We used diced siRNA pools (d-siRNAs) [4
] to suppress the expression of cyclin A2 (the main A-type cyclin present in somatic cells [6
]) and cyclins B1 and B2 (the best-studied B-type cyclins) in HeLa cells. Diced siRNA pools were chosen because of their high efficacy [4
] and low propensity for both off-target effects [7
] and interference with endogenous miRNA-processing pathways [8
]. We synchronized cells by double thymidine block and transfected them with the cyclin d-siRNAs singly or in combination during the first thymidine washout. Twenty-four hours after transfection, cells were released from the second thymidine block. We checked levels of the cyclin proteins by immunoblotting () and followed mitotic progression between 5 and 19 h by automated epifluorescence microscopy, using a tdimer2-RFP-tagged fluorescent mitotic biosensor (MBS) [9
]. The MBS localizes to the nucleus during interphase and translocates to the plasma membrane upon NEB, signaling the start of prometaphase [10
Figure 1 The timing of mitotic entry in HeLa cells treated with cyclin d-siRNAs. (A–D) Cells were synchronized by double thymidine block and transfected with a mitotic biosensor (MBS) and various diced siRNAs. (A) Cyclin levels of d-siRNA-treated HeLa (more ...)
All of the d-siRNA pools (25 nM) were found to be effective and specific in silencing cyclin expression (). Silencing was estimated to be 82–100% by quantitative western blotting ( and data not shown) and 75–80% by histone H1 kinase assays of cyclin immunoprecipitates (, B, and data not shown). No cross-silencing was detected ().
Figure 4 Rescue of normal NEB timing by expression of a constitutively-nuclear cyclin B1. (A) Expression levels of endogenous cyclin B1, WT-cyclin B1-YFP and NLS-cyclin B1-YFP in GL3-treated and cyclin A2-knockdown cells (top blots). The efficacy of the cyclin (more ...)
Surprisingly, there was less than an hour’s delay in NEB in cells with cyclin B1, cyclin B2, or both B-type cyclins knocked down to low levels (). In contrast, there was a marked delay in NEB in the cyclin A2 knockdown cells (). Thus cyclin A2 appears to be critical for the progression of double thymidine-arrested cells into prometaphase. The cyclin A2/B1 double knockdown showed a substantial additional delay in mitosis (). Thus, cyclin B1 becomes important for NEB when cyclin A2 function is compromised.
These findings differ from those of a previous study [12
], where high concentrations of two cyclin B1 siRNAs caused HeLa cells to accumulate in G2 phase. However, the siRNAs used in that study substantially reduced the levels of cyclin A2 as well as cyclin B1 [12
]. It is possible that the effects seen in the previous study were mediated by the combined knockdown of cyclins A2 and B1.
To determine whether the delay in NEB seen here in the cyclin A2-knockdown cells was due to the loss of cyclin A2, rather than some off-target mRNA, we made use of a second cyclin A2 d-siRNA pool, generated from the 3’-untranslated region (3’-UTR) of cyclin A2. The 3’-UTR cyclin A2 d-siRNA pool knocked down cyclin A2 levels () and delayed mitotic entry (). Moreover, a cyclin A2 expression construct lacking the 3’UTR restored cyclin A2 levels () and restored the timing of NEB to nearly that seen in the GL3 knockdown/sham rescue cells (, blue vs. black points), while having little effect on the timing of mitosis in cells transfected with control (GL3) d-siRNAs (, grey vs. black points). Thus the delay in NEB seen in the cyclin A2 knockdown cells was not attributable to off-target effects or nonspecific toxicity.
