Phosphorylation-mediated stabilization of Bora in mitosis coordinates Plx1/Plk1 and Cdk1 oscillations

Abstract

Cdk1 and Plk1/Plx1 activation leads to their inactivation through negative feedback loops. Cdk1 deactivates itself by activating the APC/C, consequently generating embryonic cell cycle oscillations. APC/C inhibition by the mitotic checkpoint in somatic cells and the cytostatic factor (CSF) in oocytes sustain the mitotic state. Plk1/Plx1 targets its co-activator Bora for degradation, but it remains unclear how embryonic oscillations in Plx1 activity are generated, and how Plk1/Plx1 activity is sustained during mitosis. We show that Plx1-mediated degradation of Bora in interphase generates oscillations in Plx1 activity and is essential for development. In CSF extracts, phosphorylation of Bora on the Cdk consensus site T52 blocks Bora degradation. Upon fertilization, Calcineurin dephosphorylates T52, triggering Plx1 oscillations. Similarly, we find that GFP-Bora is degraded when Plk1 activity spreads to somatic cell cytoplasm before mitosis. Interestingly, GFP–Bora degradation stops upon mitotic entry when Cdk1 activity is high. We hypothesize that Cdk1 controls Bora through an incoherent feedforward loop synchronizing the activities of mitotic kinases.

Introduction

The eukaryotic cell cycle consists of mitosis, during which chromatids are separated and cells divide, and interphase, when cells grow in size and replicate their genome. Normally interphase is at least 20 times longer than mitosis. The earliest embryonic stage consists, however, of a series of rapid cell divisions termed cleavages, which radically differ from those at any other developmental stage. A significantly shortened interphase is achieved by at least 2 major modifications of the cycle. The first is the elimination of cell growth, dispensing of most of the gap phases. The second is the acceleration of S phases enabled by employing many more origins of replication than employed later in life. Another striking discrepancy between the somatic and the cleavage cycles is that the former consists of a series of interdependent events, while the latter runs as a free oscillator.¹ The most obvious demonstration of this difference is the fact that extracts prepared from oocytes and triggered by fertilization will cycle even in the absence of DNA or nuclei. The discovery of MPF (maturation-promoting factor)² and the finding that its activity oscillates³ provided the biochemical manifestation of these oscillations. The discovery of mitotic cyclins that are periodically degraded⁴ identified the molecular mechanism of this oscillator. These 2 discoveries culminated in the breakthrough that showed that MPF activity results from the activation of Cdk1, previously described in fission yeast as cdc2,⁵ by mitotic cyclins.^6-8 Cdk1 activity is negatively regulated by its phosphorylation by Swe1 and Myt1 kinases. This phosphorylation is reversed by the Cdc25 phosphatase, which activates Cdk1.

Importantly, active Cdk1 stimulates degradation of mitotic cyclins through activation of the ubiquitin ligase APC/C^Cdc20. Thus, by a negative feedback loop, Cdk1 activation leads to Cdk1 inactivation. Because of a time delay between Cdk1 activation and cyclin degradation, Cdk1 activity starts to oscillate.⁹ However, in somatic cells and in oocytes before fertilization, the duration of mitosis may need to exceed the time delay between Cdk1 activation and cyclin degradation. In somatic cells, APC/C^Cdc20 activation depends also on the mitotic checkpoint, which monitors bi-orientation of chromosomes, thereby perturbing the oscillator. In oocytes, where the arrest can be very long, an activity termed CSF (CytoStatic factor) keeps APC/C^Cdc20 inactive until fertilization. Thus, Cdk1-APC/C negative feedback forms the basis of a cell cycle oscillator, where mitotic duration is determined by external perturbation of the feedback loop.

Numerous events in the embryonic cycle demonstrate a periodic nature.¹⁰ Many of them are clearly regulated by the oscillating Cdk1 activity. Some periodic events are, however, not downstream of Cdk1, implying that cyclin degradation is not the sole mechanism that regulates oscillations. These events need to be coordinated with Cdk1 activity, both during embryonic oscillations and during a prolonged mitosis.

Plx1(Xenopus)/Plk1(mammals)/Polo kinase 1(Drosophila)^11,12 is a kinase that plays essential roles in mitosis such as activating Cdc25¹³ to enable entry into mitosis, and activating the APC/C^Cdc20 ubiquitin ligase to degrade mitotic cyclins¹⁴ to enable mitotic exit.¹⁵ The subsequent inactivation of Plx1 is essential for progress into interphase.¹⁶ Plx1 activity is thus essential and tightly intertwined with both Cdk1 activation and inactivation. In somatic cells, Plk1 is degraded upon mitotic exit by the APC/C^Cdh1. Early embryos lack Cdh1, and the level of Plx1 remains constant; however, its activity, regulated by activating phosphorylations, oscillates between mitosis and interphase.¹⁷ The molecular mechanism underlying oscillations in Plx1 activity remains unclear.

More recently, Plk1 has been shown to be activated by the mitotic kinase Aurora A.^18,19 Aurora A, another mitotic kinase, is an APC/C^Cdh1 substrate in somatic cells, and its level is thus constant during early embryonic cycles. The activity of Aurora A in these cycles does, however, oscillate, presumably due to activating phosphorylation,²⁰ as is the case for Plx1.

