Structural Insights into Ca2+-dependent Regulation of Inositol 1,4,5-Trisphosphate Receptors by CaBP1*

Calcium-binding protein 1 (CaBP1), a neuron-specific member of the calmodulin (CaM) superfamily, modulates Ca2+-dependent activity of inositol 1,4,5-trisphosphate receptors (InsP3Rs). Here we present NMR structures of CaBP1 in both Mg2+-bound and Ca2+-bound states and their structural interaction with InsP3Rs. CaBP1 contains four EF-hands in two separate domains. The N-domain consists of EF1 and EF2 in a closed conformation with Mg2+ bound at EF1. The C-domain binds Ca2+ at EF3 and EF4, and exhibits a Ca2+-induced closed to open transition like that of CaM. The Ca2+-bound C-domain contains exposed hydrophobic residues (Leu132, His134, Ile141, Ile144, and Val148) that may account for selective binding to InsP3Rs. Isothermal titration calorimetry analysis reveals a Ca2+-induced binding of the CaBP1 C-domain to the N-terminal region of InsP3R (residues 1-587), whereas CaM and the CaBP1 N-domain did not show appreciable binding. CaBP1 binding to InsP3Rs requires both the suppressor and ligand-binding core domains, but has no effect on InsP3 binding to the receptor. We propose that CaBP1 may regulate Ca2+-dependent activity of InsP3Rs by promoting structural contacts between the suppressor and core domains.

Neuronal Ca 2ϩ -binding proteins (CaBP1-5 (17)) represent a new sub-branch of the CaM superfamily (18) that regulate various Ca 2ϩ channel targets. Multiple splice variants and isoforms of CaBPs are localized in different neuronal cell types (19 -21) and perform specialized roles in signal transduction. CaBP1, also termed caldendrin (22), has been shown to modulate the Ca 2ϩ -sensitive activity of InsP 3 Rs (13,14). CaBP1 also regulates P/Q-type voltage-gated Ca 2ϩ channels (23), L-type channels (24), and the transient receptor potential channel, TRPC5 (25). CaBP4 regulates Ca 2ϩ -dependent inhibition of L-type channels in the retina and may be genetically linked to retinal degeneration (26). Thus, the CaBP proteins are receiving increased attention as a family of Ca 2ϩ sensors that control a variety of Ca 2ϩ channel targets implicated in neuronal degenerative diseases.
CaBP proteins contain four EF-hands, similar in sequence to those found in CaM and troponin C (18) (Fig. 1). By analogy to CaM (27), the four EF-hands are grouped into two domains connected by a central linker that is four residues longer in CaBPs than in CaM. In contrast to CaM, the CaBPs contain non-conserved amino acids within the N-terminal region that may confer target specificity. Another distinguishing property of CaBPs is that the second EF-hand lacks critical residues required for high affinity Ca 2ϩ binding (17). CaBP1 binds Ca 2ϩ only at EF3 and EF4, whereas it binds Mg 2ϩ at EF1 that may serve a functional role (28). Indeed, changes in cytosolic Mg 2ϩ levels have been detected in cortical neurons after treatment with neurotransmitter (29). Other neuronal Ca 2ϩ -binding proteins such as DREAM (30), CIB1 (31), and NCS-1 (32) also bind Mg 2ϩ and exhibit Mg 2ϩ -induced physiological effects. Mg 2ϩ binding in each of these proteins helps stabilize their Ca 2ϩ -free state to interact with signaling targets.
Despite extensive studies on CaBP1, little is known about its structure and target binding properties, and regulation of InsP 3 Rs by CaBP1 is somewhat controversial and not well understood. Here, we present the NMR solution structures of both Mg 2ϩ -bound and Ca 2ϩ -bound conformational states of CaBP1 and their structural interactions with InsP 3 R1. These CaBP1 structures reveal important Ca 2ϩ -induced structural changes that control its binding to InsP 3 R1. Our target binding analysis demonstrates that the C-domain of CaBP1 exhibits Ca 2ϩ -induced binding to the N-terminal cytosolic region of InsP 3 R1. We propose that CaBP1 may regulate Ca 2ϩ -dependent channel activity in InsP 3 Rs by promoting a structural interaction between the N-terminal suppressor and ligand-binding core domains that modulates Ca 2ϩ -dependent channel gating (8,33,34).

