The crystal structure of the novel calcium-binding protein AtCBL2 from Arabidopsis thaliana.

Arabidopsis thaliana calcineurin B-like protein (AtCBL2) is a member of a recently identified family of calcineurin B-like calcium-binding proteins in A. thaliana. The crystal structure of AtCBL2 has been determined at 2.1 A resolution. The protein forms a compact alpha-helical structure with two pairs of EF-hand motifs. The structure is similar in overall folding topology to the structures of calcineurin B and neuronal calcium sensor 1, but differs significantly in local conformation. The two calcium ions are coordinated in the first and fourth EF-hand motifs, whereas the second and third EF-hand motifs are maintained in the open form by internal hydrogen bonding without coordination of calcium ions. Both a possible site and a possible mechanism for the target binding to AtCBL2 are discussed based on the three-dimensional structure.

Calcium signaling mechanisms are widely employed by all eukaryotic organisms to regulate gene expression and a variety of cellular processes. In plants, many extracellular signals, such as light, drought, cold, salinity, and stress factors, elicit changes in cellular calcium concentration (1,2). The calcium sensor protein often changes its conformation in a calcium-dependent manner and interacts with other signaling proteins to relay the signal. Several families of calcium sensor proteins have been identified in higher plants. One of the most fully characterized sensor proteins is calmodulin, which has four EF-hand motifs for calcium binding (3).
Recently, novel calcium sensor proteins with EF-hand motifs from Arabidopsis thaliana have been identified (4,5), referred to as AtCBL 1 (A. thaliana calcineurin B-like protein). AtCBL is also referred to as SCaBP (SOS3-like calcium-binding proteins), where SOS3 (salt overlay sensitive 3) is the first sensor protein identified in this family. These proteins show substantial sequence similarity with the regulatory B subunit of cal-cineurin (CNB) and neuronal calcium sensor 1 (NCS-1) from animals ( Fig. 1). AtCBLs are encoded by a multigene family of at least 10 members in Arabidopsis and are predicted to have three EF-hand motifs for calcium binding with lower affinity than calmodulin (6). Furthermore, other members of the AtCBL family, such as AtCBL 1, 4, and 8, contain a putative N-terminal myristoylation motif. In fact, AtCBL4 (SOS3) requires N-myristoylation for plant salt tolerance (6).
A family of sucrose nonfermenting-like protein kinases has been identified as targets for CBL proteins (7)(8)(9). These are serine-threonine protein kinases, referred to as AtCIPK (A. thaliana CBL-interacting protein kinase), that form a novel family of proteins found so far only in plants. AtCIPK is equivalent to PKS (SOS2-like protein kinase), where SOS2 is identified as a SOS3 target protein. CIPK consists of regulatory and catalytic domains that interact with each other to keep the kinase inactive, presumably by preventing substrate access to the catalytic site (10). Binding of CBL to the regulatory domain leads to the opening of the catalytic domain's active region and appears to disrupt the intramolecular domain interaction of CIPK. Biochemical data show that a 21-or 24-amino acid region in the regulatory domain is necessary and sufficient for CBL binding. This region contains several conserved residues and is referred to as the FISL motif (10) or NAF domain (11). The Arabidopsis genome contains ϳ25 CIPK protein genes (12). Considering the number of genes in the CBL and CIPK families, one would expect that more than one CBL would interact with more than one CIPK. In addition, CBL/CIPK systems are widely distributed among higher plants, because proteins homologous to CBL and CIPK are identified in other higher plants.
The novel CBL/CIPK system functions in a large array of plant signal transduction in response to stress stimuli. For example, the expression of AtCBL1 is induced remarkably by drought, cold, and wounding stresses, suggesting that the calcium sensor protein is responsive to the respective signaling cascades. AtCBL4 (SOS3) binding activates AtCIPK24 (SOS2) in a calcium-dependent manner, and AtCBL4/AtCIPK24 (SOS3/SOS2) is involved in sodium and potassium ion homeostasis and salt tolerance in Arabidopsis (4,13,14). Furthermore, recent reports show that AtCBL1/AtCIPK15 (ScaBP5/ PKS3) are parts of a calcium-responsive negative regulatory loop that controls sensitivity to abscisic acid (ABA) (15) and that AtCIPK3 regulates ABA response during seed germination (16).
