Isolation of a Calmodulin-binding Transcription Factor from Rice (Oryza sativa L.)*

Calmodulin (CaM) regulates diverse cellular functions by modulating the activities of a variety of enzymes and proteins. However, direct modulation of transcription factors by CaM has been poorly understood. In this study, we isolated a putative transcription factor by screening a rice cDNA expression library by using CaM:horse-radish peroxidase as a probe. This factor, which we have designated OsCBT (Oryza sativa CaM-binding transcription factor), has structural features similar to Arabidopsis AtSRs/AtCAMTAs and encodes a 103-kDa protein because it contains a CG-1 homology DNA-binding domain, three ankyrin repeats, a putative transcriptional activation domain, and five putative CaM-binding motifs. By using a gel overlay assay, gel mobility shift assays, and site-directed mutagenesis, we showed that OsCBT has two different types of functional CaM-binding domains, an IQ motif, and a Ca2+-dependent motif. To determine the DNA binding specificity of OsCBT, we employed a random binding site selection method. This analysis showed that OsCBT preferentially binds to the sequence 5′-TWCG(C/T)GTKKKKTKCG-3′ (W and K represent A or C and T or G, respectively). OsCBT was able to bind this sequence and activate β-glucuronidase reporter gene expression driven by a minimal promoter containing tandem repeats of these sequences in Arabidopsis leaf protoplasts. Green fluorescent protein fusions of two putative nuclear localization signals of OsCBT, a bipartite and a SV40 type, were predominantly localized in the nucleus. Most interestingly, the transcriptional activation mediated by OsCBT was inhibited by co-transfection with a CaM gene. Taken together, our results suggest that OsCBT is a transcription activator modulated by CaM.

In plants as well as in animals, Ca 2ϩ has a vital role in mediating cellular responses to external stimuli of both abiotic and biotic origins (1)(2)(3)(4). The transient elevation of cytosolic free calcium concentration that occurs in response to specific stimuli differs in amplitude, frequency, and duration depending on the exact nature of the stimulus (5).
Thus, Ca 2ϩ triggers a myriad of cellular processes that influence growth (6), development (7), and physiology (8), which allow organisms to adapt to the changing environment. Ca 2ϩ -dependent modulation of cellular processes occurs via intracellular Ca 2ϩ -binding proteins. Calmodulin (CaM) 5 is a highly conserved and well characterized primary Ca 2ϩ sensor in eukaryotes and is involved in Ca 2ϩ -mediated signal transduction pathways by regulating the activities of its targets/binding proteins (9 -12).
Although CaM-binding proteins identified to date are mostly cytosolic, accumulating evidence suggests that CaM may be involved in transcriptional regulation. In animals, Ca 2ϩ /CaM indirectly regulates transcription through kinase cascades and the phosphatase calcineurin (13). Ca 2ϩ /CaM can bind to some members of the basic helix-loop-helix transcription factor family, inhibit their DNA binding, and directly influence downstream gene transcription (14 -16). In plants, Zielinski and co-workers (17) showed that TGA3, a bZIP transcription factor, displayed enhanced binding activity to the Arabidopsis CaM-3 promoter element when CaM was bound to TGA3. However, the CaMbinding domain (CaMBD) of TGA3 has not been identified. Recently, Yoo et al. (18) showed that AtMYB2 bound to soybean CaM (GmCaM) in a Ca 2ϩ -dependent manner, and the GmCaM isoforms GmCaM-1 and GmCaM-4 differentially regulated AtMYB2 DNA binding activity.
A family of CaM-binding/DNA-binding proteins containing a DNAbinding domain homologous to parsley CG-1 has been identified from Arabidopsis and designated AtSRs/AtCAMTAs (19,20). Parsley CG-1 is a novel DNA-binding protein whose binding site includes the CGCG motif, and whose transcript accumulates rapidly and transiently upon UV irradiation (21). AtSR/AtCAMTA homologues are also found in tobacco (NtER1), tomato (LeER66), and rapeseed (BnCAMTA) (19,22,23). All of these homologues contain the conserved CG-1 homology DNA-binding domain, the TIG domain, several ankyrin (ANK) repeats, and a variable number of CaM-binding motifs. Yang and Poovaiah (20) showed that the NtER1 homologues AtSR1-6 are induced by various environmental signals, and all of them bind to Ca 2ϩ /CaM. Bouchè et al. (19) showed that one Arabidopsis homologue, AtCAMTA1, binds to CaM in a Ca 2ϩ -dependent manner and verified its ability to activate transcription in yeast. Moreover, they suggested that this family of proteins is conserved in a wide range of multicellular eukaryotes, including human and Drosophila, based on a bioinformatics analysis (19). Even though the physical interaction between CaM and this family of pro-teins has been extensively characterized, the mechanism by which CaM regulates their activity has not yet been elucidated.
In this study, we isolated a putative CaM-binding transcription factor (having structural features and amino acid sequences with a high degree of similarity to the AtSRs/AtCAMTAs family) by screening a rice cDNA expression library using horseradish peroxidase-conjugated CaM (CaM:HRP) as a probe. We designated this cDNA clone OsCBT (Oryza sativa CaM-binding transcription factor). By using CaM overlay assays with several deletion mutants, a gel mobility shift assay, and site-directed mutagenesis, we showed that OsCBT possesses two distinct types of CaM-binding domains, Ca 2ϩ -dependent CaM dissociation and Ca 2ϩ -dependent CaM-binding motifs, and that CaM binding to OsCBT has an inhibitory effect on transcriptional activation. Our results suggest that OsCBT encodes a transcriptional activator regulated by CaM binding.