Bora, short for Aurora Borealis, was first identified as an Aurora A activator in Drosophila, where it is required for asymmetrical cell divisions.²¹ We and others^18,19,22 have demonstrated that Bora serves as an adaptor or scaffold that enables Plk1 activation by Aurora A. Bora knockdown, overexpression, as well as expression of non-degradable Bora, all have a perturbing effect on several aspects of mitosis.^22,23 Bora depletion resulted in late activation of Plk1, thus delaying Cdk1 activation, presumably due to a delay in Cdc25 activation and delayed Wee1 degradation.¹⁸ Moreover, during checkpoint recovery, a special situation where Plk1 is essential for Cdk1 activation, Bora depletion abolished both Plk1 and Cdk1 activation.¹⁹ Bora knockdown also lengthened metaphase by activating the mitotic checkpoint and causing occasional multipolar spindles.^22,23 Elevated levels of Bora, by overexpression, or by expression of non-degradable Bora, also lengthened metaphase and perturbed the localization of both Aurora A and Plk1.

Bora degradation is mediated by its ubiquitination by the SCF^β-TrCP and requires both Plk1 and Cdk1. It seems that Cdk1 initially phosphorylates Bora on S252,²³ which primes phosphorylation of S497 and T501 by Plk1. The latter form a phosphodegron recognized by the SCF^βTrCP, leading to Bora ubiquitination and subsequent degradation by the proteasome.²² Bora is thus degraded, like mitotic cyclins, by the machinery it activates. However, when Bora is degraded, whether Bora is involved in embryonic oscillations of Plx1, and how Plx1 activity is sustained in a prolonged mitosis remain unclear.

Here we demonstrate that, like in somatic cells, embryonic Bora degradation requires phosphorylation by both Cdk1 and Plx1. Bora degradation modulates inactivation and activation of Plx1 at the exit from mitosis. In CSF extracts, Bora is phosphorylated on the Cdk1 consensus site T52, which blocks degradation until its calcium-dependent dephosphorylation by Calcineurin. This observation illustrates how the calcium surge caused by fertilization jump-starts the oscillations of the embryonic cycles. The role of Calcineurin in the initiation of cycling has been recently described,²⁴ and we provide here evidence that Bora could be a relevant substrate of its activity. We identified 13 additional novel phosphorylation sites on Bora, whose roles remain to be determined.

We further followed the dynamics of GFP–Bora degradation in somatic cells. We observed that most of the Bora is degraded about 2 h before mitosis, when Plk1 activity spreads to the cytoplasm. Strikingly this degradation stops when Cdk1 is activated during mitotic entry. Taken together, our work proposes that Cdk1 activity controls Bora by an incoherent feedforward loop²⁵ that synchronizes the activities of the mitotic kinases Cdk1 and Plk1/Plx1.

Results

Bora degradation by SCF^β-TrCP in CSF-arrested extracts is triggered by calcium

Bora has been identified as an adaptor that greatly enhances the activation of Plk1 by Aurora A to promote entry into mitosis. Active Plk1 targets Bora for degradation, but when Bora is degraded, and how Plk1 activation can be preserved after Bora degradation, remains unclear. In addition, Bora activity and degradation has so far been studied only in somatic cell cycles that considerably differ from those of early embryonic ones.

To study the role of Bora and its degradation in embryonic cycles we transcribed and translated ³⁵S labeled Bora in an in vitro-coupled reticulocyte lysate reaction (IVT) and added it to CytoStatic factor (CSF)-arrested extracts prepared from Xenopus oocytes. These extracts represent mature, metaphase II-arrested oocytes prior to fertilization that start to “cycle” when treated with calcium, which mimics fertilization. Figure 1A shows that Bora rapidly undergoes a considerable shift in its electrophoretic mobility in these extracts. Bora is extremely rich in serine and threonine, which make up 15.2% and 6.6%, of its residues, respectively. We therefore suspected that this mobility shift is caused by phosphorylation, and Figure 1A shows that phosphatase treatment indeed fully reversed the shift. In somatic cells the phosphorylation of Bora by Plk1 triggers its ubiquitination by the SCF^β-TrCP ubiquitin ligase, mediating its degradation by the proteasome. It has further been reported that Plk1 binding and phosphorylation of Bora depend on priming by Cdk1. CSF-arrested extracts express high levels of both active Cdk1 and Plx1; nevertheless, Bora remained largely stable (Fig. 1B). Once the CSF extracts were treated with calcium, Bora was rapidly degraded (Fig. 1B). The partial decrease in the levels of Bora in the absence of calcium can be explained by the slight leakiness of the frozen extracts. In mammalian cells, Bora degradation is mediated by its ubiquitination by the SCF^β-TrCP following the phosphorylation of its degron on S497 and T501. Consistent with these reports, Figure 1B shows that the Bora^S497A mutant was not degraded in calcium-treated CSF extracts.

