AL, WvZ, CKR, and RMFW conceived and designed the experiments and analyzed the data. AL, WvZ, and RMFW performed the experiments. AL and RMFW wrote the paper.
¤ Current address: Department of Medical Oncology, University Medical Center, Utrecht, The Netherlands
The authors have declared that no competing interests exist.
Activation of cyclin B1–cyclin-dependent kinase 1 (Cdk1), triggered by a positive feedback loop at the end of G2, is the key event that initiates mitotic entry. In metaphase, anaphase-promoting complex/cyclosome–dependent destruction of cyclin B1 inactivates Cdk1 again, allowing mitotic exit and cell division. Several models describe Cdk1 activation kinetics in mitosis, but experimental data on how the activation proceeds in mitotic cells have largely been lacking. We use a novel approach to determine the temporal development of cyclin B1–Cdk1 activity in single cells. By quantifying both dephosphorylation of Cdk1 and phosphorylation of the Cdk1 target anaphase-promoting complex/cyclosome 3, we disclose how cyclin B1–Cdk1 continues to be activated after centrosome separation. Importantly, we discovered that cytoplasmic cyclin B1–Cdk1 activity can be maintained even when cyclin B1 translocates to the nucleus in prophase. These experimental data are fitted into a model describing cyclin B1–Cdk1 activation in human cells, revealing a striking resemblance to a bistable circuit. In line with the observed kinetics, cyclin B1–Cdk1 levels required to enter mitosis are lower than the amount of cyclin B1–Cdk1 needed for mitotic progression. We propose that gradually increasing cyclin B1–Cdk1 activity after centrosome separation is critical to coordinate mitotic progression.
When active, the enzyme cyclin B1–cyclin-dependent kinase 1 (Cdk1) commits a growing cell to the process of mitotic cell division and chromosome separation. Cyclin B1–Cdk1 activation is controlled in many ways, but once its activity rises above a certain level, further activation of cyclin B1–Cdk1 is catalyzed by a positive-feedback loop. This generates highly active cyclin B1–Cdk1 and triggers the start of mitosis, which can only be completed when cyclin B1–Cdk1 activity is properly shut off again. However, it is not clear how cyclin B1–Cdk1 activity develops in human cells or how the switch between its inactive and active states is controlled. Our work combines activation measurements with a kinetic model to study how cyclin B1–Cdk1 activity accumulates just before and during mitosis. We show that cyclin B1–Cdk1 activity develops gradually in early mitosis and that different activity levels are required for initiation of, and progression through, mitosis. We also demonstrate that once cyclin B1–Cdk1 activation is truly launched, it is bound to continue and will not lightly drop back again. We propose that the successive cyclin B1–Cdk1 activity levels by themselves may coordinate the progression through the distinct phases of mitosis.
The gradual increase of cyclin B1-Cdk1 activation in human cells is proposed to be critical for the progression of mitosis.
Mitotic entry is catalyzed by the kinase activity of cyclin-dependent kinase 1 (Cdk1) in complex with cyclin B1 [
Cyclin B1–Cdk1 activation contributes to the separation of centrosomes in late G2 [
It has been proposed that such a cyclin B1–Cdk1–APC/C module could function as an autonomous oscillator governing cell cycle progression [
Although it is clear that cyclin B1–Cdk1 is mostly inactive in G2 and highly active in mitosis, methods to follow the development of cyclin B1–Cdk1 activity during the rapidly successive mitotic stages have been lacking. As a result, experimental data of Cdk1 activation kinetics in human cells are very limited. Such data could, for instance, reveal whether cyclin B1–Cdk1 activation, once initiated, is sufficiently robust to drive mitotic progression even in the cytoplasm, when a large fraction of cyclin B1–Cdk1 translocates to the nucleus in prophase. This would be indicative of a bistable Cdk1 activation response and further indicate that precise regulation of cyclin B1–Cdk1 activity could control mitotic events per se.
