Inhibition of S/G2 Phase CDK4 Reduces Mitotic Fidelity*

Cyclin-dependent kinase 4 (CDK4)/cyclin D has a key role in regulating progression through late G1 into S phase of the cell cycle. CDK4-cyclin D complexes then persist through the latter phases of the cell cycle, although little is known about their potential roles. We have developed small molecule inhibitors that are highly selective for CDK4 and have used these to define a role for CDK4-cyclin D in G2 phase. The addition of the CDK4 inhibitor or small interfering RNA knockdown of cyclin D3, the cyclin D partner, delayed progression through G2 phase and mitosis. The G2 phase delay was independent of ATM/ATR and p38 MAPK but associated with elevated Wee1. The mitotic delay was because of failure of chromosomes to migrate to the metaphase plate. However, cells eventually exited mitosis, with a resultant increase in cells with multiple or micronuclei. Inhibiting CDK4 delayed the expression of the chromosomal passenger proteins survivin and borealin, although this was unlikely to account for the mitotic phenotype. These data provide evidence for a novel function for CDK4-cyclin D3 activity in S and G2 phase that is critical for G2/M progression and the fidelity of mitosis.

mutations that disrupt p16 binding are carried in melanoma-prone kindred (for review, see Ref. 3). In addition, somatic p16 INK4A loss or mutation occurs in up to 30% of melanomas (3,4), and a small proportion of melanomas carry somatic CDK4 mutations that disrupt p16 INK4A binding (4,5). This points to the importance of the cyclin D/CDK4/Rb pathway in controlling G 0 /G 1 cell cycle progression. However, the role of CDK4-cyclin D in the phosphorylation and inactivation of Rb can in some cases be effectively substituted by cyclin E-CDK2 (6), and knockout mouse models have shown redundancy of CDK4 and CDK6, whereas knock-out of both had only minimal effect on the cell cycle of cells (7,8). These data indicate that CDK4 is not essential for cell cycle progression, so then why is the CDK4-cyclin D pathway often de-regulated in cancer? An explanation may be that CDK4 has important functions in addition to the G 1 phase phosphorylation of Rb. A peak of CDK4-cyclin D3 activity has been demonstrated during S/G 2 phase (9), suggesting that there are likely to be additional functions and substrates for CDK4 outside of G 1 phase. This hypothesis is strengthened by the finding that CDK4-cyclin D3 can phosphorylate nucleolin, a protein that is implicated in the transcription and processing of ribosomal RNA and nucleolar structure as well as additional unidentified proteins (10).
To define the role for CDK4-cyclin D, we have produced small molecule inhibitors of CDK4-cyclin D and used siRNA knockdown of cyclin D3 to demonstrate specificity of the inhibitors. Using these, we demonstrate here that CDK4 plays a critical role in G 2 /M phase progression. Loss of CDK4-cyclin D3 activity produced a transient delay in early G 2 phase before to the activation of CDK2-cyclin A in G 2 phase and CDK1cyclin B1 at G 2 /M. This delay does not appear to operate through the known G 2 /M checkpoints mechanisms, the ATM/ATR, and the p38 MAPK mediated pathways that operate in response to a variety of cellular stresses. Inhibition of CDK4 also delayed transit through mitosis, with 30% of cells progressing through an aberrant mitosis to produce cells with multiple nuclei or containing micronuclei. We also observed a delay in the G 2 /M phase expression of two gene products critical for the fidelity of partitioning of the replicated genome in mitosis, the chromosomal passenger proteins survivin and borealin. However, these proteins accumulated to relatively normal levels in cells that reached mitosis, suggesting that these were unlikely to contribute significantly to the aberrant mitosis observed. Our data indicate that inhibition of CDK4 activity during S and G 2 phase delays G 2 /M progression and reduces the fidelity of mitosis, providing strong evidence that CDK4 plays an important role outside of G 1 and Rb phosphorylation.

EXPERIMENTAL PROCEDURES
Materials-Caffeine, the p38 MAPK inhibitor SB203580, and etoposide were purchased from Sigma-Aldrich. CDK4 inhibitors (RO 0506220 and RO 0505124) and two structurally similar compounds that show no CDK4 kinase inhibition (RO 0507574 and RO 0507304) were synthesized by Discovery Chemistry Group (Hoffmann-La Roche). All other chemicals used were of analytical grade.