AtCBL2 is a CBL that has no myristoylation motif in the N terminus, and its expression profile suggests a role in lightsignal transduction (17). AtCBL2 interacts prominently with AtSR1, which is identical to AtCIPK14 in the CIPK family. AtSR1 accumulates in response to light (18), evidence that AtCBL2 and AtSR1 are located downstream of the light signal. Despite the great importance of intracellular signaling pathways in plants, no structural information has been obtained for the CBL/CIPK system. Here, we report the crystal structure analysis of calcium-bound AtCBL2 and show that the structure of AtCBL2 is similar in overall folding topology to the structures of CNB and related proteins but differs in the local conformation involved in the target recognition mechanism.

MATERIALS AND METHODS
Expression, Purification, CD Spectroscopy, and Electrospray Ionization Mass Spectra-Details on the expression, purification, and crystallization of AtCBL2 have been reported previously (30). In brief, the AtCBL2 encoding A. thaliana 1-226 was expressed as a fusion protein with glutathione-S-transferase. It was then purified by three column chromatography steps using glutathione Sepharose, HiTrap Q, and Superdex 75 (Amersham Biosciences).
The CD spectra of calcium-bound AtCBL2 were measured on a JASCO J-720W spectrometer using a 0.1-mm quartz cuvette in a stock solution (5 mM Tris-HCl buffer, pH 7.5, 100 mM NaCl, 1 mM dithiothreitol, and 1 mM CaCl 2 ). Successively, the CD spectra were also recorded after the addition of 1.25 mM EGTA.
Electrospray ionization mass spectra were acquired on a Micromass Q-Tof2 mass spectrometer. A solution of 50 M calcium-bound AtCBL2 in the stock solution was applied on a hand-filled short size-exclusion chromatography column packed with Amersham Sephadex G-25, and rapid in-line desalting was achieved prior to electrospray ionizationmass spectroscopy analyses (49). After the equilibration of the sizeexclusion chromatography short column with 10 mM ammonium acetate, 5 l of the sample solution was injected. The mobile phase was introduced at 5 l/min, and electrospray mass spectra were obtained.
Crystallization and Data Collection-Crystals of AtCBL2 were obtained using polyethylene glycol 8000 as a precipitant. The crystals belong to space group C222 1 with unit-cell parameters of a ϭ 83.9 Å, b ϭ 118.1 Å, and c ϭ 49.1 Å. A summary of data collection statistics is given in Table I. All the sets of diffraction intensity data for structural analysis (data-sets of Native-2, trimethyllead acetate, LuCl 3 , and K 2 Pt(NO 2 ) 4 in Table I) were collected at 100 K using Cu-K␣ radiation with a Rigaku R-AXIS IV ϩϩ Imaging Plate diffractometer equipped with Osmic confocal mirror optics, the whole being mounted on a Rigaku FR-D ultra-high brilliant rotating-anode x-ray generator operated at 50 kV and 60 mA. High-resolution native data (data set of Native-1 in Table I) were collected at 100 K using a Jupiter 210 (Rigaku MSC) on the BL45XU beam line at SPring-8, Harima, Japan. The wavelength was set to 1.02 Å with a crystal-to-detector distance of 190 mm and an exposure time of 30 s per degree of oscillation. All the data sets were processed using Crystal-Clear (31).

