The ATP-mediated regulation of KaiB-KaiC interaction in the cyanobacterial circadian clock.

The cyanobacterial circadian clock oscillator is composed of three clock proteins--KaiA, KaiB, and KaiC, and interactions among the three Kai proteins generate clock oscillation in vitro. However, the regulation of these interactions remains to be solved. Here, we demonstrated that ATP regulates formation of the KaiB-KaiC complex. In the absence of ATP, KaiC was monomeric (KaiC(1mer)) and formed a complex with KaiB. The addition of ATP plus Mg(2+) (Mg-ATP), but not that of ATP only, to the KaiB-KaiC(1mer) complex induced the hexamerization of KaiC and the concomitant release of KaiB from the KaiB-KaiC(1mer) complex, indicating that Mg-ATP and KaiB compete each other for KaiC. In the presence of ATP and Mg(2+) (Mg-ATP), KaiC became a homohexameric ATPase (KaiC(6mer)) with bound Mg-ATP and formed a complex with KaiB, but KaiC hexamerized by unhydrolyzable substrates such as ATP and Mg-ATP analogs, did not. A KaiC N-terminal domain protein, but not its C-terminal one, formed a complex with KaiB, indicating that KaiC associates with KaiB via its N-terminal domain. A mutant KaiC(6mer) lacking N-terminal ATPase activity did not form a complex with KaiB whereas a mutant lacking C-terminal ATPase activity did. Thus, the N-terminal domain of KaiC is responsible for formation of the KaiB-KaiC complex, and the hydrolysis of the ATP bound to N-terminal ATPase motifs on KaiC(6mer) is required for formation of the KaiB-KaiC(6mer) complex. KaiC(6mer) that had been hexamerized with ADP plus aluminum fluoride, which are considered to mimic ADP-Pi state, formed a complex with KaiB, suggesting that KaiB is able to associate with KaiC(6mer) with bound ADP-Pi.

Here, we investigated the interaction of KaiC with KaiB in vitro and found that KaiC associates with KaiB via its Nterminal domain and that ATP regulates the KaiB-KaiC interaction. We propose a model for ATP regulation of KaiB-KaiC interaction.

Assay for formation of the KaiB 1-94 -KaiC 6mer complex by native-PAGE and 2-dimensional sodium dodecyl sulfate (SDS)-PAGE
We incubated reaction mixtures containing 10 μM KaiB 1-94 and 6 μM KaiC 6mer at 4, 25, or 40 °C in reaction buffer and subjected aliquots to native PAGE. We confirmed the presence of KaiB  and KaiC in the complex bands by native PAGE followed by SDS-PAGE on 18 % gels, as described previously [8].

Immunoblot analysis
Because we could not easily detect the KaiB 1-94 contained in a putative complex between KaiB 1-94 -KaiC N 6mer by CBB staining of SDS-PAGE gels, we subjected the gels to immunoblotting. We incubated reaction mixtures containing KaiB 1-94 and/or KaiC N 6mer at 4 °C for 6 h as described above, and then subjected them to gel filtration chromatography on a Superdex 200/HR 10/30 column equilibrated with reaction buffer containing 0.5 mM ATP and 5 mM MgCl 2 at 4 °C. We subjected fractions containing a putative KaiB 1-94 -KaiC N 6mer complex to SDS-PAGE, blotted the proteins onto Immobilon-P Transfer Membrane (Millipore), and visualized them using the ECL Western Blotting Analysis System (GE Healthcare) with a rabbit anti-KaiB antiserum (diluted to 1/2000) as a primary antibody and a donkey anti-rabbit Ig antibody (GE Healthcare) as a secondary antibody, as described previously [8].

Assay for the time course of formation of the KaiB 1-94 -KaiC DD 1mer complex
We incubated KaiC DD 1mer (5 μM) with 5 μM KaiB 1-94 in reaction buffer at 4 °C for various periods and subjected aliquots of the reaction mixtures to native PAGE on 10 % gels, staining the gels with CBB as described above. We estimated the amount of KaiB 1-94 -KaiC DD 1mer complex from the intensity of the band by densitometry as described above.