Figure 1. Bora degradation by the SCF^β-TrCP in CSF-arrested extracts requires Plx1 and Cdk1 activities and is triggered by calcium. (A) IVT Bora was added to CSF extracts and incubated for 5 min. Samples were then diluted and incubated in the presence of lambda phosphatase (Upstate) and analyzed. (B) IVT Bora^wt or Bora^S497A were added to calcium-activated or -arrested CSF extracts, and aliquots were taken at indicated time points. (C) IVT Bora was added to calcium-activated CSF extracts in the presence of DMSO or the Cdk inhibitor roscovitine. (D) IVT Bora was added to calcium-activated CSF extracts in the presence of DMSO or the Plx1 inhibitor BI2536. (E) Flag-tagged β-TrCP was expressed in HEK-293T cells and immunoprecipitated. Beads with β-TrCP were used to bind IVT Bora^wt or Bora^S497A that were incubated in activated CSF extracts. Since the precise time of phosphorylation was unknown, and prolonged incubation would lead to degradation of Bora^wt, we took 1-μl aliquots of the mixture every minute for 8 min, and added them to the beads to perform the co-immunoprecipitation. (F) Plx1 was immunoprecipitated from calcium-activated or not activated CSF extracts with anti-Plk1 antibodies. The immunoprecipitated was blotted with Bora antibodies (top panel) or with Plk1 antibodies (bottom panel). XlBora co-immunoprecipitated with Plx1 from CSF extracts, and its level was considerably reduced upon calcium activation.

The requirement of Cdk1 and Plx1 for Bora degradation was tested by their respective inhibition with Roscovitine and BI2536. Figure 1C and D show that both drugs prevented Bora degradation. To test the requirement of S497 of Bora for its binding to β-TrCP we conducted a co-immunoprecipitation experiment depicted in Figure 1E. We expressed flag-tagged β-TrCP in HEK-293T cells and bound it to protein A beads with an anti-flag antibody. In parallel, we incubated in vitro-translated Bora^wt or Bora^S497A in CSF extracts to allow it to be phosphorylated. Since the precise timing of the phosphorylation required for binding is not known, we took 1 μl aliquots of the mixture every minute for 8 min and added them to the beads to perform the co-immunoprecipitation. As a control for the extracts, samples were taken at times 0, 8, and 30 min and run in parallel to the co-immunoprecipitation results. The results show that Bora incubated in CSF extracts binds β-TrCP, while the Bora^S497A mutant does not bind it.

The experiments described so far were performed with Bora transcribed and translated in vitro and added to CSF extracts. Bora has recently been identified in mouse oocytes;²⁶ however, we wondered whether Xenopus oocytes also express endogenous Bora. According to unigene transcript data (http://www.ncbi.nlm.nih.gov/UniGene/library.cgi?ORG=Xl&LID=6801) Bora transcript is expressed in oocytes at significant levels. According to this data set oocytes express about 1000 Bora transcripts per million, which is less than Plx1 (3500) but more than Aurora A (500). Bora is known to co-immunoprecipitate with Plk1 in mammalian cells.¹⁸ To confirm that CSF extracts express Bora, we immunoprecipitated Plx1 from CSF extract and immunobloted the precipitates with a Bora antibody.²⁷ Figure 1F shows that Xenopus Bora indeed co-precipitated with Plx1, indicating that the protein is present in the extract. Moreover when CSF extracts were activated by calcium, the level of Bora was considerably reduced. The reduction is presumably due to Bora degradation, as observed for the in vitro-expressed Bora (Fig. 1B).

We thus conclude that, similar to what happens in somatic cells and mouse oocytes,²⁶ Bora is present in CSF extract, and its degradation is mediated by Cdk1, Plx1, and the SCF^β-TrCP. However, in contrast to somatic cells, degradation of Bora in CSF extracts apparently requires an additional step triggered by calcium.

Phosphorylation of T52 of Bora protects it from degradation

We next addressed the requirement of calcium for initiating Bora degradation in CSF extracts. The calcium/Calmodulin-activated phosphatase Calcineurin, was the most likely candidate to initiate this degradation. Figure 2A shows that treatment of calcium-activated CSF extracts with the Calcineurin inhibitors cyclosporine (CsA) or FK-506 prevented Bora degradation. In contrast, treatment with Okadaic acid (OA), a protein phosphatase 1 and 2A inhibitor, did not inhibit degradation.

Figure 2. Phosphorylation protects Bora from degradation and is reversed by Calcineurin. (A) IVT Bora was added to calcium-activated CSF extracts treated with the indicated inhibitors, and aliquots were taken at indicated time points. (B) IVT Bora was phosphorylated in CSF extracts for 5 min and then supplemented with roscovitine and beads loaded with constitutively active or phosphatase-dead Calcineurin. The active Calcineurin de-phosphorylated Bora, while the phosphatase-dead one did not. Bora was not degraded in this experiment due to inhibition of Cdk1 activity.

To test whether Bora is a Calcineurin substrate we expressed and purified recombinant Calcineurin subunits A and B²⁴ and tested its ability to dephosphorylate Bora that was previously phosphorylated in CSF extracts. Figure 2B shows that incubation of phosphorylated Bora with constitutively active Calcineurin, but not with the phosphatase-dead Calcineurin, resulted in the reversal of the gel shift of Bora. Perhaps not surprisingly, this effect was apparent only when the experiment was performed in the presence of the Cdk inhibitor roscovitine, since presumably otherwise Cdk1 overrides Calcineurin, and Bora is degraded. This demonstrates that Bora is a target for opposing phosphorylation and dephosphorylation by Cdk1 and Calcineurin, respectively.

