Cell, Vol. 66, 1217-1226,
September
20, 1991, Copyright
0 1991 by Cell Press
Human Cyclin E, a New Cyclin That Interacts with Two Members of the CDC2 Gene Family Andrew Koff,” Fred Cross,t Alfred Fisher, * Jill Schumacher, l Katherine Leguellec,* Michel Philippe,* and James M. Roberts’ *Department of Basic Sciences Fred Hutchinson Cancer Research Center 1124 Columbia Street Seattle, Washington 98104 tRockefeller University New York, New York 10021-6399 *Laboratoire de Biologie et Genetique du Developpement Centre National de la Recherche Scientifique Unite Associee 256 Universite de Rennes I Campus de Beaulieu 35042 Rennes Cedex France
Summary A new human cyclin, named cyclin E, was isolated by complementation of a triple c/n deletion in S. cerevisiae. Cyclin E showed genetic interactions with the CDC28 gene, suggesting that it functioned at START by interacting with the CDC28 protein. Two human genes were identified that could interact with cyclin E to perform START in yeast containing a cdc28 mutation. One was CDC2-HS, and the second was the human homolog of Xenopus CDK2. Cyclin E produced in E. coli bound and activated the CDC2 protein in extracts from human Gl cells, and antibodies against cyclin E immunoprecipitated a histone Hl kinase from HeLa cells. The interactions between cyclin E and CDCZ, or CDKP, may be important at the Gl to S transition in human cells. Introduction Molecules that regulate the onset of mitosis have been highly conserved in evolution (reviewed in Nurse, 1990; Cross, 1990). In Saccharomyces cerevisiae the product of the CDC28 gene and in Schizosaccharomyces pombe the cdc2+ gene product is required for entry into mitosis (Nurse and Bisset, 1981; Piggot et al., 1982; Reed and Wittenberg, 1990). The cdc2 and CDC28 gene products are structurally similar (Lorincz and Reed, 1984; Hindley and Phear, 1984) and functionally homologous (Beach et al., 1982; Booher and Beach, 1987). They encode a serinel threonine protein kinase that is the homolog of the 34 kd protein kinase in vertebrate and invertebrate MPF (Lee and Nurse, 1987; Arion et al., 1988; Dunphy et al., 1988; Gautier et al., 1988; Labbe et al., 1988). The activity of the p34 CDC2 kinase oscillates during the cell cycle (Mendenhall et al., 1987; Draettaand Beach, 1988; Labbe et al., 1989b; Moreno et al., 1989; Pines and Hunter, 1990). This is not controlled by variations in the
amount of the CDC2 gene product (Durkacz et al., 1986; Simanis and Nurse, 1986; Draetta and Beach, 1988) but rather by its interactions with other proteins, including the cyclins (Rosenthal et al., 1980; Evans et al., 1983; Swenson et al., 1986; Draetta et al., 1989; Meijer et al., 1989; Minshull et al., 1989; Labbe et al., 1989a; Solomon et al., 1990; Gautier et al., 1990; reviewed in Murray and Kirschner, 1989; Hunt, 1989). Association between the p34 CDC2 protein and the B-type cyclins is necessary for the activation of the p34 kinase at the onset of mitosis in a wide variety of organisms including yeast (Booher and Beach, 1987; Hagan et al., 1988; Moreno et al., 1989; Solomon et al., 1988; Booher et al., 1989; Surana et al., 1991; Ghiaraetal., 1991)andhumans(Draettaand Beach, 1988; Pines and Hunter, 1989; Riabowol et al., 1989). In yeast the Gl IS transition, called START, also requires the CDC28 or cdc2 gene product (Hartwell et al., 1973, 1974; Nurse and Bisset, 1981). The biochemical pathways that activate CDC28 at START are not completely understood but the recent identification of Gl-specific cyclins in yeast, the CLNl, -2, and -3 proteins, suggests that Gl/S activation and G2/M activation of CDC28 share common features (Sudbery et al., 1980; Nash et al., 1988; Cross, 1988,199O; Hadwigeret al., 1989; Richardson et al., 1989; Wittenberg et al., 1990). Cells deficient in all three CLN proteins arrest at START; they continue to grow but are