Cyclin E2, a Novel Human G1 Cyclin and Activating Partner of CDK2 and CDK3, Is Induced by Viral Oncoproteins

Home , Cyclin E, Cyclin E1, Cyclin E2

Oncogene (1998) 17, 2787 ± 2798 ã 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00 http://www.stockton-press.co.uk/onc Cyclin E2, a novel human G1 cyclin and activating partner of CDK2 and CDK3, is induced by viral oncoproteins

Maimoona Zariwala1, Jidong Liu2 and Yue Xiong*,1,2,3

1Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-3280; 2Department of Biochemistry and Biophysics; University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599- 3280; 3Program in Molecular Biology and Biotechnology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-3280, USA

G1 cyclin E controls the initiation of DNA synthesis by complexes with CDK4 or CDK6 to prevent the CDKs activating CDK2, and abnormally high levels of cyclin E from binding with and becoming activated by D-type expression have frequently been observed in human cyclins. The main function of CDK inhibitors is cancers. We have isolated a novel human cyclin, cyclin believed to couple diversi®ed growth inhibitory signals E2, that contains signi®cant homology to cyclin E. to the cell cycle clock. Cyclin E2 speci®cally interacts with CDK inhibitors of The decision to enter the replicative DNA synthesis the CIP/KIP family and activates both CDK2 and (S) phase or arrest in G1 is linked to diverse cellular CDK3. The expression of cyclin E2 mRNA oscillates processes such as signal transduction, cell differentia- periodically throughout the cell cycle, peaking at the tion, senescence, and oncogenic transformation G1/S transition, and exhibits a pattern of tissue (Hunter and Pines, 1994). In the budding yeast speci®city distinct from that of cyclin E1. Cyclin E2 Saccharomyces cerevisiae, the G1 cyclins Cln1 and encodes a short lived protein whose turnover is most Cln2 play the key role in determining the rate and fate likely governed by the proteasome pathway and is of G1 phase progression. Accumulation of CLN1 and regulated by phosphorylation on a conserved Thr-392 CLN2 proteins in G1 results in activation of Cdc28 residue. Expression of the viral E6 oncoprotein in normal kinase activity, thereby driving cells to passage through human ®broblasts increases the steady state level of a G1 restriction point known as START and to cyclin E2, but not cyclin E1, while expression of the E7 committing cells to initiate DNA synthesis (Reed, oncoprotein upregulates both. These data suggest that 1992). In mammalian cells, two CDK enzymes, CDK4 the expression of these two G1 E-type cyclins may be or CDK6 in combination with three D-type cyclins similarly regulated by the pRb function, but distinctly by (D1, D2 and D3) and CDK2 in association with cyclin the p53 activity. E, play the principle roles in regulating G1 progression (reviewed in Sherr, 1993, 1994). Although overexpres- Keywords: cyclin E2; CDK; G1 cell cycle control; cell sion of both D- and E-type cyclins similarly shorten transformation the G1 interval and reduce mitogen dependency (Ohtsubo and Roberts, 1993; Ohtsubo et al., 1995; Resnitzky et al., 1994), the function of these two cyclins are distinctly dierent in regulating G1 Introduction progression. While the function of cyclin D-associated kinase is largely dependent on the wild type pRB Progression through eukaryotic cell division cycle is (Guan et al., 1994; Lukas et al., 1995a, b; Koh et al., controlled by a family of protein serine/threonine 1995; Serrano et al., 1995; Medema et al., 1996), the kinases known as cyclin-dependent kinases (CDKs) function of cyclin E remains essential in Rb-negative which consist of an activating subunit, a cyclin, and a cells. The expression of cyclin D genes and the pRb catalytic subunit, a CDK. Sequential activation and kinase activity of the cyclin D-CDK4/6 enzymes are inactivation of CDK enzymatic function constitutes the induced during the delayed early response to mitogenic molecular basis for orderly progression through the cell stimulation, prior to the accumulation and kinase cycle. The primary positive and negative regulation of activity of cyclin E (Matsushime et al., 1991, 1994; CDK activity is mediated by the binding of a cyclin Meyerson and Harlow, 1994). The level of D cyclins and of a CDK inhibitor, respectively (Hunter and ¯uctuate little during the rest of the cell cycle. On the Pines, 1994). In mammals, two families of seven CDK other hand, the expression of cyclin E, but not CDK2, inhibitors have been identi®ed that dier in both is periodic during the cell cycle, peaking during G1 and structure and mechanism of action. Members of the declining during S phase (Lew et al., 1991; Dulic et al., CIP/KIP family contains three genes, p21CIP1/WAF1, 1992; Ko et al., 1992) and is necessary for S phase p27KIP1 and p57KIP2, that inhibit CDK activity by entry (Sherr, 1993; Duronio et al., 1996; Ohtsubo and forming a ternary p21-cyclin D-CDK4 complex. The Roberts, 1993; Ohtsubo et al., 1995; Pagano et al., INK4 family inhibitors which include four closely 1992; Resnitzky et al., 1994). These features suggest related ankyrin repeat containing genes, p16INK4a, that CDK4/6-cyclin D and CDK2-cyclin E enzymes p15INK4b, p18INK4c and p19INK4d, selectively form binary have distinct functions in regulating G1 progression, with cyclin D coupling extracellular growth signals to the cell cycle machinery while cyclin E controlling the initiation of DNA replication. *Correspondence: Y Xiong; E-mail: [email protected] Received date 11 September 1998; revised 22 October 1998; accepted The regulation of periodic accumulation and 22 October 1998 diminish of cyclin E is achieved largely through Novel human G1 cyclin E2 M Zariwala et al 2788 transcriptional activation of E2F during G1 (Duronio proteins, cyclin H has not been previously reported to and O'Farrell, 1995; Ohtani et al., 1995; Degregori et interact with CDK2. Whether in vivo CDK2 associates al., 1995; Botz et al., 1996; Geng et al., 1996), and with cyclin H and whether cyclin H can activate CDK2 subsequent degradation of cyclin E by the ubiquitin- kinase activity remains to be determined. In addition, a mediated proteolysis during S phase (Clurman et al., novel protein, encoded by clone HT376, was identi®ed 1996; Won and Reed, 1996). Deregulated cyclin E and con®rmed to strongly interact with CDK2K33M after expression could contribute to and may even be re-transformation of the rescued plasmid into yeast necessary for entering into S phase in cells which cells with CDK2K33M as bait (Figure 1a). This clone, would otherwise arrest in G1. Indeed, altered which codes for a novel human cyclin most closely expression of cyclin E has been frequently observed related to cyclin E (termed as cyclin E2), was chosen in human cancers. While genetic alteration of cyclin E for further study. gene rarely occurred in human cancer, abnormally high Cyclin E2 was cloned from the screen with mutant levels of cyclin E protein, resulted from either gene CDK2 suggesting that interaction between CDK2 and ampli®cation or deregulated degradation by the cyclin E2 may be facilitated by the mutation in CDK2. ubiquitin-mediated proteolysis pathway, have fre- To test this directly, cyclin E2 expressing HF7c yeast quently been observed in both primary tumors and cells were sequentially transformed with either mutant cancer-derived cell lines (reviewed in Hunter and Pines, or wild type CDK2. As shown in Figure 1, catalytic 1994; Sherr, 1996; Dou et al., 1996; Hall and Peters, inactive CDK2K33M, but not wild type CDK2, is 1996). Such frequent quantitative and qualitative capable of interacting with cyclin E2 as determined alteration of cyclin E expression and its correlation by growth on a selective media. These observations with increasing stage and grade of the tumor has led suggest that CDK2 and cyclin E2 may form a catalytic the proposal of using cyclin E as a potential prognostic active complex in yeast cells, causing either cyclin E2 marker for human cancer (Keyomarsi et al., 1997). degradation or interference with yeast CDC28 func- These observations suggest that although it is not a tion. Subsequent sequencing analysis revealed an in- proto-oncogene like cyclin D1, cyclin E overexpression frame stop codon between the GAL4 DNA binding might override certain growth inhibitory signals and domain and cyclin E2. This in-frame stop codon might thus contribute to cell transformation. In this report, reduce, but not completely terminate the synthesis of we describe a novel cyclin that shares signi®cant cyclin E2 so that there is sucient amount to interact sequence homology and many characteristic features with CDK2 bait. Removal of this stop codon with cyclin E, including activation of CDK2 and essentially arrested yeast cell growth (MZ, unpub- periodic oscillation during the cell cycle with a peaking lished observation). The basis for such intolerance of at the G1/S boundary. Expression of the papilloma cyclin E2 expression by yeast cells is not clear, but viral E6 oncoprotein in normal human ®broblasts could be caused by the ability of human cyclin E2 to increases the steady state level of cyclin E2, but not bind to and interfering the function of yeast CDC28. cyclin E1, suggesting distinct regulation mechanisms CDK3 is the closest relative to CDK2 among the for the two G1 E-type cyclins. nine mammalian CDK genes identi®ed thus far and its activating cyclin subunit has yet to be identi®ed. When tested in yeast cells, both the catalytically inactive CDK3K33M mutant and wild type CDK3 interact with Results cyclin E2 (Figure 1a), suggesting that cyclin E2 could potentially function as an activating cyclin subunit of Cloning of cyclin E2 CDK3 and that CDK3-cyclin E2 is functionally We used a catalytically inactive mutant of CDK2, distinct from the CDK2-cyclin E2 complex. Cyclin E2 CDK2K33M, to search for additional CDK2-interacting was unable to interact with other human CDK proteins protein(s) by the yeast two-hybrid assay. Use of (CDK4 ± 8), whether wild type or catalytically inactive mutant CDK2 as bait was prompted by its potential mutants, reinforcing the speci®city of the cyclin E2- to stabilize kinase-substrate interaction and to prevent CDK2 and the cyclin E2-CDK3 interaction. interference with endogenous yeast CDC28-cyclin activity and CDK-dependent cyclin degradation The human cyclin E2 gene (Clurman et al., 1996; Won and Reed, 1996). Of an estimated 36106 transformants from a human HaCaT Clone HT376 contains a 2801 bp cDNA insert which keratinocyte cell cDNA library screened with the predicts a 404 amino acid residue open reading frame CDK2K33M mutant, approximately 200 clones were with a calculated molecular weight of 47 kDa (Figure obtained that proliferated on media lacking histidine 1b). It contains an in-frame translational termination and were positive for staining of an independent codon 12 nucleotides upstream from the initiation reporter gene, b-galactosidase. As determined by methionine codon and a poly (A) tail (34 adenine DNA sequencing and Southern hybridization, the residues), indicating that clone HT376 contains the full majority of the positive clones corresponded to either length gene. The most obvious structural feature of this the ubiquitously expressed CDK-associated, small Suc/ open reading frame is its signi®cant sequence similarity Cks1 protein (Richardson et al., 1990), or the CDK- to cyclin proteins. The most closely related cyclins are inhibitory phosphatase, KAP (Gyuris et al., 1993; E-type cyclins from dierent species, with human cyclin Hannon et al., 1994). Sequencing of the remaining E sharing 75% identity within the 107 residues of the positive clones identi®ed several previously character- cyclin box and 47% identity throughout the 389 ized cell cycle regulatory proteins including cyclin D1, residues common to both cyclin E proteins (Figure cyclin D2, cyclin H, the CDK inhibitor p21, and the 1c). As a comparison, HT376 shares with its next dual speci®city phosphatase CDC25B. Among these closest related human cyclin, cyclin A, a 44% sequence Novel human G1 cyclin E2 M Zariwala et al 2789