Cell Culture and Synchrony-All cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 3 mM HEPES and 10% (v/v) Serum Supreme (BioWhittaker) in tissue culture incubators set at 5% CO 2 and 37°C. Hydroxyurea and double thymidine block release synchronizations were performed as described in Gabrielli et al. (11). Cells were treated with CDK4 inhibitors 1 h after release from synchrony unless stated otherwise and harvested at the indicated times. Growth assays were performed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay as described previously (12). Assays for mycoplasma were carried out monthly to ensure that all of the cultured cells were free of contamination.
CDK4 Inhibitor Kinase Screen-Kinase assays were performed by FlashPlate TM assays (PerkinElmer Life Sciences). FlashPlate assays were performed using recombinant human cyclin B-CDK1, human cyclin E-CDK2, or human cyclin D1-CDK4 complexes. Glutathione S-transferase (GST)-cyclin E, CDK2, GST-cyclin B, CDK1, GST-CDK4k, and cyclin D1 cDNA clones in baculovirus vectors were provided by Dr. W. Harper at the Baylor College of Medicine, Houston, TX. Proteins were co-expressed in High Five TM insect cells, and the complex was purified on glutathione-Sepharose resin (Amersham Biosciences). A 6ϫhistidine-tagged truncated form of Rb protein (amino acid 386 -928) was used as the substrate for the cyclin D1-CDK4, cyclin B-CDK1, and the cyclin E-CDK2 assays (the expression plasmid was kindly provided by Dr. Veronica Sullivan, Department of Molecular Virology, Roche Research Centre, Welwyn Garden City, UK). Protein expression was under the control of an isopropyl 1-thio-␤-D-galactopyranoside-inducible promoter in an M15 Escherichia coli strain. Cells were lysed by sonication, and purification was carried out by binding lysates at pH 8.0 to a nickel-chelated agarose column pretreated with 1 mM imidazole. The resin was then washed several times with incrementally decreasing pH buffers to pH 6.0 and eluted with 500 mM imidazole. Eluted protein was dialyzed against 20 mM HEPES, pH 7.5, 30% glycerol, 200 mM NaCl, and 1 mM dithiothreitol. Purified Rb fusion protein stocks were quantitated for protein concentration, separated into aliquots, and stored at Ϫ70°C.
For all three kinase assays, 96-well FlashPlates were coated with Rb protein at 10 g/ml using 100 l/well. Plates were incubated at 4°C overnight or at room temperature for 3 h on a shaker. To control for nonspecific phosphorylation, one row of wells was coated with 100 l/well coating buffer (20 mM HEPES, 0.2 M NaCl). Plates were then washed twice with wash buffer (0.01% Tween 20 in phosphate-buffered saline). Compounds to be tested were added to the wells at 5ϫ final concentrations. Reactions were initiated by the immediate addition of 40 l of reaction mix (25 mM HEPES, 20 mM MgCl 2 , 0.002% Tween 20, 2 mM dithiothreitol, 1 M ATP, 4 nM [ 33 P]ATP) and a sufficient amount of enzyme to give counts that are at least 10-fold above background. Plates were incubated at room temperature on a shaker for 30 min. Plates were washed four times with the wash buffer, sealed, and counted on the TopCount Scintillation Counter.
The protein kinase B, p38, and MAPK and tyrosine kinases were assayed in-house at Roche. Assays for these kinases are based on IMAP Technology TM (Molecular Device Corp.) that enables quantitation of kinase activity via preferential binding of phosphorylated fluorescent peptide substrates to immobilized metal beads. These reactions were carried out at ATP concentrations of 3ϫ the K m for the respective enzyme. Counter-screening against the other kinases was carried out using fluorescence resonance energy transfer assays at the K m for ATP.