EXPERIMENTAL PROCEDURES
Expression and Purification of CaBP1-CaBP1 has two splice variants expressed in the brain, termed l-CaBP1 and s-CaBP1 (17). Both variants regulate Ca 2ϩ channels with similar efficacy (14) and the extra residues in the long variant can be deleted without affecting CaBP1 binding to InsP 3 Rs. The short splice variant (19.4 kDa and 167 residues) is more soluble and amenable for NMR structural analysis and was used throughout this study. Recombinant CaBP1 and mutants were expressed and purified from Escherichia coli strain BL21(DE3) as described previously (28).
Construction of CaBP1 N-domain and C-domain Fragments-cDNAs coding for the CaBP1 N-domain (residues 1-91; CaBP1-N) and C-domain (residues 96 -167; CaBP1-C) were cloned into protein expression vectors pET-28a(ϩ) and pET-3a(ϩ), respectively. Recombinant CaBP1-C protein was expressed and purified by the same method as full-length CaBP1. The His 6 -CaBP1-N protein was purified first by nickel-Sepharose (Amersham Biosciences), and then by using Superdex 200 size exclusion chromatography.
Construction of CaBP1 Mutants-The D35A, D37A, D39A, D46A, ⌬L132, H134E, and V148A mutants of CaBP1 were generated by using the QuikChange site-directed mutagenesis kit (Stratagene) and the presence of these mutations was confirmed by DNAsequencing.Themutantexpression and purification procedures were the same as that for wild type. Expression and Purification of IP 3 R sup - , IP 3 R core -(224 -604), and InsP 3 R-(1-587)-The recombinant suppressor domain (InsP 3 R sup residues, 2-223) and ligand-binding core domain (InsP 3 R core residues, 224 -604) containing a GST tag were cloned, expressed, and purified as described by Ref. 35. The GST tag was removed by adding 1 g of thrombin to the purified GST fusion protein sample that was then applied to a Superdex-75 size exclusion column to remove the GST tag and other impurities. A recombinant dual domain construct containing both InsP 3 R sup and InsP 3 R core (InsP 3 R-(1-587)) was cloned, expressed, and purified as described (35). The recombinant InsP 3 R-(1-587) protein contained a C-terminal intein-CBD-His 9 tag that was first purified with nickel-nitrilotriacetic acid resin (Qiagen) and the CBD-His 9 tag was cleaved by treatment with 20 mM dithiothreitol for 24 h. The cleaved protein was released from chitin beads, concentrated, and then chromatographed on a Superdex 200 column.
NMR Spectroscopy-Samples for NMR analyses were prepared by dissolving unlabeled, 15  -bound), or 5 mM CaCl 2 , 5 mM MgCl 2 (Ca 2ϩbound). All NMR experiments were performed at 30°C on a Bruker Avance 600 MHz spectrometer equipped with triple resonance cryoprobe and z axis gradient. Backbone and side chain assignments were described previously (36,37). All NMR data were processed and analyzed by using the programs NMRPipe and nmrView.
NMR Structure Calculation-The structures were calculated with XPLOR-NIH (38) that employed the YASAP protocol (39). Distance restraints derived from inter-proton NOEs and dihedral angles ( and ) from chemical shift index data are summarized in Table 1. Distance constraints involving Ca 2ϩ bound to loop residues 1, 3, 5, 7, and 12 in EF3 and EF4 (27), and Mg 2ϩ bound to loop residues 1, 3, and 5 in EF1 were introduced as described previously (40). Fifty independent structures were calculated and the 15 lowest energy structures were selected. The final structural statistics are summarized in Table 1 and coordinates were deposited into the RCSB Protein Data bank (accession numbers 2k7b, 2k7c, and 2k7d).

CaBP1
Has Two Independent Domains-A critical first step in the NMR structural analysis of CaBP1 was to identify whether the four EF-hands in CaBP1 combine to form two separately folded domains: N-domain (EF1 and EF2) versus C-domain (EF3 and EF4) like what is seen in CaM (42). Alternatively, the four EF-hands might interact to form a single globular domain like what is observed in NCS-1 (43) and CIB1 (44). First, we analyzed the NOESY-HSQC spectra of CaBP1 and were unable to detect NOE-based contacts between the two domains, consistent with this protein having non-interacting domains.