We use novel assays to determine how cytoplasmic cyclin B1–Cdk1 activation proceeds in human cells. We find that cyclin B1–Cdk1 is activated in late G2 and that the activity gradually increases between centrosome separation and prometaphase. Moreover, we investigated the effect of partially reducing Cdk1 levels on mitotic progression. We show that whereas low levels of Cdk1 are sufficient for mitotic entry, higher levels are needed for normal mitotic progression and initiation of anaphase. Altogether, we show that the development of cyclin B1–Cdk1 activity in vivo proceeds as if it were governed by bistability. Our results explain how cyclin B1–Cdk1 can be responsible for cytoplasmic rearrangements even during nuclear translocation. We propose a model in which different thresholds of cyclin B1–Cdk1 activity, in combination with a gradual increase in activity, help to coordinate early and late mitotic events.
Cyclin B1–Cdk1 is initially activated on centrosomes in late G2, shortly before they start to migrate apart [
We first wanted to test the specificity of these antibodies, and therefore we used short hairpin RNA (shRNA) to reduce the levels of cyclin B1 or Cdk1 in cells. The cytoplasmic staining of the Cdk1-P antibody in G2 almost completely disappeared by microinjecting cells with an shRNA to cyclin B1 or Cdk1, showing the cytoplasmic signal represents phosphorylated Cdk1 (
Fluorescence images are undeconvolved maximum intensity projections of unsynchronized cells from a single cover slip. Cells are presented from G2 (left) to anaphase (right). Top row: DNA, second row: phosphorylated Cdk1, third row: cyclin B1, bottom row: merge of all three fluorescence channels.
Subsequently, we analyzed numerous undeconconvolved fluorescent images of G2 and mitotic cells, to carefully quantify the cytoplasmic cyclin B1 and Cdk1-P staining in various phases of G2 and mitosis (see
After correction for changes in morphology, we subsequently plotted cytoplasmic Cdk1-P levels as a function of cytoplasmic cyclin B1 levels. We observed that both Cdk1-P and cyclin B1 increased in a roughly linear fashion when cells proceeded through G2, with little change in the Cdk1 phosphorylation (P) state, in agreement with cyclin B1–Cdk1 complexes displaying low activity in early G2 (
(A) Quantification of immunofluorescence labeling. Cytoplasmic Cdk1-P (
(B) Relative cyclin B1–Cdk1-P during nuclear translocation. Ratio of cytoplasmic Cdk1-P and cyclin B1 signal (
(C) Average values of cytoplasmic cyclin B1 and phosphorylated Cdk1. Values are shown as percentage of the highest average. Error bars indicate SD. Neg, cells used for background subtraction; E G2, G2 cells with cyclin B1 levels between average anaphase levels and lowest value of cell with separated centrosomes; L G2, G2 cells with cyclin B1 levels above lowest value of cell with separated centrosomes; Sep, separated centrosomes; Tra, translocated cyclin B1; Pro, prometaphase; Met, metaphase; Ana, anaphase.
(D) Estimated cytoplasmic cyclin B1/Cdk1 activity of individual cells. The degree of dephosphorylation multiplied with the level of cyclin B1 (
(E) Average values of estimated cytoplasmic cyclin B1–Cdk1 activities within the distinct phases, shown in
(F) Average values of cytoplasmic Cdk1-P/cyclin B1 ratio and estimated cytoplasmic cyclin B1–Cdk1 activities in the different phases. Values are shown as percentage of the highest average. Error bars indicate SD. Labels are as in
(A) Deconvolved maximum projections of unsynchronized HeLa cells. The cell-cycle stage is indicated above the figure. Top row: merge, second row: DNA, third row: cyclin B1, bottom row, APC3 S426P.