Figure 1 Interaction of human cyclin E2 with CDK proteins. (a) Yeast HF7c cells expressing human cyclin E2 plasmid was sequentially transformed with plasmids expressing the indicated CDKs. Cells were streaked onto nonselective medium containing histidine (7LW) and then replica plated onto selective medium lacking histidine (7LWH) to assay for the activation of the histidine reporter gene. (b) Nucleotide and amino acid sequences of human cyclin E2. The stop codon is indicated by an asterisk. (c) Sequence Comparison of human cyclin E1 and E2. The entire sequences of both cyclins were aligned and identical residues are in bold. Underlined sequences correspond to the cyclin Box. Thr-380 of cyclin E1 and Thr-392 in cyclin E2, whose phosphorylation play an important role in regulating the stability of both proteins are indicated by italics Novel human G1 cyclin E2 M Zariwala et al 2790 identity within the cyclin box, and barely detectable blotting with an antibody speci®c to human CDK2 scattered similarity outside the cyclin box region. detected CDK2 protein in the anti-cyclin E2 precipi- Because it contains such high sequence similarity to tates derived from CT10, but not HSF43 cells, and shares many characteristic features with cyclin E, demonstrating an in vivo association between cyclin we have named this novel cyclin as cyclin E2. E2 and CDK2. Accordingly, the previously characterized cyclin E (Ko et al., 1991; Lew et al., 1991) can be designated as cyclin E1. Cyclin E1 and E2 represent two distinct genes and should not be confused with the several alternatively spliced forms of cyclin E1 (Sewing et al., 1994; Mumberg et al., 1997), nor with several reported Xenipus cyclin E sequences (named as cyclin E1, cyclin E2 and cyclin E3; Rempel et al., 1995; Howe and Newport, 1996; Chevalier et al., 1996) whose diver- gence are likely to re¯ect sequence polymorphisms resulting from the tetraploidy of the Xenopus genome, rather than represent distinct cyclin E genes.