siRNA Design and Transfection-The Ambion Silencer siRNA construction kit (Ambion) was used to produce siRNAs against cyclin D3 and D1. The following target sequences were identified and appropriate oligonucleotides were constructed as per manufacturer's instructions. Three cyclin D3 target sequences were identified, and Blast searches were performed to ensure that the sequences did not contain significant homology to any other known genes. The sequences were: Sequence 1, AAGGATCTTTGTGGCCAAGGA; sequence 2, AAGATGCTGGCT-TACTGGATG; sequence 3, AATTGGATACATACACCAGCA. All three produced similar levels of specific cyclin D3 knockdown, and siRNA against sequence 1 was then used for all remaining experiments. In addition, a scrambled version of sequence 1 (NS sequence, AAGAAGTTAGTCCGATGCGTG), which contained no homology to CDK4 or any known genes, was used as a control.
siRNA transfection were performed as per the manufacturer's instructions. Where indicated, cells were synchronized by blocking overnight with 2 mM hydroxyurea, and released into S phase the following day. They were then analyzed by live cell imaging, immunofluorescence, and immunoblotting.
RNA Isolation and cDNA Synthesis-Cells were lysed using TRIzol reagent (Invitrogen), and total RNA was extracted as per the manufacturer's instruction. The RNA was quantified by a Smart Spec spectrophotometer (Bio-Rad) at a wavelength of 260/280 nm, and the extract was stored at Ϫ70°C until cDNA synthesis. cDNA was synthesized using Superscript III (Invitrogen) as per the manufacturer's instruction. An equal amount of total RNA was used for cDNA synthesis in each set of experiment.
Immunofluorescent Staining-Cells were grown on glass coverslips. For immunostaining, cells were washed with phosphate-buffered saline, fixed with 100% Ϫ20°C methanol, and stored at Ϫ20°C until required. Coverslips were allowed to air-dry and then rehydrated with phosphatebuffered saline containing 0.1% Tween 20 and 3% bovine serum albumin for 1 h at room temperature. The cells were stained with anti-␣-tubulin, survivin, human autoimmune serum (h-ACA) to detect kinetochores, and 4,6-diamidino-2-phenylindole (0.1 g/ml) for DNA. Coverslips were mounted on glass slides supplemented with Vectashield (Vector Laboratories, Burlingame, CA) to maintain fluorescence. Cells were visualized using a Carl Zeiss Axioskop 2 plus microscope, and images were captured using a Carl Zeiss AxioCam 2 camera and Axiovision and Adobe Photoshop software.
Flow Cytometry-For flow cytometric analysis (FACS), floating and attached cells were collected for analysis. Cells were fixed in ice-cold 70% ethanol and stored at Ϫ20°C. Samples were then washed once in phosphate-buffered saline and re-suspended in DNA staining solution consisting of 5 g/ml propidium iodide and RNase A (0.5 mg/ml) in phosphate-buffered saline. The stained cells were filtered through 37-m silk gauze, and the single-cell suspensions were analyzed on a FACSCalibur (BD Biosciences) using Cell Quest, with the proportion of cells in each cell cycle stage and subdiploid population calculated using the ModFit analysis package (Verity Software House, Topsham, ME). Bromodeoxyuridine (BrdUrd) staining was performed as described previously (14) as was MPM-2 staining (15).
Live Cell Imaging-Cells were seeded the day before the experiment on 6-or 12-well plates at ϳ50% confluence. Live cell imaging was performed using a Zeiss Cell Observer. Images were captured every 10 min for the duration of the cell cycle using Zeiss Axiovision software. Single cell analysis was performed by individually following each cell through each frame and observing the number of frames required to complete each distinct phase. Entry into mitosis was identified when the cells rounded up, and exit from mitosis was when anaphase and cytokinesis were observed.

