Cyclin A–CDK1 suppresses the expression of the CDK1 activator CDC25A to safeguard timely mitotic entry

Cyclin A and CDC25A are both activators of cyclin-dependent kinases (CDKs): cyclin A acts as an activating subunit of CDKs and CDC25A a phosphatase of the inhibitory phosphorylation sites of the CDKs. In this study, we uncovered an inverse relationship between the two CDK activators. As cyclin A is an essential gene, we generated a conditional silencing cell line using a combination of CRISPR-Cas9 and degron-tagged cyclin A. Destruction of cyclin A promoted an acute accumulation of CDC25A. The increase of CDC25A after cyclin A depletion occurred throughout the cell cycle and was independent on cell cycle delay caused by cyclin A deficiency. Moreover, we determined that the inverse relationship with cyclin A was specific for CDC25A and not for other CDC25 family members or kinases that regulate the same sites in CDKs. Unexpectedly, the upregulation of CDC25A was mainly caused by an increase in transcriptional activity instead of a change in the stability of the protein. Reversing the accumulation of CDC25A severely delayed G2–M in cyclin A-depleted cells. Taken together, these data provide evidence of a compensatory mechanism involving CDC25A that ensures timely mitotic entry at different levels of cyclin A.

Cyclin A and CDC25A are both activators of cyclindependent kinases (CDKs): cyclin A acts as an activating subunit of CDKs and CDC25A a phosphatase of the inhibitory phosphorylation sites of the CDKs. In this study, we uncovered an inverse relationship between the two CDK activators. As cyclin A is an essential gene, we generated a conditional silencing cell line using a combination of CRISPR-Cas9 and degron-tagged cyclin A. Destruction of cyclin A promoted an acute accumulation of CDC25A. The increase of CDC25A after cyclin A depletion occurred throughout the cell cycle and was independent on cell cycle delay caused by cyclin A deficiency. Moreover, we determined that the inverse relationship with cyclin A was specific for CDC25A and not for other CDC25 family members or kinases that regulate the same sites in CDKs. Unexpectedly, the upregulation of CDC25A was mainly caused by an increase in transcriptional activity instead of a change in the stability of the protein. Reversing the accumulation of CDC25A severely delayed G 2 -M in cyclin A-depleted cells. Taken together, these data provide evidence of a compensatory mechanism involving CDC25A that ensures timely mitotic entry at different levels of cyclin A.
The cell cycle is choreographed by an evolutionarily conserved engine composed of a family of protein kinases called cyclin-dependent kinases (CDKs) (1). The current paradigm states that in human cells, CDK1 is activated by the mitotic cyclins (cyclin A and B) and drives G 2 cells into mitosis (2). Another CDK family member, CDK2, associates mainly with cyclin E and cyclin A, and the complexes formed are critical for G 1 -S transition and in S phase, respectively (3). CDK4 and CDK6 are partners of cyclin D, functioning in G 1 -S transition before cyclin E-CDK2 (4).
The activities of CDKs are stringently regulated by proteinprotein interactions and phosphorylation. For example, binding to a mitotic cyclin subunit is necessary for full activation of CDK1. On binding to cyclin B, the kinase activity of CDK1 is initially suppressed by inhibitory phosphorylation on CDK1 T14/Y15 by MYT1 and WEE1 (5). At the end of G 2 , the stockpile of inactive cyclin B-CDK1 complexes is activated by members of the CDC25 dual-specificity phosphatase family (6). Active CDK1 then activates more CDC25 and inactivates WEE1 by directly phosphorylating these proteins. This autocatalytic loop enables rapid and complete activation of all the cyclin B-CDK1 complexes by an initially small amount of active CDK1.
The presence of three isoforms of CDC25 (A, B, and C) differing in cell cycle regulation, localization, and mood of regulation suggests that they may play nonoverlapping roles in the cell cycle. CDC25A appears to be particularly important, at least in mice, as knockout of Cdc25A results in early embryonic lethality (7). By contrast, mice lacking both Cdc25B and Cdc25C are generally normal (8). The prevailing view is that while CDC25B and CDC25C regulate mainly the G 2 -M cyclin-CDK complexes, CDC25A is involved in the control of both G 1 -S and G 2 -M cyclin-CDK pairs (9). Another unique feature of CDC25A distinguishing it from other isoforms is its rapid degradation in response to DNA damage or stalled replication forks. This mechanism is dependent on the ATM/ATR-CHK1/CHK2 pathway and is critical for the checkpoints that halt the cell cycle in response to genotoxic stresses (10).
Expression of CDC25A is periodically regulated during the cell cycle by both transcription and proteolysis. Transcription starting from late G 1 is mediated by transcription factors including MYC and E2F (11)(12)(13). Unlike CDC25C or CDC25B, which is expressed throughout the cell cycle (14) or is targeted to proteasome-dependent degradation during mitosis (15), respectively, CDC25A accumulates during mitosis in a phosphorylated state (16). CDC25A is then targeted to ubiquitinmediated degradation by APC/C CDH1 during mitotic exit and by SCF βTrCP during interphase (17)(18)(19). DNA damage enhances the SCF βTrCP -mediated degradation of CDC25A through phosphorylation by CHK1/CHK2 (20,21). Similar to CDC25A, cyclin A also functions at multiple points in the cell cycle. During S phase, phosphorylation of various components of the prereplicative complexes by cyclin A-CDK2 complexes is involved in both the firing of DNA replication origins as well as preventing the re-firing of the same origins within the same cell cycle (22). Cyclin A also functions during G 2 -M, but its precise role is less well-defined. One hypothesis is that cyclin A itself is a component of M phase-promoting factor, the engine that drives cells into mitosis. An alternative hypothesis is that cyclin A is part of the network that triggers the activation of M phase-promoting factor (2,6). For example, cyclin A-CDK has been implicated in turning on PLK1, which then activates CDC25C to allow cyclin B-CDK1 activation (23,24).
In this study, we uncovered an additional relationship between cyclin A and CDC25A. Downregulation of cyclin A induced an accumulation of CDC25A through an increase of transcription. We provide evidence of a compensatory mechanism that ensure timely entry into mitosis at varying levels of cyclin A.

