|Home | About | Journals | Submit | Contact Us | Français|
A circadian clock coordinates physiology and behavior in diverse groups of living organisms. Another major cyclic cellular event, the cell cycle, is regulated by the circadian clock in the few cases where linkage of these cycles has been studied. In the cyanobacterium Synechococcus elongatus, the circadian clock gates cell division by an unknown mechanism. Using time-lapse microscopy, we confirm the gating of cell division in the wild type, and demonstrate the regulation of cytokinesis by key clock components. Specifically, a state of the oscillator protein KaiC that is associated with elevated ATPase activity closes the gate by acting through a known clock output pathway to inhibit FtsZ ring formation at the division site. An activity that stimulates KaiC phosphorylation independently of the KaiA protein was also uncovered. We propose a model that separates the functions of KaiC ATPase and phosphorylation in cell division gating and other circadian behaviors.
In organisms from microscopic cyanobacteria and fungi to plants and animals, genetically programmed daily cycles, known as circadian rhythms, pervade various aspects of physiology and behavior (Bell-Pedersen et al., 2005). In the cyanobacterium Synechococcus elongatus, the timing of cell division (Mori et al., 1996), global patterns of gene expression (Liu et al., 1995), and compaction of the chromosome (Smith and Williams, 2006) are all controlled by a circadian clock that exhibits the same properties as in eukaryotic organisms. However, the S. elongatus clock is distinct in components, mechanism, and evolutionary history from eukaryotic clock systems (Dong and Golden, 2008; Mackey and Golden, 2007).
In S. elongatus, three neighboring genes, kaiA, kaiB, and kaiC, encode proteins of the central oscillator. Inactivation of any of them abolishes the clock, as does overexpression of KaiA or KaiC (Ishiura et al., 1998). KaiC is an autokinase, autophosphatase, and ATPase; in complex with KaiA and KaiB, KaiC displays a daily rhythm of phosphorylation at residues Ser 431 and Thr 432 (Nishiwaki et al., 2004; Xu et al., 2004) both in vivo and in vitro (Nakajima et al., 2005; Tomita et al., 2005). KaiA stimulates KaiC autophosphorylation and KaiB opposes KaiA's stimulatory activity (Iwasaki et al., 2002; Kim et al., 2008; Rust et al., 2007; Williams et al., 2002). The oscillation of KaiC phosphorylation in a mixture of the three Kai proteins and ATP in vitro (Nakajima et al., 2005) suggests that the phosphorylation cycle is the fundamental timekeeping mechanism in cyanobacteria. However, the ATPase activity of KaiC also oscillates in a circadian manner, is intrinsically temperature-compensated, and determines circadian period length, suggestive of a timekeeping role that may be separable from the phosphorylation cycle (Terauchi et al., 2007). A gene expression rhythm persists in the absence of a KaiC phosphorylation rhythm (Kitayama et al., 2008); thus, other aspects of KaiC, such as the ATPase activity, may underlie the basic timing mechanism instead of, or in addition to, KaiC phosphorylation.
Temporal information from the cyanobacterial oscillator is broadcast to downstream genes via the histidine protein kinase SasA, whose autophosphorylation is stimulated by interaction with KaiC (Iwasaki et al., 2000; Smith and Williams, 2006). SasA then transfers the phosphoryl group to RpaA, a response regulator with a DNA binding domain. Disruption of either sasA or rpaA results in severely damped rhythms or arrhythmia, depending on growth conditions (Iwasaki et al., 2000; Takai et al., 2006b).
An input pathway that includes CikA, LdpA, and Pex relays environmental information to the oscillator for synchronization (Dong and Golden, 2008). Both CikA and LdpA sense light indirectly through cofactors that perceive changes in the cellular redox state, which varies with photosynthetic activity (Ivleva et al., 2005; Ivleva et al., 2006). CikA is found in a complex with LdpA, KaiA, KaiC, and SasA in vivo, but no direct biochemical interaction has been detected between CikA and the oscillator. A cikA null mutant exhibits short-period, low-amplitude gene expression rhythms, and fails to reset the phases of rhythms after an environmental cue (Schmitz et al., 2000); additionally, it is defective in cell division, resulting in elongated cells (Miyagishima et al., 2005).
