Structural Insights into Activation of the Retinal L-type Ca2+ Channel (Cav1.4) by Ca2+-binding Protein 4 (CaBP4)*

Background: Cav1.4 is regulated by CaBP4, which is required for continuous release of neurotransmitter in retinal photoreceptor cells. Results: CaBP4 contains two separate EF-hand lobes that bind Ca2+ and form a collapsed structure around the IQ motif in Cav1.4. Conclusion: CaBP4 is suggested to activate Cav1.4 by disrupting an interaction between IQ and ICDI. Significance: CaBP4 mutations associated with congenital stationary night blindness impair its binding to IQ. CaBP4 modulates Ca2+-dependent activity of L-type voltage-gated Ca2+ channels (Cav1.4) in retinal photoreceptor cells. Mg2+ binds to the first and third EF-hands (EF1 and EF3), and Ca2+ binds to EF1, EF3, and EF4 of CaBP4. Here we present NMR structures of CaBP4 in both Mg2+-bound and Ca2+-bound states and model the CaBP4 structural interaction with Cav1.4. CaBP4 contains an unstructured N-terminal region (residues 1–99) and four EF-hands in two separate lobes. The N-lobe consists of EF1 and EF2 in a closed conformation with either Mg2+ or Ca2+ bound at EF1. The C-lobe binds Ca2+ at EF3 and EF4 and exhibits a Ca2+-induced closed-to-open transition like that of calmodulin. Exposed residues in Ca2+-bound CaBP4 (Phe137, Glu168, Leu207, Phe214, Met251, Phe264, and Leu268) make contacts with the IQ motif in Cav1.4, and the Cav1.4 mutant Y1595E strongly impairs binding to CaBP4. We conclude that CaBP4 forms a collapsed structure around the IQ motif in Cav1.4 that we suggest may promote channel activation by disrupting an interaction between IQ and the inhibitor of Ca2+-dependent inactivation domain.

CaBP4 3 is a 35-kDa Ca 2ϩ -binding protein expressed in retinal photoreceptor cells, localized primarily at the synaptic bulb (1,2). CaBP4 controls the continuous release of the glutamate neurotransmitter in dark state photoreceptor cells (3), by regulating L-type calcium channels (Cav1.4) (1,4). The C-terminal regulatory region of Cav1.4 binds to CaBP4 (4), which enables channel activation at high cytosolic Ca 2ϩ levels and hyperpo-larized voltages (1). A distinctive characteristic of Cav1.4 is the lack of Ca 2ϩ -dependent inactivation (CDI), because calmodulin (CaM) does not bind to its C-terminal region (5). Instead, Cav1.4 contains a stretch of residues in the C-terminal tail called the inhibitor of Ca 2ϩ -dependent inactivation (ICDI) that prevents binding of CaM to the IQ motif. In addition, CaBP4 binding to Cav1.4 can prevent Ca 2ϩ -dependent inactivation even in the absence of ICDI (4,6).
Ca 2ϩ -dependent regulation of Cav1.4 mediated by CaBP4 is genetically linked to congenital stationary night blindness (1). Particular mutations in both the CACNA1F gene that encodes the ␣ 1 -subunit of an L-type Ca 2ϩ channel (Cav1.4␣) (7-10) and the CABP4 gene (11,12) were associated with this autosomal recessive retinopathy. These findings are further underscored by observations that mice lacking either CaBP4 or Cav1.4␣ display a CSNB2-like phenotype (1,13,14). The first two mutations identified in the CABP4 gene were c.800_801delAG and c.370C3 T. The first c.800_801delAG mutation causes a frameshift, p.Glu267fs, which extends the protein by 91 novel amino acids and deletes its C terminus (11). In both humans and mice, mutations in CABP4 lead to defective signaling between rods and cones with bipolar cells, wherein cones appear to be more affected. Thus, the phenotype presents more as a cone-rod dystrophy with a color vision deficit than congenital stationary night blindness.
