The Oncogenic Activity of Cyclin E Is Not Confined to Cdk2 Activation Alone but Relies on Several Other, Distinct Functions of the Protein*

We have previously shown that cyclin E can malignantly transform primary rat embryo fibroblasts in cooperation with constitutively active Ha-Ras. In addition, we demonstrated that high level cyclin E expression potentiates the development of methyl-nitroso-urea-induced T-cell lymphomas in mice. To further investigate the mechanism underlying cyclin E-mediated malignant transformation, we have performed a mutational analysis of cyclin E function. Here we show that cyclin E mutants defective to form an active kinase complex with Cdk2 are unable to drive cells from G1 into S phase but can still malignantly transform rat embryo fibroblasts in cooperation with Ha-Ras. In addition, Cdk2 activation is not a prerequisite for the ability of cyclin E to rescue yeast triple cln mutations. We also find that the oncogenic properties of cyclin E did not entirely correspond with its ability to interact with the negative cell cycle regulator p27Kip1 or the pocket protein p130. These findings suggest that the oncogenic activity of cyclin E does not exclusively rely on its ability as a positive regulator of G1 progression. Rather, we propose that cyclin E harbors other functions, independent of Cdk2 activation and p27Kip1 binding, that contribute significantly to its oncogenic activity.

We have previously shown that cyclin E can malignantly transform primary rat embryo fibroblasts in cooperation with constitutively active Ha-Ras. In addition, we demonstrated that high level cyclin E expression potentiates the development of methyl-nitroso-urea-induced T-cell lymphomas in mice. To further investigate the mechanism underlying cyclin E-mediated malignant transformation, we have performed a mutational analysis of cyclin E function. Here we show that cyclin E mutants defective to form an active kinase complex with Cdk2 are unable to drive cells from G 1 into S phase but can still malignantly transform rat embryo fibroblasts in cooperation with Ha-Ras. In addition, Cdk2 activation is not a prerequisite for the ability of cyclin E to rescue yeast triple cln mutations. We also find that the oncogenic properties of cyclin E did not entirely correspond with its ability to interact with the negative cell cycle regulator p27 Kip1 or the pocket protein p130. These findings suggest that the oncogenic activity of cyclin E does not exclusively rely on its ability as a positive regulator of G 1 progression. Rather, we propose that cyclin E harbors other functions, independent of Cdk2 activation and p27 Kip1 binding, that contribute significantly to its oncogenic activity.
E-type cyclins (cyclin E and cyclin E2) and their associated catalytic subunits Cdk2 constitute part of the regulatory proteins that control progression through the G 1 /S border and the entry of cells into S phase of the cell cycle. Both cyclins were found to be expressed in the late G 1 phase through S phase, and it has been shown that the level of cyclin E is the limiting factor for G 1 phase progression and S phase entry (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11). One of the important events in this phase of the cell cycle is the phosphorylation of the tumor suppressor protein Rb, which is one of the substrates of E-type cyclin-Cdk complexes (12)(13)(14). The Rb protein is active in its nonphosphorylated or hypophosphorylated form and becomes hyperphosphorylated and inactivated by cyclin-Cdk complexes during G 1 phase (12)(13)(14). This inactivation by phosphorylation leads to the release of E2F family transcription factors from a complex with the Rb protein. E2F transcription factors in turn activate a set of target genes that are essential for entry into S phase (15)(16)(17). Cyclin-Cdk complexes are regulated by a number of inhibitors called Cdk inhibitors, among them the proteins p16 Ink4a , p21 Waf1 , and p27 Kip1 (18 -24). Both p21 Waf1 and p27 Kip1 bind to cyclin E-Cdk2 complexes and block their activities, whereas p16 Ink4a specifically inhibits the catalytic subunits of D-type cyclins Cdk4 and Cdk6 (25).
A link between cyclin E and malignant transformation has been established by a number of independent studies. Cyclin E was found to be expressed at high levels in a number of human tumors (for an overview see Refs. 26 and 27), and the overexpression of the cyclin E protein is in many cases the result of an amplification of the locus of the cyclin E gene at chromosome 19q12-13 (27)(28)(29)(30)(31). In particular, high levels of cyclin E and low levels of p27 Kip1 have been correlated with poor prognosis in breast cancer (32)(33)(34), and one study with transgenic mice suggested a functional role of elevated cyclin E levels in malignant transformation of mammary tissue (35). In addition, in several types of leukemias, in chronic lymphocytic leukemia, Hodgkin's and non-Hodgkin's lymphoma cyclin E overexpression was detected and appeared to correlate with low rates of complete remission and disease free survival (36,37).