Structural Determination and Refinement-No successful results
were obtained by the molecular replacement method, in which AMoRe (32) was applied to the structure of the human calcineurin-B subunit (Protein Data Bank code 1AUI) (22) or on the structure of NCS-1 (Protein Data Bank code 1G8I) (19) as a search model. This was attributed mainly to a rather low sequence homology with either of these models, as well as to structural differences. The structure was therefore solved by a multiple isomorphous replacement method. Experimental phases were calculated up to 2.5 Å resolution with SOLVE (33) and improved by solvent-flattening with RESOLVE (34). An initial model built with O (35) was refined with crystallography NMR software (36) to an R value of 33.5%. After several cycles of rebuilding and refinement with REFMAC (37), the model finally converged, resulting in a crystallographic R value of 20.4% and a free R value of 24.8% for all diffraction data up to 2.1 Å resolution. The Ramachandran plot of the final model, containing 189 amino acid residues from Asp-32 to His-220 plus two calcium ions and 138 water molecules, shows that all of the amino acid residues are in the most favored and allowed region defined by the program PROCHECK (38). The structural determination and refinement statistics are summarized in Table I. Accessible surface areas were calculated using the Protein-Protein Interaction server (39). The figures are displayed by GRASP (40)

RESULTS
Overall Structure-The polypeptide chain of AtCBL2 is folded into two globular domains (N-terminal and C-terminal domains) composed of an ␣-helical structure with nine ␣-helices (␣A-␣I), two 3 10 -helices (␣J and ␣K) and four short ␤-strands. These two domains are connected by a short linker (Figs. 1 and 2a) and superimposed with a root mean square deviation (r.m.s.d.) of 2.0 Å for C␣ atoms corresponding to four ␣-helices (␣B-␣E, ␣F-␣I). The structure of AtCBL2 is similar in overall folding topology to the structures of the related proteins, CNB and NCS-1 (respective 23 and 22% sequence identities with AtCBL2), neurocalcin and recoverin (Fig. 1). However, AtCBL2 contains an additional helix (␣A) in the N terminus and a long C-terminal region including ␣J and ␣K, compared with the structure of CNB. In addition, significant differences are observed in local conformation at the domaindomain interface between AtCBL2 and these related proteins. In fact, the superposition of AtCBL2 on CNB for C␣ atoms corresponding to eight ␣-helices yields the large r.m.s.d. of 2.3 where the free reflections (5% of total used) were held aside for R free throughout refinement.
Å, whereas the N-terminal and C-terminal domains are both well superimposed, with r.m.s.d of 1.1 and 1.6 Å, respectively. This indicates that the different domain-domain hinge motions occur between AtCBL2 and the related proteins. In fact, when the N-terminal domains of AtCBL2 and CNB are superimposed ϳ30°swiveling of the C-terminal domain is observed (Fig. 2b). Likewise, ϳ15°rotations are observed between AtCBL2 and NCS-1 (19) and between AtCBL2 and neurocalcin (20). Loop structures connecting ␣-helices at the opposite side of the EF-hands are significantly different between AtCBL2 and CNB. In the loop connecting ␣C and ␣D, the residues located at the end of ␣C and the loop (Leu-74 and Phe-77) make intimate interactions with the residues in the C-terminal region (Tyr-206, Pro-215, and Phe-217). As a result, ␣C and the loop orient inward (Fig. 2c). In contrast, CNB contains a short ␣-helix in the corresponding loop region and the loop protrudes toward the solvent. These structural differences induce a large conformational change in the loop region and a significantly different interhelical angle of ␣C and ␣D (AtCBL2, 50°; CNB, 132°). Although the sequence and length of the loop connecting ␣G and ␣H is similar in AtCBL2 and CNB, the loop conformation and interhelical angle of ␣G and ␣H are different (AtCBL2, 97°; CNB, 126°). This is mainly because of the interactions among the residues (Met-147 and Glu-154) in the ␣G and the loop, and the C-terminal region (Leu-198 and Asn-201) and the loop ␣C-␣D (Lys-78). EF-1 (residues 58 -71; Fig. 4a) displays a large conformational change due to a characteristic sequence in AtCBL2 (see "EF-hand").
Compared with the amino acid sequences of CNB and NCS-1, AtCBL2 has a ϳ30-residue segment extending to the N terminus. This segment is disordered in the crystal and considered to be functionally unimportant, because the deletion mutant of the first 30 amino acids from the N terminus retains the interaction activity (21). Hydrophobic Crevice Shielded by C Terminus Region-The C-terminal regions, including 3 10 -helices ␣J and ␣K of the related proteins, adopt various conformations. The conformation affects the degree to which the hydrophobic crevice is exposed. The crevice is located at the opposite sides of EFhands (Figs. 2a and 3a) and is involved in target recognition by the CNB in complex with calcineurin A (CNA) (22,23). The helices ␣J of neurocalcin and recoverin extend toward this crevice, and the helix ␣K of recoverin is exposed to the solvent region (24,25). The helix ␣J of NCS-1 differs from the helices of neurocalcin and recoverin and is oriented along the edge of the crevice.