Cell division is a cyclic event that is tightly regulated by and coordinated with other cellular activities. Few studies to date have focused on the interaction between the cell and circadian cycles, with even fewer molecular details. For example, in regenerating liver cells of mice, circadian clock proteins directly control the expression of Wee1, a kinase that inhibits the entry into mitosis (Matsuo et al., 2003). Cell division is also gated by the clock in mouse fibroblast cells cultured in vitro (Nagoshi et al., 2004) and in S. elongatus (Mori et al., 1996). The rate of DNA synthesis is constant in the cyanobacterium and not phase-dependent, suggestive of regulation further downstream—such as cytokinesis (Mori et al., 1996). The mechanism of cell division gating in S. elongatus has remained unknown in the face of rich molecular details of the cyanobacterial circadian clock. Elucidation of this pathway would tie the oscillator to a key fitness component of cell physiology.
Here, we show that elevated ATPase activity of KaiC closes the cell division gate, and demonstrate a linear signal transduction pathway from the input components to the central oscillator and to the output pathway in the regulation of cell division. We also show that localization of the bacterial tubulin homolog FtsZ is a target of clock control. This work revealed the action of a novel KaiA-independent, but CikA-suppressed, activity that stimulates KaiC autophosphorylation. A model of the relationship of KaiC ATPase and phosphorylation activities, and how they are incorporated with the input and output pathways of the clock, emerges from this work.
A previous report showed the gating of cell division in a population of S. elongatus cells measured over several circadian cycles (Mori et al., 1996). That work predated the identification of molecular components of the cyanobacterial clock, and did not address the process in individual cells. CikA is the only clock component that has been reported to play a role in cell division (Miyagishima et al., 2005); therefore, we tested the requirement of cikA for the gating of cell division. Using time-lapse microscopy, we directly monitored growing cells for three days, recording events of cell division and the circadian rhythm of kaiBC promoter activity as reported by a destabilized yellow fluorescent protein, YFP-SsrA(LVA) (Chabot et al., 2007).
Individual cikA mutant cells show rhythmic gene expression, with a period of 22.0 ±1.1 h, whereas the WT cells oscillate with a period of 24.9±1.0 h (Figure 1A), consistent with results from luciferase reporters (Schmitz et al., 2000). To address whether and how the circadian clock gates cell division, all division events were assigned to their corresponding circadian phases, normalized into one circadian period 0 ~ 2 π, and plotted as a histogram. To avoid sampling bias, we ensured that the initial circadian phases were evenly distributed; i.e. the cells examined are unsynchronized (data not shown). The occurrence of cell division in the WT is apparently suppressed around the peak of fluorescence (Figure 1B), indicating that cell division is gated. In the cikA mutant, a similar dip in the histogram was seen, although the overall occurrences of cell division during this window are higher and the duration of the inhibition is longer.
As a control we monitored cell division in an arrhythmic kaiC null mutant. Thus, the division events cannot be assigned to specific circadian phases. However, when we plotted the distribution of doubling time of all cells monitored (Figure 1C), the kaiC mutant had only a single peak in contrast to two peaks in the WT and cikA mutant. We reasoned that when gating is active, the doubling time of cells whose life span covers the window of gating is prolonged compared to those that do not, resulting in two populations of cells with distinct doubling times. On the other hand, the doubling time of arrhythmic mutants follows a Gaussian distribution pattern in the absence of a gate. Both of these complementary types of analysis are able to detect gating of cell division. In conclusion, cikA is not essential to the gating of cell division, although its absence causes elongated cells and affects characteristics of the gate.
The phenotypic effects of a cikA null mutation on both circadian rhythms and cell length could be due to CikA's involvement in two pathways that separately govern circadian rhythms and cell division; alternatively, CikA may affect a clock pathway that regulates the gating of the cell cycle. If the second hypothesis is true, mutation of some clock genes would suppress the cikA null cell elongation phenotype. We made double mutants of cikA and central oscillator genes. Cell length of various mutants, grown under the same conditions simultaneously to minimize variations caused by cell density and/or light intensity (Figure S1), was measured and compared. As expected, cikA mutant cells are significantly longer than the WT (Figures 1D and 1E). In the ΔkaiCΔcikA mutant, however, the absence of KaiC suppresses the cell division defect of cikA. The ΔkaiAΔcikA and ΔkaiBΔcikA mutants are still elongated. These data suggest that cikA regulates cell division through the clock, and that KaiC lies downstream of CikA in this pathway.
