ATPase activity of KaiC determines the basic timing for circadian clock of cyanobacteria

Abstract

Self-sustainable oscillation of KaiC phosphorylation has been reconstituted in vitro, demonstrating that this cycle is the basic time generator of the circadian clock of cyanobacteria. Here we show that the ATPase activity of KaiC satisfies the characteristics of the circadian oscillation, the period length, and the temperature compensation. KaiC possesses extremely weak but stable ATPase activity (15 molecules of ATP per day), and the addition of KaiA and KaiB makes the activity oscillate with a circadian period in vitro. The ATPase activity of KaiC is inherently temperature-invariant, suggesting that temperature compensation of the circadian period could be driven by this simple biochemical reaction. Moreover, the activities of wild-type KaiC and five period-mutant proteins are directly proportional to their in vivo circadian frequencies, indicating that the ATPase activity defines the circadian period. Thus, we propose that KaiC ATPase activity constitutes the most fundamental reaction underlying circadian periodicity in cyanobacteria.

The circadian clock is an intracellular timing mechanism in living organisms that coordinates their lives with environmental changes (1). Cyanobacteria are the simplest organisms that are known to exhibit circadian rhythms. In the cyanobacterium Synechococcus elongatus PCC 7942, the kaiABC gene cluster is essential for the generation of the circadian rhythms (2). Recently, we reconstituted a self-sustainable circadian oscillation of KaiC phosphorylation by combining KaiA, KaiB, and KaiC with ATP in vitro (3), demonstrating that this cycle is the basic timer of the cyanobacterial circadian clock. Using this in vitro system, we examined the biochemical mechanisms underlying the generation of the circadian rhythms and found that KaiA and KaiB rhythmically associate with KaiC (4). These interactions are essential for the generation of the robust phosphorylation rhythm. In addition to these analyses, we sought the molecular basis by which the circadian period length is defined. Because the circadian period length and its stability over a range of temperatures are essential characteristics for the circadian clock to be adaptive (1), the molecular mechanism that defines the period length is a key aspect of circadian biology. The molecular mechanism that defines the essential characteristics of circadian clocks is likely to lie in the biochemical reactions catalyzed by KaiC protein. KaiC has two ATP-binding motifs (2), but ATPase activity of KaiC has not been reported. In this study, we examined ATP consumption during the KaiC phosphorylation cycle and found that an extremely low and temperature-compensated ATPase activity of KaiC defines the period of the cyanobacterial circadian system.

Results

We measured the ATP concentration in the reaction mixtures under the standard conditions (see Materials and Methods) used to reconstitute cyclic KaiC phosphorylation and found that it dropped by ≈60 μM ATP in 24 h (Fig. 1A). Furthermore, the phosphorylation cycle could not occur unless that amount of ATP was present; when the starting ATP concentration was set to 100 μM, KaiC underwent only one phosphorylation cycle (Fig. 1B). Returning the concentration of ATP in the damped mixture to 1 mM promptly restored the cycling (Fig. 1C). This result indicated that one molecular KaiC required 15 to 25 ATP molecules per cycle for cyclic phosphorylation to continue. To evaluate in more detail the kinetics of ATP-mediated cyclic phosphorylation of KaiC, we examined the kinase and phosphatase activities of KaiC by using [γ-³²P]ATP. As shown in Fig. 2A, radioactivity was incorporated into KaiC only when the level of phosphorylated KaiC was increasing (peaks at 0, 24, and 48 h); activity was barely evident at all other times. The maximum activity (at 24 and 48 h) was ≈20-fold higher than the minimum activity (at 12 and 36 h). Next, we labeled KaiC with [γ-³²P]ATP and chased the labeled KaiC to monitor KaiC dephosphorylation. As shown in Fig. 2B, the amount of radio-labeled KaiC collected at various time points was stable and diminished gradually over several hours. Because the dephosphorylation activity was rather slow and constant, the circadian oscillation of KaiC phosphorylation must be primarily due to the rhythmic kinase activity, not the dephosphorylation one.

Fig. 1.