We assumed that some of Bora’s phosphosites are involved in this regulation. We therefore expressed GST-tagged Bora in bacteria, incubated it in CSF extracts, concentrated it on a gel, and subjected it to mass spec analysis. This analysis revealed a total of 15 phosphorylated residues (Table S1). Only one of these sites (S252) has previously been described, and the other 14 were all novel. We used site-directed mutagenesis to mutate 7 of the identified sites to alanine and tested the effect on Bora degradation. Mutation of 2 of these, S252 and T52, affected Bora degradation in CSF extracts. Bora^S252A was no longer degraded (Fig. 3A). This result is consistent with the report of Chan et al.²³ that S252 phosphorylation by Cdk1 mediates Plk1 binding, leading to degradation of Bora.

Figure 3. Bora degradation is triggered by T52 dephosphorylation by Calcineurin. (A) IVT Bora^wt, Bora^T52A, or Bora^S252A mutants were added to arrested or calcium-activated CSF extracts, and aliquots were taken at indicated time points. (B) Cdc13 (positive degradation control), Bora^wt, and Bora^T52A were in vitro translated and added to calcium-activated CSF extracts, untreated or treated with the Calcineurin inhibitors cyclosporine (CsA) and FK-506, and aliquots were taken at indicated time points.

Strikingly the mutation of T52 had an opposite effect on Bora degradation, rendering Bora^T52A constitutively unstable in CSF, even in the absence of calcium (Fig. 3A). This suggests that phosphorylation of Bora on T52 protects it from degradation in the presence of Plx1 and Cdk1. We next demonstrated that dephosphorylation of T52 by Calcineurin is essential for Bora degradation in calcium-activated CSF extracts. Figure 3B shows that Bora^T52A is degraded in such extracts even in the presence of Calcineurin inhibitors. This strongly implies that T52 is indeed the target of Calcineurin.

Bora degradation is essential for inactivation of Plx1

Plx1 protein levels are constant during the early cell cycles of Xenopus embryos. Plx1 is active, however, only in mitosis and gets inactivated upon entry into interphase. Moreover Plx1 inactivation is essential for entry into interphase.¹⁶ To follow Plx1 level and activity we used 2 antibodies. The first antibody is a mouse anti-Plk cocktail (Zymed) that identifies Plx1 and the second, AZ44, identifies only the phosphorylated active form of Plx1.²⁸ Following calcium activation, CSF extracts enter interphase, and Plx1 is dephosphorylated and inactivated. As we have shown above, Bora degradation requires Plx1 activity (Fig. 1D). Strikingly, when Bora was added to interphase extracts it got rapidly degraded despite the apparent absence of active Plx1. Bora degradation in interphase was sensitive to inhibition of both Plx1 and Aurora A (Fig. 4A and B). This suggests that either a low level of Plx1 activity is present in interphase extracts, or that addition of Bora can stimulate a transient Plx1 activation. Bora degradation also requires Cdk activity²³ and, indeed, degradation in interphase was also sensitive to the Cdk1/Cdk2 inhibitor roscovitine (data not shown). While Cdk1 is inactive in interphase due to cyclin B degradation, Cdk2-cyclin E is active these extracts²⁹ and is likely to phosphorylate Bora.

Figure 4. Bora degradation is essential for inactivation of Plx1 in interphase. (A) IVT Bora^wt was added to interphase extracts treated with DMSO or Plx1 inhibitor BI2536. (B) IVT Bora^wt was added to interphase extracts treated with DMSO or the Aurora inhibitor VX-680. (C) Bacterially expressed GST-BRCT (control), GST-Bora^wt, or non-degradable GST-Bora^S497A fusion proteins were added to CSF extracts with or without calcium. Aliquots were taken at indicated time points and probed with the specified antibodies.

To test the effect of Bora degradation on Plx1 inactivation during mitotic exit, we activated CSF extracts supplemented with bacterially expressed GST-BRCT³⁰ as negative control, GST-Bora^wt, and non-degradable GST-Bora^S497A with calcium. Aliquots were taken at the time points indicated in Figure 4C, and the levels of activated Plx1 were probed with the AZ44 antibody, which identifies only active Plx1.³¹ In extracts with GST-BRCT and GST-Bora, Plx1 was almost completely inactivated 40 min after calcium treatment. In stark contrast, extracts with non-degradable GST-Bora^S497A still had considerable amounts of active Plx1 at this time point, which remained clearly detectable even by 60 and 90 min.

These experiments show that Bora degradation can modulate the oscillations of Plx1 activity in early embryonic cycles. In interphase, Bora-Plx1 will reach a steady-state with low Plx1 activity and low Bora levels. Modification of Bora levels, at least in part by phosphorylation of the Cdk consensus site T52 provides a mechanism to control both presence and duration of Plx1 activity oscillations.

Non-degradable Bora retards the embryonic cell cycle

Inactivation of Plx1 is essential for exit from mitosis and injection of the constitutively active Plx1^T201D into Xenopus 2-cell embryos leads to cleavage arrest.¹⁶ Our observations of the role of Bora degradation in Plx1 inactivation suggested that non-degradable Bora^S497A would similarly perturb the embryonic cell cycle. To test this hypothesis, we injected 1 cell of fertilized 2-cell Xenopus embryos with in vitro-transcribed mRNA of Bora^wt, non-degradable Bora^S497A, or control mRNA (YFP-histone H2A) at 3 different concentrations. We followed and compared the development of the injected embryo halves with the non-injected halves. By 6 h, all of the embryos injected with control mRNA and most of those injected with Bora^wt survived and developed normally. However, viability of embryos that were injected with non-degradable Bora^S497A was considerably reduced, and in many of the surviving ones, the injected halves had much larger cells compared with their un-injected counterparts (Fig. 5). A minority of the embryo halves injected with the highest concentration of Bora^wt also had enlarged cells. The larger cells suggest cleavage retardation, indicating that Bora degradation is essential for normal cell cycle progress.