MATERIALS AND METHODS
Screening of a Rice cDNA Expression Library-Screening of the rice cDNA expression library using HRP-conjugated soybean CaM (GmCaM-1) was carried out as described in previous reports (24 -26).
Construction of C-terminal Deletion Mutants of OsCBT cDNA and Site-directed Mutagenesis-To map the CaM-binding domain, several C-terminal deletion constructs of OsCBT were generated in a pAK-SS vector derived from pGEX-2T (Amersham Biosciences). First, a SmaI site was introduced downstream of the TGA codon of a full-length cDNA clone (named OsCBT.B) by PCR. The OsCBT.B clone was digested by SmaI, MfeI, NdeI, or EcoRI and blunted by T4 DNA polymerase. The fragments were cut with NcoI, and the resulting fragments were subcloned into a pAK-SS vector digested with BamHI and SmaI. The glutathione S-transferase (GST) fusion constructs were named according to the fragments they contained as follows: D0 for NcoI/SmaI (amino acids 1-927); D1 for NcoI/MfeI (amino acids 1-720); D2 for NcoI/NdeI (amino acids 1-540); and D3 for NcoI/EcoRI (amino acids 1-253). The five putative CaMBDs (amino acids 440 -476) of OsCBT were amplified using a 5Ј primer containing a BamHI site and a 3Ј primer containing a SmaI site. The 5Ј primer and 3Ј primer sequences used are as follows: for Da, 5Ј primer, 5Ј-GGC CCC ATG GCT CTA  GCA GCC TAT CGC-3Ј, and 3Ј primer, 5Ј-TAT CCC GGG GCA GAT  GCT TCA ATC TCA GG-3Ј; for Db, 5Ј primer, 5Ј-CGG GAT CCG  CTA ATC CTG AGA TTG AAG C-3Ј, and 3Ј primer, 5Ј-TAT CCC  GGG GCC TGT ATT CGT GCA GCA GC-3Ј; for Dc, 5Ј primer,  5Ј-CGG GAT CCG CAA TGA GAG CTG CTG CAC G-3Ј, and 3Ј  primer, 5Ј-TAT CCC GGG CGG TAC GCA GCT TGT ATC C-3Ј; for  Dd 5Ј primer, 5Ј-CGG GAT CCA TGC GAA GAC AAG TTA TCA  GG-3Ј, and 3Ј primer, 5Ј-TAT CCC GGG TCA ACA ATT CCA ACA  GAC C-3Ј; for De, 5Ј primer, 5Ј-CGG GAT CCC AAC AAG CCG AGG ACA GG-3Ј, and 3Ј primer, 5Ј-CGG GAT CCC AAC AAG CCG AGG ACA GG-3Ј. Each amplified fragment was also subcloned into a pAK-SS vector (GST:OsCBT.CaMBD a-e, denoted by Da-De). To identify the critical domains in the interaction between CaM and OsCBT, we introduced several point mutations into the GST:OsCBT CaMBD b and e clones. Substitutions of single amino acids were performed using the QuickChange TM site-directed mutagenesis (Stratagene, La Jolla, CA). The forward (F) and reverse (R) primer sequences used are as follows: for I764A of CaMBD I, F, 5Ј-GAG ATA GTT GCT GCT TTG AAG  GCT CAA CAT GCA TTT CGG AAC TAC-3Ј, and R, 5Ј-GTA GTT  CCG AAA TGC ATG TTG AGC CTT GAA AGC AGC AAC TAT  CTC-3Ј; for V829R of CaMBD II, F, 5Ј-GGT AAT ATG GTC TGT  TGG AAT TCG TGA GAA AGC AAT TTT GCG ATG G-3Ј, and R,  5Ј-CCA TCG CAA AAT TCG TTT CTC ACG AAT TCC AAC AGA   CCA TAT TAC C-3Ј; for W836R of CaMBD II, F, 5Ј-GAG AAA GCA  ATT TTG CGA CGG AGA AAG AAG AGA AAA GCC-3Ј, and R,  5Ј-GGC TTT TCT CTT CTT TCT CCG TCG CAA AAT TGC TTT  CTC-3Ј. Expression of Recombinant Proteins in Escherichia coli and the CaM Binding Assay-All the clones were introduced into E. coli BL21 (DE3) pLysS and expressed. Expression of the GST fusion proteins was induced by application of 1 mM isopropyl 1-thio-␤-D-galactopyranoside for 5 h at 25°C. The cells were harvested, resuspended in lysis buffer (50 mM Tris-HCl (pH 7.5), 2 mM phenylmethylsulfonyl fluoride, 1 mM DTT, 100 g/ml lysozyme), and incubated on ice for 20 min. The mixture was sonicated for 1 min at a 50% pulse and then centrifuged at 6,000 ϫ g for 10 min to remove cell debris. The supernatant was used for Western blotting and a CaM:HRP gel overlay assay. E. coli crude cell lysates (20 g of protein) were separated on 10% SDS-polyacrylamide gels and transferred onto Immobilon-P membranes (PVDF, Millipore, Billerica, MA). GST fusion proteins were detected by a polyclonal GST-specific antiserum. To examine the CaM binding ability of the recombinant proteins, a duplicate blot was probed with OsCaM-1:HRP in the presence of 1 mM CaCl 2 or 5 mM EGTA. The CaM:HRP overlay assay was carried out as described previously (26). The bound CaM was visualized using ECL (Amersham Biosciences).
