Flexibility of the C-terminal, or CII, ring of KaiCgoverns the rhythm of the circadianclock of cyanobacteriaYong-Gang Changa,1, Nai-Wei Kuoa,b,1, Roger Tsenga,c, and Andy LiWanga,c,d,2

aSchool of Natural Sciences, University of California, 5200 North Lake Road, Merced, CA 95343; bDepartment of Biochemistry and Biophysics, Texas A&MUniversity, College Station, TX 77843-2128; cQuantitative and Systems Biology, University of California, 5200 North Lake Road, Merced, CA 95343; anddHealth Sciences Research Institute, University of California, 5200 North Lake Road, Merced, CA 95343

Edited* by Susan S. Golden, University of California, San Diego, La Jolla, CA, and approved June 27, 2011 (received for review March 15, 2011)

In the cyanobacterial circadian oscillator, KaiA and KaiB alternatelystimulate autophosphorylation and autodephosphorylation ofKaiC with a periodicity of approximately 24 h. KaiA activatesautophosphorylation by selectively capturing the A loops of KaiCin their exposed positions. The A loops and sites of phosphoryla-tion, residues S431 and T432, are located in the CII ring of KaiC.We find that the flexibility of the CII ring governs the rhythmof KaiC autophosphorylation and autodephosphorylation and isan example of dynamics-driven protein allostery. KaiA-inducedautophosphorylation requires flexibility of the CII ring. In contrast,rigidity is required for KaiC-KaiB binding, which induces a confor-mational change in KaiB that enables it to sequester KaiA by bind-ing to KaiA’s linker. Autophosphorylation of the S431 residuesaround the CII ring stabilizes the CII ring, making it rigid. In con-trast, autophosphorylation of the T432 residues offsets phos-pho-S431-induced rigidity to some extent. In the presence ofKaiA and KaiB, the dynamic states of the CII ring of KaiC exe-cutes the following circadian rhythm: CIISTflexible → CIISpTflexible →CIIpSpTrigid → CIIpSTvery-rigid → CIISTflexible. Apparently, these dynamicstates govern the pattern of phosphorylation, ST → SpT →pSpT → pST → ST. CII-CI ring-on-ring stacking is observed whenthe CII ring is rigid, suggesting a mechanism through which theATPase activity of the CI ring is rhythmically controlled. SasA, acircadian clock-output protein, binds to the CI ring. Thus, rhythmicring stacking may also control clock-output pathways.

Most organisms anticipate daily swings in ambient light andtemperature by involuntarily adjusting their metabolism

and physiology according to an approximately 24-h biochemicalrhythm generated by the oscillator component of an endogenouscircadian clock (1, 2). The cyanobacterial circadian oscillator(3), composed of only three proteins—KaiA, KaiB, and KaiC(4)—sustains an ordered pattern of KaiC autophosphorylationand autodephosphorylation (5). This oscillator governs circadianrhythms of gene expression (6, 7) and chromosome compaction(8–10), and gates cell division (11–13).

Uniquely, the oscillator of Synechococcus elongatus can bereconstituted in vitro, free from the milieu of the cell (14).The ordered pattern of KaiC autophosphorylation is as follows(15, 16): ST → SpT → pSpT → pST → ST, with S/pS and T/pTcorresponding to the unphosphorylated/phosphorylated formsof residues S431 and T432, respectively (17, 18). KaiC is a two-domain homohexamer, with each domain forming a ring (19).The A loops of free KaiC are in a dynamic equilibrium that favorsthe buried positions, thereby maintaining KaiC in a hypopho-sphorylated state (20). KaiA selectively captures the A loopsof KaiC in their exposed positions, thereby shifting the popula-tion of KaiC proteins to autophosphorylate (ST → SpT → pSpT)(20–22). However, the allosteric mechanism by which A-loopexposure activates autophosphorylation at sites >15 Å awaywas unclear. When KaiC is phosphorylated at the S431 residues,KaiB binds to KaiC (15, 16). KaiB then sequesters KaiA fromthe A loops in a KaiCB(A) complex (15, 23–25), inducing KaiC

autodephosphorylation (pSpT → pST → ST). The A loops,KaiB-binding site, and sites of phosphorylation (residues S431and T432) are located in the C-terminal, or CII, ring of KaiC(26, 27). It was assumed that large phosphorylation-dependentstructural changes in the CII ring underpin the timing mechanismof this circadian oscillator. Surprisingly, recent X-ray crystalstructures of several different phosphomimics of KaiC werevirtually identical (26). Thus, the mechanism driving the rhythmof phosphorylation of KaiC is not structurally based.

Recent advances in NMR spectroscopy (28, 29) allow itsapplication to supramolecular protein particles (30, 31). KaiC,with a mass of 345 kDa, belongs to the supramolecular class. Itis demonstrated here by methyl-transverse relaxation-optimizedspectroscopy (TROSY) NMR and biochemical experiments thatrhythmic flexibility of the CII ring governs circadian interactionsamong KaiC, KaiA, and KaiB. A flexible CII ring is required forKaiA-stimulated KaiC autophosphorylation, as well as to preventpremature KaiC-KaiB interactions. Autophosphorylation at resi-dues T432 plays an important role in this regard. In contrast,autophosphorylation at residues S431 induces CII ring rigidity,allowing the formation of a stable KaiCB complex. The structureof KaiB changes significantly upon formation of the KaiCB com-plex, which causes it to bind tightly to the linker segment of KaiA.CII ring flexibility/rigidity, moreover, provides an explanationfor the cyclic order of phosphorylation, ST → SpT → pSpT →pST → ST. Additionally, when the CII ring is rigid, it stackson the N-terminal or CI ring, suggesting a mechanism by whichthe ATPase activity of CI is regulated (32). It is also found herethat SasA specifically interacts with the CI but not the CII ring.SasA is a clock-output protein that transmits the oscillatorrhythm downstream, regulating processes such as cell divisiongating (11). A unifying model is presented in which the flexibilityof the CII ring of KaiC underpins the mechanism of this circadianoscillator, as well as regulating clock output. The crucial role ofKaiC dynamics in the absence of a significant change in its struc-ture is yet another example in a growing list of systems exhibitingso-called dynamic allostery (33–35).