Assay for Mg-ATP-induced dissociation of the KaiB 1-94 -KaiC 1mer complex
We assayed the ATP-induced hexamerization of KaiC DD with reaction buffer at 4 °C. We added 1 mM ATP with or without 5 mM MgCl 2 to the obtained KaiB 1-94 -KaiC DD 1mer complex and then incubated the reaction mixtures in reaction buffer at 4 °C for 6 h. Then, we subjected the reaction mixtures to gel filtration chromatography on a Superdex 200/HR 10/30 column equilibrated with reaction buffer at 4 °C. We analyzed all peak fractions by SDS-PAGE on 18 % gels, and then stained the gels with CBB.
Preparation of KaiC 6mer formed with ADP and aluminum fluoride (KaiC 6mer (ADP-AlF X )) and assay for the KaiB 1-94 -KaiC 6mer (ADP-AlF X ) complex We incubated KaiC 1mer (20 μM) in reaction buffer containing 6 mM ADP, 30 mM MgCl 2 , 2.5 mM NaF, and 2.5 mM AlCl 3 at 25°C for 2 h to form KaiC 6mer (ADP-AlF X ) [31]. We then mixed the reacion mixtures with an equal volume of 0 or 60 μM KaiB 1-94 in the same buffer and incubated them further at 25 °C for 6 h. We subjected the reaction mixtures to gel filtration chromatography on a Superdex 200/HR 10/30 column equilibrated with reaction buffer at 4 °C. We subjected peak fractions to SDS-PAGE and visualized the proteins by staining with CBB.

Assay for the KaiA-enhanced autophosphorylation of KaiC 6mer
We incubated reaction mixtures containing 0.5 μM KaiC 6mer , 0.5 μM KaiA, and Mg-ATP in reaction buffer at 40 °C for various periods. We then subjected aliquots of the reaction mixtures to SDS-PAGE on 12.5 % gels (acrylamide: bisacrylamide = 144: 1), and stained the gels with CBB. We estimated the amount of the protein from the intensities of bands by densitometry as described above.

Association of KaiC 1mer with KaiB 1-94
At 40 °C, KaiB WT formed a complex with KaiC DD 6mer ( Figure  1B) but the complex formation took more than 9 h to reach a plateau ( Figure 1B). KaiB 1-94 -KaiC DD 6mer complex formation, on the other hand, reached a plateau within 6 h with a time (t 1/2 ) of 1.2 h where the half maximal KaiB 1-94 -KaiC 6mer complex formation occurred, and even at 4 °C showed no lag ( Figure  1B). Thus, to analyze the formation of the KaiB-KaiC complex in detail, we used KaiB    Figure 1C and Table S1) as KaiC DD 6mer ( Figure 1B), whereas it showed only weak complex formation with KaiB WT ( Figure 1D). The complex formation occurred without delay and reached a plateau at 6 h ( Figure  1E). The time (t 1/2 ) where the half maximal KaiB 1-94 -KaiC 1mer complex formation occurred at 4 °C was about 2.5 h ( Figure  1E). The concentration of KaiB 1-94 where the half maximal KaiB 1-94 -KaiC 1mer complex formation occurred was about 0.7 μM ( Figure 1F). This value falls within the concentration range at which clock oscillations occur in the in vitro KaiABC clock system [5].  Table S1). Even when the reaction mixtures were incubated at 25 °C, we did not detect any KaiB 1-94 -KaiC C/DD 1mer complex ( Figure 2B). Thus, the KaiB-interacting site of KaiC was located on the N-terminal domain of KaiC.

KaiB-interacting domain of KaiC
When we incubated 2.5 μM KaiC N 6mer with 7.5 μM KaiB 1-94 in the presence of Mg-ATP, KaiC N 6mer showed a weak association with KaiB 1-94 that was detected by gel filtration chromatography but not by Native PAGE ( Figure 2C and Table S2). Unexpectedly, most KaiC N 6mer became monomeric when it formed a complex with KaiB 1-94 ( Figure 2C). KaiC C/DD 6mer , on the other hand, did not form a complex with KaiB 1-94 (Table S2). KaiC 6mer , therefore, also associates with KaiB 1-94 via its Nterminal domain.