CaM Mobility Shift Assay with a Synthetic Peptide-The 826 VGIVE-KAILRWRKKRKALRG 845 peptide, corresponding to a stretch of 20 amino acids in the CaMBD II of OsCBT, was synthesized at a peptide synthesis facility (EURO-GENETEC S. A., Belgium). The CaM binding ability of the synthetic peptide was measured by determining the relative mobility shift of CaM in the presence of this peptide (27). CaM (303 pmol) was incubated with increasing concentrations of the peptide (molar ratios: 0.0, 0.25, 0.5, 0.75, 1.0, 1.5, 2.0, and 3.0) in a binding buffer (100 mM Tris-HCl (pH 7.2) plus 0.1 mM CaCl 2 or 2 mM EGTA) at room temperature for 1 h. The samples were separated by gel electrophoresis at a constant voltage of 100 V in an electrode buffer (25 mM Tris-HCl (pH 8.3), 192 mM glycine containing either 0.1 mM CaCl 2 or 2 mM EGTA) and stained with Coomassie Brilliant Blue.
Electrophoretic Mobility Shift Assay-DNA probes were generated by annealing oligonucleotides spanning the regions of interest and filling in the 5Ј overhangs with a Klenow fragment polymerase (Takara, Tokyo, Japan), [ 32 P]dATP, and unlabeled dCTP, dGTP, and dTTP. The DNAbinding reactions were allowed to proceed at 25°C for 20 min in 10 l of 1ϫ binding buffer (20 mM HEPES/KOH (pH 7.9), 0.5 mM DTT, 0.1 mM EDTA) (28), 50 -150 mM KCl, 15% glycerol, 1-5 g of poly[dI⅐dC], and 0.5 g of bacterially produced fusion protein that had been purified with glutathione-Sepharose. The mixture was incubated at 25°C for 30 min after adding 40,000 cpm of 32 P-labeled DNA probe. The reaction mixture was subjected to electrophoresis on an 8% polyacrylamide gel in 0.5ϫ TBE buffer at 80 V for 3 h. The gel was dried onto 3MM paper and mounted for autoradiography at Ϫ70°C with intensifying screens.
Random Binding Site Selection (RBSS)-The in vitro binding site selection procedure was derived from Pollock and Treisman (29), and we modified some of the procedures as described below. The oligonucleotide synthesized for the binding selection was TN66 (5Ј-CGC TAC GTC GGA AGA CAA GCT TGT AA(N) 15 ATA GGA TCC CTC ACC TCA GAC AGA C-3Ј), containing a randomized sequence of 15 nucleotides flanked by 26 nucleotides at the 5Ј-end containing a HindIII site and 25 nucleotides at the 3Ј-end containing a BamHI site. For PCR amplification, the oligonucleotides TNF20 (5Ј-CGC TAC GTC GGA AGA CAA GC-3Ј) and TNR20 (5Ј-GTC TGT CTG AGG TGA GGG AT-3Ј) were the forward and reverse primers, respectively. In order to produce double-stranded random binding DNA sequences, the oligo-nucleotides TN66 (400 pmol) and TNR20 (800 pmol) were annealed, and the second strand was synthesized with the Klenow fragment polymerase in the presence of dNTPs (final concentration, 0.25 mM each). OsCBT DNA-binding (OsCBT. DB) protein was preincubated with 5 g of poly[dI⅐dC] as a nonspecific competitor for 10 min at room temperature in 20 mM HEPES (pH 7.9), 50 mM KCl, 0.1 mM EDTA, 0.5 mM DTT, and 15% glycerol prior to addition of the 32 P-labeled probe (40,000 cpm) to give a final volume of 30 l. After additional 15 min of incubation, the DNA-protein complex was separated on a 5% polyacrylamide gel (29:1 acrylamide/bisacrylamide) containing 0.5ϫ TBE and 3% glycerol for 1.5-2 h at 15-20 V/cm. The DNA-protein complex was excised from gel and electroeluted into a dialysis bag. DNA was extracted with phenol/chloroform and precipitated with ethanol. The recovered DNA was resuspended in an appropriate volume of deionized water. PCR amplifications were carried out in a final volume of 100 l containing 80 pmol of each primer, 20 nmol of dNTPs, 10 mM Tris-HCl (pH 8.3), 1.5 mM MgCl 2 , 50 mM KCl, and 2.5 units of Taq polymerase. After an initial denaturation at 92°C for 30 s, selected binding sequences were amplified with 30 cycles of 30 s at 92°C, 1 min at 55°C, and 30 s at 72°C in a thermal cycler (Amersham Biosciences). The amplified DNA of 66 bp in length was purified on a 2% agarose gel, eluted as described above, and used as the template for the following round of selection. Out of five selection cycles (binding/gel shift/elution/PCR amplification) carried out, the first three selections were performed using binding conditions of 50 mM KCl, whereas the latter two selections were carried out at 50 and 150 mM KCl independently. The final pool of oligonucleotides were digested with BamHI and HindIII, ligated into pBluescript II SK(ϩ) (Stratagene), and transformed into XL-1 Blue MRF. Out of the transformants, 100 clones were randomly chosen from the 150 mM KCl selected transformants. The sequences of the inserted DNA were deter-mined by dideoxynucleotide chain termination using an automatic DNA sequencer (ABI 377, Applied Biosystems Inc., Foster City, CA).