Our second approach was to examine the backbone flexibility of the two domains and the central linker. In Fig. 2A, { 1 H}-15 N NOE measurements indicate relatively low heteronuclear NOE values (ϳ0.5) for residues in the central linker region (residues 92-98), suggesting that CaBP1 does indeed contain a flexible inter-domain linker. By contrast, much higher heteronuclear NOE values (ϳ0.8) are found for residues in each domain and indicate the two domains are separately folded.
A final test for the existence of two independent domains was to analyze NMR spectra of individual domain fragments of CaBP1: N-domain (residues 1-91) and C-domain (96 -167). The 1 H-15 N-HSQC spectra of the domain constructs (Fig. 2, B and C) indicate that each domain is separately folded without having the other domain present. In addition, the backbone amide chemical shifts for each residue in the domain fragments are nearly identical to the corresponding chemical shifts of the full-length protein. Thus, the structures of the isolated domain fragments must remain intact in the full-length protein, consistent with two non-interacting domains.
On the basis of our NMR analyses above, CaBP1 has two independently folded domains (N-domain, EF1 and EF2, and C-domain, EF3 and EF4) separated by a flexible linker. The structures of each domain were analyzed separately below. The C-domain structure was solved by analyzing NMR spectra of a peptide fragment (CaBP1-C, residues, 96 -167), whereas the structure of the Mg 2ϩ -bound N-domain was solved by analyzing NMR spectra of full-length CaBP1. The Ca 2ϩ -bound N-domain was not studied because it does not bind Ca 2ϩ under physiological conditions (28). In summary, we present below three separate NMR solution structures of CaBP1: 1) Mg 2ϩbound N-domain (PDB 2k7b), 2) Mg 2ϩ -bound C-domain (PDB 2k7c), and 3) Ca 2ϩ -bound C-domain (PDB 2k7d). The statistics for these structures are summarized in Table 1.
Structure of Mg 2ϩ -bound CaBP1-The first 21 N-terminal residues of CaBP1 exhibited weak NMR intensities and could not be accurately analyzed. The remaining residues (22-167) exhibited strong 1 H-15 N-HSQC peaks and their sequence-specific NMR assignments were analyzed and described previously (36) (BMRB number 15197). The assigned resonances in the HSQC spectrum represent main chain and side chain amide groups that serve as fingerprints of the overall conformation. Three-dimensional protein structures derived from the NMR assignments were calculated on the basis of NOE data, slowly exchanging amide NH groups, chemical shift analysis, and 3 J NH␣ spin-spin coupling constants (see "Experimental Procedures"). The final NMR-derived structures of Mg 2ϩ -bound CaBP1 are illustrated in Fig. 3, A and B.
The NMR structure of Mg 2ϩ -bound CaBP1 indicates that Mg 2ϩ is bound at EF1 as evidenced by Mg 2ϩ -dependent amide chemical shift changes for residues in the EF1 binding loop (Asp 35 , Asp 37 , Asp 39 , and Gly 40 ). Mg 2ϩ -binding caused a large downfield amide proton chemical shift for Gly 40 due in part to formation of a strong hydrogen bond between its main chain amide proton and the carboxylate side chain of Asp 35 . To identify possible chelating interactions with the bound Mg 2ϩ , we made the following point mutations (D35A, D37A, D39A, and D46A) to residues in the EF1 loop at positions 1, 3, 5, and 12. Mg 2ϩ binding to each mutant versus wild type was monitored by analyzing the Gly 40 amide resonance. The Mg 2ϩ -binding analysis revealed that Asp 35 , Asp 37 , and Asp 39 are each essential for high affinity Mg 2ϩ binding, suggesting that their carboxy-late side chains might form coordinate covalent bonds with the bound Mg 2ϩ . A similar Mg 2ϩ binding geometry involving acidic side chains from residues at positions 1, 3, and 5 was also observed in the structure of Mg 2ϩ -bound calbindin (49). The stereochemical geometry and chelation of the bound Mg 2ϩ at EF1 (magenta sphere, Fig. 3A) was modeled like that described by Ref. 50. The EF2 loop in CaBP1 does not bind Ca 2ϩ or Mg 2ϩ and is structurally distorted by the presence of Gly 75 at the fifth position in the binding loop.