Mg-ATP-induced dissociation of the KaiB 1-94 -KaiC 1mer complex
Addition of ATP as well as Mg-ATP was able to hexamerize KaiC DD 1mer and KaiC N 1mer ( Figure 3A) as described previously [33]. When the KaiB 1-94 -KaiC DD 1mer complex was incubated in the presence of 1 mM ATP or 1 mM ATP plus 5 mM MgCl 2 (Mg-ATP) at 4 °C for 6 h, Mg-ATP, but not ATP, induced the hexamerization of KaiC DD ( Figure 3B) and the concomitant release of KaiB   (Figures 3B and 3C). We also detected a small amount of the KaiB 1-94 -KaiC DD 6mer complex ( Figure 3C), which was likely formed from the KaiC DD 6mer hexamerized by Mg-ATP, and the KaiB 1-94 released from the KaiB 1-94 -KaiC DD 1mer complex during incubation. These results suggest that Mg-ATP but not ATP reduced the affinity of KaiC 1mer for KaiB 1-94 to dissociate the KaiB 1-94 -KaiC DD 1mer complex. When we examined the ATP-and Mg-ATP-induced oligomerization of KaiC N in the KaiB 1-94 -KaiC N 1mer complex, we obtained essentially the same results; Mg-ATP, to a much greater extent than ATP, induced oligomerization of KaiC N in the KaiB 1-94 -KaiC N 1mer complex ( Figure 3B). These results suggest that Mg-ATP (and ATP) inhibited the association of KaiB 1-94 with the N-terminal domain of KaiC.

Effects of mutations in the ATPase motifs of KaiC on KaiB 1-94 -KaiC complex formation
Both the N-and C-terminal domains of the KaiC subunit have a series of ATPase motifs (a Walker's motif A, a Walker's motif B, and a CatE [2]). When we examined the effects of mutations in those motifs on formation of KaiB 1-94 -KaiC 1mer complexes-using KaiCs with K53H and K294H mutations in Walker's motif A and CatE1 -and CatE2 -mutations in CatEs [32]-we found that all the mutants we examined (KaiC K53H/DD , KaiC CatE1 -/DD , KaiC K294H/DD , and KaiC CatE2 -/DD ) formed complexes with KaiB 1-94 ( Figure 4A and , which have an N-terminal ATPase motif mutation, did not ( Figure 4B). Native PAGE followed by SDS-PAGE revealed that all the candidate complex bands examined contained both KaiB 1-94 and KaiC (a typical example is shown in Figure 4C). These results indicate that KaiC's N-terminal ATPase motifs were responsible for formation of the KaiB 1-94 -KaiC 6mer complex that occurred in the presence of Mg-ATP. This observation is consistent with the finding described above that KaiC associates with KaiB 1-94 via its N-terminal domain ( Figure 2) and that Mg-ATP (and ATP) inhibits the association (Table S2).

KaiC 6mer (Mg-ADP-AlF X ) forms a complex with KaiB 1-94
Because ADP-AlF X mimics ADP-P i [34], we used gel filtration chromatography to determine whether Mg-ADP-AlF X hexamerizes KaiC DD 1mer and confirmed that it did ( Figure 5B). Next, we examined the possible complexes formed by KaiC DD 6mer (Mg-ADP-AlFx) with KaiB 1-94 . KaiC DD 6mer (Mg-ADP-AlF X ) formed a complex with KaiB 1-94 ( Figure 5B), but we also detected a KaiB 1-94 -KaiC DD 1mer complex under conditions  1mer complex fractions were collected. With (gray) or without (black) addition of Mg-ATP to the complex, the reaction mixtures were further incubated in reaction buffer at 4 °C for 6 h and then subjected to gel filtration chromatography on a Superdex 200/HR 10/30 column equilibrated with reaction buffer. The peak fractions were subjected to SDS-PAGE. Other conditions were the same as described for Figure 1C. Left and right gels, the 1st and 3rd peak fractions of the reaction products with addition of Mg-ATP, respectively; middle gels, the peak fraction products without addition of Mg-ATP corresponding to the 2nd peak fraction of the reaction products with addition of Mg-ATP. Other conditions were the same as described for Figure 1C. C. A typical 2D SDS-PAGE gel from the native-PAGE gel shown in Figure 4B. The protein bands were excised, and the proteins were extracted from them and subjected to SDS-PAGE. Other conditions were the same as described for Figure 1C wherein a substantial amount of KaiC DD 1mer was not hexamerized ( Figure 5B). This observation supports our above conclusion that ATP hydrolysis is required for formation of the KaiB 1-94 -KaiC DD 6mer complex. Therefore, the conformation of KaiC DD 6mer (Mg-ADP-P i ) that allows complex formation with KaiB 1-94 may differ from that of KaiC DD 6mer (Mg-ATP) without ATP hydrolysis and that of KaiC 6mer (ATP).