Subcellular Localization and Transient Expression Assays-For preparation of protoplasts, leaf tissue (5 g) of 3-4-week-old Arabidopsis plants grown on MS medium were digested with 50 ml of an enzyme solution (1.2% cellulase R-10 (Yakult, Japan), 0.3% macerozyme R-10, 10 mM MES-KOH (pH 5.6), 5 mM ␤-mercaptoethanol, 0.1% (w/v) bovine serum albumin, 8 mM CaCl 2 , and 0.4 M mannitol, at a final pH 5.8) at 22°C with gentle agitation (50 -75 rpm) overnight. All centrifuges were carried out at 40 -60 ϫ g for 5 min at room temperature. The protoplasts were isolated from the undigested material with a sieve (100 mesh), and the protoplasts were pelleted by centrifugation and resuspended in 5-10 ml of W5 solution (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl, 5 mM glucose, 1.5 mM MES-KOH (pH 5.6)). The protoplasts overlaid 20 ml of 21% sucrose and were centrifuged for 10 min at 78 ϫ g. The intact protoplasts at the interface were transferred to a new Falcon tube containing 20 ml of W5 solution (30). The protoplasts were pelleted again by centrifugation for 5 min at 55 ϫ g and resuspended in 20 ml of W5 solution. For the transfection experiments, polyethylene glycol-mediated DNA uptake methods were performed following the protocol of Jin et al. (31).
The reporter plasmid was a pUC19-derived plasmid containing the ␤-glucuronidase (GUS) reporter gene under the control of a 35S minimal promoter (32). A synthetic OsCBT-binding site (OsCBT BS C4) was inserted in front of the CaMV 35S minimal promoter. For effector plasmids, full-length fragments of OsCBT or OsCaM1 cDNA were inserted into a plant expression vector (HBT95) containing the 35S C4PPDK promoter and nos terminator (33). Transient expression of these constructs was performed as described previously (34). In each transfection, 2 ϫ 10 6 protoplasts were transfected with 30 g of plasmid DNA carrying a reporter construct alone or with plasmid DNA carrying an effector construct or a vector DNA control. The transfected protoplasts were incubated for 16 h in the dark. A construct carrying the 35S promoter fused to the luciferase gene was used as an internal control in each transfection where the OsCBT BS-m35S-GUS reporter construct was tested. In each sample, the GUS activity of the cell lysate was divided by the luciferase activity, thereby normalizing the data to control for variations in transfection efficiency.

Molecular Cloning of a cDNA Encoding a CaM-binding Transcription
Factor-To understand the roles of Ca 2ϩ /CaM in plant sensing and response to environmental signals and to identify the molecular components of Ca 2ϩ /CaM-mediated signaling pathways, we isolated CaMBPs by screening a fungal elicitor-treated rice cDNA expression library using HRP-conjugated soybean CaM-1 (GmCaM-1:HRP) as a probe (24 -26). From the in vitro protein-protein interaction-based screening of 5 ϫ 10 5 recombinant phages, we isolated a 3,134-bp fulllength cDNA clone encoding a 927-amino acid protein with a predicted molecular mass of ϳ103 kDa.
Analysis of the deduced amino acid sequence shows that this putative CaMBP contains several structural features of transcriptional regulators (Fig. 1). This clone has two different types of nuclear localization sequences (NLS), a bipartite NLS in the N terminus (amino acids 73-90), and an SV40-type NLS in the C terminus (amino acids 837-840). The N-terminal 147 amino acids show high homology to the CG-1 domain of parsley CG-1, which is a sequence-specific DNA-binding protein (21). A TIG domain, involved in nonsequence-specific DNA contacts in various transcription factors (19,35), follows the DNA binding domain. Three ankyrin repeat domains, each composed of 33 amino acids and known to mediate protein-protein interactions, are found in the central region of the protein (amino acids 374 -462) (36). At the C-terminal end (amino acids 861-920), acidic amino acids such as glutamate and aspartate, as well as proline and glutamine, are enriched, suggesting that this region functions as a transcriptional activation domain (Fig. 1B) (37). Moreover, the C-terminal region of the protein (amino acids 726 -845) contains a cluster of several putative CaM-binding domains, four Ca 2ϩ -dependent CaM dissociation domains (IQ motifs), and a Ca 2ϩ -dependent CaM-binding domain (38,39). These structural features suggest that this clone is a transcription factor whose activity can be modulated by CaM binding. Thus, we designated this clone OsCBT (GenBank TM accession number AF499741).
OsCBT Contains Two Distinct Types of CaMBDs-Comparisons of the CaM-binding regions from previously reported CaM-binding proteins have shown that there are multiple sequence motifs that can mediate complex formation between CaMs and their targets (38,39). Based on comparisons with the structural characteristics of known CaMbinding domains, we were able to map putative CaM-binding regions in the C-terminal region of OsCBT, between Arg 726 and Glu 845 , consisting of four putative Ca 2ϩ -dependent CaM dissociation domains (IQ motifs) and one Ca 2ϩ -dependent CaM-binding domain in which hydrophobic amino acids are present at positions 1 (Val 829 ), 5 (Ile 833 ), 8 (Trp 836 ), and 15 (Leu 843 ), interspersed by several basic residues (four lysines and four arginines) between these hydrophobic residues (Fig. 1B).
To determine whether putative CaM-binding domains predicted by sequence information are actually involved in interactions with CaM, we made a series of GST fusion constructs containing the full-length cDNA and three C-terminal deletion mutants (Fig. 2A). The GST fusion recombinant proteins were expressed in E. coli and verified by Western blotting of cell lysate with an anti-GST antibody. Binding to CaM was tested by CaM overlay assay in the presence or absence of Ca 2ϩ , using OsCaM:HRP as a probe (Fig. 2B). As discussed in a previous report (26), GmCaM-1, which was used to probe the rice cDNA expression library, is nearly identical to the conserved rice CaMs, suggesting that they might share target/binding proteins. Indeed, GmCaM-1 and OsCaM (GenBank TM accession numbers X65016) showed similar binding to OsCBT in the in vitro CaM binding assays (data not shown). Therefore, we used OsCaM as a probe in the following experiments.