ATP consumption in the in vitro KaiC phosphorylation cycle. (A) Decrease in the level of ATP in the reaction mixture during the in vitro KaiC phosphorylation cycle. The ATP concentration (closed circles) was measured every 12 h during the KaiC phosphorylation cycle under standard conditions (see Materials and Methods) with a luciferase assay. The KaiC phosphorylation rhythm in vitro also is shown (open circles). (B) KaiC phosphorylation rhythms under various ATP concentrations in vitro. Three Kai proteins were incubated at 30°C in the absence of ATP (open squares). In this case, the mixture contained ≈40 μM ATP derived from the purified KaiC protein solution. The mixture was supplemented with 1 mM (open circles), 0.5 mM (closed circles), or 0.1 mM (closed squares) ATP and incubated at 30°C. (C) ATP-induced restoration of the KaiC phosphorylation rhythm. The three Kai proteins were incubated at 30°C in the absence of ATP (open squares). In this case, the mixture contained ≈40 μM ATP derived from the purified KaiC protein solution. Then 1 mM ATP was added to the incubation mixtures at 28 h (closed circles) and 44 h (open circles).

Fig. 2.

Kinase and dephosphorylation activities of KaiC in the in vitro KaiC phosphorylation cycle. (A) The kinase activity of KaiC was assayed by using [γ-³²P]ATP. KaiC was incubated with KaiA and KaiB under standard conditions; every 4 h [γ-³²P]ATP was added to an aliquot of the reaction mixture followed by a 30-min incubation. (Top) Coomassie brilliant blue-stained gel with P-KaiC and NP-KaiC denoting the phosphorylated and nonphosphorylated KaiC bands, respectively. (Middle) Autoradiograph of ³²P-labeled KaiC. (Bottom) Quantification of the radioactive signals from the autoradiograph (closed circles) and the relative intensities of the phosphorylated KaiC bands in Top (open circles) plotted against the durations of the in vitro incubations. (B) The dephosphorylation activity of KaiC in the in vitro KaiC phosphorylation cycle. Radioactively labeled KaiC was prepared by incubating KaiC with KaiA and KaiB under standard conditions with [γ-³²P] ATP and was chased for up to 4 h after removing [γ-³²P]ATP from the reaction mixture at 20, 24, 28, and 32 h.

We then estimated the absolute amount of incorporated phosphate by densitometric comparisons with standard amounts of [γ-³²P]ATP (data not shown). The apparent incorporation rate of phosphate onto KaiC was as low as 0.3 to 0.4 phosphates per KaiC molecule per day. Because this estimate corresponded to a span (peak trough) of the cyclic KaiC phosphorylation rhythm (Fig. 2A) and the results of the chasing experiment ruled out the possibility of a fast turnover rate for the phosphate, this low incorporation rate is likely to be real. On the basis of the uptake of phosphate by KaiC during the kinase reaction, the ATP concentration in the standard reaction mixture for KaiC phosphorylation is likely to drop by <2 μM in 24 h, which is much lower than the ATP concentration required for the phosphorylation cycle (Fig. 1A). This result indicated that the KaiC phosphorylation cycle in vitro consumes ATP mainly for its ATPase activity.

To examine the ATPase activity in more detail, we monitored ADP production in the reaction mixture for 60 h by using an HPLC with a titanium dioxide column (Fig. 3A) (5). The activity was estimated as the number of ADP molecules produced by each KaiC monomer per day (Table 1). In the absence of KaiA and KaiB, the basal activity remained constant at 14.5 molecules per day (1.7 × 10⁻⁴ molecules per sec) throughout the 60-h incubation period. As was the case for the phosphorylation activity of KaiC (6, 7), KaiA stimulated the ATPase activity (Table 1 and Fig. 3A). The initial ATPase activity of the KaiC–KaiA mixture (29.8 molecules per day) was followed by a gradual decrease to a level close to the basal activity. It is notable that, by itself, KaiB lowered the ATPase activity to 8.9 molecules per day (Table 1 and Fig. 3A), although it depressed the phosphorylation of KaiC only by inhibiting the activity of KaiA (8, 9). We also found that a truncated variant of KaiC (KaiC-CI; residues 1–250 of KaiC) that lacked the phosphorylation sites showed 70% of the basal activity (Table 1), which is consistent with the results shown in Figs. 1 and 2, demonstrating that the ATP hydrolysis was not entirely dependent on kinase activity.

Fig. 3.

ATPase activity of KaiC. (A) Increase in the level of ADP due to the Kai proteins. KaiC was incubated under standard conditions with or without KaiA or KaiB at 30°C. The ADP concentration in the reaction mixture was measured by HPLC as described (5). The ADP level in a mixture of only KaiA and KaiB (open triangles) is shown as a negative control. (B) ATPase activity determined by Fig. 3A, the kinase activity (Fig. 2A), and the KaiC phosphorylation rhythm (Fig. 2A) during the KaiC phosphorylation cycle.

Table 1.

