Functionally Important Substructures of Circadian Clock Protein KaiB in a Unique Tetramer Complex*

KaiB is a component of the circadian clock molecular machinery in cyanobacteria, which are the simplest organisms that exhibit circadian rhythms. Here we report the x-ray crystal structure of KaiB from the thermophilic cyanobacterium Thermosynechococcus elongatus BP-1. The KaiB crystal diffracts at a resolution of 2.6 Å and includes four subunits organized as a dimer of dimers, each composed of two non-equivalent subunits. The overall shape of the tetramer is an elongated hexagonal plate, with a single positively charged cleft flanked by two negatively charged ridges whose surfaces includes several terminal chains. Site-directed mutagenesis of Synechococcus KaiB confirmed that alanine substitution of residues Lys-11 or Lys-43 in the cleft, or deletion of C-terminal residues 95–108, which forms part of the ridges, strongly weakens in vivo circadian rhythms. Characteristics of KaiB deduced from the x-ray crystal structure were also confirmed by physicochemical measurements of KaiB in solution. These data suggest that the positively charged cleft and flanking negatively charged ridges in KaiB are essential for the biological function of KaiB in the circadian molecular machinery in cyanobacteria.

Structure Determination-The details of crystallization and data collection have been previously described (19). The KaiB crystals belonged to the monoclinic space group P2 1 , with unit-cell dimensions a ϭ 90.0, b ϭ 67.0, c ϭ 105.8 Å, and ␤ ϭ 101.3°. The T64C crystals belonged to the monoclinic space group P2 with unit-cell dimensions a ϭ 63.7, b ϭ 33.4, c ϭ 93.7 Å, and ␤ ϭ 100.1°. The diffraction data were collected at SPring-8 beamlines, BL41XU and BL38B1, and the statistics of the diffraction data were reported previously (19). Initial phase was calculated from multiwavelength anomalous dispersion data of the osmium-derivative crystals of KaiB using the SOLVE program (20). After density modification, the protein backbone was traced using the graphics program O (21), but only 50% of the molecule could be constructed due to the low quality of the obtained map. To improve the map, experimental phase was combined with the partial model phase, and after several iterations of model building and phase improvement, Ͼ90% of the molecule was built. Because the T64C crystal showed higher diffraction limit, the model was used to solve the crystal structure of T64C using the molecular replacement method with the program MOLREP (22). By using the dimer model as a search model, we obtained the best solution with a correlation coefficient of 0.487 and an R-factor of 0.478. We refined the model using X-PLOR (23) and, for the final stage, REFMAC5 (22). We randomly selected 5% of the reflections and set them aside as a test set for cross-validation. During the refinement process, we performed manual modification using "omit map." The refinement converged to an R-factor of 22.7% and a free R-factor of 28.9% for all data at a resolution of 2.6 Å. The final model contained 404 residues and 93 solvent atoms. A Ramachandran plot showed 86.5% and 13.5% residues located in the most favorable and allowed region, respectively. Refinement statistics are summarized in TABLE ONE.
Gel Filtration Chromatography-The molecular weight of KaiB was estimated by gel filtration chromatography on a Superdex 75 HR 10/30 column (Amersham Biosciences) equilibrated with 20 mM Tris-HCl (pH 7.5) and 150 mM NaCl at 4°C. For T64C, 1 mM dithiothreitol was included in the elution buffer. Bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa), and cytochrome c (12.4 kDa) were used as molecular mass standards. Proteins were detected at A 280 .
Chemical Cross-linking Experiments-Chemical cross-linking experiments were performed as previously described (15). A total of 6 g of KaiB was incubated at 25°C with 20 mM 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (Pierce) and 20 mM N-hydroxysulfosuccinimide (Pierce) in 20 mM sodium phosphate buffer (pH 7.0) containing 0.15 M NaCl. After 0, 15, 60, or 120 min, the reaction was terminated by the addition of 1 M Tris-HCl (pH 8.0) and 20 mM ␤-mercaptoethanol at final concentrations of 333 mM and 6.7 mM, respectively. Then the products were precipitated by addition of trichloroacetic acid to a final concentration of 6% (v/v), collected by centrifugation, and dissolved in 10 l of a solution containing 6 M urea, 1% SDS, 0.22 mM bromphenol blue, 1% ␤-mercaptoethanol, and 0.1 M sodium phosphate (pH 7.0). The reaction products were subjected to SDS-PAGE according to the Weber-Osborn method (24) and stained with a Silver Stain Kit (Wako). Low Molecular Weight Marker (Amersham Biosciences) was used as a molecular size marker.