At least two mechanisms could account for the effect of cyclin A2 knockdown on NEB. Cyclin A2 could be required for the completion of S-phase, with the delayed NEB being secondary to checkpoint activation, or it could be important for the G2/M transition per se [13
]. To determine which was the case, we synchronized cells, knocked down cyclin expression () and then measured DNA content by propidium iodide staining and flow cytometry (). At the start of the thymidine washout (0 h), the control GL3 cells were well synchronized with a G1-phase 2N DNA content (). The cells had mostly entered S-phase by 5 h post-release and had nearly completed S phase by 8 h (). By 11 h, S phase was completed and post-mitotic G1 cells were beginning to appear. Mitosis was largely completed by 14 h (). The cyclin A2 knockdown cells showed no change in S-phase progression, which was largely completed by 8 h post-release, and completed by 11 h (). These findings were surprising in light of previous reports showing that cyclin A antibodies and antisense oligonucleotides inhibit DNA replication in REF52 cells [13
], HeLa cells, and human foreskin fibroblasts [14
]; perhaps the previous studies achieved an even lower level of cyclin A2 function than that shown in . However, the cyclin A2 knockdown cells were delayed in the progression from a G2/M 4N DNA content to a G1 2N DNA content (). This indicates a role for cyclin A2 in mitotic entry or completion. There was a slight delay in the completion of DNA replication in the cyclin A2/B1 double-knockdown cells and A2/B1/B2 triple-knockdown cells, but again the main delay was in the disappearance of the G2/M peak and consequent appearance of a post-mitotic G1 peak. Thus, cyclin A2 function is important for mitotic entry per se; the delays in mitosis cannot be accounted for by delayed DNA replication. These results agree well with a previous report showing that microinjection of cyclin A2/Cdk2 into G2-phase HeLa cells causes premature entry into mitosis, and microinjection of the amino terminus of p21Cip delays mitotic entry [15
Figure 2 DNA content in d-siRNA-treated HeLa cells released from double thymidine block. (A) Cyclin levels 20 h after release, as assessed by immunoblotting. (B) DNA content. Cells treated with various cyclin d-siRNAs were harvested at various times after release (more ...)
We also looked for phenotypic defects and changes in the duration of mitosis in those cyclin knockdown cells that did enter M-phase. As described in the Supplemental Data, knocking down cyclin B1 or cyclin B2 individually had no apparent effect on the duration of prometaphase, metaphase, or anaphase. However, the cyclin B1/B2 double knockdown cells showed a modest increase in duration of prometaphase as well as qualitative defects in mitosis, suggesting a redundant role for cyclins B1 and B2 in mitotic progression (Figure S1
The fact that knocking down cyclin A2 caused a marked delay in NEB, but knocking down cyclin B1 did not (), suggests that cyclin A2 does not normally regulate NEB through the intermediacy of cyclin B1. However, in Xenopus
egg extracts, the cyclin A-, cyclin B1-, and cyclin B2-Cdk complexes are activated and inactivated in temporal succession [16
], which raises the possibility that cyclin A2 might regulate some aspect of cyclin B1 function, even if cyclin B1 is not essential for NEB. In support of this hypothesis, it has been reported that cyclin A2 siRNAs and dominant-negative Cdk2 inhibit cyclin B1-Cdk1 activation in human primary fibroblasts and U2-OS cells [17
]. To test whether cyclin A2 regulates cyclin B1-Cdk activation, we compared the timing of cyclin A2-Cdk and cyclin B1-Cdk activation in control cells and cells transfected with cyclin d-siRNAs. Knocking down cyclin A2 caused a marked reduction in cyclin A2-associated histone H1 kinase activity, as expected (). However, cyclin B1-associated histone H1 kinase activity was not significantly decreased or delayed (). This suggests that cyclin A2 protein is not required for activation of cyclin B1-Cdk complexes, and suggests that NEB is delayed in the cyclin A knockdown cells despite normal mitotic levels of cyclin B1-Cdk activity. In cyclin B1 knockdown cells, the cyclin B1-associated histone H1 kinase activity was markedly reduced () and the cyclin A2-associated H1 kinase activity was unaffected (). These observations suggest that the cyclin B1-Cdk and cyclin A2-Cdk complexes are activated independently of each other.
Figure 3 Cyclin-Cdk activities and localization in cells treated with cyclin d-siRNAs. (A, B) Timing of cyclin A2-Cdk and cyclin B1-Cdk activation. Cells were treated with the indicated d-siRNAs, lysed at various times after release from double thymidine block, (more ...)
The function of the cyclin B1-Cdk complex depends not only upon its activity, but also upon its location in the cell. During interphase, cyclin B1-Cdk shuttles between the cytoplasm and nucleus, with its equilibrium favoring the cytoplasm in general and the centrosome in particular [18
]. Early in prophase, phosphorylated cyclin B1, which is thought to indicate active cyclin B1-Cdk1 complexes, becomes detectable at the centrosomes [3
] and presumably triggers centrosome separation [19
]. Sometime later cyclin B1 abruptly moves to the nucleus, and shortly thereafter NEB occurs [21
]. This prompted us to ask whether the G2-delayed cyclin A2 knockdown cells possessed cytoplasmic or nuclear cyclin B1, and whether centrosome separation occurred with normal timing in cyclin A2 knockdown cells.