Structure of Ca 2ϩ -bound CaBP1-The NMR-derived structure of the Ca 2ϩ -bound CaBP1-C is shown in Fig. 3C (37) (BMRB number 15623). The secondary structure of Ca 2ϩbound CaBP1 is nearly identical to that determined above for Mg 2ϩbound CaBP1 (Fig. 1). By contrast, the overall tertiary structure of Ca 2ϩ -bound CaBP1-C (Fig. 3C) is quite different from that of Mg 2ϩbound CaBP-C (Fig. 3B), reminiscent of the Ca 2ϩ -induced closed to open transition seen previously in CaM (45) and troponin C (48). The structures of EF3 and EF4 in Ca 2ϩbound CaBP1 resemble the familiar "open" conformation of Ca 2ϩ occupied EF-hands in CaM (27) and many other EF-hand proteins. The interhelical angles are 100.6°(EF3) and 110.6°(EF4) for Ca 2ϩ -bound CaBP1 (see Table 2). The overall main chain structure of Ca 2ϩbound CaBP1-C (Fig. 3C) is very similar to that of Ca 2ϩ -bound CaM with a root mean square deviation is 1.2 Å when comparing their main chain atoms.
The NMR structure of Ca 2ϩbound CaBP1 confirms Ca 2ϩ binding at EF3 and EF4, as evidenced by characteristic Ca 2ϩ -dependent amide chemical shift changes assigned to Gly 117 in EF3 and Gly 154 in EF4. Ca 2ϩ -binding caused large downfield amide proton chemical shifts for Gly 117 and Gly 154 due in part to formation of a strong hydrogen bond between its main chain amide proton and the carboxylate side chain of Asp 112 (EF3) and Asp 149 (EF4), respectively. The geometry of the coordinate covalent bonds formed between chelating amino acid residues in CaBP1 and the bound Ca 2ϩ could not be observed directly in our NMR study. Instead, the stereochemical geometry and chelation of Ca 2ϩ bound at EF3 and EF4 (orange spheres, Fig. 3C) was modeled using structural constraints derived from the x-ray crystal structure of Ca 2ϩ -bound CaM (27), which closely resembles the binding site geometry conserved in other EF-hand proteins (51).
Dimerization of CaBP1?-Previous hydrodynamic analyses of CaBP1 suggested a monomer-dimer equilibrium under NMR conditions (28). Indeed, our 15 N NMR relaxation analysis (T 1 and T 2 ) of CaBP1 in this study suggests an average rotational correlation time of ϳ12 ns (consistent with a protein dimer) that decreased somewhat when the protein concentration was lowered 10-fold. But, we did not observe any significant chemical shift changes in NMR spectra recorded as a function of protein concentration (50 M to 1 mM). Also, intermolecular NOEs could not be detected in 13 C-filtered NOESY-HMQC spectra of CaBP1 recorded from a mixed labeled sample. Thus, the CaBP1 monomer-dimer equilibrium must have an exchange rate that is much faster than the chemical shift time scale and the structure of the dimer cannot be resolved by NMR. Such fast exchange kinetics and hence low affinity for dimerization (K d ϳ 100 M) is not likely to be physiologically relevant and was not characterized further.
Surface Properties of CaBP1 Versus CaM-Space filling representations of Mg 2ϩ -bound and Ca 2ϩ -bound CaBP1 are illustrated and compared with those of CaM (Fig. 4). The surface residues of Mg 2ϩ -bound CaBP1 are similar to those of apo-CaM (Fig. 4, A and B). The N-domain of Mg 2ϩ -bound CaBP1 contains mostly negatively charged residues on the protein surface (highlighted red in Fig. 4A) that remain invariant in CaM, and the overall surface charge is nearly the same between the two. The C-domain surface of Mg 2ϩ -bound CaBP1 also looks similar to that of apo-CaM (Fig. 4B). The N-domain has a few exposed hydrophobic residues in Mg 2ϩ -bound CaBP1 (Met 57 , Met 61 , Met 72 , and Leu 74 ) not conserved in CaM (Fig. 1) that might serve a functional role in target recognition.