To address the roles of other clock genes in cell cycle control, single mutants were examined. Cells of a kaiB mutant average almost twice as long as WT, whereas mutants of kaiA, kaiC, or ldpA have normal cell lengths (Figures 1D, 1E and data not shown). The extensive evidence of KaiB-KaiC interaction suggested that inactivation of kaiC would also suppress the kaiB phenotype. Indeed, the cell length of a kaiBC double mutant is like WT (Figures 1D and 1E). We propose that the role of KaiB in determining cell length is the same as its function in modifying KaiC phosphorylation state.
KaiB is known to inactivate KaiA to initiate KaiC dephosphorylation (Rust et al., 2007). Loss of kaiB abolishes circadian rhythms (Ishiura et al., 1998) and results in constitutively phosphorylated KaiC both in vivo (Figure 2C) and in vitro (Iwasaki et al., 2002; Kitayama et al., 2003). CikA also affects KaiC phosphorylation state (Ivleva et al., 2006). Therefore, we asked whether constitutively phosphorylated KaiC is the cause of cell division inhibition. To stimulate KaiC phosphorylation, we overexpressed KaiA from an IPTG (isopropyl-β-D-thiogalactopyranoside)-inducible trc promoter in a WT background, which abolishes circadian rhythms of gene expression and KaiC phosphorylation (Ishiura et al., 1998; Kitayama et al., 2008), and in a kaiC null background in parallel. Cells overexpressing KaiA in the presence of KaiC are significantly longer than those in the kaiC null background and in WT cells after IPTG addition, whereas no difference was observed before induction (Figures 2A and 2B). Note that, during the 5-day induction period, WT cells also lengthened, which is due to higher cell density and/or lower light penetration and not related to IPTG (Figure S1). Interestingly, the cikA mutant's cell length is much more affected by culture density than the WT (Figure S1A). We confirmed that KaiC becomes and stays highly phosphorylated after IPTG induction (Figure 2C and data not shown). We further demonstrated that the cell elongation phenotype is reversible by removing IPTG after days of KaiA-YFP overexpression (Figure S2).
Recently, several KaiC variants were created that are constitutively hyper- or hypophosphorylated independent of KaiA or KaiB (Kim et al., 2008). KaiC487, a truncation mutant missing the last 32 aa, and the missense mutant KaiCE444D are both highly phosphorylated, whereas KaiC497 remains unphosphorylated. We expressed these alleles in a kaiC null mutant and, consistent with the hypothesis that phosphorylated KaiC is an inhibitor of cell division, both KaiC487 and KaiCE444D caused cell elongation whereas KaiC497 did not (Figures 2D and 2E).
Cells of the ΔkaiAΔcikA mutant are elongated (Figures 1D and 1E), which was intriguing because all evidence had suggested that phosphorylated KaiC inhibits cell division, yet KaiA is required for KaiC phosphorylation. We examined the KaiC phosphorylation state in this strain and found it ~60% phosphorylated, in contrast to being fully unphosphorylated in the ΔkaiA strain (Figure 2C). No oscillation in gene expression or KaiC phosphorylation is observed in the ΔkaiAΔcikA mutant (Figure S2). We concluded that an unknown activity is able to stimulate KaiC phosphorylation independently of KaiA, and that this cryptic activity is inhibited by CikA. Furthermore, KaiC does not need to be fully phosphorylated to inhibit cell division. Either the absolute amount of phosphorylated KaiC is sufficient, or some state of KaiC other than phosphorylation is the underlying mechanism.
KaiC phosphorylation is a dynamic process and four forms of KaiC cycle in a stepwise fashion (Figure 3A) (Nishiwaki et al., 2007; Rust et al., 2007). By replacing S431 and T432 with Glu, Asp, or Ala, Nishiwaki et al. (2007) mimicked different phosphoforms of KaiC and characterized certain properties of the Kai oscillator in vitro. We hypothesized that the KaiCS431D, T432E (KaiC-DE) mutant, which mimics fully phosphorylated KaiC, would inhibit cell division whereas KaiC-AA, mimicking unphosphorylated KaiC, would have no effect. We made other phosphomimetics to further dissect the relationship between KaiC phosphorylation and cell division. The results, however, were opposite to our predictions: the strain expressing KaiC-DE has WT cell length, but those expressing either KaiC-AA or KaiC-AE show inhibition of cell division (Figures 3B and 3C). All variants except KaiC-SE were expressed at levels comparable to those in WT (Figure 3D), ruling out the possibility that KaiC-DE abundance is insufficient to inhibit cell division; KaiC-SE was undetectable and thus uninformative (data not shown). We have clearly shown that phosphorylated KaiC or a specific activity coupled with it causes cell elongation, whereas unphosphorylated native KaiC has no effect. Therefore, these KaiC phosphomimetics are not completely faithful representations of the different KaiC phosphoforms. The fact that KaiC-AA cannot be phosphorylated, does not mimic phosphorylated KaiC, and yet still inhibits cell division suggests that KaiC phosphorylation is not required for cell elongation.