CaBP4 belongs to a family of neuronal Ca 2ϩ -binding proteins (CaBP1-5 (15)) and contains four EF-hands like those found in CaM and CaBP1 (16) (Fig. 1). By analogy to CaM (17), the four EF-hands are grouped into two domains connected by a central linker that is four residues longer in CaBP4 than in CaM. In contrast to CaM, CaBP4 contains ϳ100 non-conserved amino acids upstream of the EF-hands in the N-terminal region. Another distinguishing property of CaBP4 is that the second EF-hand lacks critical residues required for high affinity Ca 2ϩ binding (15,18).
Despite extensive studies on CaBP4 (1,4,6,19), little is known about its structure and interaction with Ca 2ϩ channel targets. Here, we present NMR solution structures of both Mg 2ϩ -bound and Ca 2ϩ -bound conformational states of CaBP4 and characterize the CaBP4 structural interaction with Cav1.4. CaBP4 NMR structures reveal important residues essential for its Ca 2ϩ -dependent binding to Cav1.4. ITC and NMR analyses demonstrate that Ca 2ϩ -saturated CaBP4 (but not the Ca 2ϩfree/Mg 2ϩ -bound state) binds to the IQ motif in Cav1.4 (residues 1579 -1605). All four EF-hands in CaBP4 contact the helical IQ motif and form a collapsed structure in the complex. Exposed residues in CaBP4 (Phe 137 , Glu 168 , Leu 207 , Ile 209 , Met 251 , Phe 264 , and Leu 268 ) make specific contacts with conserved residues in the IQ motif (Phe 1586 , Ile 1592 , Tyr 1595 , and Arg 1597 ) that are essential for binding. We propose that CaBP4 activates Cav1.4 by binding to the IQ motif, which we suggest may disrupt an interaction between the IQ and ICDI domains.

EXPERIMENTAL PROCEDURES
Expression and Purification of CaBP4 -Full-length mouse CaBP4 (residues 1-271) could not be concentrated beyond 1 mg/ml and therefore was not soluble enough for high resolution structural analysis by NMR. The first 99 residues from the N terminus of CaBP4 were shown to be unstructured, because this region was extensively cleaved in limited proteolysis studies (4). Removal of the first 99 residues of CaBP4 markedly improved protein solubility and did not affect target or Ca 2ϩ binding. Thus, NMR experiments in this study were performed with the N-terminal deletion construct of CaBP4 (residues 100 -271, called CaBP4(100 -271)). Recombinant murine CaBP4(100 -271) and mutants were subcloned into pET28 vector, expressed in Rosetta2(DE3) cells, and purified as described previously (20,21).
Construction of CaBP4 Mutants-The M251A, F264E, and L268A mutants of CaBP4(100 -271) were generated by using the QuikChange site-directed mutagenesis kit (Stratagene), and the mutations were confirmed by DNA sequencing. Mutant expression and purification procedures were the same as those used for wild type CaBP4.
NMR Spectroscopy-Samples for NMR analyses were prepared by dissolving unlabeled, 15 N-labeled, 15 N, 13 C-labeled, or 2 H, 13  Avance 800-MHz spectrometer equipped with a triple resonance cryoprobe and z axis gradient. Backbone and side chain NMR assignments for Ca 2ϩ -bound CaBP4(100 -271) and Mg 2ϩ -bound N-lobe and C-lobe were determined as described previously (20,21). All triple-resonance NMR experiments done for making resonance assignments were performed, processed, and analyzed as described (24) on a sample of 13  (F3) 512, 64 ms). 15 N-edited and 13 C-edited NOESY-HSQC (mixing time of 120 ms) and TOCSY-HSQC experiments were also performed as described previously (25). 1 H-15 N residual dipolar coupling constants (D NH ) were measured with a 15 Nlabeled CaBP4 (ϳ0.3 mM) containing 10 mg/ml Pf1 phage (Asla Biotech) and using a two-dimensional in phase/antiphase 1 H-15 N HSQC experiment (26). Heteronuclear { 1 H}-15 N NOE experiments were performed using standard pulse sequences described previously (27). Steady-state { 1 H}-15 N NOE values were obtained by recording two sets of spectra in the presence and absence of a 3-s proton saturation period. The NOE experiments were repeated three times to calculate the average and S.D. for the NOE values. All NMR data were processed and analyzed with the programs NMRPipe and nmrView.