We have previously shown that cyclin E can cooperate with constitutively active Ha-Ras in the malignant transformation of primary rat embryo fibroblasts (REFs) 1 (38). Moreover, we have established a transgenic mouse model in which high levels of ectopically expressed cyclin E was targeted to the T-lymphoid lineage. These animals were significantly more susceptible to the induction of T-cell lymphomas upon treatment with methyl-nitroso-urea (39). To explore in more detail the molecular mechanisms by which cyclin E overexpression can lead to malignant transformation, we have used cyclin E mutants and the REF transformation assay. We demonstrate here that the oncogenic activity of cyclin E relies on different domains of the protein, which are distinct in their function to mediate binding and/or activation of Cdk2, p27 Kip1 , and p130 and their ability to drive cells into S phase.

EXPERIMENTAL PROCEDURES
Plasmid Construction-cDNAs for cyclin E, cyclin E mutants, and Myc were subcloned into a pLTR⌬7bp vector (40), which was cut, PstI/T4-DNA-polymerase-blunted, and modified with a part of the multiple cloning site of pBluescript (Stratagene), that was cut NotI/Asp 718I/Klenow-blunted, to gain pLTR-MCSϩ. The vector pEGFP-N3 (CLONTECH) was cut with BamHI/XbaI, and oligonucleotides encoding the FLAG epitope (GATCCATGGACTACAAAGACGATGACGATA-AATAGTCTAGAGGC and CTAGGACTATTTATCGTCATCGTCTTTG-TAGTCCATG) were inserted into these sites to replace the enhanced green fluorescent protein gene and gain a new vector, pFLAG-N3, which expresses the gene of interest with a C-terminal FLAG epitope under the control of the cytomegalovirus promoter. Cyclin E mutants were PCR-amplified with the respective primers to generate 3Ј-ends of the cDNAs without stop codons and subsequently subcloned into pFLAG-N3 and pEGFP-N3 to obtain cDNAs coding for C-terminal fusion proteins. pM and pVP16 plasmids (CLONTECH) were cut with EcoRI/HindIII and modified with oligonucleotides encoding a new multiple cloning site (AATTCCTCGAGGGATCCTGCAGATCTA and AGCTTAGATCTGACGGATCCCTCGAGG) to gain pM-MCSnew and pVP16-MCSnew. The cDNAs for WT cyclin E and the cyclin E mutants were subcloned into pVP16 or pVP16-MCSnew. cDNAs for p27 Kip1 , Cdk2, Cdk3, and Cdk4 were subcloned into pM or pM-MCSnew. All cDNAs used in this study are human in origin. (All primer information is available upon request.) Rat Embryo Fibroblast Transformation Assay-Essentially, experiments were carried out as described earlier (38,40,41). Pregnant Fisher rats were killed at day 14 after mating, the embryos were removed, and carcasses were prepared in sterile PBS. Cells from the carcasses were isolated using trypsin, EDTA, 1% serum. The cells were separated from debris using low speed centrifugation and subsequently plated in Dulbecco's modified Eagle's medium, 10% fetal calf serum at 3 ϫ 10 6 cells/145-cm 2 dish. After the cells had reached confluence, they were either frozen for storage or replated at 1 ϫ 10 6 per 56.7-cm 2 dish and used for transfection 6 h later. Transfections were carried out using the conventional calcium phosphate precipitation method overnight with 30 g of plasmid DNA per dish. This included 15 g of pGEJ6.6 (41) (containing constitutively active Ha-ras (Ha-Ras-V12) from a human bladder carcinoma) and 15 g of the respective pLTR-MCSϩ plasmid, carrying the cDNA for the cooperating oncogene (Myc as positive control). 16 h after the start of transfection, cells were washed with PBS, and fresh medium was supplied for another 24 h. Cells were then washed with PBS, trypsinized, and replated on four dishes and assayed for focus formation after 12-14 days.
Mammalian Two-hybrid Analysis-NIH 3T3 and HeLa cells were seeded in 12-well dishes (50,000 cells/well) and transfected using a standard Ca 3 (PO 4 ) 2 technique with 1 g of plasmid DNA/well consisting of plasmids pM, pVP16, 5ϫGal4-E1b-luc, and pZeolacZ (Invitrogen) in the ratio 5:5:1:5, respectively. 16 h after transfection, cells were washed with PBS, and fresh medium was supplied for another 24 h. Cells were then washed with PBS and lysed in 100 l of lysis buffer (100 mM Tris, 10 mM magnesium acetate, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100). 10 l were subjected to measurement of the luciferase activity using a Lumat LB9507 luminometer (EG & G Berthold). As an internal standard, the activity of ␤-galactosidase coded for by pZeolacZ was determined using the same lysates and the ␤-galactosidase reporter gene assay kit (Roche Diagnostics). Luciferase RLU were normalized to ␤-galactosidase RLU. As a positive control, a Gal4dbd-VP16-fusion protein was expressed from pM3-VP16-plasmid (CLONTECH). Each experiment was performed at least three times in triplicate and in both cell lines. The correct expression of Gal4-dbd and VP16 fusion proteins was controlled by Western blotting using ␣-VP16, ␣-Gal4-dbd, and ␣-cyclin E antibodies (14-5, RK5C1, and HE12, respectively (all from Santa Cruz Biotechnology, Inc., Santa Cruz, CA) and ␣-cyclin E (Biozol)).