The protein surface of Ca 2ϩ -bound CaBP1-C is somewhat different from that of Ca 2ϩ -bound CaM (Fig. 4C). The front face of Ca 2ϩ -bound CaBP1-C exhibits a striking solvent-exposed hydrophobic surface (highlighted yellow in Fig. 4C) that is wider and more expansive than that of Ca 2ϩ -bound CaM. The solvent-exposed hydrophobic patch in Ca 2ϩ -bound CaBP1-C contains non-conserved residues located in the loop between EF3 and EF4 (Leu 132 , His 134 , Val 136 , and His 13 8), the helix of EF4 (Ile 141 , Ile 144 , and Val 148 ), and the domain linker (Ile 99 and Val 101 ). The exposed hydrophobic patch is surrounded by a ring of charged residues that might form electrostatic contacts with target proteins. Basic residues in CaBP1 (Lys 102 , His 134 , His 138 , and Arg 139 ) are replaced by negatively charged residues in CaM that might confer specific electrostatic contacts. We suggest that these nonconserved residues on the surface of Ca 2ϩ -bound CaBP1 may form a target binding site and help explain the highly selective binding of CaBP1 to InsP 3 Rs.
CaBP1 Interaction with N-terminal Cytosolic Residues in InsP 3 R1-CaBP1 was shown previously to regulate Ca 2ϩ -induced channel activity of InsP 3 Rs and the CaBP1 interaction site was localized to the N-terminal cytosolic region of InsP 3 R1 (residues 1-604) (13). We performed a series of target binding studies using isothermal titration calorimetry (ITC) and NMR to characterize the structural interaction between CaBP1 and InsP 3 R1 (Fig. 5). An N-terminal peptide fragment of InsP 3 R1 (residues, 1-587, called InsP 3 R-(1-587)), saturated with InsP 3 , exhibited strongly exothermic binding (⌬H ϭ Ϫ1.96 kcal/mol) to Ca 2ϩ -bound CaBP1 with a 1:1 stoichiometry and a dissociation constant (K d ) of 3 M (Fig. 5A). By contrast, InsP 3 R-(1-587) exhibited about 10-fold weaker binding to Mg 2ϩ -bound CaBP1 (⌬H ϭ Ϫ1.55 kcal/mol and K d ϭ 30 M). InsP 3 R-(1-587) showed no detectable binding to either Ca 2ϩ -free or Ca 2ϩbound CaM under these same conditions, which was somewhat surprising given that Ca 2ϩ -CaM has been suggested to bind InsP 3 R1 and negatively regulate channel gating (12). Also, apo-InsP 3 R-(1-587) binds to CaBP1 with approximately the same affinity as the ligand-bound receptor. Thus, ligand-bound InsP 3 R-(1-587) exhibits Ca 2ϩ -induced binding to CaBP1 with high selectivity over CaM. The lack of InsP 3 R-(1-587) binding to CaM was also verified by using NMR in which the 1 H-15 N-HSQC spectrum of 15 N-labeled CaM remained unaffected as a function of adding excess, unlabeled InsP 3 R-(1-587). The 1 H-15 N-HSQC spectrum of 15 N-labeled CaBP1, by contrast, exhibited striking peak broadening and chemical shift changes upon addition of saturating InsP 3 R-(1-587), further demonstrating that CaBP1 binds to InsP 3 R-(1-587). Unfortunately, because all NMR peaks in the HSQC spectrum of CaBP1 are severely broadened and uniformly affected by InsP 3 R-(1-587) binding, it was not possible to identify any specific binding site residues by chemical shift mapping.
The InsP 3 R1-binding site on CaBP1 was investigated by performing ITC experiments separately on N-domain and C-do-   To probe the CaBP1-binding site within InsP 3 R-(1-587), ITC studies were performed separately by using the suppressor domain (residues 1-224, InsP 3 R sup ) and the InsP 3 binding core domain (residues 236 -604, InsP 3 R core ). No heat signal could be detected upon individually adding either the suppressor domain and/or core domain to CaBP1, suggesting either ⌬H ϭ 0 or a lack of binding in the micromolar range (K d Ͼ Ͼ 10 Ϫ4 M). A lack of binding under NMR conditions was verified by the 1 H-15 N-HSQC spectrum of 15 N-labeled CaBP1 (fulllength) that remained unaffected as a function of adding excess suppressor and/or core domain. Thus, CaBP1 does not exhibit high affinity binding to either the suppressor or core domain alone, but rather the two domains must be linked together to have high affinity binding to CaBP1.