ResultsThe Thermosynechococcus elongatus oscillator serves as a modelof the S. elongatus oscillator. Although most is known about the

Author contributions: Y.-G.C., N.-W.K., and A.L. designed research; Y.-G.C., N.-W.K.,and R.T. performed research; Y.-G.C., N.-W.K., R.T., and A.L. analyzed data; and Y.-G.C.,N.-W.K., R.T., and A.L. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

See Commentary on page 14377.1Y.-G.C. and N.-W.K. contributed equally to this paper.2To whom correspondence should be addressed. E-mail: aliwang@ucmerced.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1104221108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1104221108 PNAS ∣ August 30, 2011 ∣ vol. 108 ∣ no. 35 ∣ 14431–14436

BIOCH

EMISTR

YSE

ECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

circadian oscillator of the mesophile S. elongatus, its KaiA andKaiC proteins were poorly suited for NMR spectroscopy, inour hands. The highly conserved homologs from the thermophileT. elongatus, on the other hand, were superior with respect toexpression, stability, and solubility. The T. elongatus and S. elon-gatus Kai proteins are highly similar with respect to sequence,structure, and biochemistry (21, 27, 36–39). Thus, the resultsherein were obtained on T. elongatus proteins, which will hence-forth be referred to as simply KaiA, KaiB, and KaiC. S. elongatushomologs will be referred to as S. elongatus KaiA, etc.

As shown in Fig. 1A, the phosphorylation profiles of KaiC inthe presence and absence of KaiA and KaiB were the same asthose published on the S. elongatus homologs (20, 40–42): KaiAinduced autophosphorylation of KaiC, whereas KaiC alone or inthe sole presence of KaiB autodephosphorylated. The role ofKaiB is to inhibit KaiA, but only after KaiC becomes hyperpho-sphorylated. Shown in Fig. 1B are the KaiA-induced phosphor-ylation profiles of KaiC-AT, KaiC-SE, KaiC-SA, and KaiC-ET,where X and Y in KaiC-XY represent the amino acyl residuesat positions 431 and 432, respectively. Alanyl and glutamyl sub-stitutions at these positions have been shown to mimic, respec-tively, the dephosphorylated and phosphorylated states of S431and T432 (16, 20, 23). KaiA-induced autophosphorylation ofKaiC-AT and KaiC-SE, respectively, mimicked the reactionsST → SpT and SpT → pSpT, whereas that of KaiC-SA andKaiC-ET mimicked the ST → pST and pST → pSpT reactions.The observation that KaiC-ATautophosphorylated more rapidlythan KaiC-SA strongly suggests that T432 autophosphorylatesfirst, as occurs in S. elongatus KaiC (15, 16). The observation thatKaiC-SE autophosphorylated to a steady-state level of approxi-mately 81% and KaiC-ET remained unphosphorylated in thepresence of KaiA suggests that S431 autophosphorylates second.The basis for the kinetics of autophosphorylation, ST → SpT →pSpT, was not clear, however.

Progressive CII Ring Phosphorylation Is Inhibited by Phospho-S431.KaiC autophosphorylation is probably sequential rather thanconcerted, as several crystal structures of S. elongatus KaiC wereonly partially phosphorylated (26). Accordingly, the observationthat the autophosphorylation profile of KaiC-SA reached anattenuated steady-state level of approximately 56% (Fig. 1B) sug-gests that autophosphorylation at S431 is autoinhibitory beyond3∕6–4∕6 residues. S. elongatus KaiC-SA only partially phosphory-lated in the presence of KaiA as well (15, 16). A comparison of the

steady-state phosphorylation levels of KaiC-SE (approximately81%) and KaiC-SA (approximately 56%) (Fig. 1B) suggests thatthe role of phospho-T432 is to offset this autoinhibitory effectsomewhat, thereby enabling nearly complete autophosphorylationof the six S431 residues within the CII ring. Comparing the steady-state phosphorylation levels of KaiC-AT (approximately 83%)and KaiC-ET (approximately 0%) shows that phospho-S431 alsoinhibits autophosphorylation at T432. The virtually identical crys-tal structures of KaiC (26) suggest that these different kineticsprofiles were not controlled by phosphorylation-dependent struc-tural differences.

Phospho-S431 Rigidifies the CII Ring of KaiC. Methyl-TROSY lineshapes of several resonances, including those assignable to theCII ring, were significantly broader for dephospho-S431 KaiCvariants, KaiC and KaiC-SE, than for phospho-S431 mimics,KaiC-EE and KaiC-ET (Fig. 2A, panels 1–4). Thus, phospho-S431 restricts microsecond–millisecond time scale motions of theCII ring. Crystal structures of S. elongatus KaiC show that thephosphate group of phospho-S431 hydrogen bonds to residueH429 of the same subunit, which in turn interacts with D417 fromthe adjacent subunit, thereby stabilizing the CII ring, presumably(26). Although the crystal structures are of the S. elongatus homo-log, residues D417 and H429 are conserved. The positions andintensities of several NMR peaks of ST-KaiC and the SpT, andpSpT phosphomimics were more similar to each other than tothose of the pST phosphomimic (Fig. 2B, panels 1–4). This latterobservation suggests that phospho-T432 offsets the effects ofphospho-S431 to some extent and is thus consistent with thekinetics data in Fig. 1B.