Analytical Ultracentrifugation-The molecular weight of KaiB was determined by sedimentation equilibrium analysis at 20°C on an Optima XL-A analytical ultracentrifuge (Beckman) with various rotation speeds, using 0.49 or 0.73 mg/ml KaiB in 20 mM Tris-HCl buffer (pH 7.5) containing 150 mM NaCl. The molecular mass was analyzed using the program Optima TM XL-A/XLI version 4.0 (Beckman).
In Vitro Mutagenesis, Gene Transfer, and in Vivo Rhythm Assay-We mutated kaiB genes by PCR-mediated site-directed in vitro mutagenesis and introduced them into TS2 in the genomes of a kaiB-null strain of Synechococcus carrying a P kaiBC ::luxAB reporter gene (2). We measured the circadian bioluminescence rhythms of the transformed cells as previously described (5,18) at 30°C by using a bioluminescence monitoring apparatus (18,25) and program RAP (26).
For in vivo rhythm assay in T. elongatus, we replaced the kaiABC loci and its downstream tlr0484 locus (17) in the genome of a T. elongatus carrying a P psbA1 ::Xl luxAB reporter gene (18) with a DNA segment containing both the T. elongatus kaiABC loci whose kaiB gene was mutated and a Cm-selective marker gene (17) inserted into the tlr0484 gene by homologous recombination (17). Therefore, the tlr0484 gene was disrupted in the transformed cells carrying the mutated kaiB gene. We measured the bioluminescence rhythms of the transformed cells at 41°C as described previously (18).
Measurement of the CD 4 Spectra of KaiB-CD spectra were determined using a spectropolarimeter equipped with a thermally jacketed quartz cuvette of a 1-mm path length (Jasco JA-720W). The CD spectra of KaiB (10 M) were determined at 25°C in 20 mM Tris-HCl (pH 7.5) containing 150 mM NaCl over a wavelength range of 200 -270 nm.

X-ray Crystal Structure of the KaiB Tetramer: A Dimer of Dimers-KaiB
is highly conserved among twelve cyanobacterial species, with the exception of a short C-terminal region (Fig. 1). KaiB from T. elongatus consists of 108 amino acid residues and has a deduced molecular mass of 12,025 Da. T. elongatus KaiB was crystallized (see "Experimental Procedures"), and the protein structure was determined at 2.6 Å by x-ray crystallography. The asymmetric unit of the KaiB crystal contains four KaiB monomers (labeled A-D in Fig. 2A). The final model includes Lys-6 to Ala-101 for subunit A, Leu-4 to Glu-108 for subunit B, Ala-2 to Ala-101 for subunit C, and Arg-5 to Leu-107 for subunit D. Subunits A and B and subunits C and D form two independent dimers (AB and CD), respectively, in each of which subunits are related by local 2-fold symmetry. Each of these two dimers, AB and CD, has a partner dimer related by a crystallographic 4 The abbreviation used is: CD, circular dichroism.
where F o and F c are the observed and calculated structure factors, respectively. b R free is the R-factor calculated using 5% of the reflections data chosen randomly and omitted from the start of refinement.
