Structures of Anabaena Calcium-binding Protein CcbP

Ca2+-binding proteins play pivotal roles in both eukaryotic and prokaryotic cells. CcbP from cyanobacterium Anabaena sp. strain PCC 7120 is a major Ca2+-binding protein involved in heterocyst differentiation, a process that forms specialized nitrogen-fixing cells. The three-dimensional structures of both Ca2+-free and Ca2+-bound forms of CcbP are essential for elucidating the Ca2+-signaling mechanism. However, CcbP shares low sequence identity with proteins of known structures, and its Ca2+-binding sites remain unknown. Here, we report the solution structures of CcbP in both Ca2+-free and Ca2+-bound forms determined by nuclear magnetic resonance spectroscopy. CcbP adopts an overall new fold and contains two Ca2+-binding sites with distinct Ca2+-binding abilities. Mutation of Asp38 at the stronger Ca2+-binding site of CcbP abolished its ability to regulate heterocyst formation in vivo. Surprisingly, the β-barrel subdomain of CcbP, which does not participate in Ca2+-binding, topologically resembles the Src homology 3 (SH3) domain and might act as a protein-protein interaction module. Our results provide the structural basis of the unique Ca2+ signaling mechanism during heterocyst differentiation.

Whereas the significance of Ca 2ϩ ions in eukaryotic cells has been well recognized for a long time (1), its importance in prokaryotic cells has only gained increasing interests recently (2)(3)(4). There is growing evidence that the intracellular Ca 2ϩ concentration is tightly regulated in prokaryotes, and Ca 2ϩ signaling is involved in cell structure maintenance, gene expression, cell cycle and cell differentiation processes, including the regulation of heterocyst formation in cyanobacteria (2)(3)(4).
Cyanobacteria are a group of ancient prokaryotes that appeared on earth at least 2ϳ3.5 billion years ago. Some cyanobacteria can simultaneously carry out oxygenic photosynthesis and nitrogen fixation, which are two biochemically incompatible processes. When combined nitrogen is scarce, some photosynthetic vegetative cells differentiate into specialized nitrogen-fixing cells called heterocysts (5)(6)(7)(8)(9). The signaling network during heterocyst differentiation is highly complex and recalls that of eukaryotic cells. One of the essential triggering signals for heterocyst formation is the increase of intracellular free Ca 2ϩ concentration, and it could represent an earliest example of calcium required cellular differentiation in evolution (9 -11).
Protein CcbP (cyanobacterial calcium binding protein) from cyanobacterium Anabaena sp. strain PCC 7120 (Anabaena sp.) was identified as a major Ca 2ϩ -binding protein involved in Ca 2ϩ sequestration and the regulation of heterocyst differentiation (10,11). At the early stages of heterocyst differentiation, CcbP is degraded by a serine-type protease HetR, leading to a Ca 2ϩ release and subsequent differentiation processes (11). Nevertheless, CcbP shares low sequence identity with proteins of known structure, and its Ca 2ϩ -binding sites remain unknown. Although CcbP shows certain biochemical and biophysical similarity to the sarcoplasmic Ca 2ϩ -binding protein calsequestrin in vertebrates, its Ca 2ϩ -binding capacity (ϳ2 Ca 2ϩ per molecule) differs from calsequestrin (ϳ40 -50 Ca 2ϩ per molecule) (10 -12). These results strongly suggest that CcbP may represent a novel class of Ca 2ϩ -binding proteins. Therefore, the structural information of CcbP is essential for elucidating the molecular mechanism of its role in Ca 2ϩ signaling during heterocyst differentiation.
Here, we report the solution structures of CcbP in both Ca 2ϩ -free and Ca 2ϩ -bound forms determined by nuclear magnetic resonance (NMR) 3 spectroscopy. The structures of CcbP in both forms reveal an overall new fold with an ␣-subdomain and a ␤-barrel subdomain. Ca 2ϩ titration experiments by NMR and mutagenesis analysis identified two Ca 2ϩ -binding sites. The stronger Ca 2ϩ -binding site I locates at an ␣-turn-␤ region, whereas the weaker Ca 2ϩ -binding site II resembles a single EFhand motif with defects. Furthermore, the ␤-barrel subdomain of CcbP unexpectedly reveals an SH3-like topology that might act as a protein-protein interaction module during the degra- dation of CcbP by HetR. Our study provides the structural basis for understanding the mechanism of Ca 2ϩ signaling during heterocyst differentiation and further extends our knowledge of Ca 2ϩ -binding proteins.