We co-transfected synchronized cells with a cyclin B1-YFP construct, cyclin A2 or control d-siRNAs, and the MBS, and monitored individual cells after release from thymidine block. The fluorescent cyclin B1 allowed us to assess the timing of centrosome separation, nuclear accumulation of cyclin B1, and NEB (). The timing of NEB was confirmed with the MBS (data not shown). Expression of cyclin B1-YFP at levels comparable to endogenous cyclin B1 () was sufficient for reliable assessment of centrosome separation and cyclin B1-YFP redistribution.
Two individual cells are shown in . The GL3-treated cell shown underwent centrosome separation 11.8 h after release from thymidine block. Over the next 50 min, the separated centrosomes migrated to opposite sides of the nucleus and became more conspicuous. By 55 min, the cyclin B1 fluorescence had abruptly moved to the nucleus, with NEB following 5 min later (). The cyclin A2 knockdown cell shown underwent centrosome separation earlier—6.5 h after release from the thymidine block—but then remained with separated centrosomes for more than 8 hours before undergoing nuclear accumulation of cyclin B1 and NEB (). Thus, centrosome separation had become decoupled from NEB and the nuclear accumulation of cyclin B1 in the cyclin A2 knockdown cells.
Overall there was no apparent delay in centrosome separation in the cyclin A2 knockdown cells (, red circles vs. black circles). However, the nuclear accumulation of cyclin B1 and NEB were substantially delayed in cyclin A2 knockdown cells (). The median interval between centrosome separation and NEB was 1.2 h in the GL3 knockdown cells vs. more than 8 h in the cyclin A2 knockdown cells (). Note that the nuclear accumulation of cyclin B1-YFP was always temporally abrupt, no matter how long the G2-delay, and was always promptly followed by NEB.
If the nuclear translocation of cyclin B1 is important for cyclin B1-Cdk1 function, then a constitutively-nuclear form of cyclin B1 might be better able to rescue a cyclin A2 knockdown than wild-type cyclin B1. To test this hypothesis, we expressed wild-type cyclin B1-YFP and a cyclin B1-YFP chimera fused to an SV40 large T antigen nuclear localization sequence (NLS-cyclin B1) in synchronized HeLa cells. At the level of expression achieved (), the WT-cyclin B1-YFP was present in the cytoplasm and concentrated in the centrosome (, ), and the NLS-cyclin B1-YFP was constitutively nuclear (). We then examined the timing of NEB in individual cells released from the thymidine block. As shown in , WT-cyclin B1-YFP had no apparent effect on the timing of NEB in the GL3 knockdown cells (grey points vs. black points), in agreement with a previous report [22
]. The NLS-cyclin B1-YFP was also unable to accelerate NEB, again in agreement with a previous report [22
] (Fig. 5C, blue points vs. grey and black points). Thus, neither the WT-cyclin B1-YFP nor the NLS-cyclin B1-YFP acted as a general accelerant of entry into mitosis; NLS-cyclin B1-YFP evidently still becomes activated at a normal time despite being constitutively nuclear.
In contrast, transfection of the NLS-cyclin B1-YFP restored the timing of NEB nearly to normal in the cyclin A2 knockdown cells (, red points vs. blue points). Thus, when cells expressed modest levels of NLS-cyclin B1-YFP (), cyclin A2 function became largely dispensable for NEB. WT-cyclin B1-YFP was less effective at rescuing NEB (, red points vs. yellow points), despite higher expression levels (). This indicates that nuclear cyclin B1 can substitute for cyclin A2 in triggering NEB, but cytoplasmic cyclin B1 cannot.