In vivo association of cyclin E2 and CDK2 To examine in vivo association of cyclin E2 and CDK2 proteins, we ®rst analysed a series of established human cell lines for the level of cyclin E2 mRNA by Northern analysis. Among a number of cell lines in which expression of cyclin E2 mRNA was detected are two SV-40 virally transformed cell lines, CT10 and VA13 (Figure 2a). Notably, the level of cyclin E2 mRNA is almost undetectable in their respective un- transformed parental cell lines, HSF43 or WI-38, suggesting that the expression of cyclin E2 may be upregulated by the SV-40 viral transformation. Alteration of cyclin E2 expression by viral oncoproteins was further investigated later. Notably, cyclin E2 mRNA was also detected in Saos-2 (human osteosarcoma cell line) which has deletion in Rb gene and contains a mutation in p53 gene. Another osteosarcoma derived cell line U-2 OS which has wild type Rb and p53 had no detectable levels of cyclin E2. Rabbit polyclonal antibody speci®c to human cyclin E2 was raised using a synthetic peptide corresponding to the C-terminus of human cyclin E2 (residues 391 ± 404) which greatly diverged from cyclin E1 sequence as an immunogen. This antibody is capable of precipitat- ing in vitro translated cyclin E2, but not cyclin E1 (data not shown), and recognizes bacterially produced recombinant cyclin E2 (rE2) protein in immunoblotting (lane 1, Figure 2b), con®rming its speci®city. Total Figure 2 Association and activation of CDK2 by cyclin E2. (a) cell lysates were prepared from CT10 and HSF43 and Expression of cyclin E2 mRNA. Ten micrograms (mg) of total sequentially immunoprecipitated with the cyclin E2 RNA were prepared from two pairs of normal (HSF43 and WI- antibody and immunoblotted (IP-Western) with either 38) and virally transformed (CT10 and VA13) human ®broblasts and human osteosarcoma (U-2 OS and Saos-2) cells and were anti-cyclin E2 or anti-CDK2 antibody (Figure 2b). loaded on a 1% agarose gel. Resolved RNA was transferred to a Two bands at approximately 41 kDa were detected by nylon membrane and the blot was hybridized with a probe the cyclin E2 antibody in CT 10 cells (lane 4). The derived from the full length of human cyclin E2. (b) In vivo faster migrating band has an electrophoretic mobility association of cyclin E2 and CDK2. Total cell lysates were close to that of bacterially produced cyclin E2 protein. prepared from HSF43 or CT10 cells and immunoprecipitated with an anti-cyclin E2 antibody. After SDS ± PAGE, proteins Judging by their absence in HSF43 cells (lane 2), and were transferred to nitrocellulose, and the upper portion was absence in immunoprecipitates from CT10 cell lysate immunoblotted with anti-cyclin E2 antibody and the lower containing molar excess of a competing cyclin E2 portion of the same blot was immunoblotted with anti-CDK2 antigen peptide (lane 5), we believe that both bands antibody. Bacterially produced recombinant cyclin E2 (rE2) protein was included as a positive control. (C) Total cell lysate correspond to dierently modi®ed forms of cyclin E2 in were prepared from HeLa cells and immunoprecipitated with the vivo. The nature of this modi®cation and its eect on indicated antibodies with or without prior incubation with a cyclin E2's association with other cellular proteins molar excess of a competing antigen peptide. Kinase activity of including CDK2 has not been determined. Cyclin E1 each immunoprecipitate was assayed in vitro using histone H1 as protein has been shown to be a phosphoprotein whose a substrate. (d) Activation of CDK2 and CDK3 by cyclin E1 and E2. Sf9 insect cells were infected with the baculovirus expressing faster migrating band corresponds to the active, indicated protein. Cell lysates derived from infected cells were CDK2-associated form (Ko et al., 1991). Immuno- assayed for kinase activity in vitro using histone H1 as a substrate Novel human G1 cyclin E2 M Zariwala et al 2791 remove the trans-activating activity of the GAL4BD-p21 Activation of CDK2 and CDK3 by cyclin E2 fusion protein or so-called `self-activation'), p27 and Association of cyclin E2 with CDK2 led us to p57. All three CIP/KIP inhibitors interact with cyclin determine whether it can activate CDK2 kinase E2 as determined by the two-hybrid assay in yeast activity. Anti-cyclin E2 precipitates were prepared cells, with p21 and p27 interacting with cyclin E2 more from HeLa cells and assayed in vitro for kinase strongly than p57 (Figure 3a). To further con®rm the activity using histone H1 as a substrate. Cyclin E2 speci®city of this interaction, we generated a series of (lane 3, Figure 2c), as well as cyclin E1 (lane 1) and mutants of p27 that deleted either the cyclin (p27DCYC), CDK2 (lane 5) immunocomplexes exhibit readily both the cyclin and CDK binding sequences (p27DCYC/ detected kinase activity toward histone H1 protein CDK), or the C-terminal deletion (p27[DC]) and tested that were prevented by their respective competing their ability to interact with cyclin E2. Removal of antigen peptides (lanes 2, 4 and 6), demonstrating cyclin or cyclin and CDK binding sequences, but not that cyclin E2 acts as an activating cyclin in vivo. the C-terminal portion, from p27 abolished its CDK2 can form catalytically active enzyme complexes with other cyclins, including cyclin E1 and cyclin A, and conversely cyclin E2 may assemble active kinase complexes with other CDK subunits as well. These complications prevent us from concluding that cyclin E2 can act as an activating subunit of CDK2 based on the detection of both cyclin E2 and CDK2 kinase activity in vivo. To directly demonstrate the ability of cyclin E2 in activating CDK2, we produced baculoviruses expressing cyclin E2 and CDKs and assayed for kinase activity in insect cells infected with individual viruses or their combinations. Lysates derived from insect cells infected with either cyclin E1, cyclin E2 or CDK2 alone exhibited very low background levels of H1 kinase activity (lanes 1 ± 3, Figure 2d). In contrast, lysates derived from insect cells expressing the combination of CDK2 with either cyclin E1 or cyclin E2 displayed high levels of H1 kinase activity (lanes 5 and 6), demonstrating that at least in insect cells cyclin E2, like cyclin E1, can bind to and activate the kinase activity of CDK2. In addition to CDK2, cyclin E2 also interacts strongly with CDK3 as determined by the yeast two- hybrid assay (Figure 1), leading us to determine if cyclin E2 can also activate CDK3. As shown in Figure 2d, lysates derived from insect cells co-infected with baculoviruses expressing cyclin E2 and CDK3 (lane 8) as well as cyclin E1 and CDK3 (lane 7), but not CDK3 alone (lane 4), contained high levels of H1 kinase activity, indicating that cyclin E2 may also be an activating subunit of CDK3. Lack of suitable antibodies to CDK3 at present and the low level of CDK3 expression in most cell lines (Meyerson et al., 1992) hampered us from demonstrating an in vivo cyclin E2-CDK3 association and from comparing the contribution of CDK3 activation by cyclin E1 and cyclin E2.