NMR Structure Calculations-The structures were calculated with XPLOR-NIH (28), which employed the YASAP protocol (29). Distance restraints derived from interproton NOEs and dihedral angles ( and ) from chemical shift index data are summarized in Table 1. Distance constraints were introduced for Ca 2ϩ bound to loop residues 1, 3, and 5 in the EF1 closed state (30) and Ca 2ϩ bound to loop residues 1, 3, 5, 7, and 12 in EF3 and EF4 (17). Fifty independent structures were calculated and refined using D NH restraints as described (31). The final structural statistics are summarized in Table 1, and coordinates were deposited into the RCSB Protein Data bank (accession numbers 2M28 and 2M29).
CaBP4 binding to the Cav1.4 IQ motif was measured as described previously (18). CaBP4 was exchanged into buffer containing 15 mM Tris-HCl, pH 7.5, 500 mM NaCl, and 1 mM tris(2-carboxyethyl)phosphine with the addition of either 2 mM EDTA (apo-state), 5 mM MgCl 2 (Mg 2ϩ -bound state) or 5 mM Ca 2ϩ (Ca 2ϩ -bound state). The IQ peptide (residues 1579 -1605) at a concentration of 10 M in the sample cell was titrated with CaBP4 (0.2 mM, delivered in 50 steps of 5 l each). Data were analyzed with a one-site model using the MicroCal Origin 7 for ITC (33). The term "one-site model" refers to one type of site (having a particular K d ) with n number of these sites per protein molecule.
Docking Calculations-Separate structures for the Ca 2ϩbound CaBP4 N-lobe and C-lobe were docked onto the helical structure of the IQ motif using HADDOCK (34 -36). The helical structure of the Cav1.4 IQ motif was generated by SWISS-MODEL using the coordinates of the Cav1.2 IQ domain (Protein Data Bank code 2B6E) as a template. Docking was done with the Guru interface at the HADDOCK Web server (34 -36). Mutagenesis data (IQ mutants: I1592A (4) and Y1595E, which weaken CaBP4 binding) and chemical shift perturbation data from 15 N-1 H HSQC (CaBP4 residues Leu 122 , Leu 130 , Phe 137 , Leu 157 , Phe 186 , Leu 207 , Phe 214 , Leu 235 , Met 251 , Glu 263 , Phe 264 , and Leu 268 ) were used to define active and passive restraints (34). All together, 1000 rigid body docking runs, 200 structure calculation runs with torsion angle dynamics, and 200 refinements in explicit solvent were carried out. Resulting structures were clustered according to intermolecular energy terms and an RMSD from the lowest energy structure. The 15 lowest energy structures had an interface RMSD of 1.7 Å.

RESULTS
The N-terminal Region of CaBP4 Is Unstructured-Fulllength CaBP4 (residues 1-271) tended to aggregate at protein concentrations needed for NMR (10 mg/ml) and was not soluble enough for structural analysis. The CaBP4 N-terminal region (residues 1-100) contains residues predicted to be unstructured (37). Indeed, the first 99 residues from the N terminus of CaBP4 were extensively cleaved in limited proteolysis studies (4), and an isolated 100-residue peptide fragment from CaBP4 (residues 1-100) is unstructured in solution based on having random coil NMR chemical shifts. Because the first 100 residues from the N terminus are unstructured, an N-terminal deletion construct was generated (residues 100 -271 called CaBP4(100 -271)) that is more soluble than full-length CaBP4 and adopts a well folded protein conformation in solution (21). The calorimetry and NMR structural studies described below were performed using CaBP4(100 -271).