Gel-filtration chromatography experiments on isolated CIIdomains of KaiC showed a distribution between monomeric andhexameric populations, depending on the amino acyl substitu-tions at positions 431 and 432 (Fig. 2D, panels 1–4). The depho-spho-S431 samples, CII and CII-SE, existed predominantly asmonomers. In contrast, the pST-CII phosphomimic, CII-ET,existed predominantly as a hexamer. Between these two extremeswas the pSpT-CII phosphomimic, CII-EE, which had significantmonomeric and hexameric populations. Thus, trends on isolateddomains by gel-filtration chromatography correlated with thoseobserved by NMR on full-length KaiC phosphomimics. Together,these observations suggest that phospho-S431 stabilizes/rigidifiesthe CII ring and that this rigidity is reduced somewhat by thepresence of phospho-T432. Insights into the destabilizing effectof phospho-T432 on the CII ring can be obtained from the crystalstructures of KaiC (26). Although phospho-T432 appears toform stabilizing intersubunit interactions with R385 (26), itsphosphate group may also have repulsive intersubunit interac-tions with E318 and, to a lesser extent, with D417 (SI Appendix,Fig. S3). Thus, the dynamic states of the CII ring of KaiC appearto cycle as follows: CIISTflexible → CIISpTflexible → CIIpSpTrigid →CIIpSTvery-rigid → CIISTflexible. The different interactions of D417with phospho-T432 and phospho-S431 may explain at least in parttheir different roles in CII ring dynamics.

CII Ring Flexibility Regulates KaiA-Induced KaiC Autophosphorylation.The A loops of KaiC (residues 487–497) exist in a dynamic equi-librium between buried and exposed positions, the latter of whichpromotes KaiC autophosphorylation and the former of whichpromotes autodephosphorylation (20). In the absence of KaiAbinding, the equilibrium is toward the buried position, therebymaintaining KaiC in a hypophosphorylated state. (Note thatKaiC by itself does not completely autodephosphorylate, becauseof this A-loop dynamic equilibrium.) KaiA selectively captures Aloops in their exposed positions, thereby shifting the populationof KaiC proteins to autophosphorylate. The crystal structures ofKaiC show that the A loops, in their buried positions, form a ring

Fig. 1. Phosphorylation kinetics profiles of KaiC and KaiC phosphomimics.Each data point represents the average of two experiments. The error barsrepresent the SEM. (A) Phosphorylation of KaiC in the presence and/orabsence of KaiA and KaiB. KaiC alone (red circle); KaiC+KaiA (dark bluediamond); KaiC+KaiB (green triangle); KaiC+KaiA+KaiB (purple square).(B) Phosphorylation of KaiC variants KaiC-AT (dark blue diamond), KaiC-SE(green triangle), KaiC-SA (purple cross), and KaiC-ET (cyan square) in thepresence of KaiA. These profiles were obtained by densitometric analysis ofSDS-PAGE gels (SI Appendix, Fig. S1). Phosphorylation profiles of these KaiCvariants in the absence of KaiA are shown in SI Appendix, Fig. S1.

14432 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1104221108 Chang et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

of hydrogen bonds (20, 26), from which it can be inferred that inthese positions they help reduce CII ring flexibility in cooperationwith phospho-S431. Mimicking A-loop exposure by their trun-

cation after residues E487 in isolated CII-EE and CII-ET do-mains (CII-EE487 and CII-ET487) destabilized ring formation(i.e., hexamerization) (Fig. 2D, panel 5). Similarly, by NMR itcould be seen that truncation of the A loops in KaiC-EE(KaiC-EE487) enhanced microsecond–millisecond time scalemotions, including those of residues in the CII ring (Fig. 2A, pa-nel 5). These observations suggest that the dynamic equilibriumof the A loops increasingly shifts toward the buried position withincreasing CII ring rigidity as follows: STflexible ≈ SpTflexible →pSpTrigid → pSTvery-rigid. That KaiA could not promote the popu-lation of KaiC-SA beyond approximately 56% steady-state phos-phorylation at S431 (Fig. 1B) was probably because the stabilityof A-loop burial exceeded that of A-loop binding by KaiA. Like-wise, KaiA was unable to induce autophosphorylation of thepST phosphomimic, KaiC-ET (Fig. 1B), because the CII ring,being maximally rigid in this “phosphorylation” state, presumablytrapped the A loops in the buried positions. Thus, CII ringflexibility regulates KaiA-induced KaiC autophosphorylation, be-cause it is coupled to the energetics of A-loop exposure/burial.

According to the crystal structures of KaiC (26), in which the Aloops reside in the buried positions, the bound ATP moleculesin the CII ring are too far from S431 and T432 for phosphoryltransfer (43). It is reasonable to expect, therefore, that CII ringdynamics enhanced by KaiA-mediated A-loop exposure is criticalfor autocatalysis by bringing the ATPs closer to S431 and T432residues, stochastically, on the microsecond–millisecond timescale. Unique spectral features from A-loop exposure (Fig. 2C,panel 5) probably reflect conformational and dynamic changescritical to the induction of KaiC autophosphorylation. The fasterkinetics of autophosphorylation at T432 relative to S431 (44)can now be rationalized, based upon their opposite effects onCII ring flexibility/rigidity. The ST → pST reaction is autoinhibi-tory because it promotes CII ring rigidity and thus A-loop burial.In contrast, ST → SpT maintains CII ring flexibility and thusA-loop dynamics. Consequently, ST → SpT is inherently fasterthan ST → pST. That the second autophosphorylation step isSpT → pSpT and not pST → pSpT is now also understood fromthe same perspective. The SpT → pSpT reaction is favoredbecause phospho-T432 prevents the A loops from becomingimmobilized in their buried positions even as S431 residuesprogressively autophosphorylate.