Identification of Highly Selective Small Molecule Inhibitors of CDK4-
The diaminothiazole series of compounds were identified as a result of a screen for cyclin-cyclin-dependent kinase inhibitors (see "Experimental Procedures" for details of the screen). Modifications to the compound class produced two potent and selective CDK4 inhibitors with IC 50 values of 20 nM for recombinant CDK4-cyclin D in vitro that were Ͼ100-fold more potent for CDK4 compared with CDK1 and 2 in vitro ( Fig. 1, Table 1). The inhibitors are reversible and bind the ATP pocket of the kinases. A screen of 128 protein kinases was performed commercially using RO0505124 at a 1 M concentration to assess the potential number of kinases that may be inhibited by these compounds. Eleven kinases from different families were inhibited Ͼ90% at this dose, demonstrating the relatively high degree of selectivity of the drugs but indicating that other kinases may also be targets. To demonstrate that these compounds inhibit CDK4 in cells, a number of cell lines with or without functional Rb were tested. Inhibition of CDK4 should produce a G 1 delay and reduction in Rb phosphorylation in cells that contain functional Rb. HeLa cells that do not contain functional Rb did not display a G 1 arrest after 24 h of treatment with 2 M RO0506220, whereas SK-Mel-13 cells that have functional Rb had a strong G 1 arrest ( Fig. 2A). Similar data were obtained with both RO0506220 and RO0505124, whereas the inactive analogues had no effect on either cell lines. Later experiments were performed with RO0506220 and RO0505124 and are denoted as CDK4 inhibitor (CDK4i). Asynchronously growing HCT-116 cells, which contain functional Rb, were treated for 24 h with increasing doses of CDK4i. Cells were then either pulsed with BrdUrd and harvested for FACS analysis, or samples were collected to examine Rb phosphorylation status. A dose of 3.9 M was required to reduce S phase to 50% of control levels (IC 50 ), with a significant reduction in the level of Rb phosphorylation also observed (Fig. 2, B and C). Higher doses further inhibited BrdUrd incorporation and Rb phosphorylation. These data demonstrated that drug treatment arrested cells in G 1 phase, blocking progression into S phase. Interestingly, the proportion of G 2 /M phase cells appeared to either remain constant or increased after CDK4i treatment, whereas it would be expected to decrease if the drugs caused only a G 1 phase arrest. An additional 9 cell lines were tested, and the IC 50 values for inhibition of S phase in these cells ranged from 0.5 to 8 M ( Table 2). The much higher drug concentrations required for the in vivo inhibition of CDK4 activity compared with in vitro is likely to be a consequence of nonspecific binding of the drugs reducing their effective concentrations. At IC 50 concentrations the drugs caused little toxicity even after 3 days of treatment. Higher doses caused a moderate increased toxicity. To avoid any potentially nonspecific toxic effects, a dose of 2-4 M CDK4i was chosen as an appropriate concentration for the remainder of the experiments, as this provided strong in vivo inhibition of CDK4 without affecting CDK1 or CDK2 activity.
Inhibition of CDK4 Delays Cells in G 2 Phase-Previous work had provided evidence that CDK4-cyclin D3 was active in S and G 2 phases and that inhibiting this activity caused a G 2 /M phase delay (9). To analyze the effects of the CDK4 inhibitors on cell cycle progression beyond the Rb-dependent G 1 arrest, cells were synchronized and then treated during early S phase with CDK4i. The progression of cells through the remainder of the cell cycle was monitored using flow cytometry, biochemical analysis, and time-lapse microscopy. Upon release, cells were   3A). A dose of 4 M increased the G 2 /M delay to Ͼ4 h. The ability of CDK4i to inhibit G 2 /M progression was also observed with two other cell lines, the melanoma cell lines A2058 and SK-Mel-13 (Fig. 3B). The latter has a functional Rb-dependent G 1 arrest, whereas the former is defective for the checkpoint. There was little effect on S phase progression observed with any of the cell lines examined. The G 2 /M delay was dependent on the addition of CDK4i during early S phase. The addition during mid-S phase (4 h) produced a shorter delay of up to 2 h, whereas the addition in late S/G 2 phase (6 h) had no effect on progression through G 2 /M (Fig. 3A, data for A2058 and SKMel-13 not shown). Treatment with inactive analogues had no effect on cell cycle progression.