The binding of CaBP1 to InsP 3 R-(1-587) has little or no effect on ligand-binding affinity. The binding of InsP 3 to InsP 3 R-(1-587) is exothermic (⌬H ϭ Ϫ16.2 kcal/mol) with a 1:1 stoichiometry and dissociation constant (K d ) of ϳ1 M (Fig.  5B). The apparent K d measured by ITC is at least 100-fold weaker than the intrinsic ligand binding affinity measured for full-length InsP 3 R1 (52). The discrepancy could be explained in part by a protein conformational change in InsP 3 R-(1-587) coupled to InsP 3 binding. The intrinsic binding of InsP 3 (K a ϳ 10 8 M Ϫ1 ) if coupled to an unfavorable conformational change   Table 3.
(K eq ϳ 10 Ϫ2 ) would yield an overall equilibrium constant of K tot ϭ K a ϫ K eq ϳ 10 6 M Ϫ1 , consistent with the overall K d measured by ITC. Thus, InsP 3 binding to InsP 3 R-(1-587) induces a protein conformational change, consistent with predictions from small-angle x-ray scattering analysis (35). The apparent K d for InsP 3 binding to InsP 3 R-(1-587) is NOT affected by the presence or absence of saturating CaBP1 (Fig. 5B), demonstrating that CaBP1 binding to InsP 3 R-(1-587) does not block or otherwise influence ligand binding.
Structural Model of the CaBP1⅐InsP 3 R-(1-587) Complex-The relatively low solubility of the CaBP1⅐InsP 3 R-(1-587) complex has thus far hampered our efforts to directly solve the complex structure by NMR or x-ray crystallography. Instead, we used a computational docking approach that takes into account variables such as shape complementarity, desolvation energetics, and electrostatics to simulate the structure of the protein complex (53). Separate x-ray crystal structures have been solved recently for InsP 3 R sup (54) and InsP 3 R core (55). Our ITC analysis indicates that CaBP1-C binds cooperatively to InsP 3 R sup and InsP 3 R core only when both domains are connected (Fig. 5). This cooperativity suggests that CaBP1-C might contact both InsP 3 R sup and InsP 3 R core in the complex. The first step in the model calculation was to individually dock CaBP1-C to each domain and generate binary complexes: CaBP1-C⅐InsP 3 R sup and CaBP1-C⅐InsP 3 R core . Structures of the separate binary complexes were then aligned with respect to CaBP1-C to predict the disposition of InsP 3 R sup and InsP 3 R core in the ternary complex.
A total of 20 independent docking calculations were performed for each binary complex. A statistical analysis of the CaBP1-InsP 3 R sup docked structures revealed a striking tendency for CaBP1-C to bind to an exposed surface on the helical "arm" (residues 66 -110) in InsP 3 R sup , suggested previously to be functionally important (14,56). This docking model is also consistent with previous mutagenesis studies, suggesting that the arm residues interact with InsP 3 R core (55). Arm residues (66 -81) also form a potential calmodulin binding motif shown previously to inhibit CaBP1 binding to InsP 3 R1 (14). Finally, it is well known that EF-hand proteins generally bind to helical segments in target proteins (57). Thus, the docking interactions of CaBP1-C with the suppressor helical arm are plausible and well justified experimentally. The family of docked structures of the CaBP1⅐InsP 3 R core complex revealed a tendency for CaBP1-C to interact with the ␤-trefoil subdomain (residues 397-420) located on the opposite face from the ligand-binding site. The lowest energy structures of the CaBP1⅐InsP 3 R sup and CaBP1⅐ InsP 3 R core binary complexes were then aligned with respect to CaBP1. Candidate docked structures were selected that minimize any overlap between InsP 3 R sup and InsP 3 R core while maintaining a reasonably close distance (Ͻ30 Å) between the final residue of InsP 3 R sup and initial residue of InsP 3 R core .
A representative structure of the docked CaBP1⅐InsP 3 R sup ⅐ InsP 3 R core complex is shown in Fig. 6. CaBP1-C interacts primarily with the arm helix in the suppressor domain (residues 72-94, colored brown in Fig. 6), suggested previously to interact with CaBP1 (8,13,14). This CaBP1-binding site on InsP 3 R1 is located far away from the ligand binding site, consistent with CaBP1 having no effect on the ligand-binding affinity (Fig. 5B).