CII Ring Rigidity Governs Formation of the KaiC-KaiB Complex. KaiBformed stable complexes with the pSpT and pST but not STand SpT phosphomimics of KaiC (Fig. 3A, panels 1–4), as alsoobserved for S. elongatus proteins (16). This selective bindingevent, localized to the CII ring of KaiC (27), critically separatesthe phases of phosphorylation and dephosphorylation. However,as the phosphoforms of KaiC that do and do not bind KaiB havethe same quaternary structure (26) (SI Appendix, Fig. S4), selec-tive molecular recognition of pSpT- and pST-KaiC by KaiB is notstructurally based. As shown in Fig. 3A, KaiB formed a stablecomplex with the pSpT phosphomimic (panel 3), but not withits more flexible A-loop exposed state (panel 5), implying thatKaiB does not directly recognize phospho-S431. Instead, ourdata collectively suggest that KaiB recognizes the rigid and notthe flexible state of the CII ring of KaiC. Thus, the importanceof the flexibility of the ST and SpT-KaiC phosphoforms extendsbeyond promoting A-loop dynamics. It also prevents prematureformation of the KaiCB complex during the phosphorylationphase.

KaiA Probably Interacts Directly with KaiB in the KaiCB(A) Complex.Bythemselves, KaiA and KaiB interact only weakly (45, 46). How-ever, the KaiCB complex sequesters KaiA from the A loops in astable KaiCB(A) complex throughout the dephosphorylationphase (15, 23–25). It was unknown, though, how KaiA sequestra-tion is achieved, whether by KaiB and/or KaiC. Methyl-TROSY

Fig. 2. CII ring flexibility depends on the state of phosphorylation at residuesS431 and T432 and A-loop positions. (A–C) Selected regions from methyl-TROSY (54) spectra of U-½15N;2H�-Ile-δ1-½13C;1H�-labeled KaiC (panel 1) andphosphomimics of SpT-KaiC (panel 2), pSpT-KaiC (panel 3), pST-KaiC (panel4), and pSpT-KaiC487 (panel 5). Horizontal and vertical axes are 1H and 13Cchemical shifts in ppm, respectively. All panels were plotted at the samecontour level relative to the signal (SIAppendix, Table S1). Gel-filtrationexperi-ments indicated that these KaiC mutants did not aggregate. Red and bluecontours indicate peaks assigned to Ile residues of the CII and CI domains ofKaiC, respectively, whereas black contours indicate unassigned peaks. SeeFig. S2A for full spectra, Fig. S2B for peak assignments, Table S1 for plottingparameters, and Table S3 for NMR experimental details in SI Appendix.It should be noted that KaiC-SE is approximately 30% phosphorylated atresidue S431 (see SI Appendix, Fig. S1C). (D) Gel-filtration profiles of isolateddomains of CII (panel 1) and phosphomimics of SpT-CII (panel 2), pSpT-CII(panel 3), pST-CII (panel 4), and pSpT-CII487 and pST-CII487 (panel 5). The hex-americ forms of the CII domains eluted at approximately 11.7mL, whereas themonomeric forms eluted at approximately 15.4 mL. In panel 5, CII-EE487 andCII-ET487, two KaiC CII domains that were truncated just prior to the A loopsat residues 488, eluted asmonomers at approximately 16mL. The percentagesof hexameric (H) and monomeric forms (M) are provided in each panel.

Chang et al. PNAS ∣ August 30, 2011 ∣ vol. 108 ∣ no. 35 ∣ 14433

BIOCH

EMISTR

YSE

ECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

spectra of the KaiCB complex are informative in this regard. Asseen in Fig. 3B, spectra of the pSpT-KaiC phosphomimic free andbound to KaiB were highly similar. Thus, direct KaiA-KaiC inter-actions likely play a minor role, if any, in the temporal formationof the KaiCB(A) complex. In contrast, the chemical shifts of KaiBfree and bound to KaiC-EE were very different (Fig. 3C), indi-cating that the structures of these forms of KaiB were also quitedifferent. Thus, temporal sequestration of KaiA is likely dictatedby direct KaiA-KaiB interactions.

KaiCB(A) Is Formed by Direct Interactions with the Linker Segment ofKaiA. KaiA has three segments, an N-terminal domain (residues1–129), a linker segment (residues 130–179), and a C-terminaldimerization domain (residues 180–283) (47). The C-terminaldomain binds to the A loops and C-terminal tails of KaiC duringthe phosphorylation phase (20, 21), but is apparently preventedfrom doing so by KaiB during the dephosphorylation phase (15,23). The site of sequestration on KaiA was unknown. In order toidentify this binding site on KaiA, a pSpT phosphomimic of KaiCwas used that retained the A loops but was missing the solvent-exposed C-terminal tails that compose 2∕3 of the KaiA-bindingsite (KaiC-EE497). As a result, KaiA could not bind directly toKaiC-EE497 (SI Appendix, Fig. S5A, panel 2). However, thisKaiC mutant in a complex with KaiB stably sequestered KaiA(SI Appendix, Fig. S5A, panel 4), thereby providing an assay,which, in combination with a series of KaiA mutants, was usedto identify the site of sequestration on KaiA.

NMR signals of isolated N- and C-terminal domains ofKaiA were unperturbed in the presence of the KaiC-EE497:KaiBcomplex, indicating that KaiCB(A) complexes did not form(SI Appendix, Fig. S5B). However, NMR signals from a fragmentof KaiA that included the linker (KaiA-130C: residues 130–283)

broadened beyond detection upon the addition of the KaiC-EE497:KaiB complex (Fig. 4A), indicating that KaiA was becom-ing part of a large complex. Analysis of the NMR sample usinggel-filtration chromatography and SDS-PAGE revealed the pre-sence of a complex containing KaiC-EE497, KaiB, and KaiA-130C (Fig. 4 B and C). Thus, residues along the linker segmentof KaiA that connects the N- and C-terminal domains were in-ferred to make direct contacts with KaiB within the KaiCB(A)sequestration complex.