Subcellular Localization Analysis-NIH 3T3 cells were seeded on glass coverslips, transfected using a standard Ca 3 (PO 4 ) 2 method with pEGFP-N3 plasmids and either pEBFP-N3, pDsred-N3, or pCMV plasmids in an equal ratio. 16 h after transfection, cells were washed with PBS, and fresh medium was supplied for another 24 h. Cells were then washed with PBS and fixed using 1% formaldehyde, 0.2% glutaraldehyde in PBS for 20 min. After subsequent washing with PBS and nuclear staining with 4Ј,6-diamidino-2-phenylindole, coverslips were mounted and analyzed using a confocal laser-scanning microscope (LSM510) and software (both from Zeiss).
Flow Cytometry-Cell cycle distribution of cells, transiently transfected with pFLAG-N3 expression plasmids and pBB14 (Us9-GFP) as transfection marker (42), were analyzed using a flow cytometer. For cell cycle analysis with propidium iodide staining, transfected cells were trypsinized, pelleted, washed once with cold PBS, and fixed with 70% ethanol and 30% PBS overnight at 4°C. Cells were collected, incubated in PBS with RNase A (10 g/ml) and 2 g/ml propidium iodide (both from Sigma) for 30 min at room temperature in the dark, and subsequently analyzed using a FACSCalibur and Cellquest software (both from Becton Dickinson). The transfected population of cells was gated by choosing the highly GFP-positive cells that overexpress the membrane-residing Us9-GFP marker protein.
Yeast Toxicity of Cyclin E and Cyclin E Mutants-Toxicity in yeast was measured by transforming YPH 499 yeast (Mata, ura3-52, las2-801, ade2-101, trp1-⌬63, his3-⌬200leu2⌬1 (Stratagene)) with plasmids (pRS416Met25) that conferred independence of uracil and the gene of interest, here the different cyclin E mutants, under the regulatable Met 25 promoter (2,43,44). By plating on selective medium with or without methionine, the toxic phenotype of the expressed cyclin E variant could be observed by growth or absence of growth depending on methionine. To investigate the ability to rescue a triple cln mutant strain of Saccharomyces cerevisiae ⌬L1 (Mat␣, ade1, his2, leu2-3, ura3, cln1::TRP1, cln2, cln3, GAL1::Cln2 (2)), these cells were transformed with the above mentioned regulatable expression plasmids and plated on glucose (restrictive) or galactose (permissive) containing selective media with 1 mM methionine. The ability of cyclin E to rescue the triple Cln-deficient phenotype could be observed under restrictive conditions on glucose medium, when empty vector-transformed ⌬L1 yeasts do not survive.

RESULTS
To determine whether a specific domain or function of cyclin E correlates with its activity to transform REFs in cooperation with an oncogenic, constitutively active Ha-Ras (Ha-Ras-V12) (45), we constructed a series of cyclin E mutants in a modified pLTR expression vector (40) that would allow the expression of mutants that lack N-terminal, C-terminal, or central parts of the protein. These constructs were cotransfected into primary REFs together with a plasmid driving the expression of the oncogenic Ha-ras gene, and the formation of foci of transformed cells was scored. Compared with the full-length protein, mutants of cyclin E lacking up to 45 C-terminal or 129 N-terminal amino acids still produced significant numbers of foci that were able to grow as permanent cell lines (mutants 4 -8; Fig. 1a). A similar situation was found for mutants 10 and 11, which lacked part of or the whole cyclin box (Fig. 1a). Mutant 13, which carries an alanine residue instead of a threonine at position 380 and is no longer degradable via the phosphorylation-induced ubiquitin pathway, showed oncogenic activity that was similar to the WT full-length cyclin E in this assay (Fig. 1a). Only mutants 1-3 and mutant 9 containing aa 1-215, aa 1-250, aa 1-300, or aa 219 -395 lost their transforming potential and produced very few foci in REF assays that, in contrast to those generated with the longer mutants, could not b, protein extracts from foci that have been established as a stable cell line were separated by SDS-PAGE, transferred onto membranes, and probed with an ␣-cyclin E antibody. The presence of the mutant proteins in these extracts indicates that the respective mutant constructs were functional. a, cyclin E mutants that were unable to provoke the outgrowth of foci that could be established as a cell line were transiently transfected into NIH 3T3 cells. Cell extracts were separated by SDS-PAGE and after transfer on a solid support were developed with an ␣-cyclin E antibody. All mutant constructs were able to direct the expression of the respective proteins at comparable levels, demonstrating the functionality of the constructs. We observed that in many cases the expression of cyclin E and its mutants led to the appearance of a predominant band of the expected molecular weight in Western blots (arrowheads), as well as different faster migrating immunoreactive bands, suggesting either posttranslational modification, degradation, or alternative translation from a second start codon (see also Figs. 2 and 5). be established as permanent lines (Fig. 1a). Mutant 12, which carries a point mutation at aa 130 and is unable to bind Cdk2, did not give rise to any foci (Fig. 1a). Analysis of extracts from cell lines that were established from foci demonstrated that all oncogenic cyclin E mutants were expressed at similar levels at the RNA or the protein level ( Fig. 1, a and b). In addition, the constructs that were unable to provoke focus formation upon cotransfection with Ha-ras-V12 were also functional and able to express the respective cyclin E mutants in murine fibroblasts (Fig. 1c).