Depletion of cyclin A triggers rapid accumulation of CDC25A
We initially found that downregulation of cyclin A in HeLa cells with siRNA promoted an accumulation of CDC25A (Fig. 1A). It was noteworthy that siRNA-mediated depletion of cyclin A was not highly effective, and cell cycle distribution was not significantly altered after transfection (Fig. 1A, lower panel). This suggested that the striking increase of CDC25A did not require complete depletion of cyclin A or changes in cell cycle distribution.
Given the limitations associated with siRNAs, including their slow kinetics, incomplete knockdown, and nonspecificity, we next generated a conditional cell line for acute and tight silencing of cyclin A based on a recently developed tetracycline-controlled transcriptional activator (tTA)-auxininduced degron (AID) dual transcription-degron system (25,26). Concurrent with the disruption of cyclin A with CRISPR-Cas9, an AID-tagged cDNA of cyclin A (which was resistant to the CRISPR-Cas9 due to the introduction of silence mutations at the CRISPR-Cas9-targeting site) under the control of a Tet-Off promoter was integrated into the genome. Transcription of the cDNA by tTA could be turned off using doxycycline (Dox). Moreover, AID-tagged cyclin A could be targeted to rapid proteolysis when indole-3-acetic acid (IAA) was added. Figure 1B shows that cells lacking endogenous cyclin A and expressing AID-cyclin A (designated as AID Cyclin A KO herein) were able to degrade AID-cyclin A effectively in response to Dox and IAA (DI) treatment, in effect producing a cyclin Adeficient environment. The AID-cyclin A was expressed to a similar level as the endogenous cyclin A (before CRISPR-Cas9mediated disruption) and was destroyed rapidly after DI addition. Quantifying the band intensity (using a serially diluted standard curve) revealed that 5% of AID-cyclin A remained at 6 h (and 1% at 9 h) after DI treatment (Fig. S1). The destruction of AID-cyclin A was accompanied with a rapid and sustained accumulation of CDC25A (Fig. 1B). In some experiments, the expression of CDC25A decreased at later time points (48 h), probably due to the increase in DNA damage and apoptosis after prolonged cyclin A depletion (indicated by the accumulation of γ-H2AX and cleaved PARP1, respectively).
Turning off cyclin A resulted in delays in both S and G 2 /M (Figs. 1B and S2A). As CDC25A accumulated before substantial alteration of cell cycle distribution occurred, it is unlikely that the increase of CDC25A was an indirect outcome of cell cycle redistribution. This was validated later using synchronized cells (see below). Furthermore, the increase of CDC25A after cyclin A destruction was confirmed using other independently isolated clones of AID Cyclin A KO , indicating that it was not a consequence of clonal effects (Fig. S2B). Unlike CDC25A, other CDK1 T14/Y15 kinases and phosphatases including CDC25B, CDC25C, WEE1, and MYT1 were unaffected by cyclin A depletion (Fig. 1C).
Although the AID Cyclin A KO cell line was a rescue system by design, we further transfected a FLAG-tagged cyclin A into the cells before degrading AID-cyclin A. Figure 1D shows that ectopically expressed FLAG-cyclin A could reverse the accumulation of CDC25A. Although the expression of CDC25A was generally low in control cells, overexpression of cyclin A was able to further reduce CDC25A expression. Furthermore, we also generated AID Cyclin A KO in H1299 cells and found that CDC25A was enriched in the absence of cyclin A, indicating that the phenomenon was not limited to HeLa cells (Fig. S2C).
Finally, using a similar approach as AID Cyclin A KO , we generated inducible depletion of cell lines expressing AID (or mini-AID)-tagged cyclin B1, CDK1, and CDK2 in backgrounds lacking the respective endogenous genes. To avoid cell cyclerelated effects on CDC25A, the cells were first synchronized in S phase with a double thymidine block before incubated with DI for 6 h to turn off the AID/mAID proteins. By contrast to cyclin A, depletion of cyclin B1, CDK1, or CDK2 during S phase did not result in CDC25A accumulation (Fig. 2), indicating a specific role of cyclin A in this process.
Collectively, these results revealed that depletion of cyclin A promotes the rapid and specific accumulation of CDC25A.
Cyclin A regulates CDC25A throughout the cell cycle One of the difficulties in studying cyclin A is owing to the multifaceted functions of cyclin A during the cell cycle (functioning at both S phase and mitosis and binding to both CDK1 and CDK2). Moreover, CDC25A expression is also highly regulated during the cell cycle (10). To exclude the possibility that the accumulation of CDC25A triggered by cyclin A depletion was caused by a disruption of the cell cycle, we next examined the effect of cyclin A destruction at different periods of the cell cycle using synchronized samples.
We first confirmed the robust cell cycle variation of CDC25A in synchronized HeLa cells (Fig. 3A). As expected, CDC25A expression was relatively low in early S phase and started to accumulate in S and G 2 . It was highly phosphorylated during mitosis and became undetectable in G 1 , consistent with its APC/C CDH1 -dependent proteolysis (17).
As both cyclin A and CDC25A were undetectable in early G 1 , it is not surprising that turning off cyclin A during that time did not affect CDC25A expression (Fig. 3B)  destruction of cyclin A in late G 1 was already able to promote an accumulation of CDC25A (Fig. 3B). Likewise, cells synchronously released into S phase using a double thymidine block procedure also contained increased amount of CDC25A after cyclin A was turned off (Fig. 3C). It is noteworthy that CDC25A was also increased when cyclin A was destroyed in cells continuously blocked with thymidine, indicating that cell cycle progression was not required for cyclin A-mediated CDC25A accumulation. This suggested that any cell cycle delay caused by cyclin A depletion was not required for CDC25A accumulation. In fact, as cyclin A depletion delayed S and G 2 /M (Fig. S2A), it is expected that CDC25A level should become lower (instead of higher as observed) if cell cycle effects caused by cyclin A depletion plays a major role in CDC25A expression. Finally, turning off cyclin A in synchronized G 2 ( Fig. 3D) or mitotic (Fig. 3E) cells also promoted CDC25A accumulation. Taken together, these data indicate that the increase of CDC25A after cyclin A depletion occurs throughout the cell cycle and is independent on cell cycle delays caused by cyclin A deficiency.
Cyclin A regulates CDC25A independently to DNA damagemediated degradation A major mechanism of CDC25A regulation during the cell cycle is through SCF βTrCP -mediated degradation, which can be accelerated by CHK1 (20,27). As shown previously (28), inhibition of CHK1 using a small chemical inhibitor AZD7762 (CHK1i) stabilized CDC25A (Fig. 4A). Nevertheless, CDC25A was further induced after cyclin A was turned off, suggesting that the increase of CDC25A upon cyclin A depletion was independent on the SCF βTrCP pathway. Furthermore, the CDC25A in cyclin A-depleted cells could be targeted to degradation after ionizing radiation (IR), suggesting the DNA damage-mediated CDC25A degradation mechanism remained intact in the absence of cyclin A (Fig. 4B). CDC25A was rapidly degraded to a background level after irradiation irrespective of the presence or absence of cyclin A (Fig. S4A). Finally, we also examined other substrates of SCF βTrCP and APC/C CDH1 after cyclin A degradation (29). Unlike CDC25A, the expression of several substrates of SCF βTrCP (including WEE1 (Fig. 1C) and EMI1 (Fig. S4B)) or APC/C CDH1 (including cyclin B1 (Fig. 1B) and PLK1 (Fig. S4B)) was unaffected by cyclin A depletion. DI:

DNA
A I D C y c l i n . Cyclin A regulates CDC25A throughout the cell cycle. A, CDC25A expression starts at late S, accumulates in G 2 , and degrades at the end of mitosis. HeLa cells were synchronized using a double thymidine block procedure. Cells at different cell phases of the cell cycle were harvested at the indicated time points after release from the block. NOC was added to the mitotic sample (M) for 6 h before cells were isolated with mechanical shake-off. A portion of the mitotic cells were released from the block by washing and re-plating in NOC-free medium for 3 h before harvested as G 1 cells. Lysates were prepared and analyzed with immunoblotting. Histone H3 S10 phosphorylation and CDC27 mobility shifts are mitotic markers. Dephosphorylation of CDK1 Y15 occurred as cells entered mitosis. Note that CDC25A was phosphorylated and displayed a gel mobility shift during mitosis. The DNA contents of the cells were also examined with flow cytometry to validate the synchronization (Fig. S3A). B, preventing cyclin A accumulation during late G 1 promotes CDC25A accumulation. AID Cyclin A KO cells were synchronized in mitosis using a NOC block procedure as described in Experimental procedures. The cells were then released into NOC-free medium for 3 h and 6 h to obtain cells in early and late G 1 , respectively. The cells were treated with or without DI for Collectively, these data indicate that the increase of CDC25A after cyclin A destruction is not caused by inhibiting SCF βTrCPor APC/C CDH1 -dependent turnover.