The exposed hydrophobic patch of Ca 2ϩ -bound CaBP1-C (Fig.  4C) interacts with both arm helices of InsP 3 R sup . The aromatic rings of Phe 72 and Trp 73 (suppressor domain) contact the side chains of Ile 144 , Val 148 , and Met 164 of CaBP1. Also, non-conserved CaBP1 residues (Leu 132 , His 134 , and Val 148 highlighted red in Fig. 1) interact with the C-terminal arm helix that might help explain its highly specific binding to CaBP1 versus CaM. Indeed, the CaBP1 mutants (⌬L132, H134E, V148A) show ϳ2-fold weaker binding to InsP 3 -(1-587) ( Table 3). CaBP1 also makes a few contacts with residues in InsP 3 R core (residues 405-409) that are also close to InsP 3 R sup helical arm residues, which might explain in part the cooperative interaction. Nearly all exposed residues on the C-terminal arm helix (Leu 88 , Lys 91 , His 94 , Ala 95 , Leu 98 , and Thr 105 ) interact with exposed ␤-trefoil residues from InsP 3 R core , consistent with previous mutagenesis studies (55). The extensive domain interface predicted in Fig. 6 causes InsP 3 R sup and InsP 3 R core to interact in a relatively compact arrangement, consistent with previous small-angle x-ray scattering measurements on InsP 3 R-(1-587) (35). We conclude that CaBP1 binding to the receptor may stabilize a structural interaction between InsP 3 R sup and InsP 3 R core that might play a role in channel gating. This cooperative interdomain associa- FIGURE 6. Structural model of the docked complex for CaBP1-C⅐InsP 3 R sup / InsP 3 R core . CaBP1-C, InsP 3 R sup (residues 2-223), and InsP 3 R core (residues 236 -586) are colored red, yellow, and cyan, respectively. Arm helix of InsP 3 R core interacting with CaBP1 is colored brown. The docking calculation was performed using Zdock as described under "Experimental Procedures."

DISCUSSION
In this study, we determined the NMR solution structures of CaBP1 in both Mg 2ϩ -bound and Ca 2ϩ -bound states and characterized their structural interaction with InsP 3 R1. The overall main chain structure of Mg 2ϩ -bound CaBP1 (Fig. 3, A and B) is similar to that seen previously in apo-CaM (45) and troponic C (48). One important difference is that Mg 2ϩ is bound tightly at EF1 in CaBP1. The structure of Ca 2ϩ -bound CaBP1 is somewhat different from that of CaM (Fig. 4C). At saturating Ca 2ϩ levels, the CaBP1 N-domain does NOT bind Ca 2ϩ but remains in a closed conformation with Mg 2ϩ bound at EF1. The C-domain binds Ca 2ϩ at EF3 and EF4 and adopts the familiar Ca 2ϩ -bound open conformation (Fig.  3C) with an exposed hydrophobic patch (Fig. 4C). Many of the exposed hydrophobic residues in CaBP1 (Leu 132 , His 134 , Ile 144 , and Val 148 ) are not conserved in CaM and might play a role in controlling the highly specific and Ca 2ϩ -induced binding to InsP 3 R1 to CaBP1 (Fig. 6). Indeed, the CaBP1 mutants (⌬L132, H134E, and V148A) show noticeably weaker binding to InsP 3 -(1-587) ( Table 3).
Our target binding analysis indicates that Ca 2ϩ -bound CaBP1 binds tightly to InsP 3 R-(1-587) but does not bind to either InsP 3 R sup or InsP 3 R core alone. These observations seem somewhat at odds with an earlier report, suggesting that CaBP1 can bind to isolated segments of InsP 3 R sup independent of Ca 2ϩ (14). Indeed, such binding of CaBP1 to InsP 3 R sup is consistent with our proposed structural model of CaBP1-InsP 3 R (Fig. 6), showing that CaBP1 forms intimate contacts with the helical arm in InsP 3 R sup . However, the affinity of CaBP1 binding to InsP 3 R sup alone must be quite low, which would explain why this weak binding escaped detection in our ITC analysis (Fig. 5). Furthermore, we suggest that the affinity of CaBP1 binding to InsP 3 R1 is significantly enhanced by the cooperative interaction between InsP 3 R sup and InsP 3 R core as depicted in Fig. 6. This same interaction also appears to partially block the ligand binding site, which may explain why InsP 3 binds with ϳ10-fold higher affinity to an isolated fragment of InsP 3 R core than it binds to InsP 3 R1 (52).
