Expression of a Gene Cluster kaiABC as a Circadian Feedback Process in Cyanobacteria

See allHide authors and affiliations

Science  04 Sep 1998:
Vol. 281, Issue 5382, pp. 1519-1523
DOI: 10.1126/science.281.5382.1519


Cyanobacteria are the simplest organisms known to have a circadian clock. A circadian clock gene cluster kaiABC was cloned from the cyanobacterium Synechococcus. Nineteen clock mutations were mapped to the three kai genes. Promoter activities upstream of the kaiA and kaiB genes showed circadian rhythms of expression, and both kaiA andkaiBC messenger RNAs displayed circadian cycling. Inactivation of any single kai gene abolished these rhythms and reduced kaiBC-promoter activity. ContinuouskaiC overexpression repressed the kaiBCpromoter, whereas kaiA overexpression enhanced it. TemporalkaiC overexpression reset the phase of the rhythms. Thus, a negative feedback control of kaiC expression by KaiC generates a circadian oscillation in cyanobacteria, and KaiA sustains the oscillation by enhancing kaiC expression.

The circadian clock, a self-sustained oscillator with a period of about 24 hours, is found ubiquitously among eukaryotes and cyanobacteria. The clock controls a temporal program of cellular metabolism to facilitate adaptation to daily environmental changes (1). To elucidate the molecular mechanism of this oscillator, several clock genes ofDrosophila, Neurospora, and mouse have been cloned and analyzed (2). Negative feedback-loop models have been proposed for the circadian clock of Drosophila andNeurospora, in which the clock genes Drosophila per and Neurospora frq are repressed by their products PER and FRQ, respectively (3). Recently, clock components that interact with each other and enhance expression of per,mper, and frq genes have been identified inDrosophila, mice (4), and Neurospora(5).

Cyanobacteria, the simplest organisms that display circadian rhythms, provide a model system for the circadian clock. We use bacterial luciferase as a reporter for the expression of a clock-controlled gene, psbAI, in the cyanobacteriumSynechococcus sp. strain PCC 7942 (6). The rhythm of bioluminescence from this reporter gene, like that of firefly luciferase in Arabidopsis and Drosophila, reliably reflects the activity of the native psbAI gene (7). Many mutants that display a wide range of circadian phenotypes have been isolated (8). We report here the cloning of a clock gene cluster kaiABC fromSynechococcus and propose a model of the circadian oscillator of Synechococcus based on negative feedback control of kaiABC expression.

We constructed a library of wild-type (WT) genomic DNA in a plasmid vector (pNIBB7942) that targets the inserts to a specific site in the cyanobacterial genome termed NSII (9). A long-period clock mutant, C44a (period, 44 hours) (Fig. 1A, top), was transformed with the library (10), and transformants were screened by an automated apparatus (8). A few “rescued clones” (period, 25 hours) were found (Fig. 1A, middle) among 20,000 to 40,000 transformants. The library DNA from a rescued clone was recovered as a plasmid (p44N) in Escherichia coli (11). Targeting of p44N again, such that the transformant carried the mutated gene at its native locus and the WT gene or genes at NSII (12), rescued mutant C44a completely (Fig. 1A, bottom), suggesting that the plasmid contains an entire WT gene for thekaiC11 mutation (Fig. 1D) of mutant C44a and that this mutation is recessive.