SasA Selectively Binds to the CI Ring of KaiC. The CII ring of KaiC isthe oscillator component of KaiC, as it rhythmically phosphory-lates and interacts with the other oscillator proteins, KaiA andKaiB. It was not known how this biochemical rhythm is trans-duced into clock-output signals. SasA is a sensor histidine kinaseand part of a two-component signal transduction pathway forclock output (48, 49). Its N-terminal domain binds to KaiC, in-ducing SasA autophosphorylation (48) and subsequent phosphor-yl-group transfer to its cognate response regulator, RpaA (50).Recently, it was shown that SasA is part of a clock-output path-way that gates cell division (11). Here, NMR titration experi-ments between the N-terminal domain of SasA (51), and theisolated domains of KaiC show that SasA selectively binds to

Fig. 3. KaiC-KaiB binding experiments. (A) Gel-filtration profiles of mixturesof KaiB with KaiC (panel 1) and with phosphomimics of SpT-KaiC (panel 2),pSpT-KaiC (panel 3), pST-KaiC (panel 4), and pSpT-KaiC487 (panel 5). The elu-tion profiles of KaiB in each panel are from the same run. (B) Methyl-TROSYspectra of the U-½15N;2H�-Ile-δ1-½13C;1H�-labeled pSpT-KaiC phosphomimicfree (black contours) and bound (red contours) to unlabeled KaiB. Thetwo spectra were plotted at the same contour level. (C) Methyl-TROSY spec-tra of U-½15N;2H�-Ile-δ1-½13C;1H�-labeled KaiB free (black contours) and bound(red contours) to the unlabeled pSpT-KaiC phosphomimic. “x”s indicate arti-fact peaks. Please see SI Appendix, Table S3 for experimental details. The twospectra were plotted at the same contour level.

Fig. 4. Mapping the KaiCB binding site on KaiA. (A) Titration of KaiA-130C(residues 130–283) with KaiC-EE497 and KaiB. (Left) Free KaiA-130C; (Right)KaiA-130C + KaiC-EE497 + KaiB. The monomeric molar ratio of KaiA-130C,KaiC-EE497, and KaiB was 2∶3∶2. Sample conditions and NMR parametersare described in SI Appendix, Table S3. The two spectra were plotted atthe same contour level. (B) Gel-filtration chromatography showing the coe-lution of KaiA-130C, KaiB, and KaiC-EE497. The NMR sample for A, Right wasloaded onto a Superdex 200 10/300 GL column to determine whether theabsence of NMR signals was due to sample aggregation. Arrows indicatepeaks for free KaiC-EE497 (red peak) and the complex formed by KaiA-130C, KaiB, and KaiC-EE497 (black peak). The apparent molecular size ofthe complex (approximately 440 kDa) was close to the size expected for acomplex composed of a KaiC-EE497 hexamer, two KaiA-130C dimers, andtwo KaiB dimers (460 kDa) (27, 55). (C) SDS-PAGE gel of the elution peak in-dicated by the black arrow in B. Protein bands in lane 2 corresponding toKaiC-EE497, KaiA-130C, and KaiB are indicated. Molecular weight markersare in lane 1.

14434 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1104221108 Chang et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

the CI ring (SI Appendix, Fig. S6). Thus, a major role of the CIring of KaiC is apparently to transduce the circadian rhythm gen-erated by the CII ring along SasA-mediated clock-output path-ways. It was not clear, however, how the CI and CII ringscommunicate.

CII Ring Rigidity Enhances Interactions Between the CII and CI Rings.Gel-filtration chromatography experiments showed that isolatedCI rings interact with the predominantly hexameric CII-ET, butnot with CII variants that have significant monomeric popula-tions (CII, CII-SE, and CII-EE) (Fig. 5A, panels 1–4). The ob-servation that CI interacted with CII-ET, but not monomericCII-ET487 (Fig. 5A, panel 5), suggests that stability of the CIIring is required for CII-CI interactions. The trend observed bygel-filtration chromatography correlated with that by NMR. Sev-eral methyl-TROSY peaks of KaiC and KaiC-SE assignable tothe CI ring were similar in position to those of the isolated CIring (Fig. 5B). In contrast, the corresponding CI peaks fromKaiC-ETwere significantly different in position and/or intensity.Taken together, the data in Fig. 5 A and B strongly suggest thatthe rigid, not the flexible, form of the CII ring induces CII-CIring-on-ring stacking. NMR peaks from the CI ring of KaiC-EE were very similar to those of KaiC-ET, suggesting that CII-CI stacking is also present in KaiC-EE (Fig. 5B, panel 3). How-ever, isolated CII-EE domains did not interact with the CI ring bygel-filtration chromatography (Fig. 5A, panel 3). These two ob-servations suggest that the rigidity of the CII-EE ring is insuffi-cient to induce stable CII-CI interactions between isolateddomains, yet is sufficient in the context of full-length KaiC-EE. By NMR, the CI peaks of KaiC-EE487 were shifted back

to the positions of free CI peaks, indicating that CII ring rigidity,rather than phospho-S431, was directly responsible for CII-CIring-on-ring stacking (Fig. 5B, panel 5). Thus, it is reasonableto expect that the autokinase activity of SasA, and thereby clockoutput, is regulated on a circadian time scale by CII-CI ring-stacking interactions.