Depletion of cyclin A increases transcription but not protein stability of CDC25A
We next examined the protein stability of CDC25A by using cycloheximide (CHX) to abolish de novo protein synthesis. Although CDC25A was expressed at a higher level before the addition of CHX in cyclin A-depleted cells than in control cells, its protein stability was not significantly increased after cyclin A was depleted (Fig. 4C). The half-lives of CDC25A in both cyclin A-containing and -deficient cells were 17.6 min after the addition of CHX. In agreement with the above results, IR-induced DNA damage further reduced the stability of CDC25A in both cyclin A-containing and -deficient environments. Unlike that of the endogenous CDC25A, the expression of exogenous CDC25A driven by a constitutive promoter was unaltered after the destruction of cyclin A (Fig. 5A). As a control, both endogenous and exogenous CDC25A could be stabilized with a CHK1 inhibitor (Figs. 4A and S4C). This provided further evidence that the regulation of CDC25A by cyclin A was not caused by a change in protein stability.
Given that the protein stability of CDC25A was not affected by cyclin A, we next investigated if the expression of CDC25A mRNA was affected by cyclin A. Figure 5B shows that CDC25A mRNA accumulated after the destruction of cyclin A in a time-dependent manner. Similar results were obtained using an AID Cyclin A KO cell line that also lacked CDK2, which displayed a stronger G 2 -M delay than cells lacking cyclin A alone (manuscript in preparation). Transcriptome analysis using RNA sequencing confirmed that CDC25A transcript was upregulated after cyclin A was destroyed in G 2 -synchronized cells (Fig. S5). By contrast, CDC25B and CDC25C transcripts were downregulated in the absence of cyclin A.
Taken together, these results indicate that the increase in CDC25A after downregulation of cyclin A was mainly caused by an increase in transcriptional activity instead of a change in protein stability.