Effects of KaiC phosphorylation-site mutations on the N-terminal ATPase activity of KaiC
To examine the effects of mutations in the two phosphorylation sites of KaiC on the N-terminal ATPase activity of KaiC, we compared the ATPase activities of KaiC CatE2 -/AA 6mer , KaiC CatE2 -/DD 6mer , and KaiC CatE2 -6mer because KaiC CatE2 -lacks the C-terminal ATPase activity of KaiC [13]. The ATPase activities of KaiC CatE2 -/DD 6mer and KaiC CatE2 -6mer reflect the N-terminal ATPase activity while that of KaiC DD 6mer and KaiC WT 6mer reflect the total ATPase activity. The former activities were approximately half of the latter activities ( Figure 6 and Table  S4). These four KaiCs 6mer formed a complex with KaiB 1-94 (Tables S2 and S3). On the other hand, the ATPase activity of KaiC AA 6mer , which did not form a complex with KaiB 1-94 (Table  S3), showed 6 times higher ATPase activity than KaiC DD 6mer and KaiC WT 6mer ( Figure 6). In consistent with this, KaiC AA has been reported to show 2.5 times higher ATPase activity than KaiC DE , which is a mutant KaiC similar to KaiC DD [4]. The ATPase activity of KaiC CatE2 -/AA , which formed a complex with KaiB 1-94 that could only be detected by silver staining (Table  S3) ATPase activity of KaiC CatE1 -/DD 6mer reflecting the C-terminal ATPase activity of KaiC was almost the same as that of KaiC C/DD 6mer and approximately half of that of KaiC DD 6mer ( Figure  6 and Table S4), as described previously [10].
That mutations in the N-terminal but not its C-terminal ATPase motifs affected the complex formation of KaiC 6mer (Mg-ATP) with KaiB 1-94 ( Figure 4B) indicates that ATP hydrolysis by the N-terminal KaiC's ATPase motifs is responsible for KaiB-KaiC 6mer complex formation. These results are consistent with a recent report [38]. Because the KaiC N 6mer -KaiB 1-94 complex rapidly dissociated into a KaiB 1-94 -KaiC N 1mer complex ( Figure  2C), the partial dissociation (or relaxation) of the N-terminal domains of KaiC 6mer probably occurred on the interaction of KaiC 6mer with KaiB. This partial dissociation of the KaiC 6mer Nterminal domains is likely required for formation of KaiB-KaiC 6mer complex via KaiC's N-terminal domain. The relaxation of the KaiC 6mer N-terminal domains on interaction of KaiB has been revealed recently by NMR analysis [29].
The KaiB molecule, which is a homotetramer organized as a dimer of dimers (KaiB 4mer ) [24], probably dissociates into two dimers (KaiB 2mer ) on interaction with KaiC 6mer and forms a complex comprising one molecule of KaiC 6mer and two molecules of KaiB 2mer [17], as suggested by cryo-electron microscopy analysis [25]. We have proposed that the positively charged cleft (PC) of the KaiB 4mer molecule, where the functionally important KaiB residues are concentrated, is an active site(s) required for interaction with KaiA and KaiC [24,28,39]. The PC of KaiB 4mer , which is located on the dimerdimer interface [24], is probably exposed by dissociation of KaiB 4mer into dimers to interact with KaiC 6mer [28,39], whereas the corresponding region of KaiB 1-94 , a dimeric mutant of KaiB, is always exposed [17]. Two areas on KaiC 6mer molecule, on the other hand, are highly negatively charged-one around and inside the pore of KaiC 6mer N-terminal domains and the other around the inter-subunit interface of one of two adjacent KaiC 6mer N-terminal domains (Figures S1A and S1B) [37]. Interestingly, the ATP bound to the N-terminal ATPase motifs (namely, ATP-binding sites) is located adjacent to the latter area of KaiC ( Figure S1B) [37]. Electrostatic interaction between the PC on KaiB and the aforementioned area of KaiC may allow sequestration of KaiB 4mer (also KaiB 2mer such as KaiB 1-94 ) and induce dissociation into dimers (temporal weak association). Then, the dissociation of two adjacent N-terminal domains in KaiC 6mer resulting from the hydrolysis of ATP bound to the N-terminal ATPase motifs on one of the two adjacent subunits (Figures S1C and S1D), which pastes the two Nterminal domains each other [13,32,36,37], may expose the latter area of KaiC-a possible KaiB-interacting surface-to KaiB 2mer , and electrostatic interaction between the PC on KaiB and the latter area of KaiC may result in the tight association of KaiB 2mer with KaiC 6mer . The ATP bound to the N-terminal ATPase motifs inhibits the association of KaiB 2mer with KaiC 6mer via KaiC N-terminal domains, as demonstrated in KaiB 1-94 -KaiC 6mer complex formation ( Figure 3C). Thus, we calculated the surface potentials of KaiC 6mer without ATP ( Figure S1C) and with ATP ( Figure S1D) and found them to be almost the same and unlikely to affect the interaction of KaiC with KaiB. While KaiB WT 4mer formed a complex formation with KaiC 6mer ( Figure  1B), it did so only slightly with KaiC 1mer ( Figure 1D). KaiB 1-94 , in contrast, formed a complex with both KaiC 6mer and KaiC 1mer ( Figures 1B, 1C and 5A). Therefore, interaction with KaiC 6mer but not with KaiC 1mer likely enhanced KaiB 4mer dimerization, suggesting the possibility that enhancement requires the hexameric structure of KaiC 6mer N-terminal domains.