Previous studies have suggested that InsP 3 R1 binds to both the Ca 2ϩ -free and Ca 2ϩ -bound forms of CaBP1 (13,14). In this study, we confirm that InsP 3 R-(1-587) does indeed bind to both the Mg 2ϩ -bound and Ca 2ϩ -bound CaBP1. However, our more quantitative ITC analysis reveals that Ca 2ϩ -bound CaBP1 binds to InsP 3 R-(1-587) with ϳ10-fold higher affinity compared with that of Mg 2ϩ -bound CaBP1. Furthermore, the Mg 2ϩ -bound CaBP1 N-domain does NOT bind to InsP 3 R-(1-587) ( Table 3). The lower affinity target binding by Mg 2ϩbound/Ca 2ϩ -free CaBP1 (C-domain) at low, basal Ca 2ϩ levels might represent its binding to IQ-motifs in the receptor (12). Alternatively, we submit that the 10-fold stronger binding by Ca 2ϩ -bound CaBP1 may be sufficient to exclude InsP 3 R1 binding to Ca 2ϩ -free CaBP1 under physiological conditions. Thus, CaBP1 would selectively bind to InsP 3 R1 only when the cell is stimulated (at high cytosolic Ca 2ϩ levels) and modulate Ca 2ϩdependent channel gating. InsP 3 R-(1-587) binds to CaBP1 with at least 100-fold higher affinity than its binding to CaM. The highly selective binding to CaBP1 is explained in part by the large solvent-exposed surface area of the hydrophobic patch in CaBP1-C (Fig. 4C) as well as by a number of non-conserved residues on this surface (Fig. 1). Non-conserved CaBP1 residues (Leu 132 , His 134 , Ile 144 , and Val 148 ) are proposed to make unique hydrophobic contacts with the helical arm of InsP 3 R sup (Fig. 6). The highly specific binding of InsP 3 R-(1-587) to CaBP1 relative to CaM illustrates that CaBP1 is a specialized Ca 2ϩ sensor for regulating InsP 3 Rs in the brain and retina. This contrasts with CaM that is ubiquitously expressed in all tissues and has a much broader range of target interactions. The specialized target binding by CaBP1 may be augmented by CaBP splice variants and isoforms that exhibit tissue-specific neuronal expression (19 -21). We propose that the multiplicity of CaBPs in the central nervous system might play a role in fine tuning their interaction with various InsP 3 R isoforms and other Ca 2ϩ channel targets.
CaBP1 has been suggested to promote channel opening in the absence of InsP 3 (13). CaBP1 binds to InsP 3 R-(1-587) both in the presence or absence of InsP 3 , and CaBP1 has little or no effect on InsP 3 binding to InsP 3 R-(1-587) (Fig. 5). Thus, CaBP1 interacts structurally with InsP 3 R1 even in the absence of InsP 3 . This is consistent with our docking analysis in which CaBP1-C interacts primarily with the helical arm region of the suppressor domain (Fig. 6), located far from the ligand-binding site. It is also possible that CaBP1 binding to apo-InsP 3 R1 induces structural interactions between InsP 3 R sup and InsP 3 R core that may mimic structural changes caused by ligand-binding and thus explain the observed InsP 3 -independent channel opening.
Last, our structural studies suggest that the CaBP1 C-domain alone might be sufficient for promoting Ca 2ϩ -dependent channel activity because CaBP1-N does not bind to InsP 3 R-(1-587). However, the current study does not preclude the CaBP1 N-domain from interacting elsewhere in the channel. For example, Ca 2ϩ -dependent inactivation of L-type channels was shown recently to require separate binding by both the N-domain and C-domain of CaM (58). A similar bipartite interaction by the two domains of CaBP1 might also be important for regulation of InsP 3 R1. The CaBP1 N-domain might bind to either the central regulatory domain or the C-terminal cytosolic domain of InsP 3 R1. In the future, we plan to further investigate the functional interactions between InsP 3 R1 and CaBP1 by determining the atomic resolution structure of CaBP1/InsP 3 R-(1-587) and by exploring a possible role for the CaBP1 N-domain.