2-fold axis forming tetramers T1 (ABAЈBЈ) and T2 (CDCЈDЈ), respectively ( Fig. 2A). These two tetramers are essentially identical, except for the N-and C-terminal regions of the A (AЈ) and C (CЈ) subunits (Fig. 2, A and C). Subunits A and B in tetramer T1 correspond, respectively, to subunits C and D in tetramer T2. The tetramer forms a slightly deformed, elongated hexagonal plate with a size of ϳ85 ϫ 60 ϫ 30 Å. The local 2-fold symmetry axes, which run approximately perpendicular to the crystallographic 2-fold axis, do not intersect the crystallographic 2-fold axis; thereby the four subunits are not equivalent in contrast to usual tetrameric structures. But each tetramer has two pairs of equivalent subunits: A-AЈ (C-CЈ) and B-BЈ (D-DЈ) ( Fig. 2A). The environments of subunits A and B are completely different. Subunit A interacts directly with B and BЈ but not with AЈ, whereas subunit B is in close contact with all the other subunits in the tetramer ( Fig. 2A). The globular core of the KaiB subunit (Tyr-8 to Tyr-94) has a twolayer ␣/␤ sandwich structure consisting of three ␤ strands (␤1, ␤2, and ␤4) and three ␣ helices (␣1, ␣2, and ␣3) (Fig. 2B). This region adopts a novel fold that has not been reported previously. A ␤-sheet comprising parallel strands ␤1 and ␤2 and anti-parallel strand ␤4 forms a wall on one side, while anti-parallel helices ␣1, ␣2, and ␣3 form a wall on the other side (Fig. 2B). The chain connecting ␤2 and ␣2 forms an L-shaped loop (Val-47 to Ala-61) that includes ␤3, which extends away from the core structure (Fig. 2B). ␤3 extends almost perpendicular to the core ␤-sheet and contributes to dimer formation by forming an intersubunit anti-parallel ␤-strand. The core structures of the four independent subunits in an asymmetric unit can be superimposed with root mean square displacements of 0.52-0.65 Å for corresponding C␣ atoms (Fig. 2C).
In contrast to the core structure, both N-and C-terminal regions show large conformational variation. Subunits B and D, which are located in an equivalent position in each tetramer, exhibit similar Nand C-terminal structures (Fig. 2C). Their C-terminal regions from Tyr 94 fold back onto the surface of the core structure, making a loop and forming a short ␤-turn-␤ structure (␤5 and ␤6) in the last stretch from Ala-101, while their N-terminal regions run along these C-terminal chains, although the first three or four residues are missing ( Fig. 2, B and C). However, subunits A and C, which are also located in an equivalent position in each tetramer, have distinctly different terminal chain conformations from subunits B and D. Although the C-terminal chain of subunit A folds back in a conformation similar to subunits B and D, the C-terminal seven residues, which would form the short ␤-turn-␤ motif, is not visible and the N-terminal chain appears to extend away from the C-terminal chain, although the first five residues are not visible. The terminal chains of subunit C extend away from the core (Fig. 2, A and C), where the N-terminal chain is visible from Ala-2, whereas the C-terminal seven residues are not visible. The conformational variability of the terminal chains is displayed graphically in Fig. 2C. These data suggest that the N-and C-terminal regions of KaiB have significant flexibility and adaptability, and these regions may play important roles in the physiological function of KaiB.
Intra-and Interdimer Interfaces-The KaiB tetramer is a dimer of the dimeric unit AB or CD for tetramers T1 and T2, respectively ( Fig. 2A). The dimer-dimer subunit interface is shown graphically in Fig. 3A, and the monomer-monomer interface (within each dimer) is shown graphically in Fig. 3B. The interactions between subunits A and B and subunits C and D are tight. The AB interface is formed by an N-terminal segment of ␤4, a C-terminal segment of ␤1, and the L-shaped loop involving ␤3 (Fig. 3B). This interface is mostly hydrophobic involving Ala-15, Val-47, Leu-48, Pro-51, Leu-53, Ile-59, and Ile-88 (Fig. 3B), unlike the surrounding edges, which are more hydrophilic. Strand ␤3 runs on one edge of this interface, and ␤3 strands from the dimer subunits (AB or CD) form an intersubunit antiparallel ␤ strand interaction at the center of slightly concaved surface of the dimer. Subunit BЈ (DЈ) of the AЈBЈ (CЈDЈ) dimer binds tightly to this surface (Fig. 3A). The other two edges of this dimer interface also have some intersubunit hydrogen bonds. Thus, many hydrogen bonds on the edges enclose the hydrophobic interface, suggesting that this dimer interaction is highly stable in solution and in the intracellular environment.