EXPERIMENTAL PROCEDURES
Sample preparations-The ccbP gene was cloned into a pET15b vector (Novagen) with an N-terminal cleavable His-tag and expressed in an Escherichia coli BL21(DE3) strain. The culture was grown in LB medium, centrifuged, and resuspended in M9 minimal medium with antibiotics and 15 NH 4 Cl in the presence or absence of 13 C 6 -glucose for preparations of 13 C/ 15 Nlabeled or 15 N-labeled samples, respectively (13). The CcbP protein was purified by nickel-nitrilotriacetic acid column (Qiagen). The sample was digested using thrombin to remove the N-terminal His-tag and further purified by passing through the nickel-nitrilotriacetic acid column and subsequently the gel filtration column (Superdex-75) using an Ä KTA FPLC system (Amersham Biosciences). The purity was determined to be Ͼ95% as judged by SDS-PAGE. NMR samples were prepared with 1 mM CcbP dissolved in 90% H 2 O/10% D 2 O buffer containing 20 mM Tris-HCl (pH 7.4) and 220 mM NaCl. The sample for Ca 2ϩ -free CcbP was pretreated with excess EGTA and subsequently buffer-exchanged to remove EGTA. Excess Ca 2ϩ ions (40 mM CaCl 2 ) were added in the Ca 2ϩ -bound form. In addition, 2,2-dimethyl-2-silapentanesulfonic acid was added as the internal chemical shift reference. CcbP mutants E17A, D21A, E23A, D24A, E26A, E27A, D37A, D38A, T39A, and E41A were expressed in E. coli and purified following the same method as wild type CcbP.
NMR Spectroscopy-All NMR experiments for structural determination were performed on Bruker Avance 500-MHz (equipped with a cryoprobe) and 800-MHz spectrometers equipped with a triple-resonance probe with pulsed field gradients at 30°C. The spectra were processed with the software package NMRPipe (14) and analyzed using the program NMR-View (15). The resonance assignments of backbone and side chain atoms were obtained following the common procedures (16). The three-dimensional 15 N-and 13 C-edited NOESY-HSQC (mixing time of 100 ms) spectra were recorded to confirm the assignments and generate distance restraints for structure calculations. Hydrogen-deuterium exchange experiments were performed to obtain hydrogen bond information. The 1 H-15 N residual dipolar coupling (RDC) constants of CcbP were measured. The measurements were performed by dissolving the CcbP protein in a dilute liquid crystal buffer containing a mixture of alkyl-poly (ethylene glycol) C12E5 and n-hexanol (17). The C12E5/water ratio was 5.5% (w/w), and the molar ratio of C12E5 to n-hexanol was 0.92. The RDC values were extracted from the difference in 1 H-15 N splitting measured by 1 H-15 N IPAP-HSQC spectra between the weakly aligned and the isotropic samples (18).
Structure Calculations-The structures of CcbP in both forms were calculated using interproton NOE-derived distance restraints in combination with dihedral angles and hydrogen bonds information. The program TALOS (19) was used to predict dihedral angles and restraints. Hydrogen bond restraints were determined based on hydrogen-deuterium exchange experiments in conjunction with NOEs and secondary structural information. The initial structures were generated using the CANDID module of the CYANA program (20). The 20 conformers with the lowest target function were selected as the models for SANE (21) to extend the NOE assignments. Two hundred structures were iteratively calculated using CYANA (22), and the 100 structures with the lowest target function were selected for further refinement in the AMBER force field using the parm99 parameters (23). The RDC restraints were added during the AMBER refinement procedure. For the Ca 2ϩ -bound form, two Ca 2ϩ ions and distance restraints between Ca 2ϩ and CcbP also were added during the AMBER calculation steps. For each binding site, one Ca 2ϩ ion was restrained by adding Ca 2ϩ -O restraints of 1.8 -2.8 Å based on the experimental results from chemical shift perturbations and mutagenesis. Initially, ambiguous Ca 2ϩ -O restraints for the two oxygen atoms of side chain carboxyl groups were used. Based on the calculated structures with Ca 2ϩ ions, the oxygen atoms with Ca 2ϩ -O distances Ͻ3 Å were identified to resolve the ambiguous Ca 2ϩ -O restraints and to remove those that were violated. Subsequent calculations were performed using unambiguous Ca 2ϩ -O restraints. Finally, 20 of 100 structures with the lowest AMBER energy were selected as the representative structures of CcbP in the Ca 2ϩ -free and Ca 2ϩ -bound forms, respectively. Programs PROCHECK-NMR (24) and MOLMOL (25) were used to analyze the quality of the structures.