Cyclin E2 interacts with CIP/KIP CDK inhibitors The seven mammalian CDK inhibitors identi®ed thus far fall into two separate families that dier in both Figure 3 Interaction of human cyclin E2 with CDK inhibitors. (a) Yeast HF7c cells expressing human cyclin E2 plasmid were structure and mechanism of action. While INK4 sequentially transformed with plasmids expressing indicated CDK inhibitors selectively inhibit CDK4 and CDK6 by inhibitory proteins. Cells were streaked onto nonselective medium forming binary complexes, members of the CIP/KIP containing histidine (7LW) and then replica plated onto selective family form ternary complexes with cyclin and CDK medium lacking histidine (7LWH) and containing a 5 mM 3- aminotriazole to assay for the activation of the histidine reporter proteins through two separate N-terminally located gene. The expression of an independent reporter gene b- domains for independent binding to cyclin and CDK, galactocidase was also tested. (b) In vivo association of cyclin respectively (Russo et al., 1996). We tested interaction E2 and the CDK inhibitor p27. 293T cells were co-transfected between cyclin E2 and CDK inhibitors by the yeast with indicated plasmids expressing HA tagged cyclin E2 and myc two-hybrid assay. Cyclin E2 was sequentially trans- tagged p27. Total cell lysates were prepared from transfected cells and immunoprecipitated with antibodies speci®c to either cyclin formed individually with three inhibitors of CIP/KIP E2 or p27. After SDS ± PAGE, proteins were transferred to family, p21 (after deletion of the C-terminal portion to nitrocellulose and immunoblotted with anti-HA antibody Novel human G1 cyclin E2 M Zariwala et al 2792 interaction with cyclin E2 (Figure 3a). These results Dierential expression of cyclin E1 and cyclin E2 demonstrate that cyclin E2 speci®cally interacts with CDK inhibitors of CIP/KIP family through similar Northern blot analysis was carried out to compare sequences and mechanisms as other cyclin proteins. the expression of cyclin E2 and cyclin E1 mRNA in To examine in vivo association of cyclin E2 and 16 dierent human tissues. Under high stringency CDK inhibitor, 293T cells were transfected with conditions, the full length cyclin E1 and cyclin E2 plasmid expressing HA tagged cyclin E2 and myc- cDNA probes each detected a single discrete band of tagged p27 and lysate derived from transfected cells approximately 2.4 kb and 2.8 kb, respectively, without were immunoprecipitated with anti-cyclin E2 or anti- any apparent cross hybridization (Figure 5b). Both E p27 antibodies. After SDS ± PAGE, proteins were cyclins are expressed at nearly undetectable low levels transferred to nitrocellulose membrane and immuno- in majority of these terminally dierentiated, perma- blotted with anti-HA antibody. Cyclin E2 was readily nently arrested tissues, consistent with their oscillating detected in p27 immunocomplex (Figure 3b), demon- expression at the G1/S transition. The high levels of strating an in vivo association between cyclin E2 and expression of both the cyclin E genes are observed in CDK inhibitor. testis and placenta. In the several tissues where either one of the genes is expressed, they exhibited a distinct pattern of tissue speci®city. Cyclin E2 mRNA is Cyclin E2 is a nuclear protein highly expressed in brain and thymus and to a lesser We next examined the subcellular localization of cyclin extent in small intestine (Figure 5b) whereas nearly E2 protein. After transfection of HeLa-Tet cells with undetectable levels of cyclin E1 mRNA was observed plasmid expressing HA-tagged cyclin E2, cells were in these tissues. These observations suggest that these ®xed and immunostained with anti-HA antibody. two E-type cyclins may have distinct functions in Indirect immuno¯uorescence staining demonstrate vivo. that cyclin E2 is predominantly localized to nucleus (Figure 4). The speci®city of the HA antibody is Regulation of cyclin E2 stability by Thr-392 con®rmed by the negative staining of un-transfected phosphorylation through an LLnL and cells. MG132-sensitive pathway Cyclin E1 is an unstable protein whose stability is G1 cell cycle expression of cyclin E2 mRNA governed by ubiquitin-mediated proteasome pathway A cardinal feature of many cyclin genes is their and is regulated, at least in part, by autophosphoryla- periodic expression during the cell cycle. To determine tion on Thr-380 (Clurman et al., 1996; Won and Reed, the expression of cyclin E2 during the cell cycle, we 1996). Despite their relatively low sequence similarity chose early passage normal human foreskin ®broblasts in the C-terminal region, the Thr-380 is conserved for this analysis, as these cells can be readily arrested in between the two proteins (Thr-392 in cyclin E2, Figure a quiescent state by serum starvation and stimulated to 1c). To determine if cyclin E2 is also degraded by the proceed through one cell cycle synchronously. Cyclin proteasome pathway, we transfected HeLa-Tet cells E2 mRNA, like that of cyclin E1, was not detected in with a plasmid expressing cyclin E2 with the quiescent and early G1 cells (lane 1, 0 h point, Figure proteasome inhibitor LLnL or MG132 and deter- 5a), but became detectable as cells approached the mined the half life of cyclin E2. The half-life of G1/S boundary (12 h after serum stimulation), peaked transfected cyclin E2 is approximately 30 min and at the G1/S boundary (18 and 21 h after serum treatment of cells by both inhibitors, but not control addition), and started to decline as cells passed DMSO solvent, increased the half-life of cyclin E2 to through S phase (24 to 27 h points) reaching nearly about 120 min (lanes 1 ± 12, Figure 6a and quantitative background levels during G2/M phase and re- measurement in Figure 6b) indicating that the stability accumulating as cells entered the next G1 phase. The of cyclin E2 is governed most likely by the proteasome pattern of cyclin E2 mRNA expression is very close to pathway in vivo. that of cyclin E1 mRNA, indicating that both E cyclin To determine whether degradation of cyclin E2 is genes are expressed periodically during the cell cycle regulated by phosphorylation on a conserved threonine with a peak at the G1/S transition. residue, Thr-392, we changed Thr-392 to a non- phosphorylable Ala residue and transfected HeLa-Tet cells with either wild type or the Thr-392 mutant cyclin E2 (Cyclin E2T392A). Mutation of Thr-392 resulted in an increase of cyclin E2's half life from 30 ± 60 min. To con®rm the stabilization of cyclin E2 protein by mutation on Thr-392, lysate from HeLa-Tet cells transfected with plasmids expressing either wild type or T392A mutant cyclin E2 were immunoprecipitated with anti-cyclin E2 antibody. After SDS ± PAGE, the steady state level of each protein was determined by direct immunoblotting with anti-cyclin E2 antibody. Figure 4 Cyclin E2 localizes to the nucleus. Subcellular Consistent with half-life change, mutation of Thr-392 localization of cyclin E2 was examined by indirect ¯uorescence evidently increased the steady state level of cyclin E2 staining. HeLa-Tet cells were transiently transfected with HA- (Figure 6c). The speci®city of cyclin E2 was con®rmed cyclin E2. Twenty-four hours after transfection, cells were ®xed and stained with mouse anti-HA antibody or DAPI for DNA. by the speci®c block with a competing antigen peptide. The phase contrast of the same photographed ®eld is shown These results indicated that degradation of cyclin E2 Novel human G1 cyclin E2 M Zariwala et al 2793