The underlying signal that inhibits cell division must be a state or activity of KaiC that is usually coupled with, but functionally separable from, phosphorylation. Aside from autokinase and autophosphatase activities, KaiC has an unusually low ATPase activity that fits this description (Terauchi et al., 2007). Two Walker's A motifs are located at the N- and C-termini of KaiC (Figure 4A). The N-terminal motif is important in the hexamerization of the protein and the C-terminal motif is essential for autophosphorylation (Hayashi et al., 2004; Hayashi et al., 2006). KaiC hydrolyzes ATP in a circadian manner, and phosphorylation of KaiC accounts for only a negligible portion of the total ATP consumed per cycle.
About 15 ATP molecules are consumed per KaiC monomer per day when incubated alone (hereafter, 1 unit= 1 ATP × monomer-1 × day-1); this number increases about 2-fold in the presence of KaiA. The KaiC-AA mutant has a constitutively high ATPase activity of ~27 units, whereas KaiC-DE hydrolyzes ~11 units (Terauchi et al., 2007). Thus, we hypothesized that elevated ATPase activity, rather than phosphorylation of KaiC, triggers inhibition of cell division. To test this hypothesis, we made four ATPase mutant variants of KaiC-AA based on previous reports (Hayashi et al., 2004; Hayashi et al., 2003; Nishiwaki et al., 2000) and expressed them in a kaiC null strain. KaiCK52H-AA and KaiCcatE1-AA are expected to inactivate ATPase activity in the N-terminal domain and KaiCK294H-AA and KaiCcatE2-AA to inactivate activity in the C-terminal domain. All ATPase mutants have reduced cell length compared to KaiC-AA, but three out of the four (KaiCcatE1-AA, KaiCK294H-AA, and KaiCcatE2-AA) mutants are still elongated compared to the WT (Figure 4C). We then attempted to express a variant in which the catalytic Glu residues from both domains are replaced (KaiCcatE1catE2-AA). Although this mutant has WT cell length, immunoblot analysis showed that it is not detectable, nor is KaiCK52H-AA, whereas the other three mutants are expressed at levels comparable to KaiC-AA (Figure 4B).
We then purified recombinant KaiC mutant proteins and measured their ATP hydrolysis activity in vitro by using C18 reversed-phase HPLC to separate and quantify ADP from ATP. We predicted that mutations causing cell elongation would also elevate ATPase activities. We chose KaiC487, KaiC497, KaiC-AE, KaiCcatE1-AA, and KaiCcatE2-AA because their values have not been reported, and included WT KaiC, KaiC-DE, and KaiC-AA as controls. Consistently, KaiC (12.6 ± 1.1), KaiC-DE (16.7 ± 0.8) and KaiC-AA (27.1 ±1.1) values were comparable to published data (Figure 4E and Terauchi et al., 2007). KaiC487 hydrolyzes 59.6 ± 4.1 ATP units, 4-5 fold higher than the WT, whereas KaiC497 hydrolyzes 14.3 ± 0.3 units. Surprisingly, values from KaiCcatE1-AA and KaiCcatE2-AA were higher than that of KaiC-AA (Figure 4E), and KaiC-AE's ATPase activity is the same as WT. Because these data were not consistent with our hypothesis, we tested to see whether adding KaiA and KaiB to the in vitro system at appropriate ratios to mimic more closely the in vivo system would alter the ATPase activities, recognizing that we had expressed these mutants in a kaiC null background where both KaiA and KaiB were present. Indeed, KaiC-AE and KaiC-AA hydrolyze at least 330 units (which is saturation of the assay) in the presence of both KaiA and KaiB; values for KaiA and KaiB alone were negligible (0.4 units). In contrast, KaiCcatE2-AA activity is not altered, and KaiCcatE1-AA activity is stimulated only 2-fold, when mixed with the other Kai proteins (Figures 4E and 4F); both values are lower than for KaiC-AA, but higher than the WT, which could explain the elongated cell phenotype seen in these two mutants. KaiC-AA and KaiC-AT do not interact with KaiB, but do with KaiA (Nishiwaki et al., 2007); thus, we speculate that even when both KaiA and KaiB are present, KaiC-AA and KaiC-AE would likely interact only with KaiA and have their ATPase activities elevated.