unable to enter S phase. The CLNP protein (and probably CLNl and -3) forms a complex with the CDC28 kinase at START (Wittenberg et al., 1990). Like the mitotic cyclins, to which they show limited but significant homology, the abundance of the CLNl and -2 proteins oscillates during the cell cycle, but maximal levels are observed in late Gl rather than late G2 (Wittenberg et al., 1990). Very little is known about the biochemical pathways that control thestartof DNAsynthesisin highereukaryoticcells or the extent to which these pathways resemble those in yeast. In human cells, as in budding yeast, the predominant mode of control of cell proliferation occurs during the Gl phase of the cell cycle (Zetterberg and Larson, 1985; Zetterberg, 1990). The kinetics of passage through Gl in mammalian cells suggests that a single decision, called the restriction point, regulates commitment to DNA synthesis (Pardee, 1974). Prior to the restriction point, progress through Gl is very sensitive to the growth state of the cell. Reducing the rate of protein synthesis or removal of specific extracellular growth factors will delay entry into S phase and can even cause cell cycle arrest. After the restriction point the cell cycle becomes substantially less responsive to these signals (reviewed in Pardee, 1989). These features of mammalian cell growth control are very reminiscent of the size control of cell division that characterizes START in budding yeast, although in the mammalian cell cycle responsiveness to extracellular growth factors seems more prominent and strictly size-dependent proliferation less so. The similarities between the restriction point and START have suggested the hypothesis that activation of the p34
Cell 1218
GALACTOSE
Figure 1. Complementation of the Triple c/n Deletion by Human Cyclin E
GLUCOSE
589
S. cerevisiae strain 589-5 contains deletions of the chromosomal CLNl, -2, and -3 genes and contains the GAL7-CLN3 gene on a multicopy episome. It was transformed with the pADNS expression vector or the pADNS vector containing a human cyclin E cDNA. Transformants, and the parental strain, were streaked on galactose and glucose and grown for 3 days at 30%.
589-5 +pRONS
589-S
l pRONS-WC
E
589-S +pRONS-CVC
CDCP kinase plays a central role in regulating mammalian cell proliferation at the Gl to S transition. Indeed, the human CDCP gene can substitute for the S. pombe cdc2 gene for both its GllS and G2/M roles (Lee and Nurse, 1987). Cell fusion experiments offer circumstantial evidence in support of this hypothesis (Rao and Johnson, 1970). These experiments showed that a diffusible, transacting factor activated DNA synthesis when S phase cells were fused to Gl cells. The relationship between the S phase activator operationally defined by those experiments and the p34 CDCP kinase remains unclear. Some recent observations lend direct experimental support to the hypothesis that the p34 CDC2 kinase plays a necessary role in the Gl to S phase transition during the human cell cycle. We isolated cyclin-CDC2 complexes in human S phase cells and showed that they could activate SV40 DNA replication when added to extracts from Gl cells (D’Urso et al., 1990). Others have shown that depletion of CDC2 protein from Xenopus extracts can block DNA replication (Blow and Nurse, 1990). Also, antisense oligonucleotidesdirected against the human CDC2 mRNA can prevent primary human T cells from entering S phase following activation with phytohemagglutinin (Furakawa et al., 1990). The pathway that activates the p34 kinase during the Gl phase of the human cell cycle is not understood. By analogy with the CLN-dependent activation of CDC28 at START, it is expected that specific Gl cyclins play a role in regulating the human p34 kinase during the Gl to S phase transition. We have begun to test this idea by determining whether human cells