Figure 5 Cell cycle and tissue speci®c expression of cyclin E1 and cyclin E2 mRNA. (a) Normal human foreskin ®broblasts (NHF) were arrested in G0 by serum deprivation and released from quiescence by serum stimulation. Progression through the cell cycle was monitored by ¯ow-cytometry analysis (top). Total RNA was prepared from cells at dierent time points after stimulation by serum readdition. Equal amounts of RNA (5 mg) of each preparation as measured by ethidium bromide staining (lower panel) were loaded on a 1% agarose gel. Resolved RNA was transferred to a nylon membrane and the blot was hybridized with a series of probes derived from the indicated human cDNAs. (b) 2 mg of poly(A)+RNA from sixteen dierent human tissues as indicated at the top of each lane were hybridized with a 2.8 kb probe derived from full length cyclin E2. The same blot was stripped and re-hybridized with a 1.5 kb cyclin E1 probe and a b-actin probe

protein by the proteasome pathway is promoted, at inactivation of the function of two tumor suppressors least in part, by phosphorylation on Thr-392. pRb and p53. To gain further insight into the regulation of cyclin E2, we examined its expression in normal human ®broblast (NHF) cells and NHF cells Dierential alteration of cyclin E1 and cyclin E2 infected with retroviruses expressing individual papillo- expression by viral oncoprotein ma viral oncoprotein E6, E7 or E6 plus E7 (Xiong et We noticed that cyclin E2 levels were signi®cantly al., 1996). While SV40 T antigen binds to both pRb elevated in two SV-40 virally transformed cells (CT10 and p53, papilloma virus oncoprotein E6 and E7 bind and VA13) as compared to their un-transformed separately to and inactivate the functions of p53 and parental cells (HSF43 and WI-38, Figure 2a), pRb, respectively, thereby allowing further dissection suggesting the possibility that the viral oncoprotein T of the potential regulation of cyclin E2. Total cell antigen may alter the regulation of the cyclin E2 gene. lysates were prepared from NHF cells infected with A critical activity encoded by SV40 virus that is shared parental retrovirus vector carrying a neomycin resistant by other DNA tumor viruses such as adenovirus and gene (NHF/Neo), NHF/E6, NHF/E7 and NHF/E6 + papilloma virus is the binding to and consequent E7 cells. The level of cyclin E2 protein as well as cyclin Novel human G1 cyclin E2 M Zariwala et al 2794

Figure 6 Degradation of cyclin E2 by proteasome pathway is regulated by Thr-392 phosphorylation. (a) Determination of half-life of cyclin E2. HeLa-Tet cells were transfected with vectors expressing either HA tagged wild type or T392A mutant cyclin E2 protein for 24 h. To determine the eect of proteasome on cyclin E2 turnover (lanes 1 ± 12), 24 h after the cells were transfected with plasmid expressing HA tagged cyclin E2, cells were treated with proteasome inhibitors LLnL, MG132 or control solvent DMSO for 3 h. Transfected cells were pulsed with [35S]methionine for 40 min and then chased for the indicated length of time. Cell lysates were precipitated with anti-HA antibody, and resolved by SDS ± PAGE and autoradiographed. The amount of each labeled protein at each time point was quantitated on a PhosphoImager, normalized relative to the amount of radiolabeled protein in cells chased for 0 h and plotted against the chase time. (b) Quantitative measurement of half life of transfected cyclin E2. (c) Steady state level of wild type and T392A mutant cyclin E2. Twenty-four hours after HeLa-Tet cells were transfected with vectors expressing either HA tagged wild type or T392A mutant cyclin E2 protein, cell lysates were prepared from transfected cells and immunoprecipitated with cyclin E2. After SDS ± PAGE, proteins were transferred to nitrocellulose and immunoblotted with anti-cyclin E2 antibody. Equal amount of cell lysate from both transfections were loaded as veri®ed by immunoblotting with anti-tubulin antibody

E2-CDK2 complexes in these cells were determined by not shown). Consistent with increased expression of IP-Western analysis (Figure 7a). Cyclin E2 protein was cyclin E2 gene in E6-expressing cells, we have also barely detectable in NHF/Neo cells (lane 1), but was found that ectopic expression of p53 in a p53-de®cient evidently increased in cells expressing E6 (lane 3) and human pharynx carcinoma cell line, FaDu, caused a even higher in cells expressing E7 (lane 5) or both E6 decrease in the level of cyclin E2 protein (JL and YX, and E7 (lane 7). Consistently, cyclin E2-associated unpublished data). CDK2 was also increased in a similar manner in cells We further determined the level of cyclin E2 mRNA expressing E6, E7 or both. Intriguingly, although the in these cells in an attempt to elucidate the mechanism level of cyclin E1 protein was similarly elevated in cells leading to an increase in cyclin E2 protein by E6 expressing E7 over levels in the parental NHF cells expression. This revealed an approximately twofold (compare lanes 5 and 1, Figure 7b), the levels of E1 and fourfold increase in cyclin E2 mRNA in E6 and E7 were actually decreased in the E6 cells (lane 3), cells respectively as compared to the Neo control cells revealing a distinct dierence in the regulation of (Figure 7d), indicating that both E6 and E7 activity these two E-type cyclins. We con®rmed the activity of may potentially aect the transcription or stability of E6 and E7 proteins in these cells by examining the level cyclin E2 mRNA. The steady state level of cyclin E2 of p21, whose expression is signi®cantly decreased by mRNA was not further increased by the expression of the expression of E6 or E6 together with E7 as the E6 and E7 together (lane 4), suggesting that E6 and E7 result of p53 inactivation (Xiong et al., 1996, and did not have a synergistic eect on the cyclin E2 Figure 7c). As a control, the level of CDK2 protein, mRNA. Notably, while cyclin E1 protein was whose regulation is not known to be altered by viral decreased in E6 cells, the level of cyclin E1 mRNA, transformation, was seen to be the same in all four like that of cyclin E2 was also increased in E6 cells cells (Figure 7c). Similar results were obtained in (Figure 7d), providing further evidence for distinct IMR90 ®broblast expressing E6, E7 and E6+E7 (data mechanisms of regulation of these two E-type cyclins. Novel human G1 cyclin E2 M Zariwala et al 2795 Discussion

