A KaiC-associating SasA–RpaA two-componentregulatory system as a major circadian timingmediator in cyanobacteriaNaoki Takai*†, Masato Nakajima*†, Tokitaka Oyama*†, Ryotaku Kito*, Chieko Sugita*‡, Mamoru Sugita*‡,Takao Kondo*†, and Hideo Iwasaki†§¶

*Division of Biological Science, Graduate School of Science, and ‡Center for Gene Research, Nagoya University, Furocho, Nagoya 464-8602, Japan;†Solution Oriented Research for Science and Technology, Japan Science and Technology Agency, Furocho, Nagoya 464-8602, Japan;and §Department of Electrical Engineering and Bioscience, Graduate School of Engineering and Sciences, Waseda University,Shinjuku, Tokyo 169-8555, Japan

Edited by J. Woodland Hastings, Harvard University, Cambridge, MA, and approved June 15, 2006 (received for review April 11, 2006)

KaiA, KaiB, and KaiC clock proteins from cyanobacteria and ATP aresufficient to reconstitute the KaiC phosphorylation rhythm in vitro,whereas almost all gene promoters are under the control of thecircadian clock. The mechanism by which the KaiC phosphorylationcycle drives global transcription rhythms is unknown. Here, wereport that RpaA, a potential DNA-binding protein that acts as acognate response regulator of the KaiC-interacting kinase SasA,mediates between KaiC phosphorylation and global transcriptionrhythms. Circadian transcription was severely attenuated in sasA(Synechococcus adaptive sensor A)- and rpaA (regulator of phyco-bilisome-associated)-mutant cells, and the phosphotransfer activ-ity from SasA to RpaA changed dramatically depending on thecircadian state of a coexisting Kai protein complex in vitro. Wepropose a model in which the SasA–RpaA two-component systemmediates time signals from the enzymatic oscillator to drive ge-nome-wide transcription rhythms in cyanobacteria. Moreover, ourresults indicate the presence of secondary output pathways fromthe clock to transcription control, suggesting that multiple path-ways ensure a genome-wide circadian system.

biological clock � response regulator � phosphorelay � Synechococcus

C ircadian rhythms are endogenous biological timing processesthat are observed ubiquitously in organisms from cyanobac-

teria to green plants and mammals (1). Cyanobacteria are thesimplest organisms known to exhibit such rhythms (2). In thecyanobacterium Synechococcus elongatus PCC 7942 (hereafterreferred to as Synechococcus), essentially all gene promoters areunder the control of the circadian clock, as demonstrated bypromoter-trap experiments (3, 4). Generation of such transcrip-tion rhythms requires three clock genes: kaiA, kaiB, and kaiC (4,5). Recently, we demonstrated that the circadian rhythm of KaiCphosphorylation persists even in the absence of transcription�translation processes (6). Moreover, the circadian oscillation ofKaiC phosphorylation was reconstituted in vitro by incubatingthree recombinant Kai proteins with ATP (7). The period ofoscillation in vitro was stable despite temperature change (tem-perature compensation), and the circadian periods observed invivo in KaiC mutant strains were consistent with those measuredin vitro (7). Thus, a transcription�translation feedback loop is notessential for the circadian timing mechanism, but the biochem-ical network among Kai proteins must drive the core oscillation.

How does the Kai-based chemical oscillator drive circadianrhythms in genome-wide transcription in cyanobacteria? Kaiproteins do not show any similarity to eukaryotic circadian clockproteins or to any known DNA-binding motifs. One possibilitycould be KaiC-mediated activation of gene expression by meansof a two-component regulatory system, including a KaiC-bindinghistidine kinase (HK), SasA (8). In sasA (Synechococcus adaptivesensor A)-inactivated strains, kaiBC expression is lowered dra-matically, and circadian transcription rhythms are attenuated

severely. Although SasA is not necessary to drive basic oscilla-tion, it may function in a connection between the KaiC phos-phorylation cycle and transcription. To examine this possibility,the identification and characterization of the as yet unknowncognate response regulator(s) (RRs) of SasA is essential.

Here, we report a presumed DNA-binding RR, RpaA, as aSasA partner that is involved in the Kai-based circadian system.RpaA is required for genome-wide transcription rhythms. More-over, our biochemical analysis shows that autophosphorylationof SasA and phosphotransfer from SasA to RpaA are altered,depending on some circadian states of the coexisting Kai proteincomplex. Our results strongly suggest that a circadian timingsignal is primarily mediated by the SasA–RpaA two-componentregulatory system acting from the posttranslational oscillator tothe transcription machinery.