Figure 5. Non-degradable Bora retards the embryonic cell cycle. (A) Representative Xenopus embryos injected with Bora^wt (left) or Bora^S497A (right). One cell of 2-cell stage embryos was injected with indicated mRNA. Embryos were allowed to develop and photographed after 6 h. (B) Quantitation of the retarded cell cycle phenotype and of embryo’s viability. Ten embryos were injected for each mRNA concentration. Bars show the number of surviving embryos, while the number above each bar is the number of embryos displaying the retarded cell cycle phenotype.

GFP-Bora is stabilized in mitosis in somatic human cells

Since phosphorylation of Bora in CSF extracts prevented Bora degradation, we reasoned that phosphorylation-mediated stabilization of Bora could function as a general mechanism to promote mitotic progression. In somatic cells, both too high and too low levels of Bora lead to a mitotic delay, suggesting that its levels need to be carefully regulated for efficient progression through mitosis.^22,23

To address the regulation of Bora in human somatic cells, we made use of a polyclonal U2OS cell line stably expressing GFP-Bora under a tetracycline-inducible promoter. On a population level, tetracycline addition causes a considerable overexpression of GFP-Bora over endogenous Bora (Fig. 6A). However, we note that there is a large variability in GFP–Bora levels between individual cells. Since overexpression of Bora has been reported to impair mitotic spindle formation, we filmed cells after tetracycline induction and correlated the duration of mitosis to the level of GFP-Bora expression. Figure 6B shows that cells expressing high levels of GFP-Bora delay in mitosis. However, moderate levels of GFP–Bora expression do not interfere markedly with mitotic progression, indicating that these cells can be used to monitor the regulation of GFP-Bora.

Figure 6. GFP-Bora is stabilized in mitosis in somatic human cells. (A) Significant overexpression of GFP-Bora on population level. A polyclonal GFP-Bora TetON U2OS cell line was or was not treated with tetracycline. Cell lysates were probed with Bora and GAPDH antibodies. Left, equal loading of samples; right, 25-fold less loading of tetracycline treated sample. (B) Low levels of GFP-Bora expression causes no major delay in mitotic progression. Graph shows correlation of the GFP-Bora levels at the first image in mitosis to the length of mitosis in 97 cells from a single experiment. n.s. = P > 0.05; * = P < 0.05, Student t test. (C) Quantification of Plk1 target phosphorylation in the nucleus and cytoplasm using Plk1–FRET probe stably expressed in U2OS cells growing on fibronectin-coated micropatterns. Graph shows average and standard deviation of 10 cells. The image shows a time-lapse sequence of false colored inverted FRET ratio. Time between images is 20 min. (D) Quantification of GFP–Bora degradation in U2OS cells growing on fibronectin-coated micropatterns 16 h after tetracycline addition. Graph shows average and standard deviation of 10 cells, normalized at the intensities 3 h before mitosis. The image shows a time-lapse sequence of GFP–Bora levels in a single cell. Time between images is 30 min. (E) Quantification of GFP–Bora stabilization in mitosis upon nocodazole treatment. Graph shows average and standard deviation of 10 cells, normalized at the intensities 2 h before mitosis. The image shows a time-lapse sequence of GFP-Bora levels in a single cell. Time between images is 20 min.

It is well established that Bora is a target for Plk1-dependent phosphorylation, which results in its SCF^β-TrCP-dependent degradation. However, when Bora is degraded remains unclear. To allow for accurate quantification, we grew single cells on fibronectin-coated micropatterns, where cell shape and position remain constant over time. Growth on the micropatterns does not impair cell cycle progression.³² Figure 6C shows that, consistent with our previous observations,¹⁹ Plk1 activity could first be detected 5 h before mitosis in U2OS cells expressing a FRET-based probe that monitors Plk1 target phosphorylation.³³ We quantified GFP–Bora degradation in single live cells with moderate GFP–Bora expression levels, starting 16 h after tetracycline induction (Fig. 6D). Interestingly, GFP–Bora degradation started in G₂, 2 h before mitosis. GFP–Bora degradation was thus detected several hours after initial Plk1 activation, regardless of growth on micropatterns (Fig S1A). This indicates that similar to Xenopus interphase extracts (Fig. 4), Bora degradation is initiated after Plk1 activation. However, in human somatic cells, there is a 3-h delay between the initial Bora-dependent activation of Plk1 and degradation of GFP-Bora.