DiscussionThe data presented here support a model in which the differentdynamic states of KaiC govern the timing mechanism of thecyanobacterial circadian oscillator, as well as clock output (SIAppendix, Fig. S8). CII ring dynamics is critical for the phosphor-ylation phase, as it destabilizes the buried state of the A loops,creating a dynamic equilibrium between buried and exposedpositions. It is the latter position of the A loops that is selectivelycaptured by KaiA during KaiC autophosphorylation. EnhancedCII ring dynamics by KaiA-stabilized A-loop exposure probablyplays the important role of bringing CII-bound ATP moleculeswithin phosphoryl-transfer distance to S431 and T432. CII ringflexibility is also crucial in that it prevents premature formationof a KaiCB complex. In contrast, CII ring rigidity is critical for thedephosphorylation phase, as it is the basis of molecular recogni-tion between KaiB and KaiC.

The intrinsic order of autophosphorylation of KaiC, ST →SpT → pSpT (15, 16), can now be rationalized. Because autopho-sphorylation of S431 residues progressively stabilize the buriedpositions of the A loops, it has slower kinetics than autophosphor-ylation of T432, which does not rigidify the CII ring. Additionally,phospho-T432 promotes SpT → pSpT by relieving the auto-inhibition of S431 autophosphorylation. The intrinsic order ofautodephosphorylation of KaiC, pSpT → pST → ST, can be ra-tionalized using the same argument. Starting from the pSpTstate,phospho-T432 autodephosphorylates faster than phospho-S431,because pSpT → pST preserves KaiCB(A) stability that allowsfor continuing dephosphorylation to ST. In contrast, pSpT → SpTis slower because it destabilizes KaiCB(A), thereby desequester-ing KaiA to induce premature SpT → pSpT.

S. elongatus strains expressing KaiC mutants with high ATPaseactivities, as with the KaiC487 mutant, resulted in elongated non-dividing cells (11). In contrast, strains expressing KaiC mutantswith low ATPase activities, e.g., the pST-KaiC phosphomimic, un-derwent normal cycles of cell division. This pattern, together withthe results presented here, suggests that the CII ring, when rigid,suppresses the ATPase activity of the CI ring through ring-on-ringstacking interactions. The finding that SasA selectively binds tothe CI ring of KaiC suggests that a major role of the CI ring is totransduce the biochemical rhythm generated by the CII ring alongSasA-dependent clock-output pathways. An issue resulting fromthe work here is whether SasA activity is directly regulated bystacking-induced changes in the CI ring.

Our model suggests that KaiA has a higher affinity for KaiCparticles with more flexible CII rings, as their A loops, which com-prise 1∕3 of the KaiA-binding site, more frequently sample theexposed positions. The implication is that this differential affinitycauses more slowly autophosphorylating KaiC particles to speedup, and those more rapidly autophosphorylating KaiC particles toslow down. The predicted outcome is phase coherence among theensemble of phosphorylating KaiC particles. A mathematicalmodel (52) presaged this aspect of our model, but here the basisfor the differential affinity is provided.

In conclusion, the mechanism of this circadian oscillator is gov-erned by the different dynamic states of KaiC. It is an elegantexample of how dynamics-driven protein allostery can orchestratea recurring cycle of temporally separated biochemical events.

Materials and MethodsCloning, Protein Expression, and Purification. For details about cloning strate-gies and protocols, see SI Appendix. T. elongatus KaiA, KaiB, KaiC, and SasAand Saccharomyces cerevisiae Ulp1 proteins were expressed in Escherichia

Fig. 5. CII-CI interactions depend on the state of phosphorylation at residuesS431 and T432. (A) Gel-filtration experiments of mixtures of domains of CIwith CII (panel 1) and phosphomimics of SpT-CII (panel 2), pSpT-CII (panel3), pST-CII (panel 4), and pST-CII487 (panel 5). The hexameric form of theCI and CII domains eluted at approximately 11.7 mL, whereas the monomericforms eluted at approximately 15.4 mL. The shaded region in panel 4 indi-cates the shift of the CI + pST-CII phosphomimic elution profile relative tothose of the domains by themselves. (B) Selected regions of methyl-TROSYspectra of U-½15N;2H�-Ile-δ1-½13C;1H�-labeled (29) KaiC (panel 1) and the phos-phomimics of SpT-KaiC (panel 2), pSpT-KaiC (panel 3), pST-KaiC (panel 4), andpSpT-KaiC-487 (panel 5). Horizontal and vertical axes are 1H and 13C chemicalshifts in ppm, respectively. All panels were plotted at the same relativecontour level. Blue resonances are those assigned to the CI domain, whereasunassigned peaks are black. Corresponding regions of methyl-TROSY spectraof free CI are shown in panel 6.

Chang et al. PNAS ∣ August 30, 2011 ∣ vol. 108 ∣ no. 35 ∣ 14435

BIOCH

EMISTR

YSE

ECO

MMEN

TARY

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020

coli BL21(DE3) (Novagen). The U-½15N;2H�-Ile-δ1-½13C;1H�-labeled proteinswere generated by following a slightly modified version of a publishedprotocol (53). For details about protein expression and purification, pleasesee SI Appendix, Figs. S16–S18 and Figs. S19 and S20, respectively.

In Vitro KaiC Phosphorylation Reactions. KaiC proteins were incubated withKaiA and KaiB in a reaction buffer (20 mM Tris, 150 mM NaCl, 0.5 mM EDTA,5 mM MgCl2, 1 mM ATP, pH 7.0) at 30 °C. The final concentrations of KaiA,KaiB, and KaiC were 1.2, 3.5, and 3.5 μM, respectively. Thirty-seven microliteraliquots were taken at indicated time points from the reaction mixtures forSDS-PAGE analysis (4% of acrylamide/bisacrylamide for stacking gel, and6.5% of acrylamide/bisacrylamide for running gel). Gels were stained withCoomassie brilliant blue R250, and KaiC phosphorylation level was deter-mined by densitometric analysis using Image J (National Institutes of Health)and PeakFit (SeaSolve Software, Inc.).