The best characterized function of cyclin E is its role in the regulation of cell cycle progression from the G 1 to the S phase. Here, cyclin E interacts with Cdk2 and activates its kinase activity to phosphorylate substrates, the most important of them being the pocket protein Rb and p130 (47). To test which part of the cyclin E protein is required for this task, all cyclin E deletion mutants were inserted in a vector that allowed expression of the mutants as FLAG-tagged versions. These constructs were transiently transfected into HeLa cells, and the mutant proteins were precipitated with immobilized ␣-FLAG antibodies (M2-agarose). The precipitates were mixed with histone H1 or glutathione S-transferase-Rb fusion protein, and kinase assays were performed. The protein complexes from this reaction were separated by SDS-PAGE and revealed that the cyclin E WT protein and mutants 5, 6, and 13 were able to phosphorylate both substrates, whereas all other mutants failed to do so (Fig. 2). This indicated that already small changes in the cyclin E protein abrogate its ability to form a detectable active kinase complex with Cdk2. To test which cyclin E mutants can still form stable complexes with Cdk2, extracts from transformed REF cell lines that were established from foci and overexpress different cyclin E mutants were subjected to imunoprecipitations with ␣-Cdk2 antibodies. Western blot analysis of the precipitated complexes revealed that WT cyclin E as well as mutants 4, 5, 6, 7, 8, and 13 were coprecipitating and thus interacting with Cdk2 (Fig. 2c).
To be able to better quantify the binding of cyclin E mutants to Cdk2 and to detect weaker interactions, we performed a mammalian two-hybrid assay. For this assay, we generated expression constructs that allowed the production of cyclin E and its mutants as fusion proteins with the transactivating domain of VP16. The cDNAs for three known binding partners of cyclin E, namely Cdk2, p27 Kip1 , and the pocket protein p130, were expressed as fusions with a Gal4 DNA-binding domain (Gal4-dbd). The interaction between cyclin E or different cyclin E mutants and the respective binding partners was measured through the transcriptional activation of a luciferase reporter construct regulated by five Gal4-dbd-binding sites in the promoter (Fig. 3a). The correct expression of the VP16-cyclin E and the Gal4-dbd fusion constructs was determined by Western blotting (Fig. 3b and data not shown). Clearly, the cyclin E WT protein and the mutants 5 and 6 that still are able to activate Cdk2 showed the strongest interaction with Cdk2 and with all other investigated partner proteins (Fig. 3a). In addition, the cyclin E mutant 4 and to a lesser extent mutants 7 and 8 can still interact with Cdk2 (Fig. 3a), confirming our results from the immunoprecipitation experiments (Fig. 2). Using this system, we also found that cyclin E mutants 7, 8, 10, and 11 did not interact with p27 Kip1 or p130 (Fig. 3a). Whereas cyclin E was not able to bind Cdk4 in this system, we found that, in contrast, Cdk3 could readily interact with WT cyclin E and mutants 5 and 6 and, to a lesser extent, with mutant 4 (data not shown).