The full-length recombinant protein (D0) containing a cluster of putative CaMBDs interacted with the OsCaM:HRP conjugate, but GST only (C, vector control) and the GST fusion proteins lacking the predicted CaM-binding regions (designated D1, D2, and D3) did not bind to OsCaM:HRP. Additionally, OsCBT interacted with CaM in both the presence and absence of Ca 2ϩ (Fig. 2B). These results demonstrate that OsCBT possesses both Ca 2ϩ -dependent CaM dissociation domain and Ca 2ϩ -dependent CaM-binding domain in the C terminus. Sequence analysis suggests that OsCBT contains four Ca 2ϩ -dependent CaM dissociation domains (IQ motifs) and one Ca 2ϩ -dependent CaM-binding domain (Fig. 1B). To confirm further that these putative CaM-binding domains actually take part in interacting with CaM, we made a series of GST fusion constructs containing each CaM-binding domain, and we designated them Da, Db, Dc, Dd, and De (Fig. 3A). CaM-binding abilities of these recombinant proteins were determined by CaM overlay assay in both the presence and absence of Ca 2ϩ . As shown in Fig. 3B, out of the five putative CaMBDs of OsCBT, OsCaM bound to the Db domain in the absence of Ca 2ϩ and bound to the De domain in the presence of Ca 2ϩ . We named the Db and De regions CaMBD I and CaMBD II, respectively.
Most of the characterized CaM-binding proteins possess a basic amphiphilic ␣-helical region (40). A helical wheel projection of the 18 amino acid sequences corresponding to the OsCBT CaMBD II (Gly 829 to Arg 842 ) showed a typical CaMBD feature, segregation of basic and hydrophobic residues on opposite sides of the helix (Fig. 3C) (41).  DECEMBER 9, 2005 • VOLUME 280 • NUMBER 49

JOURNAL OF BIOLOGICAL CHEMISTRY 40825
To confirm further the Ca 2ϩ -dependent binding of CaMBD II to OsCaM, we used a synthetic 20-amino acid peptide, Val 826 to Gly 845 of OsCBT, for a gel shift assay, in which the peptide-bound CaM would migrate more slowly than free CaM under nondenaturing conditions (27). As shown in Fig. 3D, the intensity of the higher molecular mass band representing the peptide-CaM complex increased with increasing concentrations of the synthetic peptide in the presence of 0.1 mM Ca 2ϩ , whereas the higher molecular weight complex was undetectable when 2 mM EGTA was substituted for 0.1 mM Ca 2ϩ . This result shows that the 20-mer peptide from Val 826 to Gly 845 of OsCBT CaMBD II is sufficient for Ca 2ϩ -dependent CaM binding. Taken together, these results suggest that OsCBT contains two distinct types of CaM-binding motifs.
Critical Residues in the Interaction between OsCBT and CaM-To check the conservation of CaM-binding sequences in OsCBT homologues, we compared amino acid sequences of 27 OsCBT homologues identified from data base searches using the CaMBDs of OsCBT (between Arg 726 and Glu 845 ) (Fig. 4A). The Ca 2ϩ -dependent CaM-binding motif corresponding to OsCBT CaMBD II is highly conserved among all of the OsCBT homologues. Within the 20-amino acid stretch, hydrophobic amino acids are highly conserved at positions 1, 5, 8, and 15 (corresponding to Val 829 , Ile 833 , Trp 836 , and Leu 843 of OsCBT). The Trp residue (Trp 836 in OsCBT), which is known to play a critical role in the CaM binding of many CaMBPs (26,42), is invariable among all of the OsCBT family members. In contrast to the highly conserved Ca 2ϩ -dependent CaM-binding motif, the existence of the CaMBD I (IQ motif) is variable among the OsCBT homologues. The IQ motif was only found in three EST clones from monocots (HvBJ449986, HvBJ451016, and TaBE497619) and two Arabidopsis homologues (At3g16940 and At4g16150). Phylogenetic analysis reveals that the CaM-binding sequences of the OsCBT family could be divided into two groups, and clones having putative CaMBD I are clustered in the same group (Fig.  4B). These results suggest that although Ca 2ϩ -dependent CaM binding is a common feature of all OsCBT homologues, Ca 2ϩ -dependent CaM dissociation may be more specialized.
In order to verify the critical OsCBT residues interacting with CaM, we introduced amino acid substitutions into several of the highly conserved CaMBD residues of OsCBT homologues (Fig. 4C), and then we analyzed the variants for CaM binding in a CaM overlay assay (Fig. 4D). Substitution of the hydrophobic Ile residue in the IQ motif with Glu is sufficient to interfere with CaM binding (43). Similarly, substitution of Ile 763 of the OsCBT CaMBD I by Glu (denoted I/E) completely abolished CaM binding (Fig. 4, C and D). Replacement of both the hydrophobic residues Val 829 and Trp 836 , at the first and the eighth positions of the CaMBD II, by the basic Arg residue (denoted VW/RR) was also sufficient to abolish Ca 2ϩ -dependent CaM binding of OsCBT (Fig. 4, C  and D). These results suggest that OsCBT interacts with CaM in a sequence-specific manner, and Ile 763 in CaMBD I and Val 829 and Trp 836 in CaMBD II are crucial for Ca 2ϩ -dependent CaM dissociation and Ca 2ϩ -dependent CaM binding, respectively.