In this report, we describe a novel human cyclin gene that evidently encodes a second E-type G1 cyclin, based on several characteristics. First, cyclin E2 exhibits a signi®cantly higher sequence similarity to cyclin E1, both within and outside of the cyclin box, than to any other known cyclins. Second, it exhibits functional similarity to cyclin E1: cyclin E2 binds to members of the CIP/KIP family of CDK inhibitors, and selectively activates CDK2 and CDK3. Third, the expression of cyclin E2 mRNA oscillates periodically throughout the cell cycle, peaking at the G1/S transition and declining during S phase with very similar, if not the same kinetics as cyclin E1. Fourth, cyclin E2 is, like E1, a nuclear protein. Lastly, cyclin E2 enclodes a short lived protein whose turnover is governed by the proteasome pathway, and whose stability is increased by mutation in a threonine residue conserved between cyclin E1 and E2 (Thr- 380/392) as with cyclin E1. Isolation of a second E-type cyclin that shares similar properties with cyclin E1 calls for re- evaluation of previous studies in which activation of CDK2 kinase have been attributed solely to cyclin E1. In a number of cell lines that we have analysed (e.g. several commonly used ®broblasts and their virally transformed derivative strains, Figures 2 and 5), the expression of cyclin E2 mRNA and protein were detected at levels similar to or higher than that of cyclin E1. While conclusions drawn from experiments involving ectopic expression of cyclin E1 remain the same, interpretation of ®ndings from interference-based experiments such as antisense expression and antibody injection may be aected. The presence of two closely related cyclin E proteins also raises question concerning frequently observed high level expression of cyclin E in human tumors. We have found that some polyclonal antibodies raised against full length cyclin E1 cross reacted eciently with the closely migrating cyclin E2 protein (MZ, unpublished observation), and have noted that some studies using either monoclonal or polyclonal anti-cyclin E antibodies raised against full length cyclin protein suspiciously detected multiple `cyclin E' bands (e.g. Keyomarsi et al., 1997). These observations raise questions about whether previously observed increases in cyclin E1 expression in human Figure 7 Alteration of cyclin E levels by papilloma viral tumors might actually represent an increase in both oncoproteins. (a) Expression of cyclin E2 protein. Normal cyclin E proteins or even in cyclin E2 alone. This human ®broblasts (NHF) containing the neomycin resistance gene (Neo) or NHF cells expressing type 16 HPV E6, E7 and question is particularly pertinent in the light of our both 16E6 and 16E7 were described before (White et al., 1994; ®nding that expression of the viral oncoprotein E6, Xiong et al., 1996). Total cell lysates were prepared from which can bind to and inactivate the function of p53, individual cells and immunoprecipitated with anti-cyclin E2 increased the expression of cyclin E2, but not cyclin antibody with or without incubation with a molar excess of E1. competing antigen peptide as indicated on top. After SDS ± PAGE, proteins were transferred to nitrocellulose, and the upper The co-existence of two closely related E-type portion was immunoblotted with anti-cyclin E2 antibody and the cyclins within the same cell raises the question of lower portion of the same blot was immunoblotted with anti- whether they are fully functional redundant or have CDK2 antibody. (b) Expression of cyclin E1 protein. The distinct functions. Thus far, most of our studies have experiment is the same as in a except anti-cyclin E2 antibody was replaced with anti-cyclin E1 antibody. (c) Expression of p21 only revealed the functional similarities between these CDK inhibitor and CDK2. The same cell lysates as described in a were analysed for the expression of p21 and CDK2 by direct immunoblotting. (d) Expression of cyclin E1 and cyclin E2 mRNA. Total RNA was prepared from the same set of cells as transferred to a nylon membrane and the blot was hybridized ®rst described in a. Equal amounts of RNA (ten micrograms) of each with a probe derived from full length human cyclin E2 gene. The preparation as measured by ethidium bromide staining (lower same blot was stripped and re-hybridized with a probe panel) were loaded on a 1% agarose gel. Resolved RNA was corresponding to the human cyclin E1 gene Novel human G1 cyclin E2 M Zariwala et al 2796 two proteins, but two lines of evidence suggest a Materials and methods functional or regulatory distinction between them. First, the expression of cyclin E1 and cyclin E2 cDNA clones, plasmids constructs and yeast two hybrid assay exhibit distinctly dierent tissue speci®city (Figure 5), Catalytic inactive mutants of human CDK2K33M, suggesting that dierent cell types may preferentially CDK3K33M, CDK5K33M, CDK6K43M and mouse CDK4K35M express and utilize one of these two E cyclins. For were generated by PCR-mediated site-directed mutagenesis example, cyclin E2 mRNA was highly expressed in and con®rmed by DNA sequencing. The C-terminal thymus and brain, yet no detectable cyclin E1 mRNA domain deletion (residue 104 ± 164) mutant of human was seen in either tissue, whereas the levels of cyclin p21, human p27 cyclin binding domain deletion (1 ± 42, E1 mRNA in placenta appear to be considerably p27DCYC) mutant, cyclin and CDK binding domain double DCYC/DCDK higher than that of cyclin E2. Notably, while the deletion (residue 1 ± 70, p27 ) mutant, and C- terminal domain deletion (residue 94 ± 198) mutant were cyclin E genes are mainly detected in tissues that generated. For the yeast two-hybrid assay, wild type or contain actively dividing cells such as thymus, mutant CDK and CDK inhibitor genes were individually placenta, and testis, cyclin E2 was expressed in inserted into the yeast two-hybrid assay vector (pGBT8) as brain, a tissue that contains predominantly postmito- in-frame fusion with yeast GAL4 DNA binding protein. tic cells, perhaps suggesting a function outside of cell For yeast two-hybrid screen, pGBT8-CDK2K33M bait and a cycle control. Second, while the expression of both human cDNA library derived from immortalized human cyclin E1 and E2 are elevated by the expression of keratinocyte cell line HaCaT were sequentially transformed viral oncoprotein E7, they exhibit opposite responses into yeast HF7c cells. An estimated 36106 transformants to the expression of the E6 protein, with cyclin E1 were screened. being downregulated and cyclin E2 increased. For expression in mammalian cells, individual cDNA clones were subcloned into the pcDNA3 vector under the Although the molecular basis for the increase of control of the CMV promoter (Invitrogen), or its derivatives cyclin E2 by E6 expression is currently unknown, it pcDNA3-HA and pcDNA3-Myc, for expressing the HA or is tempting to speculate that such dierent responses myc epitope tagged fusion protein. For expression in may re¯ect the dierence in the regulation of two bacterial cells, cyclin E2 was subcloned into the pET vector cyclin E genes by p53 function. (Novagen). It has been well established that cyclin E1 is a downstream target of E2F transcriptional regulation Cell lines, culture conditions and cell transfection (Duronio and O'Farrell, 1995; Ohtani et al., 1995; Degregori et al., 1995; Botz et al., 1996; Geng et al., All mammalian cells were cultured in a 378C incubator with 5% CO . HeLa is human cervix epithelioid carcinoma 1996). As a result, the expression of cyclin E1 is 2 cell line and HeLa-Tet-O, a derivative of HeLa cell line enhanced in cells de®cient for pRB function. The (Clontech, Inc.), and is simply referred to as HeLa-Tet. U- results presented here indicate that it is likely that 2 OS and Saos-2 are human osteosarcoma cell lines and cyclin E2 is similarly regulated by E2F. In several cell 293T is SV40 transformed human kidney cell line. lines transformed by DNA tumor viruses such as the Characterization of CDK and cyclin complexes in normal SV-40 transformed VA13 and CT10 cell lines, both of and virally transformed human ®broblasts strains HSF43, which express the pRb-inactivating T antigen, we CT10, WI-38 and VA13 was described in (Xiong et al., observed a signi®cant increase in both cyclin E2 1993), and characterization of normal human foreskin mRNA and protein. Such an increase in cyclin E2 ®broblast line NHF and its E6, E7 or E6 and E7 expressing expression does not appear to be speci®cally dependent derivatives were described earlier (White et al., 1994; Xiong on viral transformation, as we have also seen a high et al., 1996). Cell transfections were carried out using the LipofectAMINE reagent according to the manufacturer's level cyclin E2 expression in several non-virally instructions (Gibco ± BRL). Procedures for pulse-chase transformed, tumor-derived cell lines, Rb7 (and experiment were as described previously (Zhang et al., p537) de®cient cells such as osteosarcoma cell line 1998). Insect Sf9 cells were cultured in suspension in SF900 Saos-2, but not in another osteosarcoma cell line U2OS II serum free medium (Gibco-BRL) containing that retain wild type Rb (and possibly p53) function 0.56penicillin/streptomycin. (Figure 2a). More direct evidence for a Rb-E2F Synchronization of NHF cells was achieved by serum regulation of cyclin E2 expression comes from the starvation and stimulation. NHF cells were cultured in observation that introduction of papilloma viral DMEM containing 15% FBS to a 30 ± 40% density, starved oncogene E7 alone, whose main activity is believed to in DMEM containing 0.5% FBS for 3 days and stimulated bind to and inactivate the function of Rb family by switching to DMEM containing 15% FBS. Progression through the cell cycle was monitored by ¯ow cytometry proteins, led to a substantial increase in cyclin E2 analysis. Procedures for isolation of total RNA and Northern expression (Figure 7). hybridization have been described before (Li et al., 1994). For Potentially signi®cant is the ®nding that the multiple tissue Northern analysis, 2 mg poly(A)+ RNA were expression of cyclin E2, but not cyclin E1, is isolated from dierent human tissues, resolved on a 1.2% increased in cells expressing papilloma viral oncogene agarose gel and transferred to a nylon membrane (Clontech, 16E6 (Figure 7). Although E6 protein may carry out Inc.). other activities in addition to inactivating p53, one obvious implication of this ®nding would be the Antibodies and immunochemistry procedures negative regulation of cyclin E2, but not cyclin E1, by 35 p53. A scenario could be envisioned in which an Procedures for S-methionine metabolic labeling, immu- noprecipitation and immunoblotting have been described increase of p53 in response to growth inhibitory previously (Xiong et al., 1993; Jenkins and Xiong, 1995). signals would downregulate the expression of cyclin Anti-CDK2 antibody is described earlier (Xiong et al., E2 and therefore contribute to G1 arrest in cells 1993). Anti-p27 polyclonal antisera was raised against whose G1 progression is largely dependent on the bacterially produced full length human p27 protein. Both function of cyclin E2. anti-cyclin E1 and anti-cyclin E2 antibodies were raised in Novel human G1 cyclin E2 M Zariwala et al 2797 rabbits using a synthetic peptide as an immunogen. The Indirect immuno¯uorescence sequence of synthetic peptides used are as follows Approximately 70% con¯uent HeLa-Tet cells were seeded (underlined cysteine residue was added to facilitate onto a 6-well plate and transfected with 1 mg of the coupling of peptide to carried KLH): anti-human cyclin appropriate plasmid DNA. All subsequent steps were E2, CMTPPKSTEKPPGKH (amino acid residues 391 ± 404 carried out at room temperature. Twenty-four hours after at C-terminus of human cyclin E2); anti-human cyclin E1. transfection, cells were washed three times with PBS and CPPQSGKKQSSGPEMA (residues 381 ± 395 at C-termi- ®xed in PBS containing 3% formaldehyde for 15 min. nus of human cyclin E1). All rabbit polyclonal antibodies After ®xation the cells were washed once with PBS and used in this study were anity puri®ed using respective permeabilized in cold PBS containing 0.2% Triton X-100 peptide columns following manufacturers instruction for 5 min. The cells were incubated with PBS containing (Sulfolink Kit, Pierce, Rockford, IL, USA). Monoclonal 0.5% BSA as a blocking agent for 30 min followed by anti-HA (12CA5, Boehringer-Mannheim), anti-myc (9E10, incubation with the blocking buer containing anity- NeoMarker), anti-tubulin (NeoMarker), Fluorescein-con- puri®ed anti-HA monoclonal antibody (1 : 30 dilution of jugated goat anti-mouse IgC (Jackson ImmunoResearch 12CA5 hybridoma culture supernatant) for 1 h. The cells Laboratory) was purchased commercially. Coupled in vitro were then incubated with Fluorescein (FITC)-conjugated transcription and translation reactions were performed anity-puri®ed goat anti-rabbit IgG or Rhodamine- using the TNT kit following the manufacturer's instruc- conjugated anity-puri®ed goat anti-mouse IgG for tions (Promega). Protein lysate concentrations were 30 min. Between each step of antibody incubation, cells determined by Bradford assay and equalized for each were washed three times with PBS. Cells were then experiment. Kinase assays in immunoprecipitates or in incubated with PBS containing DAPI stain (1 : 1000 baculovirus infected insect cell lysate were performed as dilution) for 1 min. Stained cells were examined with an previously described (Guan et al., 1994; Phelps and Xiong, Olympus IX70 microscope ®tted with appropriate fluores- 1997). cence ®lters.