ATPase activity of KaiC

Mixture			ATPase
KaiC	KaiA	KaiB	Status	(Molecules per day)	SD	n
WT KaiC	−	−		14.5	2.0	9
WT KaiC	+	−	max.	29.8	5.1	3
WT KaiC	+	−	saturated	18.1	1.3	3
WT KaiC	−	+		8.9	0.9	5
WT KaiC	+	+	max.	27.8	4.2	11
WT KaiC	+	+	min.	5.4	1.6	8
WT KaiC	+	+	average	15.8	1.9	10
KaiC-AA	−	−		26.8	2.7	5
KaiC-DE	−	−		10.9	0.3	3
KaiC-CI	−	−		10.2	1.2	3

ATPase activity was calculated as the number of ADP molecules produced in the mixture per KaiC molecule over 24 h at 30°C.

Interestingly, when both KaiA and KaiB were present in the reaction mixture, the amount of ADP increased in a stepwise fashion; that is, the ATPase activity displayed a circadian rhythm (Fig. 3A). The rate of ATPase activity alternated between that observed with the mixture of KaiC and KaiA (27.8 molecules per day) and that seen with the mixture of KaiC and KaiB (5.4 molecules per day), resulting in an average rate of 15.8 molecules per day, which was similar to the basal activity observed in the KaiC-only mixture (Table 1). Fig. 3B depicts the KaiC phosphorylation cycle, the kinase activity (Fig. 1B), and the ATPase activity of KaiC (differential plot of Fig. 3A). The phases of the kinase and ATPase activity cycles were similar and peaked ≈4 h before the peak of the KaiC phosphorylation cycle, indicating that ATP hydrolysis and the phosphorylation of KaiC are closely linked.

Because the ATPase activity of KaiC seems to be coupled with the phosphorylation state of KaiC, we examined the activities of the mutant variants KaiC-AA (S431A/T432A) (10) and KaiC-DE (S431D/T432E) (11), which mimicked the dephosphorylated and the doubly phosphorylated forms of KaiC, respectively. The ATPase activity of KaiC-AA was 1.8-fold that of the wild type, whereas that of KaiC-DE only showed 75% of the wild-type KaiC activity (Table 1). These results suggested that the phosphorylation state of KaiC modulates its ATPase activity, with the nonphosphorylated state being the more active and the fully phosphorylated state being less active.

Temperature compensation of the free-running period is an essential characteristic of circadian rhythms. In general, periods are similar at different ambient temperatures (1). In a previous study, we confirmed the temperature compensation for the period of the KaiC phosphorylation rhythm (3). In the presence of KaiA and KaiB, the ATPase activity was only slightly affected by temperature in the range of 25–35°C, with a thermal sensitivity (Q₁₀ coefficient) of ≈1.2 (Fig. 4). This finding is consistent with the thermal sensitivity of the cyanobacterial clock in vivo (12). Surprisingly, the ATPase activity in KaiC-only incubations showed considerably stronger temperature compensation; that is, the activity was almost completely constant at all temperatures examined, with a Q₁₀ coefficient of ≈1.0 (Fig. 4). This result indicated that the temperature compensation of ATPase activity of KaiC is an inherent property of the KaiC molecule. Furthermore, the ATPase activities of KaiC-AA and KaiC-DE also were temperature-compensated (Fig. 4), indicating that the temperature compensation of the ATPase activity is not dependent on the KaiC phosphorylation state.

Fig. 4.

Temperature compensation of the ATPase activity of KaiC. KaiC protein was incubated under standard conditions at 25°C, 30°C, and 35°C in the presence (open circles) or absence (closed circles) of KaiA and KaiB. KaiC-AA (squares) and KaiC-DE (triangles) also were examined under these conditions. Results are presented as means ± standard deviations from three or more independent experiments.

We previously reported that the period lengths of the expression rhythms of kaiBC in vivo were consistent with those of the in vitro KaiC phosphorylation rhythms of the respective wild-type KaiC protein (period in vivo, 25 h) and three KaiC mutants with short (S157P, 21 h; F470Y, 17 h) and long (T42S, 28 h) periods (3). Except for period length, the various rhythmic phenotypes of these mutants were almost identical to those of the wild-type strain [see supporting information (SI) Materials and Methods], which made it possible to test whether the ATPase activities correlated with the period lengths. The ATPase activities of short-period mutant proteins (S157P and F470Y) were higher than those of wild-type KaiC, whereas those of long-period mutant (T42S) were lower (Fig. 5). We plotted the ATPase activities of the KaiC variants against the frequencies of the in vivo oscillations (period⁻¹) and confirmed that the frequencies were directly proportional to the ATPase activities (Fig. 5). Moreover, we examined the ATPase activity of two other KaiC mutants. The cyanobacterial mutants harboring R393C and A251V displayed short- (15-h) and long- (46-h) period bioluminescence rhythms, respectively (unpublished data). The ATPase activities of these KaiC mutant proteins (R393C and A251V) also are proportional to their frequencies (Fig. 5). Indeed, the activity of all six KaiC variants tested in the absence of KaiA and KaiB are proportional to their frequencies. This plot clearly indicates that the basal ATPase activity of KaiC dictates the circadian period length of cyanobacteria: The higher the activity, the faster the clock ticks.