To determine whether the G 2 /M delay was during G 2 phase or mitosis, MPM-2 FACS was performed to specifically define the mitotic population. Control HeLa cells had begun to enter into mitosis by 8 h and returned to G 1 by 12 h, whereas cells treated with CDK4i delayed entry into mitosis until 12 h post-release. This indicated that CDK4i-treated cells were delayed in G 2 phase (Fig. 3C). This G 2 delay was confirmed by cyclin A-CDK2 and cyclin B1-CDK1 kinase assays. The delay corresponded to an accumulation of phosphotyrosine 15 CDK1 and CDK2 observed by immunoblotting with a phosphospecific antibody (Fig. 3D). Cyclin B1 accumulated normally in the CDK4i-treated cultures (Fig.  3D) as did cyclin A (see Fig. 7A); thus, the blockade was due to failure to dephosphorylate the inhibitory phosphotyrosine 15 residue on CDK2 and CDK1. The activation of the G 2 phase pool of cyclin A-CDK2 and G 2 /M cyclin B1-CDK1 in CDK4i-treated cultures was also delayed compared with untreated controls (data not shown). The inhibition of phosphotyrosine 15 dephosphorylation on both CDK1 and CDK2 was similar to that observed in the G 2 arrests imposed by DNA damaging agents such as ionizing radiation, etoposide, and ultraviolet radiation (16), which utilize stress checkpoint signaling pathways involving either ATM, ATR, or p38 MAPK (17,18). The addition of either caffeine to inhibit ATM/ATR signaling or the p38 MAPK inhibitor SB203580 had no effect on the G 2 delay observed with CDK4i addition. However, both these drugs did advance entry into mitosis in the controls by 1 h (Fig.  3E). The lack of effect of either inhibitor on the CDK4i-induced G 2 delay suggests that does not operate through known DNA damage or stress checkpoint pathways. When the level of the Wee1 kinase, which phosphorylates the inhibitory tyrosine 15 residue on CDK1 and CDK2, was examined, it was found to decrease in inactive compound-treated controls as they progressed into mitosis as expected, but in CDK4i-treated samples Wee1 levels appeared to increase and stabilize for the duration of the G 2 phase delay, which may represent an alternative mechanism by which the G 2 delay was imposed (Fig. 3F).
The fate of individual cells after CDK4i treatment as they progressed into and through mitosis was examined using time-lapse microscopy. Control cells entered mitosis at 8 -9 h post-release with a high degree of synchrony. In contrast, CDK4i-treated cells began entry into mitosis 11 h post-release (Fig. 4A). There was also a difference in the time the cells remained in mitosis, with control cells requiring 78 Ϯ 51 min to reach anaphase/telophase, whereas CDK4i-treated cells required 152 Ϯ 97 min. The broader spread of time required to reach anaphase was due to a subpopulation of cells that delayed longer in mitosis (Fig. 4B). In addition, 20% of cells failed to undergo cytokinesis correctly in the CDK4i-treated cultures, with the cells re-joining and forming large multinuclear cells (Fig. 4C). This was only rarely observed in the control cultures.
Specific siRNA Knockdown of Cyclin D3 Produces a Similar G 2 Delay-Although a panel of more than 80 protein kinases were examined for inhibition by the two CDK4i, it is still possible that the delay seen was due to an unknown effect of the drug. To exclude this possibility, siRNA knockdown

TABLE 2 CDK4i inhibits proliferation of cell lines
Proliferation was measured by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MMT) assay 48 h after drug addition. The data are the mean of two separate determinations performed in duplicate. of cyclin D3, the partner of G 2 phase CDK4 identified in previous studies (9), was used to inhibit CDK4 activity, and the effects on G 2 /M progression was assessed. Three siRNA were designed against target sequences in the C-terminal, N-terminal, central region of cyclin D3. All three siRNAs produced specific knockdown of cyclin D3 at a concentration of 10 -20 nM at 24 h post-transfection, with protein levels beginning to recover by 48 h (Fig.   5, A and B). The levels of cyclin D1, E, A (not shown), and B1 (Fig. 5A) remained unaffected by the siRNA transfection, indicating that the knockdown was specific for cyclin D3. The three cyclin D3 siRNA produced similar cell cycle effects so only the data for siRNA number 1 is shown. To examine the effect of cyclin D3 knockdown in synchronized, transfected HeLa cells, FACS analysis and time-lapse microscopy was performed. The  APRIL 14, 2006 • VOLUME 281 • NUMBER 15 results were essentially identical to the CDK4i-treated cells. Cyclin D3 knockdown delayed entry into mitosis by more than 2 h compared with control cells (Fig. 5, C and D). To demonstrate that the delay observed with the CDK4i was due to specific inhibition of CDK4, synchronized cultures treated with both CDK4i and cyclin D3 siRNA were analyzed by time-lapse microscopy. Simultaneous treatment with CDK4i and cyclin D3 knockdown produced an only slightly longer delay than the individual treatments, probably due to less than complete inhibition of CDK4 with the individual treatments (Fig. 5C). The lack of additive effect of the CDK4i and cyclin D3 knockdown demonstrated that the CDK4i drugs specifically inhibited CDK4-cyclin D3 activity that was responsible for the cell cycle delays observed.