To further test our hypothesis, we compared cell lengths among different KaiC period mutants, because KaiC mutants with higher ATPase activities exhibit shorter periods, whereas those with lower activity have longer periods (Terauchi et al., 2007). We picked two KaiC mutants whose ATPase activities have been quantified—KaiCR393C with a 15-h period and KaiCA251V with a 46-h period—as well as two other KaiC period mutants—KaiCR321Q (22 h) and KaiCT409A (27 h)—to compare their cell lengths. As predicted, KaiCR393C mutant cells were longer than the KaiCR321Q mutant, and both were longer than the WT and long period mutants, which were not different from each other (Figure 4D).
Elevated ATPase activity could be achieved by an increase in either the enzymatic activity of individual KaiC molecules or the abundance of the enzyme. Our results demonstrated inhibition of cell division by elevated intrinsic ATPase activity; therefore, we tested whether an increase of total ATPase activity has a similar effect. We overexpressed WT KaiC in a kaiC null background, confirmed abundance by immunoblot (data not shown), and monitored cell length. No difference was seen between cells that overexpress KaiC and those that do not (Figure S3), indicating that the intrinsic ATPase activity of KaiC, rather than the total amount of ATP hydrolyzed, signals inhibition of cell division. This result is in line with the remarkably low ATPase activity exhibited by KaiC, making it unlikely to affect the cellular ATP pool if the measured values hold true in vivo.
Mutants of the clock output pathway genes sasA or rpaA do not have cell elongation phenotypes (Figures 5A, 5B and data not shown). To test whether this output pathway governs the circadian gating of cell division, cell lengths of double mutants of cikA with sasA (ΔcikAΔsasA) and kaiB with sasA (ΔkaiBΔsasA) or rpaA (ΔkaiBΔrpaA) were compared to the WT and single mutants. In all cases, mutation of sasA or rpaA restored cells to WT length (Figures 5A and 5B). Because kaiBC expression is significantly reduced in the sasA mutant (Iwasaki et al., 2000), it was possible that the suppression phenotype in the double mutants was merely a result of a reduction in KaiC abundance. Therefore, we introduced a copy of WT kaiC expressed from the trc promoter into the ΔkaiBΔsasA background and elevated KaiC expression by adjusting the concentration of IPTG. Cells maintained WT length regardless of KaiC abundance (Figures 5C and 5D), suggesting that SasA-RpaA is indeed downstream of KaiC in the control of cell division. Thus, we hypothesized that gating of cell division would be abolished in the sasA mutant. Using the same YFP-LVA reporter from Figure 1, we examined cell division in the sasA mutant using time-lapse microscopy and did not detect rhythmic gene expression. Nonetheless, the distribution of doubling time showed a single peak (Figure 1C), indicating that no gating of cell division is present when the clock output is disrupted.
Circadian gating of cell division is proposed to occur at the cytokinesis step (Mori et al., 1996). Consistently, the midcell ring localization of FtsZ, a conserved bacterial homolog of tubulin, is significantly reduced in the cikA mutant although FtsZ abundance is unaffected (Miyagishima et al., 2005). Targeting FtsZ to regulate cell division is a common theme in bacteria (Kawai et al., 2003; Weart et al., 2007). If the circadian clock blocks FtsZ localization at the division site, an increase in the FtsZ level might overwhelm the inhibition and allow cell division. Thus, we expressed ftsZ from trc in the cikA, kaiB, and kaiC-AA mutant backgrounds. In all cases, cell length was significantly reduced even in the absence of IPTG (Figure 6A), suggesting that an elevated level of FtsZ suppresses the cell division defect caused by the clock mutants. Immunoblots confirmed that the level of FtsZ was the same between the WT and elongated mutants, but was elevated about 2-fold in the suppressor mutants that have ectopic ftsZ expression (Figure 6B).