In contrast to the related proteins, the C-terminal region that includes two 3 10 -helices of AtCBL2 is plunged into the crevice so as to shield the crevice from the solvent (Fig. 3a), whereas the crevices of NCS-1, recoverin, and neurocalcin are fully or partially exposed into the solvent region. The surface area of the interface between the crevice and the C-terminal region of AtCBL2 is estimated to be ϳ1000 Å 2 , which is within the ranges usually found for the interfaces of protein complexes (26) but is much smaller than the interface of CNB⅐CNA complex (ϳ1800 Å 2 ). The electrostatic surface potential map shows that the crevice is hydrophobic, and the interactions between the crevice and the C terminus region are formed by several hydrophobic residues. Because the residues in the C terminus region (Met-202, Leu-204, Leu-207, Phe-214, and Phe-217) are highly conserved in the AtCBL family except for AtCBL8, the shielding mechanism may be commonly shared.
EF-hand-The canonical EF-hand motif is helix-loop-helix and is characterized by a sequence with the pattern X, Y, Z, ϪY, ϪX, ϪZ, where X, Y, Z, ϪY, ϪX, and ϪZ are the ligands that participate in calcium coordination. AtCBL2 is predicted to have three or four EF-hand motifs (EF-1 ϳ EF-4) (Figs. 1 and 4a) from the amino acid sequence, and two calcium peaks can be clearly identified in EF-1 and EF-4 from the omit map contoured at the 7 level (Fig. 4b). We also confirmed the number of calcium ions coordinated in AtCBL2 by electrospray ionization mass spectra. The spectra indicated that a series of ions corresponding to uptake of 0, 1, or 2 calcium ions per mole equivalent of protein, associated with Na ϩ adducts. After the deconvolution process of the acquired spectrum, three peaks at M r 25953.4 (apo), M r 25991.5 (ϩ1 Ca 2ϩ ), and M r 26029.9 (ϩ2 Ca 2ϩ ) were recognized associated with Na ϩ adducts. No ions with three or four Ca 2ϩ adduct were observed. This suggests that the protein binds two calcium mole equivalents per mole protein, and that some calcium ions bound to the protein are removed off during the ionization or desalting process.
The Lys at the Y position of EF-4 is exceptional for a classical EF-hand structure, but the calcium ion is coordinated in a typical EF-hand fashion of a pentagonal-bipyramidal geometry with the side chain carboxylate of Asp-176 (X), Asp-180 (Z), Glu-187 (ϪZ), the main chain carbonyl of Lys-178 (Y), and that of Lys-182 (ϪY), and a water molecule (ϪX) (Fig. 4b). The sequence of EF-1 differs markedly from that of the classical EF-hand. EF-1 lacks highly conserved Asp residue at the X position, and three residues are inserted between the X and Y positions. The sequence also differs from that of the S100 protein family, which is well known to have a non-classical EF-hand structure. A calcium ion is, however, identified in EF-1 and coordinated in the fashion of a pentagonal-bipyramidal geometry with the side chain carboxylate of Asp-64 (Z), Glu-71 (ϪZ), the main chain carbonyl of Ser-58 (X), Ile-62 (Y), Leu-66 (ϪY), and a water molecule (ϪX) (Fig. 4b).