The dimer-dimer interface (AB-AЈBЈ or CD-CЈDЈ) is comprised of three segments of subunit A or C (␤3, ␣2, and a C-terminal segment of ␤4) and four segments of subunit B or D (an N-terminal segment of ␤1, a C-terminal segment of ␤4, the loop connecting ␣1 to ␤2, and a relatively long segment from ␤3 to the N terminus of ␣3). This interface is more hydrophilic than the monomer interface (Fig. 3A). However, the dimer-dimer interface is extensive and includes several hydrophobic side chain interactions. In addition, the surface area of this interface is larger than the monomer interface, burying 26.5% of the total surface area of the four subunits of KaiB. This suggests that the tetramer structure is highly stable in solution and in the intracellular environment. It may also indicate that the KaiB tetramer has an important biological function.
KaiB Tetramer in Solution-The structure of KaiB was also studied in solution using gel filtration chromatography, chemical cross-linking, and analytical ultracentrifugation. KaiB eluted as a single peak during gel filtration chromatography with an apparent molecular mass of 55 kDa, which corresponds to the molecular mass of a tetramer or pentamer (Fig. 4A). The gel filtration elution profile was the same for KaiB without or with the five residue N-terminal tag, which was added to the recombinant protein used for crystallography (Fig. 4A). This suggests that native KaiB is structurally similar to recombinant KaiB. When KaiB was cross-linked with 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide and the reaction products analyzed by SDS-PAGE, the relative density of the KaiB tetramer increased with increasing incubation time (Fig. 4B). Sedimentation equilibrium analysis confirmed that KaiB is a tetramer with approximate molecular mass of 48.6 kDa. The residual plot for the tetramer model curve fitted to the sedimentation profile showed that higher order association and dissociation does not occur under physiological conditions (Fig. 4C). Similar experiments with Anabaena sp. strain PCC 7120 (hereafter called Anabaena) KaiB confirmed that Anabaena KaiB also exists as tetramer in solution (Fig. 5). This result is not in agreement with a previous report that Anabaena KaiB migrates as a dimer during gel filtration (12). The reason for this discrepancy is not known.
The stability of the T. elongatus KaiB tetramer was examined during gel filtration or sedimentation. The tetramer was stable in solution from pH 4 to 10, from 4 to 40°C and at very low ionic strength (data not shown). This is consistent with the observation that T. elongatus exhibits circadian rhythms in the temperature range 30°C to 60°C (18).
Surface Potential and Charge Distribution for the KaiB Tetramer-The following detailed description of the structure of the KaiB tetramer     DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 applies equally to T1 (ABAЈBЈ) and T2 (CDCЈDЈ). The surface potential of the KaiB tetramer includes regions of non-random charge distribution. The most remarkable feature of the surface potential is two parallel ridges of highly concentrated negative charge surrounding a positively charged cleft. The ridges are related by a 2-fold axis running along the short diagonal of the elongated hexagonal plate (Fig. 6B). However, this structure occurs only on one surface of the tetramer plate, SBBЈT, but it does not form on the related surface, SAAЈT, on which the terminal chains of the subunits are exposed.

Important Substructure and Functional Site of KaiB
The negatively charged ridges consist of the N-and C-terminal regions and the loop connecting ␤2 to ␤3 of subunits B and BЈ. Ten acidic residues of subunits B and BЈ (Glu-55, Glu-56, Asp-57, Glu-95, Glu-96, Asp-99, Glu-102, Asp-103, Asp-104, and Glu-108) are clustered on each ridge (Fig. 6B). Three basic residues (Lys-11, Lys-43, and Lys-58) are on the bottom of the cleft, giving it a highly positively charged character (Fig. 6B). The major portion of each ridge is made up of the C-terminal region from Glu-95 to Glu-108. The segment from Glu-95 to Asp-99 extends away from the core and forms an anti-parallel ␤ strand with the N-terminal chain of the same subunit. Asp-99 is at the peak of the ridge, moving downward through Glu-102, Asp-103, and Asp-104, which lies at the base of the ridge. The most C-terminal segment, Ala-101 to Glu-108, forms a ␤-turn-␤ structure, which is anchored onto the ␤ sheet of the subunit core (Fig. 2B) by a main chain hydrogen bond between Leu-42 and Leu-107.