We then examined FtsZ localization in these strains by immunofluorescence microscopy (Figure 6A). FtsZ was frequently mislocalized in the elongated mutants, and was often concentrated at the cell poles or observed in patches, with some rings localized away from midcell along the elongated cells. In contrast, ftsZ-compensated mutants have a higher frequency of correctly localized FtsZ rings. Therefore, we conclude that FtsZ ring localization and assembly are events that are targeted by the circadian clock machinery in the regulation of the timing of cell division.
The circadian clock imposes a checkpoint on cell division by actively inhibiting cytokinesis at specfic circadian times; the absence of a clock does not affect cell division. By default the gate is open so that cell division operates independently of the clock, and closes as ATPase activity of KaiC rises above a threshold value (Figure 6C). The signal is transmitted through SasA-RpaA to inhibit midcell FtsZ ring assembly, blocking cytokinesis of cell division until ATPase activity falls below the threshold (Figure 6D). With a functional clock, KaiC variants with higher ATPase activities would spend more time per day above the threshold, resulting in longer cells. In arrhythmic mutants where KaiC is locked in a high ATPase state, the gate is constitutively closed and cells should elongate infinitely. However, a bypass pathway seems to enable cell division, albeit at a much lower frequency, which explains the non-lethal phenotype in these mutants. The cryptic bypass pathway and disruption of interactions with non-Kai cellular partners may contribute to the lack of direct correlation between cell length and the ATPase activity measured in vitro among the arrhythmic mutants (KaiC487, KaiC-AA, KaiC-AE, etc.), as other unknown factors affect cell length. For example, cell density affects ΔcikA cell length more dramatically than that of WT (Figure S1). KaiA overexpression under the microscopic monitoring conditions caused cells to grow without dividing for about 3 days and reach 40 μm (Figure S2), but in liquid culture the same genotype maintains an average cell length of ~7 μm (Figure 2). The structurally locked, or even truncated variants (KaiC487) may be differentially defective in unknown partner interactions.
Peak KaiC ATPase activity occurs 4 h before that of phosphorylation in vitro (Terauchi et al., 2007), while the in vivo phosphorylation peak is ~16 h after light onset (CT 16) (Tomita et al., 2005); we estimate that KaiC ATPase activity peaks around CT 12 in vivo (Figure 6C). The center of the gate occurs around the peak of YFP fluorescence, which is around CT 17 (data not shown). Therefore, there is a delay between KaiC ATPase and gating of cell division, which may represent multiple steps of signal transduction (Figure 6D).
We propose that KaiC ATPase and phosphorylation are closely coupled, but functionally distinct: step-wise KaiC phosphorylation and dephosphorylation (Nishiwaki et al., 2007; Rust et al., 2007) govern the interaction dynamics among the Kai proteins to form a core oscillator, whereas ATPase activity connects the oscillator with the input and output pathways to drive circadian behaviors. Specifically, elevated KaiC ATPase activity stimulates the SasA output pathway, and KaiB competes with SasA for KaiC binding and lowers KaiC ATPase activity (Iwasaki et al., 2000; Terauchi et al., 2007) to turn down the output (Figure 6D).
We propose that a specific KaiC conformational state associated with elevated ATPase activity stimulates SasA autophosphorylation and triggers the inhibition of cytokinesis; this hypothesis is consistent with KaiC's identity in the RecA/DnaB superfamily, whose members couple ATP hydrolysis with conformational changes (Wang, 2004; Ye et al., 2004). Because some KaiC mutants hydrolyze >10 times more ATP than reported previously in the presence of KaiA and KaiB (Figure 4F), it cannot be ruled out that KaiC possesses much higher ATPase activity in vivo when specific conformational states interact with other partners we have not tested in vitro.
The data are consistent with a model in which elevated ATPase activity turns on the SasA-RpaA gene expression regulatory pathway as well, whereas lower ATPase activity shuts it off, which feeds back on the transcription of kaiBC. This loop, with appropriate levels of positive (high ATPase activity of KaiC) and negative elements (such as proteases), could potentially generate an autonomous rhythm in the absence of a KaiC phosphorylation rhythm, which has been reported (Kitayama et al., 2008).