ResultsIdentification of RpaA as a Potential SasA Partner. In the completegenome sequence of S. elongatus PCC 6301, which is highlyhomologous to that of S. elongatus PCC 7942, we identified 24RR genes, including hybrid sensory kinase genes (Table 2, whichis published as supporting information on the PNAS web site)and 13 HK genes. To search for a cognate RR of SasA, weperformed genetic deletion of each RR gene individually in a S.elongatus PCC 7942 strain harboring a kaiBC bioluminescencereporter (PkaiBC::luxAB) by substituting the gene of interest witha kanamycin-resistant gene or by insertion of a spectinomycin-resistant gene. We succeeded in inactivating 22 of 24 RR genes,whereas null mutant strains lacking syc2234�d (ycf29) orsyc0104�c (ycf27, rpaB) were not obtained, suggesting that thetwo latter genes were essential under our experimental condi-tions (Table 2). Then temporal kaiBC gene expression patternswere monitored in the 22 RR mutants under continuous light(LL) conditions after 12-h dark treatment to synchronize theclock. We found that disruption of the syc1409�d gene dramat-ically affected the PkaiBC::luxAB bioluminescence profile (Fig.1A). The gene syc1409�d encodes an OmpR (outer membraneprotein regulator) family RR harboring presumed DNA-bindingmotifs and is highly homologous to the Synechocystis rpaA(regulator of phycobilisome-associated) gene (9) (Fig. 5, which ispublished as supporting information on the PNAS web site).

Conflict of interest statement: No conflicts declared.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: LL, continuous light; LD, light�dark; �E, microeinstein; HK, histidine kinase;RR, response regulator.

See Commentary on page 11819.

¶To whom correspondence should be sent at the § address. E-mail: hideo-iwasaki@waseda.jp.

© 2006 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0602955103 PNAS � August 8, 2006 � vol. 103 � no. 32 � 12109–12114

MIC

ROBI

OLO

GY

SEE

COM

MEN

TARY

Dow

nloa

ded

by g

uest

on

July

31,

202

1

Although RpaA is somehow responsible for energy transfer fromphycobilisome to photosystem I in Synechocystis sp. PCC 6803(9), its underlying mechanism and the cognate sensory kinase areunknown.

As shown in Fig. 1 A, genetic deletion of the SynechococcusrpaA gene lowered kaiBC promoter activity dramatically andnullified rhythmicity under LL conditions with standard lightintensity [50 microeinsteins (�E)�m�2�s�1], reminiscent of thesasA-null phenotype (8). Examining a larger set of clock-controlled promoters revealed that inactivation of either of thesasA or rpaA genes similarly nullified rhythmic expression of thekaiA, psbAI, purF, sasA, cikA, rpaA, srrB, rpoD6, and sigC genes(Fig. 1 A; see also Fig. 6, which is published as supportinginformation on the PNAS web site).

To confirm that the lowered bioluminescence indeed accom-panied reduced kaiBC mRNA levels, we performed northern(RNA) blot analyses under standard LL conditions. As shown inFig. 2A, circadian fluctuation in kaiBC mRNA level was notevident in either mutant strain, whereas a robust circadianchange peaking at subjective dusk was observed in the WT strain.The levels of kaiBC mRNAs were lowered in the mutants to�10% of that in WT cells at peak, correlating well with thebioluminescence profiles shown in Fig. 1 A.

In WT cells, the KaiB and KaiC levels oscillate under LLconditions in a circadian manner, whereas KaiA and SasAproteins accumulate constitutively (8, 10). Consistent with ourbioluminescence and Northern blot analyses, circadian rhythmsin the KaiB and KaiC levels were abolished with loweredexpression levels in the rpaA-null mutant and in the sasA-nullmutant (8), whereas the KaiA and SasA accumulation profileswere unaffected (Fig. 2B).

Circadian rhythm in the KaiC phosphorylation state is ob-served in WT cells by Western blot analysis (ref. 8 and Fig. 2B).

Interestingly, accumulated KaiC protein was phosphorylatedconstitutively without evident circadian modulation in both sasAand rpaA mutant strains (Fig. 2B). Because KaiA is known toenhance the accumulation of phosphorylated KaiC (11–13), theconstitutively elevated KaiA�KaiC ratio may account for theenhanced KaiC phosphorylation levels in both mutant strains.These results also suggest that both the SasA protein and theRpaA protein are necessary for accumulating the levels of KaiBand KaiC that are sufficient for driving a robust KaiC phos-phorylation rhythm.