Regulation of CDC25A by cyclin A ensures timely mitotic entry
To evaluate the biological consequences of CDC25A accumulation in response to the downregulation of cyclin A, CDC25A was downregulated with siRNA, and mitotic entry was analyzed at single-cell level using live-cell imaging. We titrated the concentration of the siRNA so that CDC25A was reduced to a level similar to that before cyclin A was depleted (Fig. 6A). The cells were synchronously released into the cell cycle from a double thymidine block before turning off the cyclin A (Fig. 6B). As expected, mitotic entry was deferred after the destruction of cyclin A. Reducing CDC25A likewise delayed mitotic entry. Mitotic entry was further delayed when CDC25A was downregulated in cyclin A-depleted cells, suggesting that the normal accumulation of CDC25A in these cells was responsible for partially compensating for the loss of cyclin A. A similar delay in mitotic entry was observed after cyclin A and CDC25A were downregulated in cells synchronized in late G 2 using the CDK1 inhibitor RO3306 (Fig. S6). In contrast to CDC25A, depletion of CDC25C did not delay G 2 -M in the presence or absence of cyclin A (Fig. S7).
In addition to the delay in G 2 -M, the duration of mitosis was also increased after downregulation of cyclin A and CDC25A (Fig. 6B). The delay in G 2 -M in cyclin A-and CDC25A-depleted cells was consistent with an elevated CDK1 Y15 phosphorylation (Fig. 6A). As γ-H2AX was not stimulated significantly, the G 2 -M delay was unlikely to be caused by DNA damage associated with the experimental procedure.
Taken together, these data suggest a mechanism in which CDC25A is involved in compensating the alteration of cyclin A expression for controlling timely entry into mitosis.

Discussion
Both cyclin A and CDC25A have been reasoned to play critical functions in controlling the cell cycle, albeit definitive evidence pinpointing their precise roles is generally lacking. For cyclin A, this is in part due to the presence of other cyclin-CDK pairs functioning at similar parts of the cell cycle (cyclin B-CDK and cyclin E-CDK for G 2 -M and G 1 -S, respectively) as well as the multiple functions of cyclin A. Likewise, the presence of CDC25B and CDC25C complicates the interpretations of experiments on CDC25A, which also functions in both G 2 -M and G 1 -S. This is compounded by the confusion of whether CDKs aside from CDK1 are actually regulated by inhibitory phosphorylation in different cell lines. For example, inhibitory phosphorylation plays a major role in the regulation of CDK1 but only a minor role for CDK2 during the unperturbed cell cycle of HeLa cells (30). However, replacing CDK2 with a nonphosphorylatable mutant of CDK2 in HCT116 accelerates S phase entry by several hours (31).
In this study, we found that downregulation of cyclin A with either siRNA (Fig. 1A) or CRISPR-Cas9 (in combination with a degron-mediated conditional depletion system, Figs. 1B and S2) promoted the accumulation of CDC25A. Conversely, overexpression of cyclin A reduced CDC25A (Fig. 1D). Interestingly, the increase of CDC25A was evident even when cyclin A was only partially depleted or before completely destroyed (see for example Fig. 1, A and B).
Although CDC25A expression is cell cycle regulated (Fig. 3A), several lines of evidence suggest that the cyclin A-mediated regulation of CDC25A was not solely due to cell cycle disruption. First, the kinetics of CDC25A accumulation was rapid, occurring before changes of cell cycle distribution was observed with flow cytometry (Fig. 1B). Second, CDC25A accumulated in cyclin A-deficient cells synchronized in late G 1 , S, and G 2 (Fig. 3, B-D). Moreover, CDC25A was increased  Figure 4. Cyclin A regulates CDC25A independently to DNA damage-mediated degradation. A, depletion of cyclin A triggers CDC25A accumulation independently of CHK1 pathway. AID Cyclin A KO cells were incubated in the presence or absence of DI and/or AZD7762 (CHK1i) for 6 h. Lysates were prepared and analyzed with immunoblotting. B, cyclin A-deficient cells remain susceptible to DNA damage-mediated CDC25A degradation. AID Cyclin A KO cells were incubated with DI for 6 h to turn off AID-cyclin A before irradiated with 15 Gy of IR. The cells were harvested either immediately or after 3 h. Cellfree extracts were prepared and analyzed with immunoblotting. C, cyclin A depletion does not increase CDC25A protein stability. AID Cyclin A KO cells were pretreated with DI for 6 h to turn off AID-cyclin A. The cells were then either mock-treated or irradiated with 15 Gy of IR. After 3 h, cycloheximide (CHX) was added to abolish de novo protein synthesis. At the indicated time points after CHX addition, the cells were harvested and the abundance of CDC25A was determined by immunoblotting. The same DI-treated sample (lanes 5 and 9) was included in different gels for normalization. The relative CDC25A band intensity was quantified using densitometry and serially diluted standard curves (not shown) and normalized either to +DI (t = 0) (lower left-hand panel) or to t = 0 of individual samples (lower right-hand panel; mean ± SEM of two independent experiments). AID, auxin-induced degron; DI, Dox and IAA.