(B) Dot plot of cytoplasmic cyclin B1 and cytoplasmic APC3 S426P staining. Each dot corresponds to one cell. Cells are arranged according to morphology and staining as in
(C) Average values of cytoplasmic cyclin B1 and phosphorylated APC3. Values are shown as percentage of the highest average. Error bars indicate SD. Bars are labeled as in
Subsequently, we focused on the development of cyclin B1–Cdk1 dephosphorylation when cyclin B1 enters the nucleus due to phosphorylation of its cytoplasmic retention signal in prophase [
When we compared cyclin B1 and phospho-Cdk1 levels before and after the onset of translocation (
Total cyclin B1–Cdk1 activity in the cytoplasm is dependent on the concentration of the complex as well as on the fraction of active complexes present. We therefore reasoned that by multiplying the “relative activity” with the cyclin B1 levels, we would acquire a rough estimate of the “total cyclin B1–Cdk1 activity.” To estimate the relative activity, we assumed the ratio between P-Cdk1 and cyclin B1 in early G2 cells represented inactive complexes (
Next, as an independent approach to determine cyclin B1–Cdk1 activity in human cells, we aimed to extend our analyses to a direct target of cyclin B1–Cdk1. As an additional benefit, this would allow us to confirm the kinetics of cyclin B1–Cdk1 activation when quantifying a positive readout signal. To this end, we used purified antibodies specifically recognizing phosphorylated APC3, a cyclin B1–Cdk1 substrate, in mitosis [
It should be noted that whereas the direct measurements in
To correlate the duration of Cdk1 activation with mitotic progression, we determined the timing of specific G2 and mitotic events: centrosome movement, DNA condensation, chromosome congression and sister chromatid separation in HeLa cells (
Average ratio of Cdk1-P and cyclin B1 staining (black solid line), average estimated cyclin B1–Cdk1 activity (gray solid line), and average APC-3 S426-P (black dotted line) plotted as a function of time (
In conclusion, cyclin B1–Cdk1 activity continued to accumulate between centrosome separation and chromosome congression in HeLa cells during a time span of approximately 45 min. The dephosphorylation of cyclin B1–Cdk1 proceeded in a gradual fashion, and the development of total cytoplasmic activity was modestly influenced by localization changes of the cyclin B1–Cdk1 complex (
If bistability and hysteresis govern cyclin B1–Cdk1 activity, Cdk1 should not be phosphorylated and inactivated when the cyclin B1 concentration drops slightly below the activation threshold [
In a recent spatial theoretical simulation, Yang and coworkers [
Abrupt and efficient Cdk1 activation at the beginning of mitosis might be required to eventually trigger normal Cdk1 inactivation, which is necessary to build an uncompromised Cdk1-APC/C cell cycle oscillator in cycling extracts [
Next, to investigate the requirement for cyclin B1–Cdk1 at different phases of mitosis in human cells, we reduced the expression levels of Cdk1 by vector-driven shRNA (
(A) Cells selected for Cdk1 shRNA expression were synchronized in G2/M by thymidine release; mitotic cells were isolated by gentle shake-off. Mitotic cells were more than 95% MPM2 positive as analyzed by immunostaining and FACS analysis (unpublished data). Separated G2 and mitotic pools were analyzed for Cdk1 expression by Western blotting. Cdc20 protein levels serve as loading control. The percentage of remaining Cdk1 protein is indicated in the figure.
(B) Cells collected by mitotic shake-off were lysed (lanes 1 and 2) or released from nocodazole and incubated in fresh medium for 3 h, recollected, and lysed (lanes 3 and4). Differences in mitotic phosphorylation shift of APC3 (human Cdc27 ortholog) and Cdc25C, depending on the Cdk1 levels, are shown (lanes 1 and 2).
(C) The impaired phosphorylation of APC3 in Cdk1-attenuated mitotic cells (lane 1) was rescued by coexpression of a Cdk1-YFP construct containing a silent mutation in the RNAi targeting region (lane 2). Lane 3 are mitotic cells transfected with a control shRNA, revealing normal endogenous Cdk1 levels.
(D) Distribution of metaphase duration, measured as time between chromosome alignment at the metaphase plate and onset of sister chromatid separation, in Cdk1 RNAi cells (right) or Cdk1 RNAi cells rescued by coexpression of non–RNAi-sensitive Cdk1-YFP (left).
(E) Time-lapse microscopy analysis of mitotic progression after entry with normal or impaired Cdk1 levels. Bottom panels are consecutive images of tubulin-YFP in a pS-control cell in mitosis; top panels show delayed chromosome alignment (frames 2 and 3) and stalled metaphase (frames 4–6) after Cdk1 shRNA.
An important consequence of our findings is that distinct thresholds for cyclin B1–Cdk1 activity exist that control successive mitotic events. This concept has recently been demonstrated in mitotic exit, which requires the passage of different thresholds of decreasing cyclin B1–Cdk1 activity for the metaphase-to-anaphase and the anaphase-to-telophase transition [
Model of relation between Cdk1 activity and mitotic progression. Cells do not enter mitosis unless a threshold concentration of active cyclin B1–Cdk1 is present (G2 arrest). After mitotic entry, the Cdk1 activity gradually increases, which enables the cell to prepare for mitotic exit (normal G2/M). If the Cdk1 activity does not develop fully after mitotic entry, the cell is delayed before anaphase (mitotic delay).