Phosphorelay from KaiC to the SasA–RpaA Two-Component System.How do the SasA and RpaA proteins modulate kaiBC geneexpression? In general, sensory kinases function with autophos-phorylation and phosphotransfer activities at the beginning of aphosphorelay system that terminates by modifying function ofthe cognate RR. The closely similar circadian phenotypes andclock protein behaviors in sasA-null and rpaA-null mutantsstrongly suggest that SasA and RpaA are bona fide cognatepartners in a two-component regulatory system. To validate thispossibility, we examined phosphotransfer between SasA andRpaA in vitro. Initially, purified recombinant SasA protein wasincubated with or without recombinant RpaA protein in thepresence of [�-32P]ATP at 25°C for 60 min and then subjected toSDS�PAGE and autoradiography. SasA was radiolabeled in thepresence of ATP because of its autophosphorylation activity(Fig. 3A Top), as shown in ref. 14. When the in vitro phosphor-ylation experiment was performed in the presence of RpaA,SasA’s autophosphorylation signal was attenuated (Fig. 3AMiddle). However, no incorporation of radioactive phosphateinto RpaA could be detected, presumably because of rapiddephosphorylation of RpaA by SasA, as demonstrated for atypical Escherichia coli two-component system, EnvZ and OmpR

Fig. 1. Genome-wide transcription rhythms are attenuated in the sasA-null and rpaA-null mutants. (A) Arrhythmia in the absence of sasA or rpaA at standardLL conditions. WT, sasA-disrupted (�sasA, green trace) and rpaA-disrupted (�rpaA, red trace) cells that carry kaiBC-, psbAI-, purF-, sigC-, and sigD6-reportercassettes were grown for 3–4 d on solid medium in LL at 50 �E�m�2�s�1 to give 30–60 colonies (0.2-mm diameter). After a 12-h dark treatment, cells weretransferred to LL at 50 �E�m�2�s�1. The bioluminescence was measured by a photomultiplier tube. For clarity, bioluminescence profiles of the reporter strainslacking sasA or rpaA are shown on magnified scales (�sasA, pale blue; �rpaA, orange; magnified scale is shown on the right). (B) Residual rhythmicity from sasA-or rpaA-null mutants under dim LL conditions at 15 �E�m�2�s�1.

12110 � www.pnas.org�cgi�doi�10.1073�pnas.0602955103 Takai et al.

Dow

nloa

ded

by g

uest

on

July

31,

202

1

(15). Interestingly, Smith and Williams (16) recently observedthat the addition of KaiC accelerates in vitro SasA autophos-phorylation activity. Thus, we tested whether addition of arecombinant KaiC protein could modify potential phosphotrans-fer activity from SasA to RpaA. KaiC activated the in vitroautophosphorylation activity of SasA dramatically and enhancedphosphotransfer from SasA to RpaA (Fig. 3A). We also testedwhether aspartate phosphorylation of RpaA by means of SasAmight be modified depending on the circadian phosphorylation�complex state of KaiC. To address this question, we reconsti-tuted a KaiC phosphorylation cycle in vitro by incubating KaiA,KaiB, and KaiC recombinant proteins with ATP (7), and we thentested the effect of the reaction mixture on the SasA-to-RpaAphosphotransfer activity (see Materials and Methods). Initially,we confirmed that the phosphorylation state of KaiC wascyclically alternated on SDS�PAGE gels in the presence of KaiA,KaiB, and ATP (see autoradiogram in Fig. 3B). These reactionmixtures were collected every 4 h, immediately frozen, andstored. Then we performed in vitro SasA–RpaA phosphoryla-tion�phosphotransfer experiments in the presence of each Kaireaction mixture (Fig. 3B). We found that both SasA autophos-phorylation and SasA-to-RpaA phosphotransfer activities weremodulated depending on the phase of the Kai reaction cycle invitro. Both activities were maximal when the KaiC phosphory-lation became active during the in vitro circadian cycle (Fig. 3B).These results strongly suggest that the Kai protein complexregulates the activity of RpaA protein cyclically in cells accord-ing to their complex�phosphorylation state and thereby activates

(or represses) expression of RpaA’s target genes in a circadianmanner.

Residual Transcription Rhythms in the Absence of SasA or RpaA. InsasA-null mutant strains, a residual unstable kaiBC biolumines-cence rhythm with a period length shortened by �3 h is observedunder LL with lower light intensity [15 �E�m�2�s�1 (8)]. Thislight-intensity-dependent attenuation of the bioluminescencerhythm in the sasA-null strain was also observed for manyclock-controlled genes (Table 1 and Figs. 1B and 6B). In the WTstrain, purF::luxAB bioluminescence rhythm peaks at subjectivedawn, in contrast to dusk-peaking genes such as kaiBC. Inter-estingly, such phase-angle difference among clock-controlledgenes was abolished in sasA-null mutant cells (Figs. 1B and 6B).Thus, SasA is involved in phase control in this cyanobacterialcircadian system, but the mechanism that drives different phasicexpression profiles remains obscure. By contrast, almost all genepromoter activities remain arrhythmic in the rpaA-null mutants(Table 1 and Figs. 1B and 6B). Most HKs have both kinase andphosphoryl-protein phosphatase activities (17). Thus, the activ-ity of its cognate RR can remain residually even in the absenceof the phosphatase activity of HK, possibly accounting for themore severe phenotype in the rpaA-null mutants than in thesasA-null mutants. Nevertheless, the residual rhythmicity inthe sasA-null mutants demonstrates the presence of some crypticoutput pathways that are sustained without SasA function.