Pulse-chase experiment Acknowledgements Approximately 46105 HeLa-Tet cells were seeded onto a We thank Dr Burno Amati and Steve Coats for commu- 60 mm plate and were transfected after overnight culture nicating results prior to publication. We also thank Tomo with a total of 4 mg of appropriate plasmid DNA for 24 h. Ohta for providing cell cycle Northern blots, Neil Coeld To determine the regulation of protein turnover by for p27 constructs and Chris Jenkins for isolation of RNA proteasome, LLnL (N-acetyl-leucinyl-leucinyl-norleucinal, to make the HaCaT library, database search, and critical 100 mM dissolved in 15 ml DMSO) or MG132 (carboben- reading of the manuscript. YX is a recipient of American zoxyl-leucinyl-leucinyl-leucinal, 50 mM dissolved in 15 ml Cancer Society Junior Faculty Award and a Pew Scholar in DMSO) or the same volume of DMSO solvent (15 ml) were Biomedical Science. This study was supported by Public added to the transfected cells during the last 3 h of Health Service grant CA65572 to YX. transfection. Cells were pulse labeled with 35S-methionine for 40 min, washed twice with pre-warmed 16PBS and chased by culturing in DMEM/10% FBS media for the Note added in proof time indicated in each ®gure. Lysates from pulse-chase The nucleotide sequence data reported in this paper will labeled cells were immunoprecipitated with anti-HA appear in the database under the GenBank accession antibody and resolved by SDS ± PAGE. number AF102778.