CDK4-inhibited Cells Undergo Aberrant
Mitosis-Time-lapse data revealed that 20% of CDK4i-treated cells failed to undergo mitosis correctly, forming large multinuclear cells, suggesting that drug treatment disrupted normal mitosis. Control and drug-and siRNA-treated cultures were fixed as they progressed through mitosis, stained for microtubules and DNA, and examined by fluorescence microscopy. Both control and CDK4i-treated cells had well-formed bipolar mitotic spindles. However, 40% of CDK4i-treated cells displayed an aberrant mitotic phenotype, characterized by lagging chromosomes, which often decorated the astral side of the spindle microtubules (Fig. 6, A and B). These lagging chromosomes were also visualized by time-lapse microscopy of GFP-H2B-expressing HeLa cells, where these chromosomes formed micronuclei after mitotic exit (Fig. 6C). A similar spectrum and level of mitotic aberrations was observed in the cyclin D3 siRNA-treated cells (data not shown). Examination of cells at 24 h after synchrony release when the majority of cells have returned to interphase revealed that 30% of CDK4i-treated cells contained multiple nuclei and micronuclei compared with 10% in control cultures (Fig. 6D). These nuclear defects are an indicator of mitotic failure and were likely to be the consequence of the failed mitosis.
Inhibition of S/G 2 CDK4/Cyclin D3 Delays Expression of Chromosomal Passenger Proteins-The failure of mitosis and defects in cytokinesis detected by time lapse and fluorescent microscopy suggested that there were possibly defects in the mechanisms controlling the fidelity of mitosis, either the spindle assembly checkpoint or the chromosomal passenger proteins such as survivin, aurora B, or borealin, which form a complex with INCENP at the centromere that  influences spindle checkpoint function (19). The mitotic delay detected after CDK4i treatment indicated that the spindle checkpoint was functioning, and CDK4i treatment did not affect the ability of nocodazole to arrest cells in mitosis (data not shown). However, the presence of cells with multiple nuclei and micronuclei is evidence that cells exited mitosis prematurely, indicating that the checkpoint eventually failed. Knockdown of survivin or borealin or inhibition of aurora B activity result in similar mitotic aberrations and failure of cytokinesis as seen with CDK4i (13,20,21). Examination of total levels of survivin revealed that the normal G 2 /M accumulation of survivin protein, which peaked at 10 h, was delayed in the drug-treated cells (Fig. 7A). In controls, survivin continued to accumulate until peak mitosis 10 h in this experiment, when cyclin A levels started to diminish. In CDK4i-treated cells, survivin levels did not reach control mitotic levels even at 14 h, which corresponded to peak mitosis. When the levels of three of the mitotic passenger proteins were examined in a separate experiment, the accumulation of both survivin and borealin were delayed and failed to reach control mitotic levels (Fig. 7B). The level of aurora B was unchanged across mitosis, and CDK4i treatment had no affect. Similar results were obtained using cyclin D3 knockdown (Fig. 7C). To determine whether the delay and reduction in survivin accumulation in CDK4inhibited cells was due to reduced transcription, real time RT-PCR analysis of the mRNA levels of survivin was performed. In control cells the expected increase in survivin mRNA was observed, with the peak corresponding to peak mitosis at 12 h (Fig. 7D). In both the CDK4i-and siRNA-treated cells there was little reduction in survivin mRNA levels, although these remained elevated longer probably due to the delay in mitosis. Borealin mRNA behaved in an identical manner (data not shown).