Ca 2ϩ Titration by NMR-For the Ca 2ϩ titrations by NMR, 15 N-labeled CcbP protein (final concentration of ϳ0.5 or 0.05 mM) was dissolved in 20 mM Tris-HCl buffer (pH 7.4) in the absence of NaCl. CaCl 2 was dissolved in the same buffer and gradually added to the protein sample. A series of two-dimensional 15 N-edited heteronuclear single-quantum coherence (HSQC) spectra were recorded, and the chemical shift changes of backbone 15 N atoms of all residues were analyzed to identify the Ca 2ϩ -binding sites. Ca 2ϩ titration by NMR was also performed in the presence of 220 mM NaCl, and the results were compared with those obtained in the absence of NaCl. In addition, Ca 2ϩ titrations by NMR in the absence of NaCl were performed similarly for CcbP mutant proteins.
Backbone { 1 H}- 15 N Heteronuclear NOE Measurements-The backbone steady-state heteronuclear { 1 H}-15 N NOE values of CcbP in the Ca 2ϩ -free and Ca 2ϩ -bound forms were measured on a Bruker Avance 600-MHz NMR spectrometer at 30°C (26). The experiments were performed in the presence and absence of a 3-s proton presaturation period prior to the 15 N excitation pulse.
Ca 2ϩ Titration into CcbP by ITC-Binding of Ca 2ϩ to CcbP was measured by isothermal titration calorimetric (ITC) using a MicroCal VP-ITC MicroCalorimeter (Northampton, MA). The protein samples used in the titration was extensively dialyzed against a buffer containing 20 mM Tris-HCl (pH 7.4), which is the same as used in NMR titration experiments. Stock solutions of CaCl 2 (2.9 and 5.8 mM), used as the titrant, were prepared the same buffer. Typical ITC experiments were performed at 25°C according to the manufacturer's instructions. A total of 272 l concentrated CaCl 2 (2.9 mM or 5.8 mM) was added into the protein solution (1.43 ml, 0.10 mM, or 0.22 mM) in 34 aliquots (8-l each). The additions were 3-min apart to allow heat accompanying each increment to return to baseline prior to the next addition. The reference experiments by titrating CaCl 2 ligand into the buffer were subtracted before data analysis. All data were analyzed by fitting with different binding models using the program Origin (version 7.0; MicroCal, Northampton, MA), and best fits were obtained using two-site binding model.

RESULTS
NMR Structure of Ca 2ϩ -free CcbP-The solution structure of CcbP in the absence of Ca 2ϩ was determined based on a total of 4,370 restraints derived from multidimensional NMR spectroscopy, including proton-proton distance restraints generated from NOE, hydrogen-bond restraints based on the hydrogendeuterium exchange experiments, dihedral angle restraints based on chemical shifts, and 1 H-15 N RDC restraints measured by weakly aligning the protein sample in a dilute liquid crystal buffer ( Table 1). The ensemble of the 20 representative structures and a ribbon diagram are depicted in Fig. 1, A and C. The stereo images of the structures are shown in supplemental Fig. 1.