OsCBT Contains Two Functional Nuclear Localization Signals-A search of GenBank TM revealed that OsCBT contains two different types of putative NLSs, a bipartite signal at the N terminus ( 73 RKVVRNFRK-DGHNWKKKK 90 ), and an SV40-type NLS at the C terminus ( 837 RKKR 840 ), which we refer to as NLS-N and NLS-C, respectively. To determine whether these two putative NLSs are able to direct OsCBT into the nucleus, we created expression plasmids for green fluorescent protein (smGFP) fused to the various mutants of OsCBT (Fig. 5A), and the subcellular localization of these GFP fusion proteins was tested in Arabidopsis protoplasts. Red fluorescent protein fused to the NLS peptide from the SV40 large T antigen was used as a positive control (cNLS::RFP) (30). First of all, we tested nuclear localization of GFP fusion proteins of the N-or C-terminal half of OsCBT (named N-term::smGFP and C-term::smGFP, respectively). As shown in Fig. 5B, N-term::smGFP was strongly expressed in the nucleus and weakly expressed in the cytosol, whereas C-term::smGFP proteins was predominantly localized in the nucleus. However, the smGFP alone and deletion mutant of both N-and C-terminal predicted NLSs (named del NLSs::smGFP) of OsCBT were equally distributed both in the nucleus and cytosol. In order to map the NLS motif(s) of OsCBT more precisely, we analyzed the nuclear targeting of GFP fusions with each predicted short NLS peptide sequences from the N and C terminus (named NLS-N::smGFP and NLS-C::smGFP, respectively) of OsCBT (Fig. 5A). As shown in Fig. 5B, both NLS-N::smGFP and NLS-C::smGFP proteins are exclusively localized in the nucleus, indicating that both predicted NLS motifs from the GenBank TM search are functional NLS sequences. Although both C-term::smGFP and NLS-C::smGFP constructs show the identical subcellular localization patterns, those of N-term::smGFP and NLS-N::smGFP constructs are slightly different, suggesting that the other region in the N-terminal part of OsCBT may also be involved in the regulation of nuclear targeting of OsCBT.
DNA Binding Ability of OsCBT-To test the DNA binding ability of OsCBT, we needed a reasonable amount of soluble protein, so we examined the expression level and solubility of each C-terminal deletion mutant (see "Materials and Methods"). Attempts to express full-length OsCBT as a glutathione S-transferase (GST)-tagged protein in E. coli were unsuccessful, because the level of expression was too low to be useful. Only the CG-1 homology region (amino acid 1-253, named OsCBT DB) of OsCBT was highly expressed in E. coli as a soluble form. Therefore, we purified the GST-fused recombinant OsCBT DB protein using a glutathione-Sepharose column and then the N-terminally fused GST was removed by thrombin digestion (Fig. 6A). We tested the DNA binding ability of OsCBT by using the purified OsCBT DB protein.
When OsCBT was incubated with an oligonucleotide probe that binds parsley CG-1 (21), we did not detect binding to OsCBT DB. However, OsCBT bound an oligonucleotide probe containing CGCG motifs, selected from the rice PAL promoter sequences (Fig. 6B) (44). This result suggests that OsCBT DB binds DNA, but its DNA binding specificity is different from that of parsley CG-1.
Identification of Consensus DNA-Binding Sequences of OsCBT-In order to determine the preferred binding sequences for OsCBT, we performed an RBSS experiment using a pool of oligonucleotides that contained a random sequence of 15 oligonucleotides flanked by primerbinding sites and restriction enzyme digestion sites for cloning (28,45). OsCBT DB was incubated with the random oligonucleotide pool in the presence of an excess amount of nonspecific competitors, poly[dI⅐dC]. The oligonucleotides recovered from each round of binding reaction were labeled and tested for binding to OsCBT DB. Electrophoretic mobility shift assays revealed clear enrichment in bound oligonucleotides after subsequent rounds of selection of preferred DNA-binding sequences from the randomized pool (data not shown). To identify the sequence motif selected by the OsCBT DB protein, oligonucleotides bound in the final round of selection at 150 mM KCl were amplified by PCR and subcloned into pBluescript II KS(Ϫ), and 79 clones were ran- The upper strand nucleotide sequence is the parsley CG-1-binding sequence, and the lower strand nucleotide sequence is the rice PAL promoter sequences that contain a CGCG motif. Both oligonucleotides were labeled with 32 P and an equal amount of probe (40,000 cpm) was incubated with 0.5 g of purified OsCBT DB protein and analyzed by gel retardation assay on a 5% polyacrylamide gel (Ϫ, labeled probe only; ؉, add 0.5 g of purified OsCBT DB protein). The poly[dI⅐dC] was added as a nonspecific competitor in increasing amounts (0 g (lanes 1 and 2) domly chosen and sequenced. The results of sequencing are shown in Fig. 7. The sequences are compiled and are divided into two groups (Fig.  7A). In the first group (group I), each of the 55 different clones contained the CGCGT sequence. In the second group (group II), 28 different clones had a CGTGT sequence. These two groups were divided into several subgroups depending on their flanking sequences. 44 clones contained a consensus flanking the CGCGT motif (classified as Group Ia). Group 1b contained clones substituted by A or T at positions 14 or 15. Group 1c contained clones modified at position 7. Group 1d contained clones that contained only the CGCG motif. Group IIb contained clones that had the CGTGT sequence but not the CG sequence at positions 14 and 15 (Fig. 7A). After tabulation, a consensus sequence was determined to be TWCG(C/T)GTKKKKTKCG from results of sequencing, where W and K represent A or C and T or G, respectively (Fig. 7B).