The localization of survivin is very characteristic, accumulating at the centromeres in early mitosis then relocating to the midline of the cells on the mid-spindle during anaphase and telophase, then decorating the midbody in cytokinesis (19,22). All of the chromosomal passenger proteins localize identically through mitosis (19). Immunofluorescent staining of mitotic cells for survivin did reveal an increased proportion of mitotic cells with low levels of survivin staining and aberrant mitosis (Fig. 8A). When the cells with low survivin levels were quantitated, the proportion displaying aberrant mitosis increased from 9% in controls to 18% in CDK4i-treated cultures, but there was a similarly increased proportion of cells with normal survivin staining and aberrant mitosis, from 7% in controls to 23% with CDK4i treatment (Fig. 8B). This suggests that reduced chromosomal passenger protein levels are unlikely to be the major determinant of the aberrant mitosis in CDK4i-treated cells.

DISCUSSION
The function of CDK4-cyclin D has been defined in terms of its G 1 phase role in regulating the activity of pocket proteins and thereby controlling the transcription of genes required for progression into S phase. However, there is evidence of a role for CDK4 later in the cell cycle. CDK4 activity has been demonstrated in S and G 2 phase (9,23,24), and this activity is inhibited in UV-irradiated G 2 phase-delayed cells, associated with increased p16 INK4A and loss of CDK4 phosphorylation (9). We have now demonstrated that inhibition of CDK4-cyclin D activity using either highly selective low molecular weight inhibitors or by siRNA knockdown of the S/G 2 phase CDK4 cyclin partner, cyclin D3, causes a delay in entry into mitosis. The specificity of the drugs for inhibiting CDK4-cyclin D3 in S/G 2 phase was demonstrated by the similarity of the effects of CDK4i addition and siRNA knockdown of cyclin D3, the CDK4 partner in S/G 2 phase. Furthermore, we have used three different cyclin D3 target sequences to knock down cyclin D3, which all produced similar cell cycle effects. Finally, CDK4i addition to cyclin D3 knockdown cells did not significantly increase the G 2 delay observed with either treatment alone, indicating they both inhibit the same target.
The G 2 phase delay corresponds to a block in the cdc25-dependent activation of CDK2-cyclin A and CDK1-cyclin B, similar to the genotoxin-induced G 2 delay reported previously (16). Surprisingly, it does not appear to be imposed by the characterized G 2 checkpoint mechanisms involving ATM/ATR or p38 MAPK. Interestingly, an increase in the level of the CDK inhibitory kinase Wee1 was detected. This simply may be a consequence of inhibition of proteasome-mediated destruction, which normally occurs as cells enter mitosis (25). However, a similar accumulation of Wee1 has been reported elsewhere, and it may represent an alternative target for G 2 checkpoint controls (26). The length of the delay is dependent on the degree of CDK4 inhibition, as doubling the dose of CDK4i significantly increased the duration of G 2 delay up to 12 h in some experiments, suggesting that CDK4 activity is rate limiting for G 2 /M progression. Thus, inhibition of CDK4-cyclin D3 activity during S and G 2 phase does delay G 2 -phase progression. However, the mechanism of the delay is at present unclear. This suggests that CDK4-cyclin D3 has an important function during S phase that directly impacts on G 2 /M progression.