A search using DALI (27) or CATH (28) did not find structural homologues with significant overall similarity, suggesting a new protein fold. The best hit by DALI is a partial region of a human cell cycle protein splindin-1 (Protein Data Bank code  2NS2, chain A) with a Z score of 4.8, and an root mean square deviation of 4.1 Å over 73 aligned C ␣ atoms. However, these aligned residues are strictly limited to the ␤-barrel subdomain of CcbP. Interestingly, a closer inspection reveals that this ␤-barrel subdomain is topologically reminiscent of the eukaryotic SH3 domain ( Fig. 2A). A structural alignment by DaliLite (29) between the SH3-like subdomain of CcbP and the structure of c-Src SH3 domain (Protein Data Bank code 1QWE) showed that the overall structures are similar, with a 2.0 Å root mean square deviation for 44 aligned backbone C ␣ atoms. The loops consisting of residues 51-58, 69 -78, and 84 -91 in CcbP correspond to the RT loop, the n-Src loop and the distal loop in the c-Src structure, respectively (Fig. 2B). The structure-based sequence alignment (Fig. 2C) shows no identity (0%) between the two protein sequences. However, several residues located on the ␤-strands are relatively similar in amino acid types and biochemical properties. Most of these residues have hydrophobic side chains and form the hydrophobic core of the ␤-barrel, indicating their importance in maintaining the overall structure. NMR Structure of Ca 2ϩ -bound CcbP-Similar to the Ca 2ϩfree CcbP, the structure of Ca 2ϩ -bound CcbP was determined with a total of 4,694 restraints, including 12 Ca 2ϩ -O restraints ( Table 1). The addition of excess CaCl 2 appears to slightly stabilize the whole structure of CcbP, as evidenced by the increased number of interproton NOE restraints. This also was supported by the fact that under a Ca 2ϩ -free condition, residues Ser 53 -Ser 56 were missing in the two-dimensional 15 Nedited HSQC spectrum, which appeared in the presence of excess CaCl 2 . Nevertheless, comparison of the overall architecture of CcbP between the Ca 2ϩ -bound and Ca 2ϩ -free forms reveals considerable similarity (Fig. 1, B and D). All secondary structural elements as well as the relative orientations between the two subdomains are retained essentially upon Ca 2ϩ bind-ing. The backbone root mean square deviation between the mean structures of the two forms is 1.0 Å for all residues and is only 0.5 Å for residues in regular secondary structures. The results indicate that Ca 2ϩ binding does not induce notable conformational changes in CcbP. In addition, we compared the motional flexibility on picosecond-to-nanosecond time scales of both forms of CcbP by measuring the backbone { 1 H}-15 N heteronuclear NOE values (for more details see supplemental "Results and Discussion"). The results demonstrated that Ca 2ϩ binding does not induce significant changes in the motional flexibility of CcbP as well (supplemental Fig. 2).
Because CcbP exists as oligomer under low ionic strength, which forbids its structure determination by NMR, the structures of CcbP in both forms were determined in the presence of 220 mM NaCl. Taking into consideration that the intracellular environments generally contain high K ϩ concentration (ϳ 100 -200 mM) (30), we compared the two-dimensional 15 Nedited HSQC spectra of both forms of CcbP in excess K ϩ with those in excess Na ϩ . The HSQC spectra with Na ϩ or K ϩ are similar, demonstrating essentially identical conformations of CcbP (supplemental Fig. 3). Therefore, under conditions near physiological environment, the binding of Ca 2ϩ ions does not induce considerable conformational changes of CcbP.
Ca 2ϩ Titration by NMR-Ca 2ϩ titration experiments by NMR were performed to identify the Ca 2ϩ -binding sites of CcbP, which were monitored by HSQC spectroscopy (additional discussions are available in supplemental "Results and Discussion" and Figs. [3][4][5][6][7][8][9][10]. During the gradual increase of Ca 2ϩ concentrations, two regions in CcbP showed responses (supplemental Fig. 10). When the molar ratio of Ca 2ϩ :CcbP changed from 0:1 to 1:1, the backbone nitrogen chemical shifts of several residues located at the C terminus of helix ␣2 were largely perturbed, suggesting that these residues constitutes the stronger Ca 2ϩ -binding site (site I). When the molar ratio Ca 2ϩ : CcbP continued to increase, another region (residues Ile 19 -Glu 27 ) located at the loop linking helix ␣1 and ␣2 showed moderate sensitivity to Ca 2ϩ ions, suggesting a weaker Ca 2ϩbinding site (site II).
Characterization of Ca 2ϩ -binding Site I-Site I locates at the segment Asp 37 -Glu 41 , which is an ␣-turn-␤ motif connecting helix ␣2 and the first ␤-strand of the ␤-barrel subdomain (Fig. 3,  A and B). Although CcbP has a large negatively charged surface, this Ca 2ϩ -binding site is highly specific and shows the highest sensitivity to the presence of Ca 2ϩ ions (Fig. 3D, supplemental Figs. 7-10). In NMR spectroscopy, metal coordination to a backbone carbonyl can cause deshielding of the backbone nitrogen atom of the succeeding residue (31). The backbone 15 N atoms of residues Asp 38 , Thr 39 , and Glu 41 at this region showed significant downfield chemical shift changes, suggest-ing the main chain carbonyls of the preceding residues Asp 37 , Asp 38 , and Leu 40 might be potential Ca 2ϩ ligands.