Regulation of CDC25A by cyclin A-CDK complexes
after cyclin A destruction even during a thymidine-induced S phase block (Fig. 3C). Finally, as depletion of cyclin A delayed interphase progression, as indicated by flow cytometry (Fig. 1B), BrdU incorporation (Fig. S2A), and live-cell imaging (Fig. 6B), it is expected that cyclin A depletion reduced instead of increased CDC25A if a cell cycle effect was involved.
Established mechanisms of CDC25A regulation by cyclin-CDK complexes center around phosphatase activity and protein turnover (10). Phosphorylation of Ser18 and Ser116 of CDC25A by cyclin B-CDK1 during mitosis stabilizes CDC25A (16). Phosphorylation of CDC25A's Ser283 by cyclin-CDK at late S/G 2 increases its G 2 -M-promoting activity without affecting its stability (32). On the other hand, cyclin D-CDK4/ CDK6 complexes are implicated in destabilizing CDC25A by phosphorylating Ser40, which primes the phosphorylation of Ser88 for SCF βTrCP -dependent degradation (33). Inhibition of cyclin A-CDK2 was also found to increase CDC25A, albeit no direct evidence of a decrease of protein turnover was obtained (34). However, we found that the accumulation of CDC25A after cyclin A destruction was not associated with an increase in protein stability (Fig. 4C). This was further supported by the lack of stabilization of exogenously expressed CDC25A (Fig. 5A). Instead, the increase of CDC25A appears to be caused by an increase of CDC25A mRNA (Fig. 5B). This is consistent with results from whole transcriptome analysis (Fig. S5). It is noteworthy that the transcriptome analysis revealed that in addition to CDC25A mRNA, many transcripts were upregulated in the absence of cyclin A. Further investigation will be needed to establish whether the accumulation of transcripts of other genes also results in an increase at the protein level.
Cyclin A affected the expression of specifically the CDC25A isoform. Several lines of evidence suggest that CDC25A is the most important isoform of the CDC25 family. There is evidence suggesting that centrosomally located CDC25B is an initiator of G 2 -M through its activation of the centrosomal subpopulation of cyclin B-CDK1, which is then able to initiate the autocatalytic loop to activate all the cyclin B-CDK1 (53). However, Cdc25B and Cdc25C are not required for mouse development (8). Furthermore, a case with homozygous deletion mutation of CDC25B in human also does not prevent live birth (with development of clinical defects including cataracts, dilated cardiomyopathy, and multiple endocrinopathies) (54). On the other hand, Cdc25A is essential for mouse development (7). Interestingly, similar to cyclin A, CDC25A has been implicated to play roles in both the G 1 -S and G 2 -M (9). Our data suggest that CDC25A is in a position to compensate for the variation of cyclin A during the cell cycle.
Both cyclin A and CDC25A were normally at their lowest levels in the cell cycle during G 1 . Hence it is not surprising that depletion of cyclin A (and the resulting increase in CDC25A) during G 1 did not significantly affect the timing of S phase  Figure 5. Cyclin A transcriptionally downregulates CDC25A. A, expression of exogenous CDC25A is independent of cyclin A. AID Cyclin A KO cells were transiently transfected with control or a plasmid expressing 3HA-CDC25A under a constitutive promoter. A plasmid expressing a blasticidinresistant gene was cotransfected. At 24 h after transfection, transfected cells were enriched by selection with blasticidin for 36 h. After recovery in normal medium for 24 h, the cells were incubated with DI for 6 h to turn off AID-cyclin A. Lysates were prepared and analyzed with immunoblotting. B, depletion of cyclin A increases CDC25A mRNA level. HeLa, AID Cyclin A KO , and AID Cyclin A KO CDK2 KO cells were cultured in the presence or absence of DI for the indicated time before harvested for RNA extraction. As a control, HeLa cells were transfected with CDC25A siRNA (siCDC25A) for 30 h before harvested. Reverse transcription and quantitative real-time PCR were performed using primers against CDC25A. Primers to actin were used as a normalization control (mean ± SEM of three independent experiments). The data were normalized to -DI for the individual time point. AID, auxininduced degron; DI, Dox and IAA.
Regulation of CDC25A by cyclin A-CDK complexes entry (our unpublished data). However, given that cyclin A is an integral component of the S phase-promoting engine, its downregulation is expected to promote replicative stress. In agreement with this hypothesis, double-strand breaks are generated after the loss of cyclin A (55). Likewise, there is evidence that overexpression of CDC25A can promote replicative stress by slowing down replication forks and inducing fork reversal (56). The resulting replication-derived DNA lesions are then carried into mitosis, causing chromosome segregation defects (57). Overexpression of CDC25A has been reported in various human cancer tissues (58). Overexpressed CDC25A can cooperate with oncogenes or loss of tumor suppressor genes in oncogenic transformation (59,60). Hence downregulation of cyclin A and the consequent CDC25A accumulation are expected to act synergistically in promoting genome instability.
We hypothesize that when cyclin A is downregulated, an increase of CDC25A could counterbalance the lowering CDK activities by activating cyclin B-CDK complexes. Conversely, overexpressed cyclin A could be compensated by a downregulation of CDC25A to delay mitotic entry. This relatively simple mechanism may ensure that mitotic entry could be fine-tuned to occur at an optimal time in spite of different expression of cyclin A (see Fig. 7 for a model). In support of this hypothesis, live-cell imaging analysis revealed that both cyclin A and CDC25A were rate limiting for G 2 -M in synchronized cells (Fig. 6B). Moreover, mitotic entry was further delayed in cyclin A-deficient cells by reducing CDC25A to a level similar to before cyclin A was destroyed (Fig. 6B).
As CDC25A is implicated to play a critical role in the DNA damage checkpoints (10), another possible route that cyclin A deficiency may lead to genome instability is through the impairment of this checkpoint. However, as CDC25A could still be rapidly degraded after irradiation in cyclin A-deficient cells (Fig. S4A), it is unlikely that genome instability associated with cyclin A downregulation was contributed by the effects of CDC25A on the DNA damage checkpoint.
Collectively, our results indicate that transcription regulation of CDC25A by cyclin A may set a threshold of CDC25 for mitotic entry. This is coupled with other CDK-dependent activation of CDC25A including direct activation of phosphatase activity, indirect activation through upstream regulators such as PLK1, as well as stabilization of the protein.