The effect of inhibition of CDK4-cyclin D3 activity on mitosis was unexpected. Inhibition resulted in a significant increase in the proportion of cells undergoing an aberrant mitosis characterized by lagging chromosomes. This generally was only one or two chromosomes as compared with the major failure of chromosome congression at metaphase observed with either mutation, inhibition of kinetochore-associated mitotic spindle checkpoint-associated proteins (21,27,28), or drugs that affect chromatin structure (14). The lagging chromosomes initiate a mitotic checkpoint delay, demonstrated by the increased time the CDK4i-treated cells require to reach anaphase. Immunostaining for the presence of MAD2, a spindle checkpoint component, revealed its localization to the kinetochores of lagging chromosomes, a marker of activation of the spindle assem- A, HeLa cells were synchronized using a double thymidine block release and treated with 2 M CDK4i in early S phase. Cells were harvested at the indicated times post-release for immunoblot analysis of the indicated proteins. Cyclin D 3 was used as a loading control. This is representative of five separate experiments. B, in an experiment similar to A, lysates were immunoblotted for the levels of the indicated chromosomal passenger proteins. In this experiment peak mitosis was at 10 -12 h in the controls and 13-14 h in the CDK4i-treated cultures. C, synchronized HeLa cells either pretreated with cyclin D3 siRNA number 1 or scramble siRNA were harvested at the indicated times after synchrony release and immunoblotted for the indicated proteins. CDK2 was used as a loading control. Peak mitosis was at 12 h in the scrambled control and 14 h in the cyclin D3 siRNA treated cultures. D, quantitative real time RT-PCR for survivin was performed using mRNA prepared from samples from the experiment shown in B. The data are the average of triplicate determination. These data are representative of three separate experiments. bly checkpoint (29). 4 However, the spindle checkpoint eventually fails as there was a significant increase in the proportion of cells that exited mitosis without properly segregating their chromosomes, evident as cells with multiple and micronuclei. This appears to be an example of mitotic slippage where the spindle checkpoint is initiated but fails to maintain the arrest, resulting in cells prematurely exiting mitosis (30) (31). The chromosomal passenger complex consisting of INCENP, aurora B, survivin, and borealin, influences the localization and operation of the kinetochore-associated mitotic spindle checkpoint components such as MAD2 and BUBR1, thereby affecting spindle checkpoint function (13, 19 -21). The chromosomal passengers accumulate at the centromere in prophase of mitosis, and each chromosomal passenger influences the localization and activity of the other passenger proteins (19). We have found that G 2 /M phase accumulation of two of the chromosomal passenger complex, survivin and borealin, was delayed, but there was little reduction in the levels of survivin or borealin that localized at centromeres of mitotic cells after CDK4i treatment, indicating that these proteins are likely to function normally. The reduced accumulation of these proteins observed by immunoblotting is likely the consequence of a loss of synchrony of the CDK4i-treated samples. The level of aurora B was unaffected after CDK4i treatment, and we found no effect on aurora B activity with normal levels of histone H3 Ser10 phosphorylation detected (data not shown). Thus, the mechanism by which S/G 2 CDK4-cyclin D3 influences the fidelity of mitosis remains unclear. However, it does appears not to involve regulating the expression of the chromosomal passenger protein.
A possible explanation for the mitotic defects is that they are a consequence of a failure to stably arrest in G 2 phase after CDK4i treatment. Mitotic defects are a common feature of cells that fail to stably arrest after exposure to ultraviolet radiation (32) or treatment with histone deacetylase inhibitors (14). In these cases it is likely that some form of either DNA damage or defective chromatin structure is responsible for the mitotic defects. The source of defects after CDK4i treatment is at present unclear.
The regulation of the CDK4-cyclin D/p16 INK4A axis is often disrupted in a wide variety of cancers, particularly melanomas, by either mutation deletion or overexpression of various components (4). This has generally been attributed to dysregulation of the G 1 -phase Rbdependent checkpoint. Although there is no doubt of the importance of this effect, it is also possible that dysregulation of CDK4 may influence other cell cycle phases. In this report we have demonstrated that CDK4 has a critical role in ensuring the fidelity of mitotic partitioning of the replicated genome. We have previously reported that in epidermal-derived cell lines that the S/G 2 phase CDK4 activity is inhibited by increased p16 INK4A levels in response to suberythemal doses of ultraviolet radiation (9) and that increased p16 INK4A expression correlated with a G 2 phase delay in irradiated human skin (33,34). Inhibition of CDK4-cyclin D3 activity can delay progression through G 2 and is likely to contribute to the G 2 delay observed after exposure to ultraviolet radiation. P16 INK4A inhibition of CDK4 in this response may also regulate pathways that influence the fidelity of mitosis. The failure to inhibit CDK4-cyclin D3 may result in cells that are capable of transiting mitosis with mutant DNA and surviving to pass their mutations to future generations, thereby contributing to melanoma.