Fig. 5.

Correlation between the ATPase activity of KaiC and the circadian period length. The ATPase activities of wild-type KaiC and the five KaiC period-mutant proteins (T42S, S157P, A251V, R393C, and F470Y) were measured in the absence of KaiA and KaiB. The ATPase activities of KaiC are plotted against the frequencies of the cycle (reciprocal of the period length of the in vivo bioluminescence rhythms). The activities are reported as relative to wild-type KaiC, which was set as 1.0. Results are presented as means ± standard deviations from four independent experiments.

Discussion

In this study, we show that the essential characteristics of circadian clocks in cyanobacteria are intrinsic to the ATPase activity of KaiC. In particular, the linear correlation between ATPase activity and circadian frequency (Fig. 5) indicates that the progress of the circadian cycle depends directly on the energy provided by the ATP hydrolysis. Quantification of this process revealed that the energy liberated by the hydrolysis of 15 ATP molecules per KaiC monomer is required for one period of the circadian cycle. This finding represents the first description of a simple biochemical reaction acting as a circadian timekeeper. Moreover, we demonstrate that the ATPase activity is temperature-independent even in the absence of KaiA and KaiB (Fig. 4). This result implies that temperature compensation of the circadian period, which is generally poorly understood in circadian biology, also is attributable to this simple biochemical reaction effected by KaiC.

KaiC belongs to the RecA superfamily (13). Many ATPases in this family are known to interact with different partners and convert the energy of ATP hydrolysis into mechanical forces (14). The RecA-type nucleotide-binding domain of these ATPases converts the chemical energy of ATP into mechanical energy that is harnessed to move itself along macromolecules. To our knowledge, the activity of KaiC (15 ATP per day) is substantially lower than any other protein in this family. RuvB, which is involved in DNA recombination, hydrolyzes 8 × 10³ ATP per day even in the inactive state in the absence of substrate DNA (15). To explain the extraordinarily weak and temperature-compensated activity of KaiC, we assume that the energy of ATP hydrolysis is transferred not to another molecule, but to KaiC to lower its activity (see Fig. 6Upper). The transition of the conformational states during ATPase cycle might occur by exerting a strain (tension) on the KaiC structure and then relieving it, as suggested in the GroEL allosteric transitions (16).

Fig. 6.

A possible model for the KaiC oscillator. See Discussion for details.

Similar to many RecA-like ATPases, KaiC forms a hexameric ring (17). ATP binding and hydrolysis induce conformational changes of RecA-like ATPases with hexameric rings (14, 18). Conformational changes of the ATPase, such as p97 or ClpB, which contains tandem two nucleotide-binding motifs and forms a hexameric ring like KaiC, have been shown during the ATPase cycle (19, 20). Thus, upon the formation of a hexamer structure, ATP hydrolysis of KaiC could immediately cause a conformational change in the hexamer to slow down a potentially higher ATPase activity to the uncommonly weak level that we observed in this study. If autoregulation of ATPase activity is sufficiently strong, it could be the basis for the temperature independence of the activity because the more the activity is elevated with higher temperature, the more the activity would be inhibited. Analysis of the submolecular mechanisms underlying the ATPase reaction in the KaiC hexamer should provide valuable information regarding the molecular basis of the circadian period.

How is the ATPase activity of KaiC expressed as a circadian period of the phosphorylation rhythm? It is important to note that ATP hydrolysis and phosphorylation of KaiC mutually influence each other because both ATPase and kinase/phosphatase activities are installed in KaiC protein (Fig. 6). In fact, the ATPase activity was influenced by KaiA, KaiB, and the phosphorylation state of KaiC (Table 1 and Fig. 3A). Moreover, these three factors could regulate the ATPase activity both negatively and positively because, in the presence of KaiA and KaiB, the associations of these proteins to KaiC and the phosphorylation of KaiC are rhythmic (4). However, the ATPase activity could influence the kinase activity of KaiC possibly by inducing a conformational change in the structure of either monomeric or hexameric KaiC, as shown by the finding that both activities of KaiC oscillate in the same phase angle (Fig. 3B). Therefore, these mutual couplings between the ATPase and kinase activities of KaiC would generate a self-sustained oscillation by two processes being resonated with the circadian period (Fig. 6).