To further characterize this binding site, plasmids carrying genes encoding CcbP mutants D37A, D38A, T39A, and E41A were constructed. The Ca 2ϩ titrations by NMR were performed similarly using these mutant proteins, and the Ca 2ϩ -binding abilities of this site were examined by comparing the backbone nitrogen chemical shift changes with the wild type CcbP. Results showed that the Ca 2ϩ sensitivity of this site was impaired largely by D37A, D38A, and E41A mutants, but not T39A mutant (data not shown). Thus, it is highly possible that the side chain oxygen atoms of Asp 37 , Asp 38 , and Glu 41 also are involved in Ca 2ϩ coordination. The local structure of site I was further calculated based on the above results using the AMBER force field (see "Experimental Procedures"). The bound Ca 2ϩ ion is in proximity (Ca 2ϩ -O distances Ͻ 3 Å) with six oxygen atoms, three from the carboxyl group of Asp 37 , Asp 38 , and Glu 41 and three backbone carbonyls of Asp 37 , Asp 38 , and Leu 40 (Fig. 3B).
Characterization of Ca 2ϩ -binding Site II-Site II locates at the loop linking helix ␣1 and ␣2 (Fig. 3, C and E, and supplemental Fig. 10B). Similarly, we investigated the possible Ca 2ϩ ligands of this site. The backbone 15 N chemical shifts of residues Val 20 , Asp 21 , and Ala 22 were most perturbed during Ca 2ϩ titration experiments by NMR, although the chemical shift changes were small ( Fig. 3E and supplemental Figs. 7F, 9F, and 10B). CcbP mutants E17A, D21A, E23A, D24A, E26A, and E27A were purified, and Ca 2ϩ titrations by NMR were performed. Results showed that the E17A and E26A mutations did not affect Ca 2ϩ binding, D21A mutation only had minor effects, whereas E23A, D24A, and E27A mutations significantly decreased Ca 2ϩ binding (data not shown). The local structure of site II (Fig. 3C) shows that the second Ca 2ϩ ion is surrounded closely by the side chains of Glu 23 , Asp 24 , Glu 27 , and the backbone carbonyl of Asp 21 .
Mutation at Ca 2ϩ -binding Site I Abolishes CcbP Function in Vivo-Because site I of CcbP binds Ca 2ϩ significantly stronger than site II, it appears that site I may contribute prominently to Ca 2ϩ sequestration in vivo. To assess the functional significance of this site in heterocyst differentiation regulation, in vivo functional assays were performed.
We constructed plasmids carrying ccbP genes encoding the wild type protein or mutants at site I under the control of the petE promoter (32). The plasmids were used to transform Anabaena sp., so that the expression of ccbP was inducible with copper. During nitrogen step down, and in the absence of added copper, the Anabaena strain transformed with plasmid carrying wild type ccbP gene was able to form heterocysts with slightly decreased heterocyst frequency (10). When the expression of wild type CcbP was induced with copper under nitrogen limiting conditions, the free Ca 2ϩ was sequestered and the heterocyst formation was suppressed completely as reported previously (Fig. 3F) (10). However, overexpressing a mutant protein CcbP-D38A failed to suppress heterocyst formation (Fig.  3G). The fact that the point mutation D38A abolishes the Ca 2ϩbinding ability of CcbP in vivo demonstrates the biological importance of site I in heterocyst differentiation regulation.

Anabaena CcbP Represents a Novel Family of Ca 2ϩ -binding
Proteins-Ca 2ϩ binds to a variety of proteins, including those that mediate cell adhesion, enzymes that need Ca 2ϩ to be activated, as well as Ca 2ϩ buffers and Ca 2ϩ sensors. The Ca 2ϩbinding protein CcbP from Anabaena displays unique structural characteristics compared with other known families of Ca 2ϩ -binding proteins. The structure of CcbP shows an overall new fold containing a triangular shaped ␣-helical region packed tightly onto a ␤-barrel subdomain. Multiple sequence alignment of CcbP proteins from different cyanobacteria species and from proteobacterium Polaromonas naphthalenivorans CJ2 (Fig. 4A) indicates that the sequence conservation is much higher in the three helices than in the ␤-barrel region. Because the three ␣-helices of CcbP together form the highly acidic surface containing the two Ca 2ϩ -binding sites (Fig. 4, B and C), the amino acid conservation in these regions might be critical for maintaining the structural scaffold in CcbP for Ca 2ϩ binding. Although binding of Ca 2ϩ ions does not result in significant conformational changes of the protein, it alters the local charge at the two sites and also partially neutralizes the nearby area, which might influence the protein-protein interaction pattern of CcbP (supplemental Fig. 11).