Next, we examined KaiB, KaiC, and SasA accumulation andKaiC phosphorylation profiles in the sasA and rpaA mutantsunder lower-light LL conditions. As shown in Fig. 6C, KaiB andKaiC protein levels in both mutants remained lower than in theWT strain. Again, the phosphorylated form of KaiC was dom-inant over the nonphosphorylated form. Therefore, the residualrhythmicity observed only under LL in both mutants cannot becaused simply by the attenuated accumulation of KaiB and KaiCproteins (see Discussion).

Growth of sasA- or rpaA-Disrupted Cells Under Light�Dark (LD) Cycles.Previously, we observed that inactivation of sasA did not affectthe growth rate of Synechococcus under various LL conditions,whereas the sasA-disrupted strain grows much slower than theWT or a kaiABC-deficient strain under LD cycles [periodicalteration of 12-h light and 12-h darkness (8)]. These resultssuggested an additional role for SasA in adapting the cell’smetabolism specifically to natural light and dark transitions. Weexamined whether RpaA is also involved in this type of SasA-mediated physiology by comparing the growth phenotypes of theWT, sasA-deficient, and rpaA-deficient strains under LL and LDconditions. As shown in Fig. 7, which is published as supportinginformation on the PNAS web site, both sasA-deficient andrpaA-deficient strains grow as well as the WT strain under LLconditions, whereas they grow much slower than the WT cellsunder LD cycles. These results further support the hypothesisthat SasA and RpaA are bona fide two-component partners,participating in both circadian output regulation and metaboliccontrol.

DiscussionFrom a Posttranslational Oscillator to Transcription Rhythms. Ourresults strongly support the hypothesis that the SasA–RpaAtwo-component regulatory system is the primary clock outputthat is necessary for coordinating genome-wide circadian geneexpression with proper phase relationship and period length andfor maintaining the robust oscillation of KaiC phosphorylation(Fig. 4).

There are LxxxExxxL, LR, and TxxGxGY DNA-binding mo-tifs that are characteristic of OmpR-type proteins in the RpaAaminoacyl sequence (ref. 9 and Fig. 5). E. coli OmpR proteinregulates only a subset of genes responding to osmotic stimuli

Fig. 2. Expression of kaiBC mRNA and KaiBC proteins in the rpaA and sasAmutants. (A) Northern blot analysis. The kaiBC mRNA levels were examined inWT, sasA-disrupted (�sasA), and rpaA-disrupted (�rpaA) strains at 4-h inter-vals during hours 0–48 in LL (80 �E�m�2�s�1). Densitometric data of the blotalso are shown. Multiple bands are caused by mRNA degradation, and sepa-rations are caused by the presence of rRNA. (B) Western blot analysis. Accu-mulated levels of KaiA, KaiB, KaiC, and SasA were examined. Cells werecollected every 4 h from hour 0 to hour 24 in LL. Proteins (0.5 �g) wereprepared from cells of WT (�); sasA-deficient (Œ); and rpaA-null strains (f);subjected to SDS�PAGE on 10% gels; and then analyzed by immunoblottingusing each anti-Kai antiserum or anti-SasA antiserum.

Takai et al. PNAS � August 8, 2006 � vol. 103 � no. 32 � 12111

MIC

ROBI

OLO

GY

SEE

COM

MEN

TARY

Dow

nloa

ded

by g

uest

on

July

31,

202

1

(15). Therefore, it is plausible that Kai-mediated activation ofthe presumed DNA-binding protein RpaA accelerates transcrip-tion of its target gene(s), although biochemical analysis of itsDNA-binding activity has not yet been reported. Considering thegenome-wide transcription rhythms regulated by the SasA–RpaA system (Fig. 1), one possible class of RpaA target genesmay include master transcription regulators such as sigma factorgenes and�or transcription factors for driving transcriptionalcascades. In this context, any dramatic reduction in the promot-ers of sigC and rpoD6 genes in the rpaA-null mutant (Fig. 1)would severely affect a number of downstream genes. Anotherclass of potential targets would be factors that regulate chro-mosome compaction (18). In any case, the search for theRpaA-binding promoter will provide a further clue to under-standing the cyanobacterial circadian system. It should be notedthat RpaA did not exhibit significant affinity to the kaiBCpromoter by the gel-shift analysis in our experimental conditions(data not shown).