Cyclin E is critical for the progression of cells from G 1 to and through S phase, and ectopic expression of cyclin E can accelerate this process (see Introduction). We have tested which part of the cyclin E protein is indispensable for the acceleration of cells by transfecting the constructs, allowing the expression of FLAG-tagged cyclin E mutants (see Fig. 2) transiently into NIH 3T3 or HeLa cells along with a GFP expression construct. After transfection, GFP-positive cells were electronically gated per fluorescence-activated cell sorting and analyzed for their relative distribution within the four cell cycle phases by propidium iodide staining. Only WT cyclin E and the cyclin E mutants 4 -6 were able to significantly enhance the relative proportion of transfected cells in S/G 2 /M phase to over 60%; all other mutants were unable to perform this task (Fig. 4). WT cyclin E and mutants 4 -6 were also positive for Cdk2 interaction in immunoprecipitation assays as well as in the mammalian two-hybrid test (Figs. 2 and 3), although only the fulllength WT cyclin E protein and the mutants 5 and 6 but not mutant 4 showed a readily detectable kinase activity (Fig. 2). This demonstrates that high levels of mutant 4 can accelerate The presence of all cyclin E mutant proteins, the WT cyclin E, and cyclin A in the transfected cells was revealed through immunoblotting of protein extracts with an ␣-FLAG antibody. b, all proteins that were expressed in HeLa cells after transfection of the indicated constructs (see a) were immunoprecipitated with an ␣-FLAG antibody and were used in kinase assays with either histone H1 or glutathione S-transferase-Rb fusion proteins as substrates. c, protein extracts from cell lines that were established from foci after transfection of REFs with the indicated cyclin E mutants or the cyclin E WT construct and activated Ha-Ras were precipitated with ␣-Cdk2 antibodies. Aliquots were separated by SDS-PAGE and, after transfer on membranes, were probed with ␣-cyclin E antibodies that were either directed against the C terminus (HE12) or the N terminus of the protein.
cell cycle progression without activating Cdk2, pointing to additional or alternate mechanisms of cyclin E action in controlling cell cycle progression.
It has been demonstrated that high level ectopic expression of human cyclin E is toxic in yeast but on the other hand that low level expression can rescue a lethal yeast cln1 Ϫ/Ϫ , -2 Ϫ/Ϫ , and -3 Ϫ/Ϫ mutation (2,9,43,44,46). We wished to test whether these functions of cyclin E overlap with its other activities as a positive G 1 regulator or as an oncoprotein and have inserted the WT cyclin E and the cyclin E mutants into the inducible yeast expression vector pRS416Met25 (43,44). This vector enables yeast cells to grow on selective media lacking uracil and allows a methionine-dependent expression. All mutant constructs and the WT cyclin E construct were used to transform the "triple Cln-deficient," conditionally lethal yeast mutant strain ⌬L1 (2, 43), which is able to grow on galactose (permissive conditions) but not on glucose (restrictive conditions). Liquid cultures of the transformants were generated, and extracts were analyzed by immunoblot. In the presence of 1 mM methionine, only low amounts of WT cyclin E were detected, whereas the expression levels rose considerably when the cells were shifted to methionine-free medium for 6 h (Fig.  5a), demonstrating the functionality of the vector system. Similar to the WT cyclin E, all cyclin E mutants used in this assay

FIG. 3. Mammalian two-hybrid assays to test the interaction of cyclin E and cyclin E mutants with different G 1 regulators. a, NIH
3T3 cells were transfected with construct directing the expression of the WT cyclin E and the indicated cyclin E mutants as fusion proteins with the VP16 domain. Co-transfected were expression constructs for Gal4-dbd-Cdk2, Gal4-dbd-p27 Kip1 , or Gal4-dbd-p130 fusion proteins as well as a reporter gene construct that contained a Gal4-binding site-dependent luciferase gene. A plasmid encoding ␤-galactosidase was cotransfected to control the transfection efficiency. The luminometrically measured enzymatic activity of the luciferase reporter gene and the ␤-galactosidase are represented as RLU of the luciferase reporter divided by RLU of the ␤-galactosidase standard. The columns represent average values with S.D. values of three independent measurements from a representative experiment of at least three independent experiments with the same qualitative result. The same results were also obtained in HeLa cells (data not shown). b, expression control for all constructs directing the expression of VP16 fusion proteins that have been used in this experiment. The indicated constructs were transfected into HeLa cells, and the respective proteins were detected by immunoblot using an antibody directed against the VP16 domain.
could be expressed at similar, elevated levels in this yeast strain in the absence of methionine (Fig. 5b). Next, all ⌬L1 transformants were plated in the presence of methionine on selective media containing either galactose (permissive) or glucose (restrictive) and were incubated at 30°C for 5 days. The full-length WT cyclin E protein was able to rescue the growth of ⌬L1 under restrictive conditions and in the presence of methionine (i.e. at low expression) (Fig. 5c). With the empty vector, no growth was achieved (Fig. 5c). Strikingly, the cyclin E mutants 4 -8 were also able to rescue the triple Cln-deficient mutants under restrictive conditions and in the presence of methionine (Fig. 5c).
To test the toxicity of WT cyclin E or the cyclin E mutants, the respective constructs were used to transform the WT yeast strain YPH 499. All transformants were plated in the absence (induction) or presence (repression) of methionine on selective media. After 3-5 days, yeast cells transformed with cyclin E mutants 4 -6 or with the WT cyclin E failed to grow in the absence of methionine (Fig. 5d), whereas colonies appeared in all transformants in the presence of methionine (Fig. 5d), indicating the toxicity of full-length cyclin E and the mutants 4 -6 in yeast.