Figure 1

Rescue of clock mutant C44a by transformation with a genomic DNA library, a map of thekai gene cluster, the deduced amino acid sequences ofkaiA, kaiB, and kaiC gene products, and mapping of clock mutations. (A) Rescue. (Top) Mutant C44a; (middle) “rescued clones” obtained by transformation with the library; (bottom) rescue by targeting of plasmid p44N into NSII in the mutant genome. Colonies (1- mm diameter) of each strain were grown on solid medium in 90- mm plastic dishes under standard conditions (8). After a 12-hour darkness to synchronize the clock, the bioluminescence rhythm of each colony was monitored under standard conditions by a luminescent colony monitor system (8). A typical trace among 100 or more independent rhythms is shown for each strain. The standard deviation of each data point is 2 to 3%. The period was determined as described (8), and the standard deviation for 100 or more rhythms was less than 0.3 hours. Zero on the ordinate of each panel indicates darkness. The bioluminescence intensities among the panels were not normalized to each other, but the maximum level of bioluminescence by PpsbAI::lux reporter varied little among these clones. (B) Restriction map and open reading frames. The six ORFs are shown as boxes. Those above the line are translated in the forward direction (left to right) whereas those below the line are translated in the opposite orientation. (C) Amino acid sequences. KaiA has 284 amino acids with a molecular mass of 32.6 kD and a calculated isoelectric point (pI) of 4.69. KaiB is composed of 102 residues(11.4 kD, pI = 7.11) and KaiC, 519 residues (58.0 kD, pI = 5.74). Walker's motif As (GXXXXGKT/S) (X, any residue) and imperfect motif Bs (ZZZZD; Z, hydrophobic residue) are boxed and reversed, respectively, and putative catalytic carboxylate Glu residues are circled. DXXG motifs are underlined. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. (D) Mutations mapped to the kai genes. Phenotypes of mutants: period in hours for period mutants; LA, low amplitude; AR, arrhythmic. Strain names are given along with the names previously published for these mutants (old names).

To localize the region essential for rescue, we subcloned genomic DNA fragments of p44N into the pSEQ1 vector (13) and introduced them into mutant C44a. Only the 4.7-kb Eco RI fragment of p44N rescued the mutant. We sequenced this fragment and analyzed putative protein-coding regions. Six open reading frames (ORFs) were found; three were known genes (“cheY,” rpL27, andrpL21), and the other three form a cluster (Fig. 1B). To find the mutation site, we amplified the cluster from mutant C44a by polymerase chain reaction (PCR) and determined its nucleotide sequence (14). A single nucleotide alteration that caused a single amino acid substitution was found in the third gene. Insertion of the Ω fragment (15) into a Bam HI, Xho I, or Cla I site in either the first or second gene in the genome disrupted rhythmicity, suggesting that the entire gene cluster cloned here comprises clock genes. Southern (DNA) blot analysis of the genomic DNA with eachkai gene probe suggested that WT Synechococcussp. strain PCC 7942 has a single copy of this gene cluster.

We named the genes kaiA, kaiB, andkaiC, respectively (kai means cycle in Japanese). The predicted proteins have 284 (KaiA), 102 (KaiB), and 519 amino acids (KaiC) (Fig. 1C). KaiC has two putative adenosine 5′-triphosphate (ATP)– or guanosine 5′-triphosphate (GTP)–binding motifs (Walker's motif A), two imperfect Walker's motif Bs, and two putative catalytic carboxylate Glu residues that are found in ATP-binding proteins (16) (Fig. 1C). KaiC also has two putative DXXG sequences conserved in GTP-binding proteins (17). No other functional motifs were found in any of the Kai proteins.

We examined the rescue by plasmid p44N of over 50 clock mutants that show short or long periods, or arrhythmia. All strains, including all mutants reported previously (8), were restored to the WT phenotype by p44N (18). Direct sequencing of thekai gene cluster from 19 mutants revealed that 14 have mutations in kaiC, three in kaiA, and two inkaiB (19) (Fig. 1D). All mutant phenotypes, even arrhythmia, were caused by single–amino acid substitutions (Fig. 1D). Thus, the kai gene cluster and Kai proteins have a crucial and central function or functions for the circadian clock ofSynechococcus.

To further test the significance of the kai genes, we deleted the whole cluster from WT psbAI-reporter cells by replacement with a kanamycin-resistance gene (20). Transformants lacking the cluster (ΔkaiABC) grew as well as WT cells and emitted comparable bioluminescence, but they completely lost rhythmicity (Fig. 2B). Thus, the cluster is essential for the circadian oscillation, whereas it is not essential for growth, and may be specific for circadian rhythmicity. To determine whether the kai genes could function from another site in the genome, we reintroduced the cluster into NSII of the ΔkaiABC strain (21). Clones that carried the transposed cluster recovered rhythmicity completely (Fig. 2C). Thus, the segment contains the entire DNA sequence required forkai function. To assess the significance of eachkai gene separately, we individually inactivated the threekai genes. The endogenous kaiC gene was disrupted by replacement with the Ω fragment and the kaiA andkaiB genes were inactivated by introduction of a nonsense mutation just downstream of each start codon (22). All three strains carrying either an inactivated kaiA,kaiB, or kaiC gene were arrhythmic (Fig. 2, D to F). Thus, all three kai genes are essential to circadian clock function (22).