Interestingly, we detected GFP–Bora degradation at the same time as we detected Plk1 activity in the cytoplasm, suggesting that cytoplasmic Plk1 activity may be responsible for GFP–Bora degradation (Fig. 6C and D). As GFP-Bora appears largely cytoplasmic in interphase, although a small nuclear pool can be detected (Fig S1B), our data suggest that nuclear exclusion may delay GFP–Bora degradation in interphase. Indeed, although we cannot exclude that GFP-Bora shuttles between the cytoplasm and the nucleus, we failed to detect a rapid nucleo–cytoplasmic transport after bleaching (Fig S1B). Moreover, the localization of GFP-Bora did not change upon addition of Leptomycin B, indicating that GFP-Bora is not continuously exported from the nucleus by Exportin1 (Fig S1C). Thus our data are consistent with a model where differential regulation of subcellular localization allows Bora accumulation until 2 h before mitosis.

Since Bora is a co-activator of Plk1, and Plk1 activity is essential throughout mitosis, degradation of Bora before Plk1 activity reaches its peak in mitosis appears paradoxical. Based on the stabilization of Bora in Xenopus CSF extracts and the previously observed requirement of Bora for unperturbed mitotic progression, we hypothesized that not all Bora was degraded on mitotic entry, and that a fraction of it could be maintained to perform mitotic functions. Indeed Figure 6D and Figure S1A show that while GFP–Bora levels dropped by 80%, the remaining GFP-Bora is sustained throughout cell division. We repeated the experiment in the presence of nocodazole, which arrests cells in prometaphase. Figure 6E shows that also under these conditions approximately 20% of GFP-Bora remains in the cells. Thus, both in somatic and embryonic cycles Bora can be stabilized in mitosis.

Discussion

Starting the cycle

A major issue in the generation of oscillating activity is how to start it off. Mature oocytes already contain all the components that generate the oscillations, cyclins and active Cdk1, active Plx1 as well as Bora. Surprisingly, despite high Plx1 activity, Bora is not degraded. The trigger that sets off the oscillations by degrading Bora is given by the calcium surge that is generated by fertilization. This surge in calcium activates Calcineurin, which de-phosphorylates T52 of Bora, enabling its degradation. This observation puts this calcium-activated phosphatase as an upstream regulator that starts the oscillations of the fertilized egg. Mochida and Hunt²⁴ have demonstrated the role of Calcineurin in activation of the first mitotic cycle 7 years ago. The relevant substrate, or substrates, of this phosphatase have, however, not been identified to this day. We show here that Bora is dephosphorylated by Calcineurin (Fig. 2B), that its degradation depends on Calcineurin activity (Fig. 2A), and identified the relevant residue as T52 (Fig. 3). Together these observations suggest that Bora could be the relevant Calcineurin substrate predicted by Mochida and Hunt, or at least one such substrate.

However, in subsequent cycles, Calcineurin is no longer active, and we speculate that at later stages, T52 gets dephosphorylated by other phosphatases that “take over” as suggested by Mochida and Hunt. As Cdk1 phosphorylations often appear in clusters,³⁴ and in combination with those mediated by other kinases,³⁵ it is very likely that a combination of phosphorylations that may differ between CSF extracts and somatic cells regulate Bora stabilization in mitosis.

The embryonic oscillator

Early embryonic cycles, as well as extracts prepared from oocytes, oscillate in an autonomous manner between mitosis and interphase. The best-characterized part of these oscillations is those of Cdk1 activity (MPF). Ever since the discovery of mitotic cyclins,⁴ these oscillations have been attributed to the periodic degradation and re-accumulation of mitotic cyclins. However, the molecular mechanism giving rise to Plx1 oscillations has remained unclear, as in contrast to somatic cells, Plx1 is not degraded during embryonic cycles. Here we show that regulated degradation of Bora might regulate oscillations of Plx1 activity and cleavage cycles in the early embryo. The molecular base for oscillations of Plx1 activity is therefore Plx1-dependent degradation of Bora, constituting a negative feedback loop.

Negative feedback can generate oscillations only if a sufficient time delay is built into it.¹⁰ In the absence of a delay, both Bora levels and Plx1 activity would presumably decay to a low steady-state levels. Here we show that Bora is stabilized in mitosis, at least in part by phosphorylation on T52, thus introducing a delay in the negative feedback. In somatic cells, regulated localization may very well also contribute to a delay, as the majority of GFP-Bora is cytoplasmic and is only degraded when Plk1 activity spreads to the cytoplasm, 3 h after initial Plk1 activation in the nucleus.

Coordinating mitotic activities

For a successful mitosis, multiple enzymes need to be activated and subsequently deactivated. The negative feedback loop, where an activating enzyme triggers its own deactivation, is a powerful motif to ensure completion of mitosis. This motif is present during early embryonic divisions, where Cdk1 activity stimulates the anaphase-promoting complex (APC/C) to target the Cdk1 cofactor cyclin B for degradation. However, during somatic cell cycles, mitosis may need to be prolonged, and in unfertilized oocytes it can last for decades. In these occasions, an external signal in the form of the mitotic checkpoint (somatic cells) or CSF (oocytes) blocks APC/C activation, thereby interrupting the feedback loop. In this way, Cdk1 inactivation can efficiently be delayed to control the duration of mitosis. However, the multiple activities required for mitosis need to somehow be coordinated. Here we show that mitotic phosphorylation of a Cdk1 consensus site ([S/T*]PX[K/R]) on Bora delays Bora degradation. At the same time, Cdk1 regulates Plk1 activation, at least partly through phosphorylation of Bora.²³ We suggest that Cdk1 controls Bora levels through an incoherent feedforward loop,²⁵ thereby synchronizing the activities of 2 major mitotic kinases under a wide range of mitotic durations (Fig. 7).