HeLa and U2OS cells were cultured in DMEM supplemented with Glutamax, 10% fetal calf serum, and antibiotics (GIBCO,
HeLa or U2OS cells were cotransfected using calcium phosphate and HBS with 5 or 10 μg of the indicated pSuper constructs, in the presence of 1 μg of α-tubulin–pEYFP as a reporter or 1 μg of pBABE-Puro as a selection marker, per 8-ml culture medium in 9-cm dishes of subconfluent cells or the equivalent for smaller dishes [
For three-dimensional time-lapse microscopy in
Cells were grown on hexametaphosphate/metasilicate-coated coverslips and fixed with PBS containing 3% paraformaldehyde and 2% sucrose for 10 min followed by permeabilization for 2 min in −20 °C methanol. Cells were incubated for 2 × 20 min in PBS containing 50 mM ammonium chloride before blocking in 2% bovine serum albumin and labeled with mouse anti–cyclin B1 (Santa Cruz, GNS1), rabbit anti-Y15P Cdk1 (Cdk1-P) (Cell Signaling,
Fifteen image z-stacks with 1-μm spacing were acquired in multiple locations of a single coverslip, using a Deltavision Spectris imaging system equipped with a ×20 objective, NA 0.7. To avoid bleaching of antibody fluorescence, cells were focused using the Hoechst labeling. The images were corrected for variation in illumination as determined by a photosensor (Applied Precision). For each z-stack, the background signal of an area without cells was subtracted. z-stacks where the background Cdk1-P signal differed more than 5% from the average background signal were not used. The average intensity of a square with five-pixel (3.3-μm) side was measured at the z-level that gave the highest average intensity in the cytoplasm, at a place where no obvious structures (e.g., centrosome, spindle, strong dotted staining) were present, of 318 cyclin B1–expressing cells using SoftWorx (Applied Precision). For examples of quantification, see
For details on the figure, see
(89 KB PDF)
HeLa cells were microinjected with 0.005 μg/μl pCFP-Golgi together with 0.1 μg/μl pS-Cdk1 or pS-cyclin B1, synchronized by a single 24-h thymidine block and fixed 10–20 h after release. Labeling was performed as in
(A) Undeconvolved maximum intensity projections of immunofluorescence staining of HeLa cells. Cells with separated centrosomes and high Cdc25B levels, indicating late G2 phase, are shown. Top, uninjected cell. Middle, cell injected with pSuper–cyclin B1 and pCFP-Golgi. Bottom, cell injected with pSuper-Cdk1 and pCFP-Golgi. All cells were growing on the same coverslip.
(B) Colocalization of the cytoplasmic Cdk1-P and cyclin B1 stainings of top cell in
(2.3 MB PDF)
Stacks of DNA, cyclin B1, and Cdk1-P stainings were acquired on multiple locations on a single coverslip. The average intensity of a square with 3.3-μm side was measured in the cytoplasm at the z-level that gave the highest cyclin B1 readout. The square was placed so that visible structures (centrosomes, mitotic spindle, strong dotted patterns, or dots that may come from secondary antibody precipitates) in both the cyclin B1 and Cdk1-P stainings were not included. Top row. G2 cell. Middle row, late prometaphase cell. Bottom row, cell with translocated cyclin B1. Note that the middle row measurements are made at a different z-level (as indicated by the z-marker in the lower left corner of each image).
(9.3 MB PDF)
(A) The NF-κB staining is cytoplasmic in untreated HeLa cells. Images are undeconvolved maximum intensity projections of NF-κB, cyclin B1, and DNA stainings.
(B) Immunofluorescence quantifications are slightly overestimated in mitotic cells. The cytoplasmic NF-κB signal was measured at similar settings as for the quantifications of cyclin B1 and phosphorylated Cdk1, as shown in
(1.0 MB PDF)
(A) Cdk1 levels do not decrease during mitosis. Images show unsynchronized HeLa cells stained for cyclin B1, Cdk1, and DNA.