Our biochemical analysis demonstrated that the magnitude ofSasA–RpaA phosphorelay changed depending on the phase ofthe coexisting Kai protein cycle (Fig. 3B). Interestingly, the timecourse profiles demonstrated that 32P incorporation into SasAand RpaA was maximal in the presence of the Kai mixture beforethe KaiC phosphorylation peak was reached. Therefore, it isunlikely that KaiC, with a minimal or maximal phosphorylationstate in itself, is most effective at accelerating the phosphorelayfrom SasA to RpaA. Instead, some biochemical reaction statesthat enhance the KaiC phosphorylation process would be moreimportant for activating the SasA–RpaA system. In this respect,under LL conditions, KaiC phosphorylation peaks at hour 16 inLL in Synechococcus, whereas transcription rhythms of mostgenes peak at LL hour 10–12, at which time the SasA–RpaAactivities presumably become maximal.

Presence of Secondary Output Pathway(s). Because the circadianKaiC phosphorylation cycle can be reconstituted with three Kaiproteins and ATP in vitro in the absence of SasA and RpaA (7),nullification of KaiC protein cycles in the sasA and rpaA mutantsseems primarily caused by dramatic reductions in the accumu-lated levels of KaiB and KaiC. Nevertheless, residual unstablekaiBC promoter rhythmicity was observed, if not always, in bothsasA and rpaA mutants exclusively under dim LL conditions(Table 1). Because the KaiC phosphorylation rhythm was ob-scure in these mutants under such low-light conditions (Fig. 6C),very weak posttranslational cycles in KaiC phosphorylationand�or Kai–protein interactions may be sufficient to drivetranscription cycles. Note that in the recently isolated kaiCpr1

mutant, the amplitude of KaiC phosphorylation rhythm becomesvery low, whereas that of the PkaiBC::luxAB bioluminescencerhythm remains less affected (19). Although we previouslysuggested that the KaiC phosphorylation cycle was the coreprocess for generating a basic oscillation (6, 7), these results mayimplicate that an additional, as yet unknown process also con-tributes to transcription rhythms. Alternatively, it is possible thatonly a small set of cells produced sufficient amounts of KaiB andKaiC proteins stochastically to gain both phosphorylation andbioluminescence rhythms, whereas bulk biochemical analysismight fail to detect such a single-cell level phosphorylationrhythm.

As documented above, sasA-null mutants produce unstable,short-period transcription rhythms, whereas rpaA-null mutantsare more arrhythmic under dim LL conditions (Table 1 and Figs.1B and 6B). This phenotypic difference between sasA and rpaAmutants may be explained by the more severe phenotype of RRmutants than that of cognate HK mutants (see above). AnotherHK may sense time signals from KaiC to mediate phosphorelayto RpaA in the absence of SasA. Furthermore, very unstable,low-amplitude rhythms with low penetrance in the rpaA-null

Fig. 3. In vitro reconstitution of SasA–RpaA phosphorelay activated by phos-phorylated KaiC. (A) Autophosphorylation of SasA and SasA–RpaA phosphorelayin the presence or absence of KaiC. (Top) SasA protein was incubated in theabsence of RpaA, in the presence of RpaA, or in the presence of RpaA and KaiCin a buffer containing [�-32P]ATP. Each reaction was stopped at the indicatedtimes and then analyzed by SDS�PAGE followed by autoradiography. (Middleand Bottom) Densitometric analyses of [32P]SasA signals in the absence (Œ) andpresence (�) of RpaA or in the presence of RpaA plus KaiC (f) and of [32P]RpaAsignals in the presence of SasA only (�) and SasA plus KaiC (�) are shown. (B)KaiC-mediated phosphotransfer signaling between SasA and RpaA in vitro. KaiCwas incubated with KaiA and KaiB, collected at the indicated times, and stored.The time-fractionated samples were then incubated with SasA and RpaA in thepresence of [�-32P]ATP at 25°C for 30 min. SDS�PAGE and autoradiography wereperformed. Below the autoradiogram, the densitometric analyses are shown asfollows. (Top) Relative ratio of phosphorylated KaiC to total KaiC. (Middle)Relative amounts of [32P]SasA. (Bottom) Relative amounts of [32P]RpaA. Twoindependent experiments were performed for each combination; the black linesshow plots of average values.

12112 � www.pnas.org�cgi�doi�10.1073�pnas.0602955103 Takai et al.

Dow

nloa

ded

by g

uest

on

July

31,

202

1

mutants (Table 1) suggest that there may be another minor clockoutput pathway functioning without the SasA–RpaA function(Fig. 4).