contain specific cyclins that can replace the S. cerevisiae CLN proteins. This assay identified a new human cyclin, cyclin E. Results Complementation of a Yeast Strain Lacking CLNI, -2, and -3 with a Human cDNA Library Our initial goal was to identify human cDNAs encoding proteins that could substitute for the yeast CLN proteins. A yeast strain, 589-5, was constructed in which all three chromosomal CLN genes were deleted and which contained the CLN3 gene under the control of the GAL1 promoter on an episome. As CLN protein is required for passage through START, this strain will grow on galactose, where the GAL7 promoter is induced; when this strain is
E
grown on glucose the GAL7 promoter is repressed, no CLN3 protein is made, and the cells arrest at START (Cross, 1990). A cDNA library (a gift of J. Colicelli and M. Wigler), using mRNA prepared from the human glioblastoma cell line U118, was constructed in an S. cerevisiae vector containing the constitutive yeast ADH promoter for expression of the human cDNAs (Colicelli at al., 1989). The library was transfected into strain 589-5, and 1 O5independent transformants were screened, by replica plating, fortheirabilitytogrowonglucose.One transformant, HU4, was isolated whose growth on glucose was dependent on expression of a human cDNA (Figure 1). HU4 Encodes a New Member of the Cyclin Protein Family The DNA sequence of the 1.7 kb HU4 cDNA is shown in Figure 2, and its homology, at the protein level, to human cyclins A and B illustrated in Figure 3. The DNA sequence predicts a protein with 395 amino acids and a molecular weight of 45,120. In vitro transcription/translation of the HU4 cDNA yields a protein with the predicted molecular weight (see Figure 5). All known cyclins have a highly conserved central domain of approximately 87 amino acids. HU4 is 49% identical to human cyclin A and 44% identical to human cyclin B within this domain. On this basis we placed the HU4 protein within the cyclin family. N-terminal to this conserved region the homology to cyclin A falls to 5% identity, and 4% identity to cyclin 8. C-terminal to this domain the identity to cyclin A is 14%, and the identity to cyclin B is 10%. This low level of homology both N- and C-terminal to the central conserved domain suggests that HU4 represents a new class of cyclin proteins, and we have designated this class as cyclin E. We cannot be certain that our cyclin E cDNA clone contains the entire cyclin E protein coding sequence, as the open reading frame extends to the 5’ end of the sequenced cDNA. However, a cyclin E cDNA clone obtained from a MANCA cell library showed a 5’ end identical to the one described here (J. S. and J. M. R., unpublished data). Human Cyclin B Complements the Triple c/n Deletion We compared the ability of human cyclin A, B, and E cDNAs to complement the triple c/n deletion in strain 5895. Full-length cDNA clones encoding human cyclins A (Pines and Hunter, 1990) and Bl (Pines and Hunter, 1989)
$~~lmgan Cyclin E
gtgctcacccggcccggtgccacccgggtccacagggatgcgaaggagcgggacacc
0
atg met
aag gag gac lys glu asp
61
act thr
gtt va:
ttt phe
121
tag gin
tgt cys
181
ccc pro
aca cct thr pro
241
att ile
att ile
gca ala
301
gtc vat
tgg trp
361
caa cat gin his
421
ggc gly
ggc gly
gcg ala
gag ttc glu phe
tcg ser
get ala
cgc arg
tee ser
egg erg
aag egg lys arg
ttg tag Leu gin
gat asp
cca pro
gat asp
gaa gaa atg glu glu met
gee ala
aaa lys
ate ile
gac asp
ggg age tag gly ser gln
cct pro
tgg trp
gac aat asp asn
gca gtc ala val
tgt cys
gca gac ccc ala asp pro
gac asp
aaa’gaa lys glu
gat asp
gat asp
gac egg gtt asp arg val
tat tyr
cca pro
sac asn
cca pro
tee ser
aga ggc arg gly
tee ser
ccg pro
gta val
ctg leu
aaa ate Lys ile
atg met
tta leu
sac asn
aag gas lys glu
aag aca tat lys thr tyr
cct pro
ctt leu
ctg leu
tag gln
cca pro
aaa atg lys met
cga gca arg ala
tgt cys
gaa gtc glu vat
tat tyr