In summary, we have shown that knocking down cyclin A2 in HeLa cells caused a delay in the nuclear accumulation of cyclin B1 () and in NEB (), but no apparent delay in DNA replication (), the activation of cyclin B1-Cdk1 (), or centrosome separation (). Knocking down cyclin B1 delayed NEB by less than an hour (), although in the background of a cyclin A2 knockdown, knocking down cyclin B1 caused substantial additional delay in NEB (). Expression of NLS-cyclin B1, but not WT-cyclin B1, mitigated the effects of cyclin A2 knockdown on the timing of NEB (). Knocking down cyclin B2 individually or together with cyclins A2 or B1 had little effect on the timing of NEB, but knocking down both B-type cyclins did delay mitotic progression and resulted in qualitative abnormalities in mitosis (Figure S1
shows a simple model to account for these observations. Cyclin B1-Cdk1 and cyclin A2-Cdk complexes are depicted as being activated independently, since knocking down cyclin B1 had no apparent affect on cyclin A2-associated H1 kinase activity or vice versa (). Active cyclin B1-Cdk1 accumulates at the centrosome [3
] and, we suppose, triggers centrosome separation in a cyclin A2-independent fashion (). Cyclin B1-Cdk1 then abruptly accumulates in the nucleus, with this accumulation requiring cyclin A2 (). The nuclear accumulation of cyclin B1-Cdk1 is regulated by phosphorylation of cyclin B1 at residues in its N-terminus [3
]; it seems plausible that cyclin A2-Cdk may regulate one or more of these phosphorylations. Shortly after cyclin B1 accumulates in the nucleus, NEB occurs. This may be the normal trigger for NEB, or it may reflect some structural change in the nuclear envelope. In either case, the accumulation of cyclin B1 is not essential for NEB as long as cyclin A2 is present (). Nevertheless, cyclin B1 can substitute for cyclin A2 if it is targeted to the nucleus (). The ability of NLS-cyclin B1 to substitute for cyclin A2, together with the lack of a strong phenotype in the cyclin B1 knockdowns, implies that there may be substantial overlap between the NEB-related targets of cyclin B1-Cdk1 and cyclin A2-Cdk.
A number of variations on this scheme are possible. For example, some small concentration of cyclin B1 may actually be required for NEB, but the extent of knockdown needed to reveal this requirement was not achieved in the present work. However, given the high efficacy of the cyclin knockdowns, one would have to assume that cyclin B1 is normally present in HeLa cells in huge excess of what is required for NEB. Perhaps a more plausible possibility is that cyclin B1 plays some role in regulating NEB, but is not strictly required, with some backup mechanism (for example cyclin A2 itself) ensuring that NEB is not delayed more than an hour if cyclin B1 function is compromised.
The present findings show that cyclin A2-Cdk, cyclin B1-Cdk1, and the substrates that mediate NEB constitute a coherent feed-forward system [25
]: cyclin A2-Cdk1 can influence NEB both directly and through stimulating the nuclear accumulation of cyclin B1-Cdk1 (). In engineering, feed-forward systems are often employed when it is desirable to prepare to respond in the face of a brief stimulus, and then actually respond in the face of a longer stimulus. Perhaps cyclin A2-Cdk normally prepares the nucleus for NEB, which is then triggered either by the abrupt relocation of cyclin B1-Cdk1 to the nucleus, or, in the absence of cyclin B1, by the sustained activation of cyclin A2-Cdk complexes themselves.
Although the functional analysis of cyclins in animal cells stretches back nearly two decades, as yet no simple consensus has emerged on which cyclins are important for mitosis. In Xenopus egg extracts cyclins B1 and B2 appear to redundantly drive NEB [26
]; it is less clear whether an Atype cyclin is required [27
]. In Drosophila embryos, cyclins B and B3 appear to redundantly trigger NEB [29
]. Cyclin A is required as well [29
], although recent work suggests that this is because cyclin A is required for the inactivation of Cdh1 and the accumulation of cyclin B, cyclin B3, and Cdc25, rather than because of a direct role in mitosis [32
]. The situation may be different in HeLa cells, since cyclin B1 and B2 levels are not low in cyclin A2 knockdown cells (, , ) as would be expected if cyclin A2 were required to suppress Cdh1 activation in G2 phase or prophase. Nevertheless, the studies presented here underscore the importance of cyclin A2 in NEB in HeLa cells, in part through regulating the localization of cyclin B1 and in part through cyclin B1/B2-independent effects. These findings help clarify the functions of and interrelationships between these important cyclin proteins.