Analytical Gel-Filtration Chromatography. All gel-filtration assays were carriedout using a Superdex 200 10/300 GL column (GE Healthcare). For details aboutsample preparation and running conditions, please see SI Appendix.

NMR Spectroscopy. SI Appendix, Table S3 contains details of each NMR sampleand experiment. All NMR experiments were run on a Bruker 600 MHzAVANCE III spectrometer equipped with a TCI cryoprobe.

ACKNOWLEDGMENTS. We thank members of the Lewis Kay lab andCharalampos Kalodimos for valuable discussions, Steve Grimaldi for workon the Bruker cryoplatform, and Yongguang Gao (University of California,Merced, CA) for providing the SUMO (small ubiquitin-like modifier) andUlp1 vectors. We gratefully acknowledge support from the US Army(W911NF-10-1-0090) and the National Institutes of Health (NIH)(GM064576). R.T. also acknowledges support from the NIH (P20 MD005049).

1. Albrecht U, ed. (2010) The Circadian Clock (Springer, New York).2. Dunlap JC, Loros JJ, DeCoursey PJ, eds. (2004) Chronobiology: Biological Timekeeping

(Sinauer, Sunderland, MA).3. Ditty JL, Mackey SR, Johnson CH, eds. (2009) Bacterial Circadian Programs (Springer,

Heidelberg).4. Ishiura M, et al. (1998) Expression of a gene cluster kaiABC as a circadian feedback

process in cyanobacteria. Science 281:1519–1523.5. Iwasaki H, Nishiwaki T, Kitayama Y, Nakajima M, Kondo T (2002) KaiA-stimulated

KaiC phosphorylation in circadian timing loops in cyanobacteria. Proc Natl Acad SciUSA 99:15788–15793.

6. Liu Y, et al. (1995) Circadian orchestration of gene expression in cyanobacteria. GenesDev 9:1469–1478.

7. Nakahira Y, et al. (2004) Global gene repression by KaiC as a master process ofprokaryotic circadian system. Proc Natl Acad Sci USA 101:881–885.

8. Smith RM, Williams SB (2006) Circadian rhythms in gene transcription imparted bychromosome compaction in the cyanobacterium synechococcus elongatus. Proc NatlAcad Sci USA 103:8564–8569.

9. Vijayan V, Zuzow R, O’Shea EK (2009) Oscillations in supercoiling drive circadian geneexpression in cyanobacteria. Proc Natl Acad Sci USA 106:22564–22568.

10. WoelfleMA, Xu Y, Qin X, Johnson CH (2007) Circadian rhythms of superhelical status ofDNA in cyanobacteria. Proc Natl Acad Sci USA 104:18819–18824.

11. Dong G, et al. (2010) Elevated ATPase activity of KaiC applies a circadian checkpoint oncell division in synechococcus elongatus. Cell 140:529–539.

12. Mori T, Binder B, Johnson CH (1996) Circadian gating of cell division in cyanobacteriagrowing with average doubling times of less than 24 hours. Proc Natl Acad Sci USA93:10183–10188.

13. Yang Q, Pando BF, Dong G, Golden SS, van Oudenaarden A (2010) Circadian gating ofthe cell cycle revealed in single cyanobacterial cells. Science 327:1522–1526.

14. NakajimaM, et al. (2005) Reconstitution of circadian oscillation of cyanobacterial KaiCphosphorylation in vitro. Science 308:414–415.

15. Rust MJ, Markson JS, Lane WS, Fisher DS, O’Shea EK (2007) Ordered phosphorylationgoverns oscillation of a three-protein circadian clock. Science 318:809–812.

16. Nishiwaki T, et al. (2007) A sequential program of dual phosphorylation of KaiC as abasis for circadian rhythm in cyanobacteria. EMBO J 26:4029–4037.

17. Xu Y, et al. (2004) Identification of key phosphorylation sites in the circadian clockprotein KaiC by crystallographic and mutagenetic analyses. Proc Natl Acad Sci USA101:13933–13938.

18. Nishiwaki T, et al. (2004) Role of KaiC phosphorylation in the circadian clock system ofsynechococcus elongatus PCC 7942. Proc Natl Acad Sci USA 101:13927–13932.

19. Hayashi F, et al. (2003) ATP-induced hexameric ring structure of the cyanobacterialcircadian clock protein KaiC. Genes Cells 8:287–296.

20. Kim Y, Dong G, Carruthers CW, Golden SS, LiWang A (2008) The day/night switch inKaiC, a central oscillator component of the circadian clock of cyanobacteria. Proc NatlAcad Sci USA 105:12825–12830.

21. Vakonakis I, LiWang A (2004) Structure of the C-terminal domain of the clock proteinKaiA in complex with a KaiC-derived peptide: Implications for KaiC regulation. ProcNatl Acad Sci USA 101:10925–10930.

22. Pattanayek R, et al. (2006) Analysis of KaiA-KaiC protein interactions in the cyanobac-terial circadian clock using hybrid structural methods. EMBO J 25:2017–2028.

23. Qin X, et al. (2010) Intermolecular associations determine the dynamics of the circa-dian KaiABC oscillator. Proc Natl Acad Sci USA 107:14805–14810.

24. Brettschneider C, et al. (2010) A sequestration feedback determines dynamics andtemperature entrainment of the KaiABC circadian clock. Mol Syst Biol 6:389.

25. Kageyama H, et al. (2006) Cyanobacterial circadian pacemaker: Kai protein complexdynamics in the KaiC phosphorylation cycle in vitro. Mol Cell 23:161–171.

26. Pattanayek R, et al. (2009) Structures of KaiC circadian clock mutant proteins: A newphosphorylation site at T426 and mechanisms of kinase, ATPase and phosphatase.PLoS ONE 4:e7529.

