Acting through a complex signal transduction pathway, α factor appears to inhibit Cln function (Cross, 1988; Nash et al., 1988; Hadwiger et al., 1989). We consequently expected that constitutive expression of cyclin C would bypass the α factor block by providing G1 function. However, CLN⁺ cells transformed with an ADH^p–DmcycC plasmid remained α factor sensitive. This suggests either that the yeast mechanisms can inhibit the Drosophila protein, that α factor inhibits a different step, or that the inhibited CLN gene products can exclude cyclin C from participating in the formation of functional complexes.

We do not understand how the Drosophila p34^cdc2 kinase rescues the Cln-deficient strain. However, the large size of the Dmcdc2-rescued cells suggests that the nature of the rescuing activity is very different from that of DmcycC. One possibility is that the Drosophila p34^cdc2 kinase has a basal level of kinase activity independent of association with a cyclin subunit. Alternatively, the Drosophila protein might be activated by a yeast cyclin–like protein other than Cln1, Cln2, or Cln3.

Yeast Strains and Media

Yh131 was derived from BF265-15D (Reed et al., 1985) and includes ura3Δns cln1::TRP1 cln3Δ (kindly provided by H. Richardson) (Richardson et al., 1989). YPL1 was constructed by replacing the chromosomal wild-type CLN2 sequence by a GAL^p–CLN2 insert in Yh131. To this end, Yh131 was transformed with a Sphl–Xhol fragment of plasmid pREP2 (see below), and chromosomal replacement was selected as LEU⁺ prototrophy. FC139 (a HMLa HMRa ura3-52 met1 lys^−a bar1-1) was kindly provided by F. Chang (Chang and Herskowitz, 1990) and used in the α factor sensitivity experiments.

In the case of the cDNA library, yeast transformations were performed by the spheroplast method of Burgers and Percival (1987). Other transformations were performed by the lithium acetate procedure (Ito et al., 1983). Cell cycle arrest by α factor was performed as described in Chang and Herskowitz (1990). YEP-galactose or YEPD (complete) media, SD (minimal) medium, and supplements were used as described by Guthrie and Fink (1991).

Recombinant DNA Manipulation

Plasmid pREP2 was designed to promote chromosome replacement of the CLN2 gene by a galactose-inducible CLN2 copy. For this purpose, a 5.2 kb EcoRI–Sall fragment from plasmid YCpG2[CLN2] (Richardson et al., 1989) containing GAL1^p-CLN2, LEU2 sequences was obtained by partial digestion and cloned into pBluescript. A 530 bp Xhol fragment corresponding to the 5′ region of the CLN2 gene was PCR amplified and cloned into a Xhol site in the pBluescript polylinker, 3′ to the LEU2 gene. A 570 bp Sspl–Xhol fragment from plasmid pSC2-Yep15 (kind gift of H. Richardson) containing 3′ flanking sequences from the CLN2 coding region was inserted into the Smal–Xhol site of pBluescript, next to the GAL1^p sequences and in the same orientation as the 5′ flanking sequences. Plasmid pREP2 was then cut with Sphl and Xbal to give a 6.3 kb fragment with free ends within the 3′ and 5′ flanking sequences of the CLN2 gene, and this was used for chromosomal replacement by the method of Rothstein (1983).

For recovery from yeast to E. coli, plasmid DNA was isolated from yeast cultures using the glass bead–phenol method (Hoffman and Winston, 1987), treated with Geneclean, and introduced into E. coli by electroporation. Bacterial colonies were tested directly by the PCR for the presence of cdc2 sequences. Reactions were carried out following conditions described by the manufacturer (Perkin-Elmer Cetus).

In situ hybridization to polytene chromosomes, Southern blots, and Northern blots were performed as described in Lehner and O’Farrell (1989).

For DNA sequencing, a cyclin C cDNA insert was subcloned from the library vector pDB20 into pBluescript SK after NotI digestion. The resulting pSKcycC and pSKcycCR plasmids, containing the insert in both orientations, were submitted to ExoIII–S1 digestion (Henikoff, 1987), and a series of unidirectional deletions were sequenced using a T7 sequencing kit (Pharmacia-LKB biotechnology).

Photomicroscopy of Yeast Cells

Unfixed yeast cells were mounted in growth medium and photographed by using a Nikon Optiphot photomicroscope with differential interference contrast (nomarski) optics. A 63 × objective was used.

We thank Daniel Lew and Steve Reed for very helpful discussions and for communicating results prior to publication. We are grateful to Helena Richardson, Fred Chang, and Ray Deshaies for providing various strains, plasmids, and extremely valuable advice. We thank Shelagh Campbell, Jim Jaynes, Danesh Moazed, Tony Shermoen, and Madhu Wahi for their comments on the manuscript, and Judy Piccini for her assistance in its preparation. This work was supported by NIH grant PO1 HL43821 (P. H. O.). P. L. acknowledges the support of a Fogarty International Center Fellowship and is on leave from the Institut National de la Santé et de la Recherche Médicale (France).