Another positive surface is observed in and around a hollow formed at the boundary between subunits A and BЈ (or AЈ and B) on the longest edge of the tetramer plate (Fig. 6B). Lys-34 and Lys-37 of subunit B (BЈ) and Lys-6, Lys-67, and Arg-74 of subunit AЈ (A) contribute to this positive surface.
Interestingly, no significant charge distribution was observed on surface SAAЈT (Fig. 6C). The N-and C-terminal chains of subunit A (AЈ) extend straight away from the edge of the tetramer plate on SAAЈT, and FIGURE 6. Electrostatic surface potential of the KaiB tetramer. A, stereo view of C␣ backbone ribbon diagram, color-coded according to the amino acid sequence in rainbow color from the N terminus in blue to the C terminus in red. Side chains of residues for which substitution mutations were examined (Fig. 7) are displayed in ball and stick representation. The figure was generated with MOLSCRIPT (28) and Raster3D (29). B, electrostatic potential of surface SBBЈT. The saturation threshold for the Grasp image is Ϫ10 and ϩ10. C, electrostatic potential of surface SAAЈT. D, electrostatic potential of surface SBBЈT after C-terminal 14 residues are truncated (KaiB 1-94 ). Electrostatic surface potential is color-coded: blue, positive; red, negative. The figures are made by using GRASP (30) and Raster3D (29). The negative ridges are outlined with dotted lines. The positive cleft and positive hollows are indicated by yellow and black arrowheads, respectively. The acidic and basic residues in the positive cleft, negative ridges, and positive hollow are labeled. the C-terminal chain was not visible from Ala-101 onward. Lys-11 and Lys-43, which are in a pocket of positive charge at the bottom of the cleft on the SBBЈT surface, were fully exposed on surface SAAЈT.

Structure-Function Analysis of Ridge and Cleft Regions of KaiB
Tetramer-The overall structure of KaiB determined here shows little similarity to other known proteins, but the electrostatic surface potential of the tetramer (Fig. 6B) provides clues as to possible structurefunction relationships in KaiB. In particular, it seems likely that the positively charged cleft surrounded by negatively charged ridges plays a role in ligand binding and the biological function of KaiB. This idea is supported by the fact that Lys-11, Lys-43, and Lys-58 and a C-terminal cluster of negatively charged residues are highly conserved among twelve strains of cyanobacteria (Fig. 1).
The functional importance of KaiB residues in the ridges and cleft region was tested directly by site-directed mutagenesis of Synechococcus KaiB followed by in vivo analysis of circadian cycling. Alanine substitution mutants of Synechococcus KaiB were constructed and introduced into a targeting site (TS2) in the genome of a kaiB-null strain carrying a P kaiBC ::luxAB reporter gene (2). Using this system, the integrity of the circadian rhythms was assessed by conducting in vivo bioluminescence assays as described previously (2,5). For convenience, in this discussion we refer to the amino acid coordinates of T. elongatus KaiB; when appropriate, the corresponding residues in Synechococcus KaiB are indicated.