The full-length cyclin E protein resides in the nucleus, but transient transfections of GFP-tagged cyclin E mutants into NIH 3T3 cells demonstrated that nuclear localization critically depends on the integrity of the protein. The cyclin E mutant 5 covering aa 1-365 was still retained in the nucleus (Fig. 6a), but the cyclin E mutants 4, 6, and 7 (Fig. 6a) and all other cyclin E mutants described here (not shown) were found to be equally distributed between cytoplasm and nucleus of transiently transfected NIH 3T3 and HeLa cells. In particular, the localization of mutant 4 indicates that the putative basic nuclear localization sequence in the N terminus of cyclin E (aa 15-17, KRK (48,49)) that is present in this mutant is apparently nonfunctional (Fig. 6a). In contrast, a different situation is observed in malignantly transformed cells from foci that were obtained after co-transformation of REFs with Ha-Ras and constructs containing GFP fusions between WT cyclin E or FIG. 4. Ability of cyclin E and cyclin E mutants to drive NIH 3T3 cells into S phase. Exponentially growing NIH 3T3 cells were transfected with constructs allowing the expression of FLAG-tagged versions of WT cyclin E or cyclin E mutants by the calcium phosphate precipitation method. 24 h after transfection, cells were harvested and fixed with ethanol. After fixation, cells were stained with a propidium iodide/RNase solution and analyzed with a FACSCalibur and Cell Quest software (both from Becton Dickinson) for their distribution in different phases of the cell cycle. At least 8000 gated, GFP-positive cells were measured in each case. Shown is a set of data representative for at least three independent experiments where the percentages of cells in G 1 and in S/G 2 /M phase is given for each cyclin E construct. Qualitatively the same results were obtained in HeLa cells (data not shown). The functionality of the constructs is demonstrated in Fig. 2a.  FIG. 5. Activity of WT cyclin E and cyclin E mutants in yeast. a, protein extracts from ⌬L1yeast cells that had been transformed with empty pRS41Met25 or with the same vector containing cyclin E and had been grown in the absence or presence of methionine were separated by SDS-PAGE, and the cyclin E expression was detected by immunoblot with an ␣-cyclin E antibody. b, protein extracts from ⌬L1 yeast cells transformed with the indicated cyclin E mutants in the pRS41Met25 vector and grown in the absence of methionine for full induction of expression were separated by SDS-PAGE and analyzed for the expression of the respective cyclin E mutant with either the ␣-cyclin E antibody that recognizes the N terminus or the monoclonal ␣-cyclin E antibody HE12 that recognizes the C terminus of the protein. We observed that in many cases the expression of cyclin E and its mutants led to the appearance of a predominant band of the expected molecular weight in Western blots (arrowheads) as well as different faster migrating immunoreactive bands, suggesting either posttranslational modification, degradation, or alternative translation from a second start codon (see also Figs. 1 and 2). c, the yeast strain ⌬L1 was transformed with constructs allowing the expression of the indicated cyclin E mutants and the WT cyclin E protein, and the cultures were plated on selective media containing either glucose (restrictive) or galactose (permissive). Both media contained 1 mM methionine to ensure low level expression of all mutants. Shown is a representative of five individual experiments. d, the yeast strain YPH499 was transformed with the same expression constructs as the strain ⌬L1 (see above), allowing the expression of the indicated cyclin E mutants and the WT cyclin E protein, and the cultures were directly plated on selective media either containing 1 mM methionine, which repressed the expression constructs, or lacking methionine, which induced high level expression. Colonies appeared after incubation for 3-5 days at 30°C. Proteins that are toxic and prevented growth are marked with a plus sign. Shown is a representative of five individual experiments. mutants 6 and 7. Here, the cyclin E mutants 6 and 7 were clearly localized in the nucleus (Fig. 6b), although they were cytoplasmic when transiently expressed in untransformed fibroblasts (Fig.  6a). This demonstrates that the nuclear localization of cyclin E mutants or even the WT cyclin E depends on the status of the cell and may correlate with the process of malignant transformation. Immunoprecipitation analysis with extracts from cell lines that were established from foci revealed that GFP-WT cyclin E and the mutants 6 and 7 fused to GFP were all capable of interacting with Cdk2 (Fig. 6c), whereas only cyclin E WT and mutant 6 as well as untagged WT cyclin E increased Cdk2-associated kinase activity (Fig. 6c). This confirmed our results from immunoprecipitations (Fig. 2) and our finding that mutant 