27. Pattanayek R, et al. (2008) Structural model of the circadian clock KaiB-KaiC complexand mechanism for modulation of KaiC phosphorylation. EMBO J 27:1767–1778.

28. Kay LE (2011) Solution NMR spectroscopy of supra-molecular systems, why bother?A methyl-TROSY view. J Magn Reson 210:159–170.

29. Goto NK, Gardner KH, Mueller GA, Willis RC, Kay LE (1999) A robust and cost-effectivemethod for the production of val, leu, ile (d1) methyl protonated 15N-, 13C-, 2H-labeledproteins. J Biomol NMR 13:369–374.

30. Sprangers R, Kay LE (2007) Quantitative dynamics and binding studies of the 20Sproteasome by NMR. Nature 445:618–622.

31. Gelis I, et al. (2007) Structural basis for signal-sequence recognition by the translocasemotor SecA as determined by NMR. Cell 131:756–769.

32. Terauchi K, et al. (2007) ATPase activity of KaiC determines the basic timing forcircadian clock of cyanobacteria. Proc Natl Acad Sci USA 104:16377–16381.

33. Popovych N, Sun S, Ebright RH, Kalodimos CG (2006) Dynamically driven proteinallostery. Nat Struct Mol Biol 13:831–838.

34. Tzeng S, Kalodimos CG (2009) Dynamic activation of an allosteric regulatory protein.Nature 462:368–372.

35. Petit CM, Zhang J, Sapienza PJ, Fuentes EJ, Lee AL (2009) Hidden dynamic allostery in aPDZ domain. Proc Natl Acad Sci USA 106:18249–18254.

36. Vakonakis I, et al. (2004) NMR structure of the KaiC-interacting C-terminal domain ofKaiA, a circadian clock protein: Implications for the KaiA-KaiC interaction. Proc NatlAcad Sci USA 101:1479–1484.

37. Hayashi F, Iwase R, Uzumaki T, Ishiura M (2006) Hexamerization by the N-terminal do-main and intersubunit phosphorylation by the C-terminal domain of cyanobacterialcircadian clock protein KaiC. Biochem Biophys Res Commun 348:864–872.

38. Iwase R, et al. (2005) Functionally important substructures of circadian clock proteinKaiB in a unique tetramer complex. J Biol Chem 280:43141–43149.

39. Uzumaki T, et al. (2004) Crystal structure of the C-terminal clock-oscillator domain ofthe cyanobacterial KaiA protein. Nat Struct Mol Biol 11:623–631.

40. Kitayama Y, Iwasaki H, Nishiwaki T, Kondo T (2003) KaiB functions as an attenuatorof KaiC phosphorylation in the cyanobacterial circadian clock system. EMBO J22:2127–2134.

41. Xu Y, Mori T, Johnson CH (2003) Cyanobacterial circadian clockwork: Roles of KaiA,KaiB and the kaiBC promoter in regulating KaiC. EMBO J 22:2117–2126.

42. Williams SB, Vakonakis I, Golden SS, LiWang A (2002) Structure and function fromthe circadian clock protein KaiA of synechococcus elongatus: A potential clock inputmechanism. Proc Natl Acad Sci USA 99:15357–15362.

43. Mildvan AS (1997) Mechanisms of signaling and related enzymes. Proteins 29:401–416.44. Rust MJ, Golden SS, O’Shea EK (2011) Light-driven changes in energy metabolism

directly entrain the cyanobacterial circadian oscillator. Science 331:220–223.45. Iwasaki H, Taniguchi Y, IshiuraM, Kondo T (1999) Physical interactions among circadian

clock proteins KaiA, KaiB, and KaiC in cyanobacteria. EMBO J 18:1137–1145.46. Mutoh R, et al. (2010) Direct interaction between KaiA and KaiB revealed by a site-

directed spin labeling electron spin resonance analysis. Genes Cells 15:269–280.47. Ye S, Vakonakis I, Ioerger TR, LiWang A, Sacchettini JC (2004) Crystal structure of circa-

dian clock protein KaiA from Synechococcus elongatus. J Biol Chem 279:20511–20518.48. Iwasaki H, et al. (2000) A KaiC-interacting sensory histidine kinase, SasA, necessary to

sustain robust circadian oscillation in cyanobacteria. Cell 101:223–233.49. Taniguchi Y, Takai N, Katayama M, Kondo T, Oyama T (2010) Three major output

pathways from the KaiABC-based oscillator cooperate to generate robust circadiankaiBC expression in cyanobacteria. Proc Natl Acad Sci USA 107:3263–3268.

50. Takai N, et al. (2006) A KaiC-associating SasA–RpaA two-component regulatory systemas a major circadian timing mediator in cyanobacteria. Proc Natl Acad Sci USA103:12109–12114.

51. Vakonakis I, Klewer DA, Williams SB, Golden SS, LiWang A (2004) Structure of theN-terminal domain of the circadian clock-associated histidine kinase SasA. J Mol Biol342:9–17.

52. van Zon JS, Lubensky DK, Altena PRH, ten Wolde PR (2007) An allosteric model ofcircadian KaiC phosphorylation. Proc Natl Acad Sci USA 104:7420–7425.

53. Tugarinov V, Kanelis V, Kay LE (2006) Isotope labeling strategies for the study ofhigh-molecular-weight proteins by solution NMR spectroscopy. Nat Protoc 1:749–754.

54. Tugarinov V, Hwang PM, Ollerenshaw JE, Kay LE (2003) Cross-correlated relaxationenhanced 1H-13C NMR spectroscopy of methyl groups in very high molecular weightproteins and protein complexes. J Am Chem Soc 125:10420–10428.

55. Hayashi F, et al. (2004) Stoichiometric interactions between cyanobacterial clockproteins KaiA and KaiC. Biochem Biophys Res Commun 316:195–202.

14436 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1104221108 Chang et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 2

8, 2

020