SasA and RpaA as Adaptive Systems for LD Cycles. Growth is severelysuppressed in a sasA-disrupted strain (8) as well as in a rpaA-disrupted strain only under LD cycles (Fig. 7). These resultssuggest that sasA is important for normal growth under naturaldiurnal conditions. In a different unicellular cyanobacterialstrain, Synechocystis sp. PCC 6803, rpaA was originally identifiedby mutant screening for abnormal energy transfer from phyco-bilisomes to photosystems (9): Disruption of rpaA and rpaB

affected energy transfer from phycobilisome to photosystem Iand photosystem II, respectively. Although both OmpR-typeRRs were originally thought to belong to the Ycf27 family of RRmembers, the same researchers showed that RpaA could not bethe member responsible by reexamining its sequence in detail(20). As documented above, we failed to obtain mutant cells inwhich rpaB was disrupted in Synechococcus. Instead, we gener-ated an inducible dominant-negative rpaB mutant and found thatthis mutant was also defective in growth under LD cycles but thatits circadian cycling remained essentially unaffected (H.I., un-published data). Therefore, the effect of sasA and rpaA disrup-tion on circadian function alone cannot explain the adaptationto diurnal growth and vice versa.

The Synechocystis rpaA gene, also known as rre31, is involvedin the induction of a subset of genes by hyperosmotic and saltstresses (21). Microarray analysis suggested that a candidatepartner of Rre31 in the stress response would be Hik33 (21).Therefore, we tried to generate a Synechococcus strain that wasdeficient in the hik33 homolog but could not obtain null mutantcells, suggesting its essential role in our standard LL and LDconditions (H.I. and R.K., unpublished data). The effects ofsignal cross-talk among His-to-Asp phosphorelay factors oncircadian systems and other cellular functions would be inter-esting topics to be further addressed.

Materials and MethodsBacterial Strains, Media, and Culture. S. elongatus PCC 7942 and itsderivatives used in this study are listed in Tables 2 and 3, whichare published as supporting information on the PNAS web site.These strains were grown in a modified BG-11 liquid medium[BG-11M (8)] for RNA and protein preparations or on BG-11solid media (1.5% agar) for bioluminescence assays at 30°Cunder LL by using white fluorescent lamps (50 �mol�m�2�s�1 or15 �mol�m�2�s�1).

Construction of Bioluminescence Reporter Plasmids. Targeting plas-mids for various bioluminescence reporters were constructed byusing the MultiSite Gateway Three-Fragment Vector Construc-tion Kit (Invitrogen, Carlsbad, CA). Briefly, the upstream regionof the gene of interest (listed in Table 3) was fused to apromoterless Vibrio harveyi luxAB operon with a spectinomycin-resistant gene (�). This promoter–reporter cassette was cloned

Table 1. Bioluminescence rhythms of PkaiBC�PpsbAI-reporter strains

Genotype Reporter cassette

Lightcondition,�E�m�2�s�1

Numberrhythmic

Averageamplitude,mean � SD

Averageperiod,

mean � SD

WT PkaiBC::luxAB 50 19�19 (100) 11.9 � 3.4 24.8 � 0.4PpsbAI::luxAB 50 19�19 (100) 9.1 � 3.9 25.6 � 0.6PkaiBC::luxAB 15 8�8 (100) 5.5 � 1.0 25.6 � 0.3PpsbAI::luxAB 15 8�8 (100) 2.3 � 0.7 25.3 � 0.7

�sasA PkaiBC::luxAB 50 0�91 (0) n.r. n.r.PpsbAI::luxAB 50 6�32 (18) 1.2 � 0.1 22.8 � 0.5PkaiBC::luxAB 15 32�32 (100) 1.7 � 0.2 23.0 � 0.3PpsbAI::luxAB 15 24�24 (100) 1.3 � 0.1 22.7 � 0.7

�rpaA PkaiBC::luxAB 50 0�35 (0) n.r. n.r.PpsbAI::luxAB 50 0�32 (0) n.r. n.r.PkaiBC::luxAB 15 3�24 (12) 1.2 � 0.1 22.6 � 0.5PpsbAI::luxAB 15 5�24 (20) 1.3 � 0.2 23.0 � 0.7

Values under ‘‘Number rhythmic’’ indicate the number of rhythmic culture plates from the total that had ananalytically extracted amplitude �1.1 (percentages are given in parentheses). Amplitude was defined as anaverage peak-to-trough bioluminescence signal ratio during the second and third cycles from each culture plate.‘‘Average amplitude’’ is the average amplitude extracted from all of the culture plates. ‘‘Average period’’ is theaverage period extracted from all of the culture plates showing a circadian period in a population. n.r., norhythmicity.

Fig. 4. Schematic representation of the SasA–RpaA functions in the cya-nobacterial circadian system. See Discussion for details.

Takai et al. PNAS � August 8, 2006 � vol. 103 � no. 32 � 12113

MIC

ROBI

OLO

GY

SEE

COM

MEN

TARY

Dow

nloa

ded

by g

uest

on

July

31,

202

1

into the XhoI site of the neutral site I (22) segment from theSynechococcus genome.