Until this work, it was unclear whether the cikA mutant's circadian rhythm and cell division phenotypes are related. We show that all known cikA phenotypes are a function of the circadian clock. We propose that in a cikA null, KaiC ATPase activity still oscillates but at a higher level, which would result in a wider window of cell division gating and longer cells (Figure 6C). Most significantly, the model predicts that CikA resets the clock by modulating ATPase activity of KaiC. During a dark pulse, CikA abundance goes up (Ivleva et al., 2006), which in turn would suppress the ATPase activity of KaiC. At the same time abundance of Pex, a transcriptional repressor of kaiA, also goes up (Kutsuna et al., 2007; Takai et al., 2006a), decreasing KaiA expression and reducing stimulation of KaiC ATPase activity. We suggest that these events work synergistically to achieve a delay in KaiC phosphorylation or an advance in KaiC dephosphorylation, with the CikA-modulated pathway playing a dominant role in resetting.
The discovery of a KaiA-independent, but CikA-repressed, factor that stimulates KaiC ATPase activity suggests that there are at least two pathways that regulate KaiC ATPase activity in vivo: a KaiA-KaiB-KaiC loop and a CikA-mediated pathway that converge on KaiC and enhance the robustness of the circadian clock. Consequently, the CikA-dependent pathway is also a part of the regulatory mechanism for gene expression rhythms. In CikA's absence, KaiA and KaiB, the main drivers of the oscillation of KaiC's ATPase activity, continue to support rhythmic gene expression, but the period is shortened and amplitude reduced. Overexpression of CikA completely abolishes circadian rhythms of gene expression (Zhang et al., 2006). Both of these phenotypes may stem from the previously unknown KaiC-stimulating pathway. This CikA-mediated pathway may also contribute to the gene expression rhythm that is observed in the absence of a KaiC phosphorylation rhythm (Kitayama et al., 2008). We predict that eliminating CikA from these strains would at least weaken, if not abolish, the residual rhythms.
We identified FtsZ as a target of the SasA-RpaA-mediated circadian checkpoint. It has been reported that promoter activity of ftsZ is rhythmic (Mori and Johnson, 2001), and suggested that circadian oscillation of FtsZ level underlies gating of cell division (Mori, 2009). However, our results indicate that polymerization or localization of FtsZ is regulated rather than its abundance. Because FtsZ plays a vital role in cell division in bacteria, we speculate that such a dedicated pathway can exert precise effects from the clock at a specific time without tying the rate of cell division to the clock at all times.
Because RpaA is predicted to be a transcription factor, and it participates in post-translational regulation of FtsZ, intermediate steps of signal transduction likely exist. Other potential targets of clock control include components of the Min system and/or a homolog of DivIVA. Although these proteins have been scarcely studied in cyanobacteria (Mazouni et al., 2004; Miyagishima et al., 2005), related proteins in other bacteria regulate the position and timing of FtsZ ring assembly (Gregory et al., 2008; Lutkenhaus, 2007; Margolin, 2005).
The biological significance of cell division gating by the circadian clock remains unknown. Perhaps a closed gate protects the clock from asymmetric distribution of components at some points in the circadian cycle, such as during monomer shuffling. The S. elongatus clock is inherited with exceptional fidelity (Mihalcescu et al., 2004), and it stands to reason that a checkpoint exists to safeguard this accuracy. Alternatively, the gate protects other cellular events from damage that might result from cytokinesis at a vulnerable time, such as when the chromosome is in a particular compaction state (Smith and Williams, 2006). However, the nucleoid occlusion effect, which is seen in many other bacteria, is absent in S. elongatus, as FtsZ ring formation can occur over signal of chromosomal DNA (Miyagishima et al., 2005). The gate coincides with the onset of darkness, when oxidative stress and UV damage would be lower than during daytime. Cell division is restricted to the reductive phase of the metabolic cycle in yeast under nutrient-limiting conditions, probably to protect genome integrity (Chen et al., 2007). However, S. elongatus does not restrict its cell division to the dark, and the rate of DNA replication, the process most sensitive to UV damage or oxidative stress, remains constant during the circadian cycle (Mori et al., 1996); cytokinesis, rather than replication, is targeted. Overall, the data suggest that gating of cell division in S. elongatus does not exist to protect the genome from damage caused by stress conditions. With limited information, we favor the hypothesis that gating protects the accuracy of the clock.
In summary, we provide mechanistic insights into the gating of cell division and the circadian clock. Moreover, this study shows an unprecedented linear flow of information from known components of the input (CikA), oscillator (KaiC), and output (SasA/RpaA) divisions of the clock (Figure 6D). Further study will not only deepen our knowledge of the circadian clock, but also will uncover the intricate interactions of two major cellular oscillators with global physiological impact that cycle with different periods.