7 can interact with Cdk2 but is unable to activate it. DISCUSSION The link between a constitutive or high level expression of cyclin E and the malignant transformation of cells has now become clear, since different lines of experimental evidence have demonstrated that cyclin E can act as a dominant oncogene (see Introduction). However, only little information exists to date on how cyclin E performs this task. One hypothesis would be that the functions that cyclin E exerts in cell cycle progression and in particular its ability to drive cells from G 1 into S phase render it oncogenic when present at high levels. Since this function of cyclin E is mediated by binding and activating Cdk2 to phosphorylate G 1 -specific pocket proteins, such a model would imply that high levels of cyclin E directly equate with constitutively high kinase activities of Cdk2. In FIG. 6. Subcellular localization of cyclin E mutants. a, cDNAs coding for the cyclin E WT protein and for the mutants 4 -7 were cloned into pEGFP-N3 (CLONTECH) and were transiently transfected into NIH 3T3 cells. 24 h after transfection, cells were harvested and fixed with ethanol. After fixation, cells were stained with Hoechst and analyzed with a Laser Scan Microscope (LSM 512; Zeiss). Besides the cyclin E-WT-protein, only mutant 5 (cyclin E-(1-365)) showed nuclear localization; all other GFP-fused cyclin E mutants were situated in the cytoplasm. The same results were obtained using HeLa cells (data not shown). b, constructs that allow the expression of the full-length cyclin E and mutants 6 and 7 as GFP fusion proteins were transfected along with the pGEJ6.6 Ha-ras plasmid into primary REFs to obtain foci of malignantly transformed cells. Foci were picked and established as continuously growing cell lines. Cells from these lines were stained with Hoechst and were analyzed using a confocal laser-scanning microscope and the appropriate software (Zeiss). c, extracts from primary REFs or from cell lines established from foci that were obtained after cotransfection of REFs with Ha-Ras and WT cyclin E (lane 2), WT cyclin E-GFP fusion construct (lane 3), mutant 6-GFP fusion (lane 4), or mutant 7-GFP fusion were precipitated with either ␣-cyclin E antibodies (left panel) or with ␣-CDK2 antibodies (right panel). The precipitates were used for a kinase assay with histone H1 as a substrate and then analyzed by SDS-PAGE and autoradiography. In addition, the precipitates were separated by SDS-PAGE, transferred onto nylon membranes, and developed with ␣-cyclin E or ␣-CDK2 antibodies. this model, a high constitutive Cdk2 activity could override control mechanisms at the restriction point R and should subsequently lead to a continuous cycling of cells, which is a prerequisite of malignant transformation. Our experiments described here with different cyclin E mutants suggest, however, that the situation is more complicated. From our findings, we have to conclude that the oncogenic activity of cyclin E does not squarely correlate with a single distinct domain. Moreover, we find that the ability of cyclin E to transform REFs in cooperation with activated Ha-Ras does not entirely depend on Cdk2 binding or its activation. Indeed, some of the cyclin E deletion mutants that are unable to activate or to bind Cdk2 still bear oncogenic activity in the REF assay. Interestingly, this is consistent with findings that human tumors with high level cyclin E expression or tumors from transgenic mice carrying a cyclin E transgene lack high Cdk2 activities as well as with a report on cyclin E-mediated stimulation of S phase progression without Rb phosphorylation (39,50,51).
Our mutational analysis also demonstrated that for the interaction of cyclin E with Cdk2 and its activation, the cyclin box alone appears to be insufficient. For instance, mutant 3 contains the entire cyclin box but is unable to bind Cdk2 or to activate it; the slightly longer variant mutant 4 can interact with Cdk2, but if at all is only able to activate it to very low levels that make its detection difficult. Given that mutants 4 -6 still are able to bind to Cdk2, it is very surprising that the point mutant R130A lacks this ability, which is consistent with earlier reports (52). We also found that the localization of cyclin E R130A is mostly cytoplasmic (not shown), which is in contrast to the recent finding of correct nuclear import of this point mutant (49). In addition, this mutant failed to interact with p27 Kip1 and p130 in the mammalian two-hybrid system (not shown). One explanation would be that the R130A mutation induces a major conformational change in this protein that abrogates multiple functions, not only Cdk2 binding and activation. This may also explain why this mutant completely lacks oncogenic activity, whereas a number of cyclin E deletion mutants including those lacking Cdk2 binding and activation still have oncogenic activity.