Figure 2

Arrhythmia of bioluminescence caused by inactivation of the kai genes. (A) WT cells. (B)kaiABC-deleted cells (ΔkaiABC) in which the entire kai gene cluster was deleted by replacement with the Ω fragment. (C) kaiABC-transposed cells (ΔkaiABC + kaiABC) in which the entire WT cluster was transposed to NSII in the genome of ΔkaiABC cells. (D) kaiA-inactivated cells in which thekaiA gene was inactivated by a nonsense mutation. (E) kaiB-inactivated cells in which thekaiB gene was inactivated by a nonsense mutation. (F) kaiC-disrupted cells in which thekaiC ORF was replaced with the Ω fragment. Monitoring of rhythms and representation of data were as described in the legend toFig. 1.

We examined expression of the kai genes by inserting a promoterless luxAB gene set just downstream of the cluster in the genome (23). The bioluminescence of the resultingkaiABC-reporter strain was rhythmic, with a period and phase like that of the previously characterized PpsbAI-reporter strain (Fig. 3, B and C). The level of bioluminescence of the reporter strain was ∼2 to 3% of that of the PpsbAI-reporter strain. To localize promoters of the kai cluster, we fused each kai upstream region (USR) with the luxAB gene set and targeted the constructs into NSII of WT Synechococcus cells (24). Both USRkaiA - andUSRkaiB -reporter strains showed bioluminescence rhythms whose period and phase were the same as those of the kaiABC-reporter strain, whereas aUSRkaiC -reporter strain showed no bioluminescence (Fig. 3A). Thus, a promoter exists upstream ofkaiA and kaiB, but none could be detected upstream of kaiC, suggesting cotranscription ofkaiBC. The average level of bioluminescence of theUSRkaiA - and USRkaiB - reporter strains was about 1 and 3%, respectively, of that of the PpsbAI-reporter strain. The waveform of thekaiB::luxAB reporter matched that from the reporter in which luxAB was inserted downstream ofkaiC (Fig. 3B), further supporting kaiBCcotranscription.

Figure 3

Expression of the kai genes. (A) USRkaiA -,USRkaiB -, andUSRkaiC -reporter strains carryingUSRs of the kaiA, kaiB, and kaiC genes, respectively, fused to a promoterless luxABgene set at the Bst EII site of NSII in the genome. (B) AkaiABC-reporter strain carrying the luxAB gene set inserted at the Nhe I site just downstream of the kaiCgene. (C) A psbAI-reporter strain carrying a PpsbAI ::lux construct in another specific genomic site (NSI) (6). Cells were grown to give 30 to 100 colonies (0.1-mm diameter) on solid medium in 40-mm plastic dishes in LL (standard conditions) (8) for 3 days. After a 12-hour dark treatment to synchronize the clock, the dishes were set in a sample changer (luminescent dish monitor, LDM) that alternated the dishes between the standard light conditions (27 min) and darkness (2 min). The bioluminescence from the dish was measured for 30 s by a photomultiplier tube (Hamamatsu R466S). Signal pulses were processed as described (31). Bioluminescence intensity was normalized to the number of colonies. A representative rhythm among three to eight replicates was shown for each reporter strain. The standard deviation at each time point was <10% of measurements. (D) Northern (RNA) blots prepared at hour 8 in LL (25). Probes were kaiA-,kaiB-, or kaiC-specific. The arrowheads indicate a 1.0-kb kaiA transcript and a 2.3-kb kaiBCtranscript. (E) Northern blots of RNA prepared at hours 0 to 48 in LL. probes were kaiA- or kaiC-specific.

The transcriptional units of the kai cluster were confirmed by Northern (RNA) blot analysis, with kaiA-,kaiB-, and kaiC-specific probes (25). The kaiA probe gave a 1.0-kb hybridization band, whereas both the kaiB and kaiC probes gave a 2.3-kb band (Fig. 3D). These data support monocistronic transcription ofkaiA and dicistronic transcription of kaiBC. Most of the kaiBC mRNA detected was degraded as shown by a smeared band (Fig. 3D). Both kaiA and kaiBC mRNAs displayed circadian cycling under continuous light conditions (LL) (Fig. 3E). The peaks of mRNA levels occurred at hours 9 to 12 and 33 to 36 in LL, which corresponded to those of the bioluminescence rhythms.