Figure 7. An incoherent feedforward loop coordinates Cdk1 and Plk1 activity. Cdk1 phosphorylates Bora at S252, leading to increased association between Plk1/Plx1 and Bora. This facilitates Aurora A-mediated activation of Plk1/Plx1 and Plk1/Plx1-mediated phosphorylation of Bora, which leads to Bora degradation. At the same time, Cdk1-mediated phosphorylation of Bora at T52 inhibits Bora degradation. Cdk1 thus counteracts its own signaling cascade by an incoherent feedforward loop.

This motif will not operate in a vacuum devoid of other signaling events. In fact, Plk1 is present in the feedback loops that regulate Cdk1 activity, together with multiple other proteins.³⁶ The considerable number of phosphorylation sites on Bora suggests that it may act as a signaling node for multiple inputs. In our analysis we have identified 15 such sites and assigned roles for 2 of them (Table S1). Several of these novel sites, including T52, have recently been confirmed by Lee et al.,³⁷ who showed that the S274/S278 sites are targets of glycogen synthase kinase 3β. Neither of these screens is thus exhaustive, and Bora is likely to yield much more exciting insights in future.

Materials and Methods

Plasmids and constructs

Human Bora was obtained from the Mammalian Gene Collection (cDNA IMAGE clone 4098541) and cloned by PCR into the vector pEsp-T (for IVT expression), pGEX-4T3 (for bacterial expression), or into pCS2 (for mRNA expression). Calcineurin subunits A (pMALc2-MBP) and B (pET16b) were a gift from T Hunt.²⁴ pCMV-flag-dnβ-TrCP was a gift of Y Ben-Neriah. All mutagenesis was performed by site directed mutagenesis with pfu turbo (Stratagene) according to the manufacturer’s protocol.

Expression of flag-dnβ-TrCP

HEK-293T cells were transiently transfected with pCMV-flag-dnβ-TrCP by CaPO₄ coprecipitation. Scraped cells were extracted in ice-cold hypotonic lysis buffer containing 20 mM Hepes pH 7.4, 0.2 mM EDTA, 0.5 mM DTT, 1 mM NaVO₄, and protease inhibitors (Sigma) and incubated on ice for 10 min. Cells were then homogenized with a syringe, incubated for additional 15 min, and added to an equal volume of lysis buffer with 1 M NaCl.

Kinase, phosphatase and proteasome inhibitors

BI2536 and VX-680 (Axon Biochemicals) were used at final concentration of 5 μM. Cyclosporine, FK-506, and Okadaic acid (Santa Cruz Biotechnology) were used at final concentrations of 1 μM, 3 μM, and 2.5 μM, respectively. Roscovitine (Biomol) and MG132 (Megapharm) were used at final concentrations of 100 μM. All inhibitors were dissolved in DMSO.

Antibodies

Mouse anti-phosphorylated Plk1 (AZ44) and mouse anti-cyclin B1 (V152) were a kind gift of J Gannon. Mouse anti-Plk1 cocktail was purchased from Zymed Laboratories, mouse anti-GST from Santa Cruz Biotechnology, and mouse anti-flag M2 from Sigma. The Bora antibodies were raised against residues 79 to 559 of human Bora.²⁷

In vitro translation

Radiolabeled substrates were transcribed and translated for 90 min in a coupled in vitro reticulocyte system (TNT Promega) at 30° C in the presence of 35S Methionine and Cysteine (Perkin–Elmer).

Degradation assays

Xenopus CSF and Interphase extracts were prepared according to standard protocols.³⁸ All assays were performed at 23° C. For degradation assays, extracts were supplemented with cycloheximide (Sigma) at a final concentration of 50 μg/ml, and indicated inhibitors were added for 10 min prior to the experiment. Where indicated, extracts were treated with CaCl₂ (0.4 mM final concentration). Aliquots were removed at the indicated time points and resolved by gel electrophoresis followed by staining with coomassie to ensure identical amounts of protein in each well. Gels were dried and autoradiographed on a phosphoimager (Fuji MACBAS 3000).

Bacterial expression

Purification of GST-Bora

pGEX-4T3 encoding GST-Bora^wt or GST-Bora^S497A was used to transform BL-21-DE3 E. coli. Bacteria were grown to optical density of 0.6, and expression was induced overnight with 0.1 mM IPTG at 16° C. The bacteria were centrifuged, suspended in PBS, spun again, and the pellet was frozen in −20° C. Pellets were later suspended in cold PBS with protease inhibitor AEBSF and sonicated. The lysate was spun for 20 min at 4°, and the supernatant was collected. Glutathione-conjugated sepharose 4B beads (GE Healthcare) were added to the supernatant and rotated at 4° for 1 h. Beads were loaded onto a column, washed and eluted with 40 mM Glutathione in PBS. Fractions containing GST-Bora were pooled and dialyzed overnight against PBS supplemented with 1 mM DTT, frozen in liquid nitrogen, and stored at −70° C.