References

Botz J, Zerfass-Thome K, Spitkovsky D, Delius H, Vogt B, Hannon GJ, Casso D and Beach D. (1994). Proc. Natl. Acad. Eilers M, Hatzigeorgiou A and Jansen-Durr P. (1996). Sci. USA, 91, 1731 ± 1735. Mol. Cell Biol., 16, 3401 ± 3409. Howe JA and Newport JW. (1996). Proc. Natl. Acad. Sci. Chevalier S, Couturier A, Chartrain I, Le Guellee R, USA, 93, 2060 ± 2064. Beckhelling C, Le Guellec K, Philippe M and Ford CC. Hunter T and Pines J. (1994). Cell, 79, 573 ± 582. (1996). J. Cell Sci., 109, 1173 ± 1184. Jenkins CW and Xiong Y. (1995). In: Cell Cycle: Material Clurman B, Shea RJ, Thress K, Groudine M and Roberts and methods Pagano M (ed.). Springer-Verlag: New York. JM. (1996). Genes & Dev., 10, 1979 ± 1990. pp. 250 ± 263. Degregori J, Kowalik T and Nevins, JR. (1995). Mol. Cell Keyomarsi K, O'Leary N, Molnar G, Lees E, Fingert HJ and Biol., 15, 4215 ± 4224. Pardee HB. (1997). Cancer Res., 54, 380 ± 385. Dou QP, Pardee AB and Keyomarsi K. (1996). Nat. Med., 2, Ko A, Cross F, Fisher A, Schumacher J, Leguellec K, 254. Philippe M and Roberts J. (1991). Cell, 66, 1217 ± 1228. Dulic V, Lees E and Reed SI. (1992). Science, 257, 1958 ± Ko A, Giordano A, Desai D, Yamashita K, Harper JW, 1961. Elledge S, Nishimoto T, Morgan DO, Franza BR and Duronio RJ and O'Farrell PH. (1995). Genes & Dev., 9, Roberts JM. (1992). Science, 257, 1689 ± 1694. 1456 ± 1468. Koh J, Enders GH, Dynlacht BD and Harlow E. (1995). Duronio RJ, Brook A, Dyson N and O'Farrell PH. (1996). Nature, 375, 506 ± 510. Genes Dev., 10, 2505 ± 2513. Lew D, Dulic V and Reed SI. (1991). Cell, 66, 1197 ± 1206. Geng Y, Eaton EN, Picon M, Roberts JM, Lundberg AS, Li Y, Jenkins CW, Nichols MA and Xiong Y. (1994). Giord A, Sardet C and Weinberg RA. (1996). Oncogene, Oncogene, 9, 2261 ± 2268. 12, 1173 ± 1180. Lukas J, Bartkova J, Rohde M, Strauss M and Bartek J. Guan K-L, Jenkins CW, Li Y, Nichols, MA, Wu X, O'Keefe (1995a). Mol. Cell Biol., 15, 2600 ± 2611. CL, Matera AG and Xiong Y. (1994). Genes & Dev., 8, Lukas J, Parry D, Aagaard L, Mann DJ, Bartkova J, Strauss 2939 ± 2952. M, Peters G and Bartek J. (1995b). Nature, 375, 503 ± 506. Gyuris J, Golemis E, Chertkov H and Brent R. (1993). Cell, Matsushime H, Roussel MF, Ashmum RA and Sherr CJ. 75, 791 ± 803. (1991). Cell, 65, 701 ± 713. Hall M and Peters G. (1996). Cancer Res., 68, 1173 ± 1180. Novel human G1 cyclin E2 M Zariwala et al 2798 Matsushime H, Quelle DE, Shurtle SA, Shibuya M, Sherr Resnitzky D, Gossen M, Bujard H and Reed S. (1994). Mol. CJ and Kato JY. (1994). Mol. Cell. Biol., 14, 2066 ± 2076. Cell Biol., 14, 1669 ± 1679. Medema RH, Herrera RE, Lam F and Weinberg RA. (1996). Richardson HE, Stueland CS, Thomas J, Russell P and Reed Proc. Natl. Acad. Sci. USA, 92, 6289 ± 6293. SI. (1990). Genes Dev., 4, 1332 ± 1344. Meyerson M, Enders GH, Wu C-L, Su L-K, Gorka C, Russo AA, Jerey PD, Patten AK, Massague J and Pavletich Neilson C, Harlow E and Tsai L-H. (1992). EMBO J., 11, NP. (1996). Nature, 382, 325 ± 331. 2909 ± 2917. Serrano M, Gomez-Lahoz E, DePinho RA, Beach D and Meyerson M and Harlow E. (1994). Mol. Cell Biol., 14, Bar-Sagi D. (1995). Science, 267, 249 ± 252. 2077 ± 2086. Sewing A, Ronicke V, Burger C, Funk M and Muller R. Mumberg D, Wick M, Burger C, Haas K, Funk M and (1994). J. Cell Sci., 107, 581 ± 588. Muller R. (1997). Nucleic Acids Res., 25, 2098 ± 2105. Sherr C. (1993). Cell, 73, 1059 ± 1065. Ohtani K, Degregori J and Nevins J.R. (1995). Proc. Natl. Sherr CJ. (1994). Cell, 79, 551 ± 555. Acad. Sci. USA, 92, 12146 ± 12150. Sherr CJ. (1996). Science, 274, 1672 ± 1677. Ohtsubo M, Theodoras AM, Schumacher J, Roberts J and White A, Livanos EM and Tlsty TD. (1994). Genes & Dev., 8, Pagano M. (1995). Mol. Cell Biol., 15, 2612 ± 2624. 666 ± 677. Ohtsubo M and Roberts JM. (1993). Science, 259, 1908 ± Won K-A and Reed SI. (1996). EMBO J., 15, 4182 ± 4193. 1912. Xiong Y, Zhang H and Beach D. (1993). Genes & Dev., 7, Pagano M, Pepperkok R, Lukas J, Baldin V, Ansorge W, 1572 ± 1583. Bartek J and Draetta G. (1992). J. Cell Biol., 121, 101 ± Xiong Y, Kuppuswamy D, Li Y, Livanos EM, Hixon M, 111. White A, Beach D and Tlsty TD. (1996). J. Virol., 70, Phelps D and Xiong Y. (1997). Methods Enzymol., 283, 194 ± 999 ± 1008. 205. Zhang Y, Xiong Y and Yarbrough WG. (1998). Cell, 92, Reed SI. (1992). Annu. Rev. Cell Biol., 8, 529 ± 561. 725 ± 734. Rempel RE, Sleight SB and Maller JL. (1995). J. Biol. Chem., 270, 6843 ± 6855.