Alanine substitution mutations of Lys-11 (K11A) or Lys-43 (K43A) disrupted circadian rhythms and mutation of Lys-58 (K58A) resulted in an unclear rhythm (Fig. 7A), as deduced from in vivo bioluminescence assays in Synechococcus. Similar results were obtained with bioluminescence assays using mutants of T. elongatus KaiB (Fig. 8). In Synechococcus, Western blots were used to confirm that the Synechococcus KaiB mutants were expressed at a similar level as wild-type KaiB (data not shown). Furthermore, the CD profiles of the Synechococcus KaiB mutants were similar to the CD profile of wild-type KaiB (Fig. 9). These data demonstrate that the KaiB alanine substitution mutants studied here are stable and expressed efficiently in vivo in Synechococcus and support the hypothesis that mutations in the positively charged cleft residues in KaiB disrupt circadian rhythms. Therefore, we propose that the positively charged cleft and flanking negatively charged ridges in KaiB may be a functional site associated with a biologically important role in vivo.
The positively charged side hollow (black arrow in Fig. 6B) is another candidate functional site of KaiB. Although Lys-6, Lys-67, and Arg-74 of subunit A (AЈ) are conserved residues, Lys-34 and Lys-37 are not (Fig.  1). Alanine substitution mutations at these residues showed that K67A mutants have low amplitude and low bioluminescence, whereas K6A mutants have normal patterns of bioluminescence (Fig. 7A). However, Lys-67 of subunit B (BЈ) contributes to the dimer-dimer interaction on surface SAAЈT by forming an intermolecular salt bridge with Glu-55 of subunit AЈ (A), suggesting that Lys-67 may play a role in tetramer stability. This possibility was tested by analyzing the gel filtration properties of Synechococcus KaiB K66A (corresponding to T. elongatus K67A mutant). (Note that gel filtration and bioluminescence studies in Synechococcus were performed at 30°C). The mutant KaiB eluted at almost the same position as wild-type KaiB at 4°C or 30°C (data not shown). Thus, it remains possible that Lys-67 plays a role on KaiB clock function. Additional experiments are needed to confirm this possibility.
Functional Analysis of the C-terminal Region of KaiB-The KaiB protein shows extensive sequence homology in twelve cyanobacterial strains (Fig. 1); however, the C-terminal region of KaiB is divergent, and it is possible that this variability confers species-specificity in KaiB function. This idea is supported by the fact that KaiB from Synechocystis sp. strain PCC 6803 (hereafter called Synechocystis) fully complements the null KaiB mutant of Synechococcus, whereas KaiB from T. elongatus and Anabaena partially complemented the mutant, generating a faint rhythm (Fig. 7B). To test this idea, T. elongatus-Synechococcus chimeric KaiB molecules were expressed in Synechococcus kaiB-null host cells, and the ability of the chimeras to complement the rhythm defect of the host was tested. Chimeric KaiB was generated with Synechococcus KaiB residues 1 -93 and T. elongatus KaiB residues 95-108; alternatively, a second chimera contained residues 1-94 from T. elongatus KaiB and residues 94 -102 from Synechococcus KaiB. The results show that the C-terminal region of Synechococcus KaiB is required for complementation of the rhythm defect in Synechococcus kaiB-null host cells (Fig. 7B). These data support the idea that the C-terminal region of KaiB plays a functionally important role and that variation in this region may confer species specificity to the circadian clock molecular machinery in vivo.
To further elucidate the role of the C-terminal region, we expressed a series of Synechococcus KaiB deletion mutants in kaiB-null Synechococcus cells and examined the circadian rhythm. Starting from KaiB 1-100 , C-terminal residues were deleted sequentially one at a time. In rhythm assays, the results showed that the rhythm amplitude was disturbed and the rhythm period was lengthened significantly at every other residue ( Fig. 7C). When C terminus was deleted up to residue 94, and the negative ridges of wild-type Synechococcus KaiB were completely removed (Fig. 6D, KaiB 1-94 ), the rhythm was strongly weakened and destabilized (Fig. 7C, KaiB 1-94 ). These results suggest that the conformation of the ridges in the Synechococcus KaiB tetramer may influence the ability of KaiB to interact with KaiA, KaiC, or some other ligands. Furthermore, the conformation of the ridges may also play a role in the species specificity of KaiB.