aaa lys
ctt leu
cat his
agg gag act arg gfu thr
tat tyr
atg met
gcg ala
aca caa thr gln
gaa aat glu asn
gtt val
541
tta \eu
ttt phe
att ile
gca ala
aaa lys
ctt teu
601
gtg val
aca thr
gat asp
gga get gly ala
tgt cys
661
gee ala
ctt leu
aag tgg tys trp
cgt arg
721
gtt val
gca tat ala tyr
eta leu
781
ata ile
tag gln
841
tat tyr
901
aag gca lys ala
gtg vat
gcg ala
agg gac arg asp
tee ser
ctg leu
ate ife
acg tgc thr cys
aag cct f.ys pro
egg arg
age tgg ser trp
gca ala
eat asn
aga gag gaa arg glu glu
tta leu
agg gat arg asp
tag gin
cat his
ttt phe
ctt leu
att ile
ctt Leu
ctg leu
tgg trp
tta leu
atg met
gag gtg glu va(
ttt phe
tat tyr
ttg leu
gca caa gat ala gin asp
ttc phe
ttt phe
gac egg asp arg
aaa act lys thr
ctt teu
tta leu
tag gin
ctt leu
att ile
ggg att gly ile
tea ser
tct ser
gag gaa gfu giu
ate ite
tat tyr
cct pro
cca pro
aag lys
ttg \eu
cat his
tag gln
ttt phe
gcg ala
tat tyr
tea ser
gga gat gly asp
gaa glu
att ile
ctc teu
act thr
atg met
gaa glu
tta leu
atg met
att ile
atg met
aag Lys
tta leu
agt ser
ccc pro
ctg act Leu thr
att ile
gtg val
tee ser
tgg trp
ctg leu
aat asn
gta val
tat tyr
atg met
tag gin
aat asn
gac asp
tta leu
cat his
gaa gtg glu val
eta leu
ctg teu
ccg pro
tag gln
tat tyr
ccc pro
tag gln
caa gin
ate ile
ttt phe
att ile
gca gag ala glu
ctg leu
ttg leu
gat asp
ctc leu
tgt cys
gtc vat
ctg leu
gat asp
gtt val
gac tgc asp cys
ctt leu
gaa ttt glu phe
cct pro
ggt gly
ata ile
ctt leu
get ala
get ala
tcg ser
gee ttg ala teu
tat tyr
cat his
ttc phe
tcg ser
tea ser
tct ser
gaa glu
ttg leu
atg met
caa gin
aag lys
gtt val
tea ser
ggg tat gly tyr
tag gin
tgg trp
tgc cys
gac asp
gag sac glu asn
tgt cys
gtc val
aag tgg lys trp
atg met
gtt vat
cca pro
ttt phe
gee ala
961
atg met
gtt val
ata ile
agg gag arg glu
acg thr
ggg age tea gly ser ser
aaa fys
ctg leu
aag cat lys his
ttc phe
agg ggc gtc arg gly val
get ala
gat asp
gaa glu
1021
gat asp
gca ala
cat his
sac asn
ata ile
tag gin
act thr
aga gac arg asp
age ser
ttg feu
gat asp
ttg leu
ctg leu
gac asp
aaa gee cga lys ala arg
gca ala
1081
aag lys
aaa gee atg fys ala met
ttg leu
tct ser
gaa caa gfu gin
aat asn
agg get arg ala
tct ser
cct pro
ctc leu
ccc pro
agt ser
ggg gly
act thr
1141
ccg pro
cm pro
aag Lys
sag tag
age age ggg ccg ser ser gly pro
tag gin
gee ala
age ggt ser gly
lys
cat his
gin
gta vat
ata ile
aat asn
ctg leu
cct pro
gas atg glu met
tea ser
gat asp
agg acg arg thr
sac asn
gcg tga ala ***
ccaccaaagacagttgcgcgcctgctccacgttctcttctgtctgttgcagcggaggcgtgcgtttgcttttacagata tctgaatggaagagtgtttcttccacaacagaagtatttctgtggatggcatcaaacagggcaaagtgttttttattga atgcttataggttttttttaaataagtgggtcaagtacaccagccacctccagacaccagtgcgtgctcccgatgctgc tatggaaggtgctacttgacctaaaggactcccacaacaacaaaagcttgaagctgtggagggccacggtggcgtggct ctcctcgcaggtgttctgggctccgttgtaccaagtggagcaggtggttgcgggcaagcgttgtgcagagcccatagcc agctgggcagggggctgccctctcc Figure 2. Sequence
of Cyclin E
DNA and predicted
protein sequence
of the cyclin E CDNA that complemented
the triple c/n deletion
tgc cys
ctc leu
ctc leu
ccaccccatccttct
gag gfu
Cell 1220
CYCLIN CYCLIN CYCLIN
E A B
CYCLIN CYCLIN CYCLIN
E A B
IlLGNSAPGP IlALR
SKINAENKAKINIiAGAKRVPTAF
CGAEFSARSRKR
E CYCLIN CYCLIN CYCLIN
E A B
E A B
CYCLIN CYCLIN CYCLIN
E A B
CYCLIN CYCLIN CYCLIN
E A B
ET
DR
RGSP-W
V CYCLIN CYCLIN CYCLIN
L
MRAILDWLV
WL
G
AKEEYPP
L
of the Protein Sequences
TD
NE
L
LA
tlYIRETG5SKLK HFRGVADEDAHNIQTHRD
Figure 3. Alignment
F
L
L
Y
&@~~&K~~EQNRAsPLP~GLL~PQ~GKKQssGPE~~A
of Human Cyclins A, B, and E
Protein sequences of cyclins A, B, and E were aligned to maximize homology. Boxes indicate identical amino acids. Amino acids shared by all three cyclins are shown in bold type above the three sequences. The area highlighted by double bold lines is a domain highly conserved among all known cyclins (the cyclin box). The domain highlighted by single bold lines is the mitotic destruction motif shared by all A- and B-type cyclins.