Autoregulation of kai expression could be a key point for generating circadian oscillation in Synechococcus. We confirmed that expression of the kai gene cluster was affected by kai mutations as observed for the PpsbAI reporter (26) (Fig. 4, A to D). We further examined the effects of inactivation of each kai gene on the bioluminescence of PkaiA- and PkaiBC-reporter strains (22). As was observed in a PpsbAI -reporter strain (Fig. 2), inactivation of either the kaiA, kaiB, orkaiC gene abolished rhythmicity of both the PkaiA- and PkaiBC-reporter strains (27). Inactivation of kaiA and kaiC lowered the bioluminescence level of PkaiBC-reporter strains to 11 ± 1 and 28 ± 5% of the WT level (n= 3), respectively, although these inactivations did not significantly affect that of a PkaiA-reporter strain.

Figure 4

(above). Expression of thekai genes in clock mutants, the effects of overexpression of the kaiA and kaiC genes on the activity of the PkaiA and PkaiBC promoters, and phase shifts by temporal overexpression of the kaiC gene. (Ato D) Expression of the kai genes in clock mutants. Bioluminescence of PkaiA(USRkaiA )-, PkaiBC(USRkaiB )-, and PpsbAI-reporter derivatives of mutants A30a (kaiA2) (A), B22a (kaiB2) (B), C28a (kaiC4) (C), and CLAb (kaiC13) (D) was monitored. Only the data for PkaiBC reporter derivatives are shown. In each period mutant (A to D), the period and phase of the rhythms in both PkaiA- and PkaiBC-reporter strains were similar to those of the PpsbAI-reporter strain. In the arrhythmic mutant (D), PkaiA- and PkaiBC-reporter strains as well as the PpsbAI-reporter strain showed arrhythmic bioluminescence. (E and F) Effects of overexpression of the kaiC gene on PkaiBC-promoter activity. WT cells (E) and cells carrying a Ptrc ::kaiC construct for overexpression (F) were treated with 1 mM IPTG. The arrows indicate the time of addition of IPTG. (G) Effects of overexpression of the kaiA gene on PkaiBC-promoter activity. Cells carried a Ptrc ::kaiAconstruct for overexpression. (H and I) Effects of overexpression of the kaiC gene on the PkaiA promoter. Cells of a PkaiA-reporter strain carrying a Ptrc ::kaiC construct were treated with water (arrows) (H) or 1 mM IPTG (I). (J to N) Phase shifting by temporal kaiC overexpression by IPTG pulses. Cells of a PkaiBC-reporter strain carrying a Ptrc ::kaiC construct were not treated (J) or treated with 1 mM IPTG for 6 hours from hours 0 (K), 3 (L), 15 (M), and 18 (N) in LL. The vertical broken line and the arrows indicate a standard phase and phase shifts, respectively. Similar treatments with water did not shift the phase at any time. Monitoring of bioluminescence and representation of data were as described in the legend to Fig. 3.

To examine the effects of overexpression of the kai genes, we fused an E. coli–inducible promoter, Ptrc, to the ORF of the kaiA orkaiC gene, targeted that construct into NSII of both the PkaiA-and PkaiBC-reporter strains, and induced overexpression of each kai gene by addition of isopropyl-β-d-thiogalactopyranoside (IPTG) (28). Overexpression of kaiC immediately nullified the bioluminescence of a PkaiBC-reporter strain, whereas addition of IPTG did not affect the rhythm of a PkaiBC-reporter strain that lacked the Ptrc ::kaiCtransgene (Fig. 4, E and F). Thus, overexpressed KaiC strongly repressed the PkaiBC promoter. In contrast, the presence of the Ptrc ::kaiA construct, even without IPTG addition, reduced the amplitude of the rhythm and markedly increased the average level of bioluminescence (Fig. 4G), suggesting that the promoter is not tightly repressed in the absence of inducer. Further kaiA overexpression induced by IPTG resulted in arrhythmicity and a further three- to fourfold increase in bioluminescence (27). In the PkaiA -reporter strain, kaiC (Fig. 4, H and I) and kaiA overexpression (27) also abolished the rhythmicity but did not change the average level of bioluminescence significantly, as was the case for inactivation experiments. All these data suggest that regulated levels of expression of all three genes is needed for circadian rhythmicity.