Genetic Deletion of RR Genes. Genetic disruption of three hybridkinase genes that contained the RR domain with the HK motifs(syc0532�d, syc0681�c, and syc2278�d) was performed by insertionof the � cassette. Briefly, each of the coding segments werecloned into the pGEM-T vector (Promega, Madison, WI). Thenthe � cassette was inserted into the RR genes on each vector byusing the GPS-M Mutagenesis System (NEB, Beverly, MA)(Table 2). Each of the other 21 RR genes (a list is shown in Table2) was deleted by substitution with a kanamycin-resistant gene(kmr) as described in ref. 23. Each of the resulting inactivationcassettes and vectors was introduced into a kaiBC::luxAB re-porter strain, NUC42 (24).

Bioluminescence Monitoring. Bioluminescence assays and analysiswere performed as described in ref. 8.

Northern Blot Analysis. Total RNA (5 �g) from Synechococcuscells was prepared by using the hot phenol method, followed byNorthern blotting analysis using a digoxigenin-labeled kaiBCprobe as described in ref. 8.

Western Blot Analysis. Western blotting analysis was performed asdescribed in ref. 6.

Bacterial Expression and Purification of the KaiA, KaiB, KaiC, SasA, andRpaA Proteins. Recombinant KaiA, KaiB, and KaiC proteins wereproduced in E. coli and purified as described in ref. 7. For theexpression and purification of SasA and RpaA, we cloned the S.elongatus sasA and rpaA genes into the BamHI–SmaI and EcoRIsites of the pGEX-6P-1 vector (Amersham Pharmacia Bio-sciences, Buckinghamshire, U.K.), respectively, and then intro-duced them to E. coli strain BL21. Cells expressing either theSasA or RpaA protein were collected and disrupted in anextraction buffer (300 mM NaCl�20 mM Tris�HCl�1 mM EDTA,pH 8.0) by sonication. After addition of benzonase nuclease(Merck, Darmstadt, Germany) and 1% Triton X-100, the ho-mogenate was centrifuged at 38,000 � g. The supernatant wasapplied to a glutathione Sepharose 4B column (AmershamPharmacia Biosciences), washed with five column volumes of thebuffer, and then applied with PreScission Protease (AmershamPharmacia Biosciences) to remove the GST tag. SasA and RpaAwere eluted with one column volume of extraction buffer, and

the eluent was diluted and applied to a Resource Q column(Amersham Pharmacia Biosciences). After washing with thebuffer (60 mM NaCl�20 mM Tris�HCl�0.5 mM EDTA, pH 8.0),proteins were eluted with a 60–450 mM NaCl gradient. SasA wasfurther purified by Superose 6 gel filtration chromatography withbuffer (150 mM NaCl�20 mM Tris�HCl�0.5 mM EDTA, pH 8.0).Protein concentration was determined by the Bradford methodwith BSA as a standard. Purity of SasA and RpaA was �95% asdetermined by SDS�PAGE.

Phosphotransfer Assays. For the experiments shown in Fig. 3A,phosphotransfer assays were performed as described in ref. 15with slight modifications. In the presence or absence of 1.7 �MKaiC and RpaA proteins, SasA (1.7 �M) was incubated in 10 �lof TEDG buffer [20 mM Tricine-NaOH�0.5 mM EDTA�0.5mM DTT�10% glycerol (vol�vol), pH 8.0] containing 0.05 mM[�-32P]ATP (10,000 cpm�pmol), 5 mM MgCl2, and 150 mM KCl.After incubation at 25°C, the reaction was stopped by additionof 5 �l of an SDS buffer. After heating at 95°C for 5 min, sampleswere subjected to SDS�PAGE by using 10% gels. The radioac-tive levels of phosphorylated proteins were analyzed by using aBAS2000 Image Analyzer (Fuji, Tokyo, Japan). For the exper-iments shown in Fig. 3B, KaiC (17 �M) was incubated with KaiA(6 �M) and KaiB (17 �M) in a reaction buffer (20 mMTris�HCl�150 mM KCl�0.5 mM EDTA�5 mM MgCl2�1 mMATP, pH 8.0) at 30°C. Aliquots (1 �l) of the reaction mixturewere collected at the indicated times and stored at �80°C. Thetime-sampled Kai reaction mixture was then incubated withSasA (1.7 �M) and RpaA (1.7 �M) in 10 �l of reaction buffer[20 mM Tricine-NaOH�0.5 mM EDTA�0.5 mM DTT�10%glycerol (vol�vol)�0.1 mM [�-32P]ATP (10,000 cpm�pmol)�5mM MgCl2�150 mM KCl, pH 8.0] at 25°C for 30 min. SDS�PAGE and autoradiography were performed as described above.