Purification of recombinant Calcineurin

pMALc2 expression vectors containing MBP fused to Calcineurin subunit A (constitutively active or protein-phosphatase dead mutants) and pET16b containing the Calcineurin subunit B, were used to transform BL-21-DE3 E. coli. Colonies were picked and grown overnight in 5 ml LB–amp media. This culture was used to inoculate 400 ml LB-amp and grown to O.D. 0.6. Expression was induced by 0.1 mM IPTG overnight at 23° C. After induction 200 ml of bacteria expressing subunit A were mixed with 50 ml bacteria expressing subunit B and centrifuged. The pellet was washed in wash buffer (WB) containing 20 mM TRIS-HCl pH 7.5, 0.2M NaCl, 10 mM ascorbic acid, and 1 mM DTT, and recentrifuged. The pellet was frozen and thawed on ice before suspension in Extraction Buffer (ExB)-WB supplemented with 1 µM CaCl₂, 0.05% Tween-20, 1 mM PMSF, and 5 mM benzamidine. Cells were sonicated, and the lysate was spun at 14 000 rpm at 4° for 15 min. The supernatant was mixed with washed amylose beads (New England Biotechnology) and rotated for 30 min at 4°. Beads were washed twice with ExB and once with WB supplemented with 1 µM CaCl₂. Beads loaded with the Calcineurin holoenzyme were directly used for experiments without elution.

Immunoprecipitation and immunoblotting

Protein extract from HEK-293T cells transfected with pCMV-flag-β-TrCP was mixed with 3 µg anti-flag antibody and rotated for 1 h at 4° C. Subsequently, 25 µl of agarose protein A (Santa-Cruz Biotechnology) were added for 30 min. Beads were washed 3 times and used to bind IVT radiolabeled Bora. Bora^wt or Bora^S497A was added to indicate extracts and aliquots were taken for 8 min every minute and added to flag- β-TrCP-loaded beads. Beads were rotated 20 min at 4°, washed and eluted by boiling in protein loading buffer. In parallel to immunoprecipitation, samples from extracts were taken to control for degradation. Samples were analyzed by SDS-PAGE followed by autoradiography. Standard methods were used for immunoblotting.³⁹

mRNA expression and embryo injection

mRNA was prepared in a 20 µl reaction mix containing 2 µg of linearized pCS2 with each insert, 1 µl cap analog (Pharmacia), 2 µl rNTP mix, 0.5µl RNA Guard, 2 µl SP6 polymerase (Roche), and the supplied buffer. Reactions were incubated for 4 h at 37° and mixed with 160 µl water, 20 µl sodium acetate, 1 µl glycogen, and 600 µl ethanol. Samples were then spun at 4° for 30 min, and the supernatant was discarded. Pellets were washed with 500 µl ethanol and recentrifuged for 15 min, then air dried for 10 min. mRNA samples were dissolved in 30 µl nuclease-free water and incubated at 60° C for 10 min. Concentrations were measured using a Nano-Drop and varied between 370–400 ng/µl.

Xenopus laevis frogs were purchased from Xenopus I or NASCO. Embryos were obtained by in vitro fertilization, injected in 1× MBSH (modified Barth solution), incubated in 0.1% MBSH, and staged according to Nieuwkoop and Faber.⁴⁰ Embryos were injected at the 2-cell stage with in vitro transcribed mRNA and visualized with a binocular microscope.

Mapping of Bora phosphorylation sites

For detection of phosphorylation sites bacterially expressed GST-Bora was incubated in CSF extracts (arrested or calcium-activated) in the presence of MG132 for 10 min at 23° C. Extracts were then diluted in cold PBS, and a GST-pull-down was performed using Glutathione beads (Amersham). Beads were washed, boiled, and resolved by SDS-PAGE. Bora bands were excised from gels after Coomassie blue staining. All the samples were digested by trypsin, analyzed by LC-MS/MS on Orbitrap, and identified by Sequest 3.31 software against the Amphibian section of the NCBInr database. The results were filtered according to the Xcore value and the probability. The results presented in Table S1 were obtained from several repetitions of this analysis both at the MS facility of the Hebrew University Medical School of Jerusalem (Ofra Moshel) and at the National Proteomic Facility of the Technion in Haifa (Arie Admon).

Inducible expression of GFP-Bora

U2OS cells stably expressing Tet repressor¹⁹ were transfected with pCDNA4/TO-GFP-Bora and selected with 400 µg/ml Zeocin. Cells were induced with 1 μg/ml tetracycline for 16 h prior to live cell imaging. For Figure 6, only cells expressing low amounts of GFP-Bora that do not markedly interfere with the cell cycle progression were selected for quantifications.

Live cell imaging

U2OS cells expressing GFP-Bora or Plk1-FRET probe were imaged in 96-well imaging plate (Falcon) or on custom-designed 80 × 15-μm fibronectin-coated micropatterns (CYTOO) on a DeltaVision Spectris imaging system, using a 20× air objective, NA 0.75. Images were quantified using ImageJ software. Imaging and quantification of ratiometric FRET was performed as previously described.³³ The Plk1-responsive FRET-based probe is described in reference 19. For Figure S1B, cells were imaged on a Zeiss LSM510 confocal microscope, using NA 1.4 oil objectives. Bleaching was performed by illumination using 100% laser power of a 488 nm argon laser.

Immunostaining

Cells were fixed and permeabilized with 3.7% formaldehyde followed by cold methanol. Mouse anti Cyclin B1 (V152) was from cell signaling and rabbit anti GFP was from Abcam.