To verify the DNA binding specificity of OsCBT DB, we carried out further DNA binding analysis using representative sequences selected from each subgroup. Clone 19 and clone 21 had sequences containing the CGCGT and CGTGT motifs with conserved flanking sequences, respectively. Clone 10 and 83 were substituted by T at positions 14 and 15. Clone 77 was substituted by C at position 7. Clone 79 had only a CGCG motif. Clone 34 had a CGTGT motif but not CG at positions 14 and 15 (Fig. 8A). The binding specificity of each clone for OsCBT DB was analyzed with electrophoretic mobility shift assays using 40,000 cpm of 32 P-end-labeled probes at 50 mM KCl and 150 mM KCl conditions (Fig. 8B). The binding affinity of each clone was determined by densitometry of the shifted bands (Fig. 8C). The binding affinity of clone 19 for OsCBT DB in 50 mM KCl was normalized to 100%, and the binding affinity of clones 10, 83, and 21 was almost 11-fold less than that of clone 19; clone 77 was 15-fold less, and clones 97 and 34 were 30-fold less. These other clones showed a considerable decrease in binding affinity to OsCBT DB when assays were conducted in the presence of 150 mM KCl. The relative binding affinity of clone 19 was not significantly different between 50 and 150 mM KCl. Because clone 19 appeared to show the highest binding affinity for OsCBT DB (Fig. 8C), the sequence of clone 19 was used as the binding sequence of OsCBT in subsequent experiments. Those results indicated that OsCBT favored a CGCGT motif rather than a CGTGT motif and that the flanking sequences of the CG(C/T)GT motif were also important in specific binding by OsCBT.
CaM Functions as a Negative Regulator of OsCBT Transcriptional Activation-To test whether the consensus binding sequence of OsCBT obtained from the in vitro RBSS also functions as an OsCBT binding sequence in vivo, we generated reporter constructs by fusion of the ␤-glucuronidase (GUS) gene and a tandem tetramer of the OsCBT binding DNA sequences (clone 19) cloned downstream of the minimal promoter pDel 151-8 (32). Also, we constructed the effector clones consisting of the CaMV 35S promoter fused to full-length OsCBT or OsCaM genes (Fig. 9A). The effects of OsCBT and/or OsCaM on the expression of the reporter gene was tested in Arabidopsis protoplasts. As shown in Fig. 9B, the expression of the reporter gene alone increased about 23-fold compared with the vector control indicating that the endogenous Arabidopsis OsCBT homologue may recognize this reporter construct. Furthermore, co-expression of OsCBT significantly increased GUS reporter expression about 261-fold compared with the vector control and about 11-fold compared with cells expressing the GUS reporter gene only (Fig. 9B). These results suggest that the binding sequence derived from the RBSS effectively functions as an in vivo binding sequence for OsCBT and that OsCBT acts as a transcription activator in vivo. In order to test the effect of OsCaM on OsCBT-mediated transcriptional activation, we co-expressed OsCaM with OsCBT and measured GUS reporter gene expression. Most interestingly, the increased expression level of the GUS reporter gene driven by OsCBT was inhibited by about half when OsCaM was co-expressed (Fig. 9B). Taken together, those results suggest that OsCaM interacts with OsCBT and inhibits transcriptional activation by OsCBT in vivo.

DISCUSSION
In this study, we identified a rice CaM-binding transcriptional regulator encoded by OsCBT, which has several conserved motifs including a CG-1 homology DNA-binding domain, a TIG domain, three ANK repeats, and two distinct types of CaMBDs, IQ motifs, and a Ca 2ϩ -dependent CaMBD. OsCBT homologues were found in various dicot species including Arabidopsis (19,20), tobacco (22), rapeseed (19), and tomato (23), as well as in human and Drosophila (46). Several EST clones showing homology to OsCBT were isolated from monocots such as wheat, barley, sorghum, and rice (19). Moreover, these genes are rapidly induced by various environmental signals such as temperature, light, hormones, and pathogen attack (20). Taken together, these results suggest the conserved function of OsCBT homologues in Ca 2ϩ -mediated signal transduction in response to environmental cues among multicellular eukaryotes.
The GenBank TM data base search using OsCBT protein sequences indicates that OsCBT contains four putative Ca 2ϩ -dependent CaM dissociation domains (IQ motifs) and a Ca 2ϩ -dependent CaM-binding domains in its C-terminal region. As shown in Fig. 4A, this cluster of putative CaM-binding sequences is well conserved among OsCBT homologues. The Ca 2ϩ -dependent CaMBD is invariable in all family members. The positions of hydrophobic amino acids in this region are well conserved at positions 1, 5, 8, and 15 in the domain, corresponding to Val 829 , Ile 833 , Trp 836 , and Leu 843 of OsCBT, and furthermore, the Trp 836 residue, which is known to play critical role in interactions with FIGURE 8. Analysis of DNA binding specificity of OsCBT DB using oligonucleotides selected by RBSS. A, sequence of characterized clones from each group used in this experiment. Every residue that does not match the consensus site is underlined. B, DNA binding specificity of OsCBT DB. Each clone was digested with BamHI and HindIII and labeled by filling in the 5Ј overhangs with a Klenow fragment in the presence of [ 32 P]dATP. 32-bp 32 P-labeled DNA fragments were excised from the polyacrylamide gel, and their radioactivities were measured with a liquid scintillation counter. We used 40,000 cpm of the each labeled probe for gel shifting assay. An electrophoretic mobility shift assay was performed with OsCBT DB in the presence of 50 mM KCl or 150 mM KCl conditions. C, relative binding affinity of OsCBT DB for each clone. The binding affinity of OsCBT DB for clone 19 in 50 mM KCl was normalized to 100%, and relative binding affinities of the other clones both in 50 mM KCl and 150 mM KCl were estimated.
CaM, is strictly conserved in all family members. Previous reports showed that CaM was able to physically interact with this sequence in a Ca 2ϩ -dependent manner (26,42). We also verified the binding of the Ca 2ϩ -CaM complex to this region of OsCBT using several independent assays, including a gel overlay assay, a peptide/CaM gel shift assay, and site-directed mutagenesis. The conservation of Ca 2ϩ -dependent CaMbinding sequences among the OsCBT family may suggest that binding of the Ca 2ϩ -CaM complex to this region is the common regulatory mechanism of all OsCBT homologues.