Materials and Methods

Bacterial Strains.

We used Escherichia coli DH5α and BL21 cells as hosts for plasmid construction and expression of recombinant proteins, respectively.

Purification of Recombinant Kai Proteins.

KaiA, KaiB, KaiC, and KaiC mutant proteins were expressed in E. coli and purified as described (9–11).

Construction of Plasmids.

The mutated kaiC genes for A251V and R393C were amplified by PCR from genomic DNA of respective cyanobacterial mutant and ligated into pGEX6P-1 vector (Amersham Biosciences, Piscataway, NJ) as described (3).

Reconstitution of KaiC Phosphorylation in Vitro.

Reconstitution of the KaiC phosphorylation cycle in vitro was performed as described (3). Briefly, KaiC was incubated at 30°C with KaiA and KaiB in buffer [20 mM Tris·HCl (pH 8.0), 150 mM NaCl, and 5 mM MgCl₂] that contained 1 mM ATP. The final concentrations of KaiA, KaiB, and KaiC were 1.2, 3.5, and 3.5 μM, respectively. The conditions described here are referred to as “standard conditions” in this study. Every 4 h, 3-μl aliquots of the reaction mixture were taken and subjected to SDS/PAGE and Coomassie brilliant blue staining. The relative amounts of phosphorylated KaiC determined by densitometric analysis are plotted against the duration of the incubation.

KaiC Kinase Assay.

Kinase assays were performed by using [γ-³²P]ATP; 3.5 μM KaiC was incubated with KaiA and KaiB under standard conditions, and aliquots of the mixture taken at each time point were incubated with [γ-³²P]ATP for 30 min at 30°C. Every 4 h, 0.5 μl of [γ-³²P]ATP (220 TBq mmol⁻¹; Amersham Biosciences) was added to a 15-μl aliquot of the reaction mixture, followed by a 30-min incubation. The reaction was terminated by adding SDS sample buffer, and the samples were subjected to SDS/PAGE, followed by blotting onto Immobilon-P membranes (Millipore, Billerica, MA) and autoradiography by using a BAS2000 image analyzer (Fuji, Tokyo, Japan). The absolute amount of incorporated phosphate was calculated by densitometric comparison with a known amount of labeled ATP. A dilution series of [γ-³²P]ATP was subjected to autoradiography and used as a standard to calculate the absolute amount of phosphorylated KaiC and estimate the kinase activity.

KaiC Dephosphorylation Assay.

Radioactively labeled KaiC was prepared by incubating KaiC with KaiA and KaiB under standard conditions with [γ-³²P]ATP (1.68 pmol/ml) for 20, 24, 28, or 32 h at 30°C. The reaction mixtures were applied to Micro Bio-Spin columns (Bio-Rad, Hercules, CA) equilibrated with buffer containing 1 mM ATP and eluted with the same buffer. The resulting mixtures, which contained labeled KaiC with 1 mM cold ATP, were incubated at 30°C. Then 3-μl aliquots were collected at each time point and mixed with 3 μl of SDS sample buffer, followed by SDS/PAGE, blotting, and autoradiography.

Measurement of ATPase Activity.

Reactions were performed under standard conditions as described previously. The ATP concentration in the reaction mixture was measured with a luciferase assay by using an ATP bioluminescent assay kit (Sigma–Aldrich, St. Louis, MO). The luciferase assay is a highly sensitive method used to quantify ATP levels, but cannot be used to precisely measure the low ATPase activity observed with the 1 mM ATP concentration used to generate the KaiC phosphorylation rhythm. Therefore, the ATPase activity of KaiC was measured by using an HPLC system (LaChrom D-7000 or D-2000; Hitachi, Tokyo, Japan) equipped with a titanium dioxide column (Titansphere TiO column; GL Sciences, Tokyo, Japan) as described (5). ATP and ADP in the reaction mixture were separated on the column at 40°C at a 1.0 ml/min flow rate. The detection wavelength was set at 260 nm. The mobile phase was composed of 50 mM NaH₂PO₄ buffer (pH 7.0) and 50% (vol/vol) acetonitrile. The ATPase activity was evaluated as a function of the amount of ADP produced.