S. elongatus PCC 7942 and its genetic variants were propagated as previously described (Clerico et al., 2007; Mackey et al., 2007), except that when multiple antibiotics were needed, concentrations of gentamycin and kanamycin were halved. Mutant alleles were made using the QuickChange (Stratagene, La Jolla, CA) protocol and clones were verified by DNA sequencing. A list of plasmids, propagated in E. coli DH10B, and cyanobacterial strains used in this study is detailed in Tables S1A and S1B.
Strains were grown in liquid medium in a shaking incubator at a light intensity of 45 μE m-2 s-1. The following routine minimized variations in cell length due to light intensity, growth phase, etc.: 5 mL of liquid culture was subcultured into 100 mL fresh BG-11 medium with appropriate antibiotics, grown for 3 days, and subcultured into 100 mL medium with a 1:100 dilution factor. Cell length was measured for all strains at the same time after 3-4 days of growth. Cells collected from 1 mL culture were resuspended in 100 μL sterile water; 2 μL samples were loaded on a glass slide, followed by 5 μL of 1% low-melting agarose dissolved in BG-11 medium, and covered with a cover glass. Autofluorescence images were taken using an Olympus IX 70 inverted microscope with a 100× oil-immersion objective.
Micrographs of cyanobacterial cells were saved as 8-bit gray images. Analysis was performed with ImageJ (http://rsbweb.nih.gov/ij/). Cells that were touching were manually separated by a 5 pt black line, or were excluded from analysis. Images were segmented using the “Otsu Thresholding 8bit” plugin and cell length was acquired by applying the “Measure Roi Curve” plugin function, with default settings, to the segmented images. Data were imported into SPSS 14.0 and compared using Dunnett's T3 algorithm in One-way ANOVA analysis with p < 0.001.
Time-lapse microscopy experiments were performed on strains that were grown under the standard conditions at a light intensity of 25 μE m-2 s-1. Cultures were diluted three days before acquisition so that their OD750 were <0.2. For each sample, 1.5 μl of cells were loaded in one of two wells of a glass chamber (LAB-TEK) and covered with a square pad of 2% low-melting agarose. Next, a liquid 2% low-melting agarose was poured on top of the first solid pad and the chamber was loaded on the microscope for imaging. Phase contrast and YFP images were taken every 40 minutes for a total of 16 positions for each sample. The acquisition period typically lasted three days, during which a constant cool-white fluorescent light source illuminated the sample with an intensity of 30 ~ 50 μE m-2 s-1 except during image acquisition. The images were analyzed with custom-written MATLAB software to track each single cell as it divided and oscillated.
Immunofluorescence microscopy on FtsZ was performed as previously described (Miyagishima et al., 2005). Immunoblots were performed as described (Ivleva et al., 2005), except that KaiC antiserum was used at 1:2000 dilution.
KaiA, KaiB, KaiC, and KaiC mutant proteins were expressed in E. coli BL21DE3 and purified as described (Iwasaki et al., 2002; Kim et al., 2008; Rust et al., 2007; Williams et al., 2002). ATPase assays of KaiC and its variants (3.5 uM) were performed in sterile 1.5-ml tubes with or without KaiA and KaiB (1.2 and 3.5 uM, respectively) in the standard buffer (150 mM NaCl, 20 mM Tris-HCl, 5 mM MgCl2, 0.5 mM EDTA, 1.3 mM ATP and pH=8.0). To separate ATP and ADP in the reaction mixture, chromatography was carried out on a Vydac C18 column with 250 mM phosphate buffer (pH 5.0) as a mobile phase. The amount of ADP produced in the reaction mixture was calculated from the area of the ADP peak. Values were calculated from two assays for WT and from three assays for all variants.
We thank P. Luitel for helpful discussions, A. LiWang for sharing unpublished data and instrumentation, and C.-C. Zhang for providing the FtsZ antiserum, and A. Suescun for technical assistance. We are grateful to J. Xiao, T. Liu and I. M. Axmann for constructive suggestions on the ATPase assays. This work was supported by grants from the NIH (R01 GM62419 and P01 NS39546 to SSG, R01 GM068957 to AVO), NSF (PHY-0548484 to AVO), and DOE (DE-FG-02-06ER15808 to KWO). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of General Medical Sciences or the National Institutes of Health.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.