In addition to the Ca 2ϩ -dependent CaM-binding motif, we also showed that OsCBT has a functional IQ motif, among the four putative IQ motifs in the C-terminal region. We verified the physical interaction between CaM and this IQ motif in the absence of Ca 2ϩ using in vitro CaM overlay assays, and we mapped its position between Ala 760 and Met 778 (Fig. 4C). In contrast to the high degree of conservation of the Ca 2ϩ -dependent CaM-binding domain, the feature of Ca 2ϩ -dependent CaM dissociation through the IQ motif may be variable among OsCBT family members. Even though it was reported previously that other OsCBT family proteins also contain various numbers of IQ motifs, these reports did not show a direct interaction with CaM (19,20). However, we cannot rule out the possibilities that some of these IQ motifs are only functional under specific physiological conditions or that some other CaM-like proteins containing EF-hands are capable of binding to these IQ motifs. Most interestingly, amino acid sequence alignments of the CaM-binding regions of OsCBT homologues showed that some EST clones from barley (HvBJ449986 and HvBJ451016) and wheat (TaBE497619) have amino acid sequences identical to the Ca 2ϩ -dependent CaM dissociation sequences of OsCBT (IQHAFRNYNRKK, designated CaMBD I), and two Arabidopsis homologues At3g16940 (AtSR3/AtCAMTA6) and At4g16150 (AtSR6/AtCAMTA5) have similar sequences with less identity (IQNAFRKYDTRR and IQHAFRNFEVRR, respectively, Fig. 4A). Although CaM binding to these sequences has yet to be verified, these data suggest that some OsCBT family proteins can be regulated by CaM in a more complex manner. For example, previous reports have shown that some animal CaMBPs, such as the ryanodine receptor (RyR1) and erythrocyte protein 4.1 (4.1R), bind CaM both in the absence and presence of Ca 2ϩ , and the activity of these proteins is differentially regulated by both apocalmodulin and Ca 2ϩ /CaM binding (47,48).
Even though CaM binding of OsCBT homologues and their gene expression patterns under various environmental conditions has been extensively characterized by several independent groups (19,20), we do not yet have any clear evidence for either their DNA binding specificity or the role(s) of CaM in the function of OsCBT homologues. In this report, we determine the preferred DNA-binding sequences of OsCBT using an RBSS method, and furthermore, from the transient assay using the Arabidopsis protoplast, we provide the first evidence that OsCBT encodes a functional transcription activator that is negatively regulated by CaM binding in vivo (Fig. 9). Moreover, we showed that OsCBT contains two functional NLS motifs (Fig. 5), a bipartite NLS (designated NLS-N) and an SV40 type NLS (designated NLS-C) in the CG-1 homology DNA-binding domain and the CaMBD II region, respectively, using GFP fusion proteins (Figs. 1 and 4A). Although the first bipartite NLS is found in all OsCBT homologues, the second SV40 type NLS is only found in some of the monocot OsCBT homologues. These observations suggest that even though the structural features are conserved among OsCBT family members, gene-specific regulation by differential CaM binding activity may occur.
Most interestingly, NLS-C overlaps the Ca 2ϩ -dependent CaMBD (CaMBD II) of OsCBT. This fact may suggest that Ca 2ϩ -dependent CaM binding can have an effect on nuclear localization of OsCBT. This can be one possible explanation for the down-regulation of the transcriptional activator activity of OsCBT by adding CaM in our transient assay (Fig. 9). However, we cannot exclude the role of CaMBD I (Ca 2ϩdependent CaM dissociation domain) from the possible regulatory mechanisms of OsCBT by CaM binding. In our transient assay system using the Arabidopsis protoplast, it is difficult to test the effect of CaM binding on the transcriptional activator activity of OsCBT in the lower cytosolic Ca 2ϩ condition because of high Ca 2ϩ concentration of transient assay solution (125 mM CaCl 2 in W5 solution) used for protoplast suspension. Even though we are not able to conclude the precise roles of two different CaMBDs, a Ca 2ϩ -dependent CaM dissociation domain (CaMBD I) and a Ca 2ϩ -dependent CaM-binding domain (CaMBD II), in the regulation of OsCBT activity in this report, it is suggested that the transcriptional activator activity of OsCBT is negatively regulated by CaM interaction with either CaMBD I or CaMBD II.
In summary, we report a rice cDNA clone OsCBT encoding a CaMbinding transcriptional activator. It contains several recognized structural features, including DNA-binding domains, ANK repeats, and OsCBT effectors is observed in Arabidopsis protoplasts. Plasmid quantities were equalized with a control plasmid, pHBT95. In addition, firefly luciferase under the control of the 35S full promoter was included in each transfection as an internal control for transfection efficiency. In this experiment, the reporter activities were normalized as GUS/luciferase activity in each transfected sample, and the relative activities were calculated relative to that of the reporter construct alone. The bars indicate the mean Ϯ S.D. for each set of three independent experiments.
CaMBDs. From the analyses of OsCBT, we show that OsCBT binds to DNA in a sequence-specific manner, and its NLS motifs are capable of driving OsCBT into the nucleus. We also provide evidence that OsCBT binds to CaM both in the absence and presence of Ca 2ϩ , and CaM binding inhibits transcriptional activation by OsCBT in vivo. We did not investigate the role of the ANK repeats in the function of OsCBT. In order to have more insight into the biological function of OsCBT, we intend to characterize both gain-of-function and loss-of-function mutants in future studies.