A role of p27 Kip1 , which is an important negative regulator of cyclin E-Cdk2 kinase, has been postulated in malignant transformation, and a number of studies correlate low p27 Kip1 expression levels with tumor progression. Although p27 Kip1 -deficient mice do not overtly develop any malignancies, they show multiorgan hyperplasia, suggesting that p27 Kip1 is critical in preventing the initiation of malignant transformation (53). Since REF cells express no or very little p21 Waf1 or p57 Kip2 , p27 Kip1 is the most important inhibitor of cyclin E-Cdk2 in our experimental system. We find that the highest numbers of foci are obtained with WT cyclin E and mutants 5 and 6, which both interact well with p27 Kip1 . All other mutants show a lower level of interaction with p27 Kip1 and also lower numbers of foci. In this respect, the oncogenic activity of cyclin E correlates with its potential to bind p27 Kip1 and would support a model in which one function of cyclin E that contributes to malignant transformation of cells is the sequestering of p27 Kip1 and thereby the inhibition of its negative effects on G 1 /S cell cycle progression. However, our data also demonstrated that cyclin E mutants that have entirely lost their ability to interact with p27 Kip1 still can malignantly transform REFs in cooperation with Ha-Ras. This indicates that sequestering of p27 Kip1 may be one but not the only important aspect of cyclin E oncogenic activity.
All G 1 -specific functions of cyclin E, such as Cdk2 binding and activation, the acceleration of G 1 /S phase progression, and the interaction with p27 Kip1 and the pocket protein p130, can be exerted by cyclin E mutants 4 -6. However, the ability of mutant 4 to activate Cdk2 kinase activity is hardly detectable. Since this mutant is still able to drive cells into S phase, it could be concluded that this acceleration of cell cycle progression is independent of Cdk2 activation. Given the evidence from previous experiments that links both Cdk2 activation and G 1 /S progression (see Introduction), it is also possible that very low levels of Cdk2 activation that are difficult to detect in our experiments are sufficient for this task. However, an alternative possibility to explain the activity of mutant 4 is provided by the finding that ectopic expression of cyclin E can override a G 1 arrest imposed by a phosphorylation-defective mutant of pRb in rat fibroblasts (51). This finding demonstrated that cyclin E can induce S phase and the completion of the cell cycle in somatic cells downstream or parallel to the phosphorylation of pRb and in the absence of E2F-mediated transactivation (51). Thus, it is not unlikely that mutant 4 accelerates G 1 /S progression and contributes to malignant transformation by exactly this function of cyclin E, which is independent of the phosphorylation of pRb, the release of E2F, and the need to activate Cdk2.
The mutants 4 -6 define a region between aa 50 and 350 that apparently mediates the G 1 -specific functions of cyclin E but extends well beyond the cyclin box. Mutants that have lost this domain have still considerable oncogenic activity, which suggests that other functions of cyclin E that are distinct from its role in G 1 /S phase progression exist that are contributing to its oncogenic activity. In particular, our data show in two different settings that cyclin E mutants 7 and 8 still bind to Cdk2 but are unable to activate it. Further, both mutants are unable to promote G 1 /S expression but have still significant oncogenic activity, are not toxic for yeast but are able to rescue a triple cln mutation. This is consistent with a view that other functions of cyclin E that are not involved in G 1 regulation exist and contribute significantly to cyclin E-mediated tumorigenesis. Indeed, it has recently been shown that cyclin E and cyclin E-Cdk2 complexes have a role in the centrosome duplication and in mitosis exit (54 -58), which might offer an explanation for this conundrum. It is possible that the critical domains in cyclin E for centrosome duplication and mitosis exit may significantly contribute to its oncogenic activity but only partly overlap with the domains responsible for G 1 -specific functions, supporting the notion that malignant transformation through cyclin E does not entirely rely on its ability to promote cell cycle progression at G 1 .
In addition, our data from the functional and biochemical assays suggest that cyclin E can bind and activate alternative Cdk partners such as Cdk3, at least at a low level, giving rise to many possible implications in other cell cycle regulatory pathways, which need further analysis. The many reported splice mutants of cyclin E (10, 46, 59, 60) also increase the presumption of further functions of the protein, since they lack different, crucial parts of the WT protein that are important for the interaction with known binding partners. Taking into account that cyclin E can also induce cell cycle progression without Rb function or E2F activation of S phase genes (51), it is very possible that cyclin E bears other functions in regulating cell cycle progression than kinase activation. According to our findings, these yet to be determined functions can also be executed in the cytoplasm, because most of the cyclin E mutants distribute throughout the cell when overexpressed. Since the potential nuclear localization signal in the N terminus of cyclin E (aa [15][16][17][18] is apparently nonfunctional (this study and Ref. 61), it is more likely that nuclear import of cyclin E via importin-␣/␤ (49) is mediated by binding of partner proteins or is guarded by a rate-limiting factor possibly in a cell cycle-dependent manner (62). In malignantly transformed cells, an exclusive nuclear localization of cyclin E seems to be required and might be even a prerequisite for the transformation step, since this is cell cycle-independent in some types of cancer cells (62). From our findings, it is conceivable that high levels of cyclin E contribute to the malignant transformation of cells not by one single mechanism or a unique pathway but rather through a combination of functions of which the well described G 1 -specific tasks of cyclin E in cell cycle progression represent only one part of the whole picture.