Ca 2ϩ -binding Sites of CcbP-The Ca 2ϩ -binding site I of CcbP locates at a short ␣-turn-␤ region, which is distinct from previously reported Ca 2ϩ -binding sites. In vivo functional assays showed that a single amino acid mutation at this site abolished the activity of CcbP in regulating heterocyst formation, demonstrating its biological importance. Furthermore, the amino acids at this site are conserved highly among the CcbP family (Fig. 4A). In particular, residues Tyr 35 , Leu 36 , Leu 40 , Pro 43 , and Phe 44 are strictly conserved. In addition, residues 37, 38, and 41, whose side chains may participate in Ca 2ϩ binding, are restricted to carbonyl-containing Asp, Glu, Asn, or Gln. The in vivo and in vitro experimental results, together with bioinformatics analysis, strongly suggest that this novel Ca 2ϩ -binding site plays essential role in Ca 2ϩ sequestration in Anabaena.
On the other hand, Ca 2ϩ -binding site II of CcbP consists of an ␣-loop-␣ region, which is similar to a single EF-hand motif. Interestingly, the 12-residue sequence of Glu 15 -Glu 26 (ETEI-IVDAEDKE) fits well to an EF-hand Ca 2ϩ -binding loop. According to the classical EF-hand, the residue Glu 26 in position 12 is the most critical and strictly conserved in the EF-hand motif. The acidic residues Glu 15 and Glu 17 at positions 1 and 3 and the water mediating Glu 24 at position 9 are ideal for Ca 2ϩ binding, whereas the amino acids at positions 5 and 6 show deviations from canonical EF-hands (33). However, the NMR titration experiments and mutagenesis results indicate that unlike canonical EF-hands, the acidic side chains of residues Glu 23 , Asp 24 , and Glu 27 in CcbP appear important in Ca 2ϩ binding.
The affinity of Ca 2ϩ binding by CcbP was determined in the previous study with the Ca 2ϩ electrode method, which identified a Ca 2ϩ -binding site with K d of 200 nM and a second Ca 2ϩbinding site with K d of 12.8 M (11). In this study, we also determined K d values of CcbP with NMR titration at both high (ϳ0.5 mM) and low (0.05 mM) protein concentrations. The results show that site I binds Ca 2ϩ with a dissociation constant K d ϳ 40 -100 M, whereas site II shows much weaker Ca 2ϩbinding affinity (K d in the millimolar range). Thus, both methods clearly identified the Ca 2ϩ -binding site with a K d at low micromolar range, demonstrating that CcbP is a Ca 2ϩ -binding FIGURE 4. Sequence similarity of the ␣-helical region of Anabaena CcbP suggests a conserved Ca 2؉ -binding scaffold. A, multiple sequence alignment of CcbP proteins from different cyanobacteria species and proteobacterium P. naphthalenivorans CJ2 by ClustalW (46). The secondary structural elements of CcbP from Anabaena sp. PCC 7120 are shown at the top. Strictly conserved residues are shown in red boxes, and highly conserved residues are shown in white boxes. The figure was generated using ESPript (47). B, mapping of the conserved residues onto the ribbon diagram of CcbP. Strictly conserved residues are depicted in red, and highly conserved residues are shown in pink. The figure was prepared by MOLMOL (25). C, surface electrostatic potential representation of Ca 2ϩ -free CcbP generated by GRASP2 (48). The calculation was performed under the salt concentration of 0.22 M. Red represents negative charges, and blue represents positive charges. The drawing in C is oriented as in B. The Ca 2ϩ -binding sites are indicated in B and C.