Whereas the C-terminal region of KaiB is poorly conserved among cyanobacterial species, the negatively charged character of this region is generally conserved (Fig. 1). To clarify whether the acidic property of the C-terminal region is essential for KaiB clock function, glutamine was substituted for Glu-95, and asparagine was substituted for Asp-98, Asp-100, and Asp-101 in Synechococcus KaiB (Fig. 7D). Each of these mutations alone had minor effects on circadian rhythm in vivo; however, when all four residues were mutated, the circadian rhythm was seriously disturbed (Fig. 7D). The electrostatic surface potential of the tetramer is altered completely in mutants that lack the C-terminal region (Fig. 6D), and a similar effect is likely for the quadruple mutant lacking negative charge in this region. These results suggest that the negative surface potential in the C-terminal region is necessary for KaiB clock function, and the disruption of circadian rhythm caused by truncation or mutation of the C-terminal region is likely to be due to the loss of this negative charge potential.

DISCUSSION
This study reports the 2.6 Å x-ray crystal structure of KaiB, which is an unusual tetramer composed of two asymmetric dimers. To our knowledge, the C-terminal domain of HrcQ B is the only other "dimer of dimers" structure that has been reported (31). In contrast, most tetrameric proteins are composed of four symmetrically related and equivalent subunits. For inherently symmetrical tetramers, non-equivalence among the subunits can be induced by substrate or ligand binding. In the KaiB tetramer, non-equivalence is intrinsic because the dimers bind in an asymmetric manner. This type of intrinsic asymmetry was observed in KaiB crystals in space group P2 1 (with three independent tetramers) and in KaiB crystals in space group P2 (with two tetramers) (19), which argues against the possibility that it is an artifact of crystal packing. In addition, Anabaena KaiB forms a similar dimer of dimers, although a recently reported study that it forms a dimer (12).
Subunits A and B in the KaiB tetramer are tightly bound via an extensive and predominantly hydrophobic interface enclosed by an intermo-  lecular ␤ strand and other hydrogen bonds (Fig. 3). In contrast, the inter-dimer interface between AB and AЈBЈ appears to be less stable and composed of more polar or charge-charge interactions: thus, the interdimer interface could potentially be destabilized by changes in pH or ionic strength (Fig. 3). This suggests that the KaiB tetramer might dissociate into two dimers, and such dissociation could play a role in regulating KaiB during cycles of circadian rhythms.
Previous studies identified two mutations in Synechococcus KaiB, L11F and R74W (corresponding to L12F and R75W, respectively, in T. elongatus KaiB), which alter circadian rhythms in Synechococcus (2). Leu-12 is in the KaiB core, and the adjacent residue, Lys-11, was identified here as essential for KaiB clock function (Fig. 6A). The impact of the L11F mutation may reflect altered side-chain packing induced by the bulky phenylalanine residue, resulting in structural change or instability. The role of Arg-75 is less clear; Arg-75 in subunits A, B, and C was exposed, but Arg-75 in subunit D formed an intermolecular hydrogen bond with Glu-96 in subunit C (Fig. 6A). Because this hydrogen bond is at the dimer-dimer interface, it may make a significant contribution to tetramer stability.
This study presents the results of in vitro mutagenesis and in vivo rhythm assays, which demonstrate that residues in the positive cleft and flanking negative ridges of KaiB are required for circadian rhythms in Synechococcus. Thus, we propose that this region is a functionally active site in KaiB. Amino acid substitutions are tolerated in the C-terminal region that form the negative ridges in KaiB, and the amino acid sequence of this region varies from species to species; however, the highly acidic property of this region is conserved and is likely to be functionally important in KaiB. It is possible that the negative ridges define the specificity of KaiB binding to target proteins, which may vary subtly from one species to another. Because the basic residues are located deep inside the cleft, ligand binding may be induced by structural changes in the more flexible ridges. This is consistent with the fact that the C-terminal region displays significant conformational flexibility and adaptability in the KaiB crystal structure. Therefore, we propose that this region may play an important role in defining and regulating KaiB binding specificity. This may be important to allow circadian rhythms to be maintained as the environment of cyanobacteria change. Thus, speciesspecific variation in the C-terminal region of KaiB may reflect adaptation of the organism to its changing environment.