Cell culture
Cells were propagated in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) calf serum (for HeLa) or fetal bovine serum (for H1299) and 50 U/ml of penicillin streptomycin (Thermo Fisher Scientific).

Cell lines
HeLa (cervical carcinoma) used in this study was a clone expressing the tTA tetracycline transactivator (64). H1299 cells were obtained from American Type Culture Collection. Cell lines AID CDK1 KO (26) and AID CDK2 KO (25) were generated as previously described. mAID Cyclin B1 KO was a HeLa cell line expressing cyclin B1-mAID without endogenous cyclin B1 (Adrijana Crncec and RYCP, manuscript in preparation). AID Cyclin A KO cells from HeLa were generated by retroviral infection (25) using the construct AID-cyclin A in pRevTRE-AID/Hyg, followed by transfection of cyclin A CRISPR-Cas9 in pX330. AID Cyclin A KO cells from H1299 were generated by transfecting H1299 cells with AID-cyclin A in pUHD-SB-AID/Hyg, pSBbi-TIR1-tTA/Pur (26), cyclin A CRISPR-Cas9 in pX330, and Sleeping Beauty transposase (pCMV(CAT)T7-SB100; a gift from Zsuzsanna Izsvak; Addgene, #34879) before selecting with hygromycin and puromycin for 2 weeks. AID Cyclin A KO cells lacking CDK2 were generated by cotransfecting CDK2 CRISPR-Cas9 in pX330 and a plasmid expressing blasticidin-resistant gene (a gift from Tim Hunt, Cancer Research UK) into AID Cyclin A KO HeLa cells. After We showed that cyclin A-CDK complexes contribute to transcriptional repression of CDC25A during interphase, preventing premature activation of cyclin B-CDK1. Downregulation of cyclin A removes this repression and promotes CDC25A accumulation. We postulate that the excess CDC25A acts as a compensatory mechanism for cyclin A-depleted cells to overcome the G 2 -M barrier by promoting cyclin B-CDK1 activation.
enriching the transfected cells with blasticidin selection for 36 h, the cells were recovered in blasticidin-free medium for 48 h. In all the above cell lines, single cell-derived colonies were obtained by limiting dilution in 96-well plates.