(B) Quantification of total Cdk1 immunofluorescence labeling. Cytoplasmic Cdk1 (
(643 KB PDF)
(A) APC3 T244 is mainly present on the centrosomes. Images are maximum intensity projections of deconvolved immunofluorescence stainings.
(B) APC3 S446 is phosphorylated both in the nucleus and in the cytoplasm. Images are maximum intensity projections of deconvolved immunofluorescence stainings.
(C) APC3 S426 and APC3 T446 show similar cytoplasmic phosphorylation patterns. The levels of APC3–S426-P (
(6.4 MB PDF)
The clamp between two bars indicates a statistical difference (
X1 − X0 ± 1.96 × √(S12/n1 + S02/n2)
(A) Cytoplasmic ratio of P-Cdk1 and cyclin B1, from
(B) Cytoplasmic APC3 S426P
(491 KB PDF)
HeLa cells growing on a glass-bottom dish (MatTek) were microinjected with 0.008 μg/μl YFP-Histone H2B (green) and 0.0016 μg/μl dsRED-γ-tubulin (red). After a 24-h thymidine block, cells were moved to a Deltavision Spectris microscope (Applied Precision) equipped with a heater and CO2 chamber (Solent Scientific). Using a ×60 objective, NA 1.4, 15 z-levels with 1-μm spacing were acquired on multiple locations in the dish every 12 min. The distance between centrosomes was measured using SoftWorx (Applied Precision). For figures, a maximum intensity projection was made of the z-levels.
(A) Maximum intensity projections of a cell entering mitosis. The centrosomes are indicated with arrows.
(B) The average time in minutes, with SD in brackets, between early mitotic events. Time is counted from the first image where the indicated event is observed. See text for details.
(1.9 MB PDF)
(A) Details of early mitosis revealing delayed spindle formation in cells entering mitosis with reduced Cdk1 levels.
(B) A cell arresting at the G2-M transition after centrosome separation. Mitotic entry after centrosome separation usually occurs within approximately 10–20 min in U2OS cells, e.g., see
(7.0 MB PDF)
(A) Mitotic U2OS cells were collected after treatment with either pSuper or pSuper-Cdk1. Extracts of these cells were blotted with indicated antibodies. Importantly, next to a reduction of Cdk1 levels, cyclin B1 levels also were reduced. This indicates that only around 40% of cyclin B1–Cdk1 complexes could be formed in pSuper-Cdk1–treated cells compared to control cells.
(B and C) Cell extracts from (A) were subjected to immunoprecipitation with cyclin B1 antibodies for 4 h at 4 °C. Next, the cyclin B1 immunoprecipitates were incubated with 32P-γATP for 30 min with Histon H1 as substrate. We separated different amounts of the reactions on gel to correct for possible differences in cyclin B1 and/or Cdk1 levels in the IPs. Western blots were performed using cyclin B1 and Cdk1 antibodies after exposures were taken using PhosphorImager films. Intensities were measured and background was subtracted using MetaMorph software and Excel. The highest signal in the pSuper samples was set at 100%, and the estimated kinase activity for every lane was plotted against protein levels of cyclin B1 (B) or Cdk1 (C) present in the same lane. In both mitotic samples, the kinase activity roughly followed the same line. Only the activity per cyclin B1 molecule from the pSuper-Cdk1 cells seemed slightly lower (B), which can be explained if cyclin B1 exceeds the remaining amount of Cdk1. Indeed, correlating kinase activity with the Cdk1 levels revealed a closer overlap (C). All together, these data show that pS-Cdk1 cells have fewer cyclin B1–Cdk1 complexes but the complexes that are formed are fully active.
(1.2 MB PDF)
(21 KB DOC)
We cordially thank our colleagues from current and past labs, and especially Rene Medema, for their support of our work. We thank Dieuwke Engelsma and Maarten Fornerod (NKI, Amsterdam, the Netherlands) for suggesting NF-κB as a cytoplasmic marker. We are very grateful to Franz Herzog and Jan-Michael Peters (IMP, Vienna, Austria) for providing APC3–S426-P and APC3-T446 antibodies.
anaphase-promoting complex/cyclosome
cyclin-dependent kinase
nuclear factor κB
phosphorylation
RNA interference
short hairpin RNA
yellow fluorescent protein