were cloned into the ADH expression vector used in construction of the library described above. The various cyclin expression plasmids were transfected into yeast by selecting for leucine prototrophy. The transformants were picked, and the ability of the human cyclin to complement the absence of the CLN proteins was tested by growth on glucose. No leucine prototrophs were obtained using the cyclin A vector, suggesting that expression of full-length human cyclin A in S. cerevisiae was lethal. Table 1 depicts the relative plating efficiency of 589-5 on glucose versus galactose when containing either cyclin B or E expression plasmids. Surprisingly, the mitotic cyclin B complemented the absence of the CLN proteins. This experiment demonstrated that complementation of CLN function was not restricted to CLN-type cyclins.
Interaction between Cyclin E and CDC28 We compared the ability of cyclin E to rescue the triple c/n deletion in isogenic strains containing either the CDC28 or the c&28-73 allele at the permissive temperature of 30% (Table 1). Table 1 shows that cyclin E substituted for the CLN proteins significantly less well in the c&28-73 background. This genetic interaction between the cyclin E and CDC28 genes suggested that the cyclin E protein performed its function by interacting with the CDC28 protein. Cyclin 6 was about as effective in cdc28-13versus CDC28 strains, suggesting that cyclin B might interact with CDC28 differently than cyclin E.
Human Cyclin E and Human CDCP Can Perform START in S. cerevisiae Strains containing c&28-13 and the endogenous Gl (CLN) and mitotic (CLB) cyclins are viable at 30%. Therefore, the cdc28-13 protein must be capable of functional interactions with both types of cyclins. As described above, a c&28-73 strain that contained cyclin E in place of the CLN genes did not grow at 30%. We speculated that this strain was defective specifically for the START function of the CDC28 protein and not for its G2/M role, presumably because the c&28-73 mutation diminished its ability to interact productively with cyclin E. These properties of the cyclin E-c&2&73 strain are shown in Table 2. We used this strain as a host to screen for human genes that could interact with cyclin E to perform START. Unlike screens requiring human genes to completely substitute for CDC28, this screen may not require that the human genes function at G2/M. We transfected this strain with the human cDNAexpression library described above, and lo5 independent colonies were tested for their ability to grow on glucose. We identified five yeast clones whose growth depended on expression of both the human cDNAs (cyclin E and the new one) (Table 3). The human cDNA within each of these clones was restriction mapped (data not shown). Four of these (S2-8a2, S2-103, S2-112, 52-227) contained the CDC2-HS gene (Lee and Nurse, 1987). The S. pombe cdc2’ gene also performed START together with human cyclin E in this strain (F. C. and A. Tinkelenberg, unpublished data). These results provide
Human Cyclin E 1221
Table 1. Efficiency of Rescue of the Triple c/n Deficiency Human Cyclins E and B in CDCPB or cdc28-73 Strains
pADNS pADANS pADNS-CYC E pADANS-CYC E pADANS-CYC B
CDC28
cdc28-13