Synchronization
Synchronization with double thymidine and NOC shake-off was performed as previously described (65). Briefly, cells were grown in medium containing 2 mM of thymidine for 14 h. The cells were then washed twice with PBS and cultured in fresh medium. After 9 h, the cells were incubated with a second round of 2 mM of thymidine for 14 h to obtain early S cells. Late S and G 2 cells were obtained at 6 h and 9 h after release from the double thymidine block, respectively.
For synchronization using RO3306 blockade, cells were first synchronized using 2 mM of thymidine for 14 h. The cells were then washed twice with PBS and cultured in fresh medium for 6 h before incubation with RO3306 for another 6 h. After washed twice with PBS, the attached cells were harvested as late G 2 cells.
For NOC shake-off synchronization, double thymidinesynchronized cells were released for 6 h before incubation with NOC for 6 h. Mitotic cells were collected by mechanical shake-off followed by centrifugation. G 1 cells were obtained by washing the mitotic cells with PBS twice and released into drug-free medium. After 3 h, attached cells were harvested as G 1 cells.
Double thymidine synchronization coupling with siRNA transfection was performed as described (66). In brief, cells were transfected with siRNA after the release from the first thymidine block. Fresh medium was then replenished before applying the second thymidine block. Cell-free extracts were prepared as described previously (67).

Ionizing radiation
IR was delivered with a caesium-137 source from a Gammacell 1000 Elite Irradiator (Nordion).

Live-cell imaging
Cells were seeded onto 24-well cell culture plates and placed into an automated microscopy system with temperature, humidity, and CO 2 control chamber (Zeiss Celldiscoverer 7). Images were captured every 5 min for up to 24 h. Data acquisition was carried out with Zeiss ZEN 2.3 (blue edition), and analysis was performed using ImageJ (National Institutes of Health). After mitosis, one of the daughter cells was randomly selected and continued to be tracked.

Flow cytometry
Flow cytometry analysis after propidium iodide staining was performed as previously described (68). Briefly, cells were trypsinized and washed with PBS. The cells were then fixed with ice-cold 70% ethanol and stained with a solution containing 40 μg/ml propidium iodide and 40 μg/ml RNase A at 37 C for 30 min. DNA contents of 10,000 cells were analyzed with FACSAria III (BD Biosciences).
For BrdU incorporation analysis, cells were pulsed with 10 μM of BrdU for 30 min before harvesting. The cells were then fixed with ice-cold 80% ethanol. After centrifugation at 2000 rpm for 5 min, the pellet was washed twice with PBS before incubated with freshly made 2 M HCl at 25 C for 20 min with gentle mixing. To neutralize the HCl, the cells were incubated with 0.1 M sodium borate buffer (pH 8.5) at 25 C for 5 min. The cell pellet obtained after centrifugation were washed twice with PBS and once with PBST (PBS with 0.5% Tween 20 and 0.05% w/v BSA). The cell pellet was resuspended in residue buffer and incubated with 2 μl of anti-BrdU antibody (DAKO) at 25 C for 1.5 h. The cells were then washed twice with PBST before incubating with 2 μl of Alexa Fluor-488 goat anti-mouse IgG antibody (Thermo Fisher Scientific) at 25 C for 1 h in the dark. After washing twice with PBST, the cells were subjected to propidium iodide staining and flow cytometry analysis.

Quantitative real-time PCR
Total RNA extraction, reverse transcription PCR, and realtime PCR were performed as previously described (25). Primers against CDC25A were: 5 0 -CCTCCGAGTCAACA-GATTCA-3 0 and 5 0 -GGGTCGATGAGCTGAAAGAT-3 0 . The expression of CDC25A mRNA was normalized to that of actin. Fold-change of sample normalized to control was calculated by 2 −ΔΔCt method.

Transcriptome analysis
Biological replicates of G 2 cells were harvested at 9 h after release from double thymidine synchronization. Immediately after harvesting, total RNA was extracted using NucleoSpin RNA kit (Macherey-Nagel). The samples were then air-dried in RNA stabilization tubes according to the manufacturer's instructions before dispatched for library preparation and RNA-sequencing (Genewiz).
For data analysis, fastq files were aligned using STAR algorithm (version 2.5.2a) using Homo sapiens GRCh38.83 as the reference genome. Reads were then counted using HTSeq-Counts (69) prior to downstream analysis. Statistical analyses on read counts were performed using the DESeq2 package (70) to identify differentially expressed (DE) genes between two experimental groups (±DI). Genes exhibited a fold change >1 or adjusted p value <0.05 were considered DE. Volcano plots were generated using DEBrowser (version 1.10.9; (71)) using R (version 3.6.3; www.R-project.org) to visualize DE genes upon cyclin A depletion.