It has been previously proposed that the phosphorylation state of KaiC was involved in its forming a complex with KaiB [18]. However, our data described here showing that KaiC K294H 6mer and KaiC CatE2 -6mer , which lack the autokinase activity ( Figure 4E) [13], formed complexes with KaiB 1-94 ( Figure  4D) indicated that the phosphorylated state of KaiC is not essential for KaiB-KaiC 6mer complex formation. Our results are consistent with the recently reported results that KaiC CatE2 -6mer formed a complex with KaiB WT [38]. The phosphorylation state of KaiC, therefore, is not directly involved in and essential for complex formation.
KaiC AA 6mer and KaiC CatE2 -/AA 6mer , which did not form a complex with KaiB 1-94 , showed much higher ATPase activity than any other KaiC ATPase motif mutants we examined ( Figure 6 and Table S4). The excessively high ATPase activity of the Nterminal ATPase motifs of KaiC AA 6mer , which bounces in and out of the ATP bound to the N-terminal ATPase motifs, may hinder formation of the KaiB-KaiC 6mer complex. In KaiC 6mer , the phosphorylation state of the C-terminal domain could affect its association with KaiB via the N-terminal domain through modulating the N-terminal ATPase activity. However, we cannot exclude a possibility that KaiC AA 6mer , which is likely not a perfect mimic for the fully unphosphorylated form of KaiC, might have a changed structure, which might enhance its ATPase activity but might reduce its association with KaiB. We propose the following model for ATP regulation of KaiB-KaiC interaction. The KaiC subunit is able to form a complex with  KaiB ( Figure 7A). The N-terminal domains of KaiC 6mer are partially dissociated (relaxed) when the ATP bound to the Nterminal ATPase motifs that pastes adjacent N-terminal domains each other in KaiC 6mer is hydrolyzed ( Figure 7B), which allows KaiB to associate with the KaiC 6mer N-terminal domains. Then, KaiC 6mer -associated KaiB suppresses the ATPase activity of KaiC 6mer [4] by inhibiting ATP binding to KaiC 6mer N-terminal domains. KaiC AA 6mer and KaiC CatE2 -/AA 6mer , which seem to mimic the unphosphorylation state of KaiC, have excessively high N-terminal ATPase activity ( Figure 6 and Table S4), and that may cause rapid interconversion of the ridged (ATP-bound) and relaxed (ATP-hydrolyzed; ADP-bound or unbound) conformations of the N-terminal domains in KaiC 6mer . We propose here that this rapid interconversion inhibits KaiB-KaiC 6mer complex formation though we cannot explain this inhibiting mechanism at present ( Figure 7C).
The C-terminal ATPase motifs of KaiC are involved in the hexamerization of KaiC C-terminal domains [13] as well as their inter-subunit autophosphorylation [8] and probably autodephosphorylation [12]. Although the KaiC N-terminal ATPase motifs are involved in the hexamerization of KaiC Nterminal domains [13], and the affinity of the N-terminal ATPase motifs for ATP is higher than that of its C-terminal ATPase motifs [13], and therefore, the N-terminal domains are likely to be more tightly connected than the C-terminal domains [13,16,29], the function of the N-terminal ATPase motifs remains unknown. In this investigation, we have succeeded in revealing that the nucleotide state of the N-terminal ATPase motifs regulates KaiB-KaiC interaction. Because KaiB and SasA competitively associate with KaiC via KaiC N-terminal domains [14,17], the nucleotide state also can regulate KaiC-SasA interaction via the KaiB-KaiC interaction. ATP acts not only as a biological fuel, but also as a physiological regulator. There are some examples for ATP regulation of the physiological function. Many different cell types release ATP in response to mechanical or biochemical stimulation, and the released ATP modulates cell function by activating nearby purinoceptors, such as ion channel P2X receptors and Gprotein-coupled P2Y receptors [40][41][42].