protein. However, there is an unexpected result from our present study; the high affinity site (K d ϳ 200 nM) was not observed in NMR titrations. Based on the facts that the Scatchard plot used to determine K d values in Ca 2ϩ electrode method could be affected by certain errors (34), we speculate that the apparent "high affinity binding site" of CcbP observed in the Ca 2ϩ -electrode method could be introduced artificially, and the much weaker Ca 2ϩ -binding site II observed in NMR titration was not observable by the Scatchard plot. This was further supported by the results from ITC experiments, which demonstrated that CcbP contains a Ca 2ϩ -binding site I with a K d value of 39 Ϯ 9 M and a Ca 2ϩ -binding site II with a K d value in the millimolar range (Fig. 5). Therefore, CcbP binds Ca 2ϩ with micromolar and millimolar range affinities. This calcium-binding ability was shown to be functionally important in vivo and underlines the biological role of CcbP in Anabaena Ca 2ϩ signaling during heterocyst differentiation.
Previous studies demonstrated that CcbP is directly associated with calcium sequestration in cyanobacterial cells and acts as a negative regulator in heterocyst differentiation. The detailed mechanism of the function of CcbP in calcium seques-tration in vivo, however, is yet unclear. Our structural study of CcbP demonstrated that it is indeed a calcium binding protein with a novel calcium-binding motif (Ca 2ϩ -binding site I), which has biological significance. A Ca 2ϩ -buffering function was suggested previously based on functional studies (11). Because the Ca 2ϩ binding affinity of site I is in the micro-molar range, whereas the intracellular concentrations of free Ca 2ϩ in cyanobacteria are between 100 nM and 200 nM (11), the calcium ions bound by CcbP alone might not be a major pool of bound calcium under nitrogen-replete conditions. CcbP could be more important in regulation of free calcium concentration during the process of heterocyst formation when calcium concentration increased significantly in heterocysts and proheterocysts (10). The increase of calcium concentration could come from an increased influx of calcium and/or a release of bound calcium ions, which remains to be further investigated.
SH3-like Subdomain in CcbP-Apart from the three acidic ␣-helices that contain the two Ca 2ϩ -binding sites, the structure of CcbP unexpectedly reveals a ␤-barrel subdomain topologically and structurally reminiscent of eukaryotic SH3 domain. The SH3 domain is a small module with 55-70 residues commonly found in eukaryotic signaling pathways, and it mediates transient protein-protein interactions with moderate affinity (35,36). It recognizes specific proline-rich sequences and prefers sequences with a PxxP core motif (where x is any amino acid residue) (35)(36)(37)(38). A close inspection found a short prolinerich sequence YPWIPGRSRIP in HetR, which contains the general consensus sequence ⌽Px⌽Pxϩ (where ⌽ is a hydrophobic residue, x is any amino acid, and ϩ is a basic residue, usually arginine) of the class II motif that interacts with the canonical SH3 domains. Moreover, this sequence in HetR also closely resembles the consensus RPx⌽P⌽RϩSxP motif recognized by the p53bp2 SH3 domain (39). Therefore, a likely scenario is that the proline-rich sequence of HetR recognizes and interacts with the SH3-like domain of CcbP, which facilitates the degradation of CcbP and the release of Ca 2ϩ ions during heterocyst differentiation.
Other prokaryotic domains that are sequentially unrelated to but topologically reminiscent of eukaryotic SH3 domains have also been discovered in recent years (40 -45). Among these, the SH3-like domain in the bacterial histidine kinase CheA mirrors the SH3 domains in mammalian tyrosine kinases and suggests the ubiquitous involvement of this common topology in cell signaling among different life kingdoms (40). Our study reveals the presence of an SH3-like subdomain in Anabaena CcbP, which demonstrates a direct association of the SH3-like domain with a Ca 2ϩ -binding protein. This represents another paradigm for the coupling of SH3 topology to prokaryotic signaling processes, and in particular, is the first example for the involvement of SH3-like domains in prokaryotic Ca 2ϩ signaling.
Conclusions-In summary, the present study reveals that Anabaena CcbP adopts an overall new fold with two Ca 2ϩbinding sites. Site I consists of an ␣-turn-␤ region unreported previously, whereas site II resembles a single EF-hand motif. Furthermore, CcbP harbors an SH3-like subdomain which might play a role in Ca 2ϩ release. Our study provides the structural basis for understanding the function of CcbP in the Ca 2ϩ signaling in Anabaena and offers novel insights for future investigations into the molecular mechanism of heterocyst differentiation regulation.