We suggest a feedback-loop model for the circadian oscillator ofSynechococcus (Fig. 5). The following four sets of data—(i) mapping of various clock mutations to the kai cluster, (ii) rhythmicity in the expression of thekai genes, (iii) alteration of the rhythmicity ofkai expression by the mutations mapped to the kaicluster, and (iv) elimination of rhythms caused by inactivation or overexpression of each kai gene—all support a model in which the kai genes are essential to the circadian clock, and the feedback regulation of the expression of the kaigenes by their gene products generates the circadian oscillation in cyanobacteria (Fig. 5). In particular, negative feedback on the PkaiBC promoter by KaiC (Fig. 4F) suggests that KaiC could play a role as a “state variable” of the circadian oscillator (29), as appears to be the case forDrosophila PER and Neurospora FRQ proteins (3). To test this hypothesis, we overexpressed thekaiC gene temporarily by pulse administrations of IPTG (Fig. 4, J to N) and found that the phase of the oscillation was reset in a way that the model predicts. Temporal elevation of kaiCexpression advanced the oscillation during a phase in which nativekaiC expression was increasing (the subjective-day phase, hours 0 to 12) and delayed the oscillation during a phase in which the expression was decreasing (the night phase, hours 12 to 24). The magnitude of the phase shifts was proportional to the phase difference between the time of kaiC overexpression and the time when the kaiC gene was maximally expressed. Thus, the level ofkaiC expression is directly linked to the phase of the oscillation.

Figure 5

(right). Feedback model for the circadian oscillator of cyanobacteria. A possible feedback loop (boldarrows) that generates a circadian oscillation is illustrated. Hatched box at left represents an unknown part of the feedback loop. X and Y are unidentified clock components. α and β are unidentified DNA binding proteins. The positive effects of KaiC on the PkaiBC promoter are not shown. See text for explanation.

Activation of clock genes is also important for the negative feedback loop to oscillate. Some threshold level of KaiC appears to be necessary for full PkaiBC-promoter activity becausekaiC expression was reduced by kaiC inactivation. Enhancement of PkaiBC-promoter activity by KaiA (Fig. 4G) might be important to sustain the circadian oscillation ofkaiC expression. Severe lowering of PkaiBC-promoter activity by kaiAinactivation is also compatible with this hypothesis. Recently, positive activators for the expression of circadian clock genes have been identified in Neurospora (5),Drosophila, and mouse (4). The Kai proteins have no DNA binding motifs as are found in some other clock-gene products (4, 5). Thus, unidentified DNA binding proteins might be involved in the KaiA- and KaiC-mediated regulation of the PkaiBC-promoter activity. At present, we cannot suggest any specific role for KaiB.

Although KaiA would be crucial to sustain the oscillation, its expression is not likely to be controlled directly by the proposed feedback loop because PkaiA-promoter activity was affected differently than PkaiBC-promoter activity by inactivation and overexpression of the kaigenes (Fig. 4, E to I). Differences in waveforms of the bioluminescence rhythms of the PkaiA- and PkaiBC-reporter strains (Fig. 3, A and B) also support this distinction.

Although the basic feedback structure of a clock model for cyanobacteria is similar to those for eukaryotes, Kai proteins do not have any similarity to known clock proteins in Drosophila,Neurospora, or mouse (2–5). InDrosophila and Neurospora, time delays in the transcription of a clock gene or genes, the processing, transport into cytoplasm, and translation of its mRNA or mRNAs, and the phosphorylation and nuclear translocation of a clock protein or proteins are postulated to account for a period as long as 24 hours (3). However, some of these processes are absent in cyanobacteria, which are prokaryotic (30). Thus, the phylogenetic relationship of clock systems among various organisms still remains to be elucidated. Putative KaiB and KaiC homologs have been found in archaeal genome databases (31). Although a physiological function for those homologs has not yet been found, this finding might uncover the evolution of circadian systems.

  • Present address: Division of Biological Informatics, Graduate School of Human Informatics, Nagoya University, Chikusa-ku, Nagoya, Aichi 464-8601, Japan.

  • Present address: Department of Cell Biology, Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA, 92037, USA.


View Abstract

Navigate This Article