The sequences of EF-2 and EF-3, which are similar to the sequence of EF-4, are almost canonical except for Lys at the Y position, but no calcium ion is identified in EF-2 or EF-3. The structures of EF-2 and EF-3 are stabilized with many electrostatic non-covalent internal interactions (Fig. 4c). In EF-2, Asp-95 at the X position is hydrogen-bonded to the main chain atom of Gly-100 and the side chain atom of Asn-99, and Glu-106 (ϪZ) is hydrogen-bonded to Asp-95 and Asn-99 through a water molecule. In EF-3, Asp-132 (X) interacts with the main chain atoms of Gln-136 and Gly-137, and water-mediated hydrogen bonds are also observed among Lys-134, Gln-136, and Glu-140. In EF-hand structures that lack calcium ions, similar structural features appear for some calcium-binding proteins. p11 (S100A10) makes hydrogen bonds between the side chain of Asp (X) and main chain atoms (27), whereas recoverin has a salt-bridge between Lys and Glu (24).
All EF-hands in CNB have calcium ions, whereas neurocalcin, NCS-1, recoverin, and guanylyl cyclase activating protein-2, which belongs to the recoverin family (28), has a disabled EF-1 hand (in addition, recoverin has a disabled EF-4 hand), because their Lys and Cys residues are unsuitable for the formation of calcium ligands.
Structural analyses of several calcium-binding proteins have shown that when both EF-hands from a domain pair are occu-FIG. 2. Overall structure of AtCBL2. a, ribbon representation with numbering scheme of secondary structure elements. ␣-Helices and 3 10 helices, colored blue and green, respectively, are sequentially labeled ␣A-␣K. ␤-strands, colored orange, are also labeled ␤1-␤4, and calcium ions are depicted by yellow balls. b, ribbon representation superimposed with CNB. AtCBL2 and CNB are represented with blue and red ribbons, respectively. Two ribbon structures were superimposed, with their corresponding ␣-helices of N-terminal domains shown in a paler shade. N-terminal regions are omitted for clarity. Rotation angle of the C-terminal domains of AtCBL2 and CNB is ϳ30°. c, superimposition of AtCBL2 (blue) and CNB (red) viewed from the opposite side of a. Segments displaying large displacements are highlighted, and the residues discussed are displayed.
FIG. 3. Hydrophobic crevice of AtCBL2 and CNB (a) positive (blue) and negative (red) potentials are mapped on the van der Waals surfaces, excluding the C-terminal region, in the range Ϫ20 k B T (red) to ϩ20 k B T (blue), where k B is Boltzmann's constant and T is the absolute temperature. The C-terminal region is represented as a tube model colored green. b, the hydrophobic crevice of CNB, represented by van der Waals surfaces with positive (blue) and negative (red) potentials. A tube model colored green is an ␣-helix protruding from CNA in CNB⅐CNA complex. pied by calcium ions, the helices of each EF-hand exhibit open conformation related by ϳ90°, and in calcium-free EF-hands the helices exhibit closed conformation with nearly anti-parallel arrangement. All the ␣-helices in the four EF-hand structures of AtCBL2 are well superimposed, with r.m.s.d. values of 1.3-1.5 Å for corresponding ␣-helices, whereas the angle between the two ␣-helices within each EF-hand structure ranges from 77°to 123°. This indicates that the four EF-hands in AtCBL2 adopt open conformations.
As shown in Fig. 5, the CD spectrum of AtCBL2 was markedly different between a holo-form and an apo-form; the latter is derived from treating the holo-form with ethylene glycol bis(2-aminoethyl ether) tetraacetic acid (EGTA), a chelating reagent. This result suggests that the AtCBL2 changes its conformation in a calcium-dependent manner and thereby functions as a molecular switch through the EF-hands. DISCUSSION The present study is the first crystal structure analysis of proteins belonging to the AtCBL family. It revealed that AtCBL2 has two calcium-loaded EF-hands despite atypical sequence for EF-hand, indicating that, among members of this family, the number of calcium ions depends on the sequence alignment (Fig. 4a). For example, the Y position of EF-2 in AtCBL7 is occupied by Asn residue, which is suitable for the coordination to calcium ion. Likewise, each Y position of EF-3 in AtCBL1, 6, or 9 is either an Asp or an Asn residue. These structural characteristics favor the coordination of calcium ions in the respective EF-hand motifs. In contrast, the fact that four amino acid residues are deleted around the Y position of EF-1 in AtCBL6 (Fig. 4a) indicates that AtCBL6 is unlikely to contain calcium ions in EF-1. Such differences in calcium coordination within the CBL family might be related to the specificity of target recognition in calcium signaling. Ishitani et al. (6) showed that the mutant of a three-amino acid deletion in the EF-2 in SOS3 (AtCBL4) revealed little or no calcium binding while showing disruption of the interaction between SOS3 and SOS2. The EF-2 in AtCBL2 has no calcium ion and SOS3 probably does not contain it, because the amino acid sequences are homologous in this region. Therefore, the deletion may not cause a loss of calcium but may instead disrupt the conformation of the EF-hand kept in the open form. This would decrease the affinity for calcium ion while also increasing disruption of the target protein.