We thank Dr. Stanly B. Williams (University of Utah, Salt Lake City) forsharing data before publication; Dr. Hirofumi Aiba (Nagoya University)for technical advice; and Ms. Hisayo Kondo, Ms. Seiko Matsuura, andMs. Hoai-Linh Vu (Nagoya University) for technical assistance. Thiswork was supported by Grants-in-Aid from the Ministry of Education,Culture, Sports, Science, and Technology of Japan (17687017, 17657002,and 17017018 to H.I.; 13206027 to M.S. and H.I.; 15GS0308 to T.K. andH.I.; and 15770025 and 17370088 to T.O.), the Japanese Science andTechnology Agency�Core Research for Evolutional Science and Tech-nology (T.K., H.I., T.O, and M.N.), Waseda University Grant for SpecialResearch Project 2005A-870 (to H.I.), the Uehara Memorial Foundationfor Systems Biology (H.I.), and the Nakajima Memorial Foundation(H.I.).

1. Dunlap, J. C., Loros, J. J. & DeCoursey, P. J. (2004) Chronobiology: BiologicalTimekeeping (Sinauer, Sunderland, MA).

2. Iwasaki, H. & Kondo, T. (2004) J. Biol. Rhythms 19, 436–444.3. Liu, Y., Tsinoremas, N. F., Johnson, C. H., Lebedeva, N. V., Golden, S. S.,

Ishiura, M. & Kondo, T. (1995) Genes Dev. 9, 1469–1478.4. Nakahira, Y., Katayama, M., Miyashita, H., Kutsuna, S., Iwasaki, H., Oyama,

T. & Kondo, T. (2002) Proc. Natl. Acad. Sci. USA 101, 881–885.5. Ishiura, M., Kutsuna, S., Aoki, S., Iwasaki, H., Andersson, C. R., Johnson,

C. H., Golden, S. S. & Kondo, T. (1998) Science 281, 1519–1523.6. Tomita, J., Nakajima, M., Kondo, T. & Iwasaki, H. (2005) Science 307, 251–254.7. Nakajima, M., Imai, K., Ito, H., Nishiwaki, T., Murayama, Y., Iwasaki, H.,

Oyama, T. & Kondo, T. (2005) Science 308, 414–415.8. Iwasaki, H., Williams, S. B., Kitayama, Y., Ishiura, M., Golden, S. S. & Kondo,

T. (2000) Cell 101, 223–233.9. Ashby, M. K. & Mullineaux, C. W. (1999) FEMS Microbiol. Lett. 181, 253–260.

10. Xu, Y., Mori, T. & Johnson, C. H. (2000) EMBO J. 19, 3349–3357.11. Iwasaki, H., Nishiwaki, T., Kitayama, Y., Nakajima, M. & Kondo, T. (2002)

Proc. Natl. Acad. Sci. USA 99, 14788–15793.12. Kitayama, Y., Iwasaki, H., Nishiwaki, T. & Kondo, T. (2003) EMBO J. 22,

2127–2134.

13. Xu, Y., Mori, T. & Johnson, C. H. (2003) EMBO J. 22, 2117–2126.14. Nagaya, M., Aiba, H. & Mizuno, T. (1993) Gene 131, 119–124.15. Aiba, H., Nakasai, F., Mizushima, S. & Mizuno, T. (1989) J. Biol. Chem. 264,

14090–14094.16. Smith, R. M. & Williams, S. B. (2006) Proc. Natl. Acad. Sci. USA 103,

8564–8569.17. Pratt, L. A. & Shilhavy, T. J. (1994) J. Mol. Biol. 243, 579–594.18. Mori, T. & Johnson, C. H. (2001) Semin. Cell Dev. Biol. 12, 271–278.19. Kiyohara, Y. B., Katayama, M. & Kondo, T. (2005) J. Bacteriol. 187, 2559–2564.20. Ashby, M. K., Houmard, J. & Mullineaux, C. W. (2002) FEMS Microbiol. Lett.

214, 25–30.21. Shoumskaya, M. A., Paithoonrangsarid, K., Kanesaki, Y., Los, D. A.,

Zinchenko, V. V., Tanticharoen, M., Suzuki, I. & Murata, N. (2005) J. Biol.Chem. 280, 21531–21538.

22. Andersson, C. R., Tsinoremas, N. F., Shelton, J., Lebedeva, N. V., Yarrow, J.,Min, H. & Golden, S. S. (2000) Methods Enzymol. 305, 527–542.

23. Kuwayama, H., Obara, S., Morio, T., Katoh, M., Urushihara, H. & Tanaka, Y.(2002) Nucleic Acids Res. 30, E2.

24. Nishimura, H., Nakahira, Y., Imai, K., Tsuruhara, A., Kondo, H., Hayashi, H.,Hirai, M., Saito, H. & Kondo, T. (2002) Microbiology 148, 2903–2909.

12114 � www.pnas.org�cgi�doi�10.1073�pnas.0602955103 Takai et al.

Dow

nloa

ded

by g

uest

on

July

31,

202

1