It is known that the ␤-sheet of the C-terminal domain of calmodulin, to which calcium binds with high cooperativity, is highly twisted in the apo form, and calcium binding removes this twist (44 -46). Therefore, this conformational change in the ␤-sheet may play an important role for the cooperative calcium binding, by which calcium binding in one EF-hand can affect the other EF-hand to form the calcium binding conformation. Cooperativity was observed for the pair of sites in each domain, but not between the N-and C-terminal domains. For example, unmyristoylated recoverin has one functional EFhand in the domain and exhibits uncooperative binding of two calcium ions (47). Because AtCBL2 has one functional EF-hand in each domain with no myristoylation, this protein would exhibit uncooperativity, and thus ␤-sheet would not be related with the cooperative binding. Moreover, AtCBL2 functions as a sensor with two EF-hands (EF-1 and -4), and this structural study shows the disabled EF-2 and -3 because of the internal interactions.
The target recognition mechanism by calcium-binding proteins with EF-hand motifs has been extensively studied for calmodulin, which is recognized by its interaction with the short helices in the calmodulin binding domain of each target protein (29). In the crystal structure of CNB in complex with CNA, the hydrophobic crevice of CNB recognizes a five-turned ␣-helix protruding from CNA (Fig. 3b). The helix J of CNB is pushed aside along the edge of the crevice. The affinity for the target protein might be stronger than that for the C-terminal region, because the interface surface area between CNB and CNA in CNB⅐CNA complex is significantly (ϳ1.8 times) larger than that between the hydrophobic crevice and the C-terminal region of AtCBL2. Assuming that the recognition mechanism between AtCBL2 and CIPK is similar to that between CNB and CNA in CNB⅐CNA complex, the C terminus region of AtCBL2 is released from the hydrophobic crevice so that it can interact with the CBL binding domain of CIPK when AtCBL2 interacts with CIPK. It is therefore indicated that the C terminus region of AtCBL2 blocks the adventitious binding of various proteins to AtCBL2 in the absence of CIPK. As stated above, the hydrophobic crevice of AtCBL2 corresponds to that of CNB and is ϳ36 Å, long enough to recognize 21 or 24 amino acid residues. It should also be noted that the NAF domain or FISL motif necessary for CBL binding consists predominantly of hydrophobic residues. These structural features suggest that the CBL binding domain of CIPK interacts hydrophobically with CBL.
Structural studies of recoverin family have shown that a hinge motion between N-and C-terminal domains is dependent on the calcium binding state (48). The binding of calcium ion to the EF-hand in recoverin leads to local structural changes within the EF-hand that alter the domain interface and cause a swiveling of the N-and C-terminal domains. Likewise, it is supposed that the calcium binding to the EF-hands of AtCBL2 induces local structural changes, and a backbone hinge rotation at the domain linker allows a re-arrangement of the domain interface. The C-terminal region in recoverin partially shields the crevice, and this region has similar conformation in apo-and holo-state. Therefore, the C terminus region of AtCBL2 would have the same conformation in apo-and holostate to shield the hydrophobic crevice.
The calcium-bound structures, guanylyl cyclase activating protein-2 (28), NCS-1 (19), and neurocalcin (20), show similar swiveling angle of the N-and C-terminal domain. Their structures were determined by different methods, NMR (guanylyl cyclase activating protein-2) and x-ray (NCS-1 and neurocalcin), and x-ray structures were solved in the different crystal packing. These facts suggest that the swiveling angle is intrinsic and not an indication of an artifact due to crystal packing interactions.