To investigate whether Mg 2ϩ competes with Ca 2ϩ binding to CaBP4, the Ca 2ϩ -binding isotherm was measured in the presence of physiological Mg 2ϩ levels (2 mM) (Fig. 2C). The Ca 2ϩbinding isotherm in the presence of Mg 2ϩ shows two separate phases: an initial endothermic binding at one site (⌬H ϭ ϩ0.5 kcal/mol) followed by exothermic binding (⌬H ϭ Ϫ1.9 kcal/ mol) at the other two sites. The endothermic Ca 2ϩ binding to CaBP4(100 -271) in the presence of Mg 2ϩ contrasts with the overall exothermic binding of Ca 2ϩ in the absence of Mg 2ϩ . Also, the total enthalpy of Ca 2ϩ binding became reduced in the presence of Mg 2ϩ , consistent with endothermic dissociation of Mg 2ϩ that precedes Ca 2ϩ binding. The Mg 2ϩ dependence of the Ca 2ϩ binding isotherm suggests that Mg 2ϩ competes with Ca 2ϩ for binding to EF1 and EF3. In the photoreceptor cell, CaBP4 exists in the Ca 2ϩ -bound state in the dark (high Ca 2ϩ ) and switches to the Mg 2ϩ -bound state upon light activation when cytosolic Ca 2ϩ levels decrease below 50 nM (38).
CaBP4 Has Two Independent Domains-CaM (39) and CaBP1 (32) both have four EF-hands divided into two structurally independent domains (an N-lobe and C-lobe) connected by a flexible linker. To test whether the EF-hands in CaBP4 form two independent domains, 1 H-15 N HSQC NMR spectra of an isolated lobe that contains EF1 and EF2 (N-lobe, residues 100 -200) and a separate lobe that contains EF3 and EF4 (C-lobe, residues 198 -271) were compared with spectra of full-length CaBP4 (Fig. 3, A and B). The backbone amide chemical shifts in the lobe fragments indicate that the individual lobes are stably folded. Also, the assigned chemical shifts in each lobe fragment were nearly identical to the corresponding chemical shifts of the full-length protein. Thus, the two isolated lobes are structurally independent, consistent with two non-interacting domains.
Heteronuclear ({ 1 H}-15 N) NOE analysis of CaBP4 reveals considerable backbone flexibility in the central linker that connects the two lobes (Fig. 3C). Relatively low heteronuclear NOE values (Ͻ0.6) were observed for residues 192-200 in the central linker region, indicating that CaBP4 does indeed contain a flexible interdomain linker. Heteronuclear NOE values (ϳ0.8) are much higher for residues within each lobe and indicate that the two lobes are separately folded as shown previously for CaBP1 (32).
Structure of Ca 2ϩ -bound CaBP4 -The sequence-specific NMR assignments of CaBP4(100 -271) were reported previously (21), and the secondary structure based on these assignments is summarized in Fig. 1. In the current study, we report the NMR structure of CaBP4(100 -271) (Fig. 4). NMR-derived structures were calculated separately for the N-lobe (Protein  Table 2. Downfield spectral regions of 1 H-15 N HSQC spectra of Mg 2ϩ bound CaBP4(100 -271) (A, inset) and Ca 2ϩ -bound CaBP4(100 -271) (B, inset) are also shown. The downfield peaks assigned to Gly 143 and Gly 220 (A, inset) indicate that Mg 2ϩ is bound at EF1 and EF3, whereas peaks assigned to Gly 143 , Gly 220 , and Gly 257 (B, inset) indicate that Ca 2ϩ is bound at EF1, EF3, and EF4. Data Bank code 2M29) and C-lobe (Protein Data Bank code 2M28) using NOE-based distances and dihedral angle restraints that served as input for restrained molecular dynamics structure calculations (see "Experimental Procedures"). Statistics for the structure calculation are summarized in Table 1 for the 15 lowest energy conformers. The NMR-derived structures of CaBP4(100 -271) were validated with PROCHECK: 82% of N-domain residues and 74.2% of C-domain residues belonged to the most favorable region on the Ramachandran plot.
CaBP4(100 -271) contains Ca 2ϩ bound at EF1, EF3, and EF4 (orange spheres in Fig. 4) as evidenced by characteristic Ca 2ϩdependent amide chemical shift changes assigned to Gly-143 in EF1, Gly-220 in EF3, and Gly-257 in EF4 (Fig. 2B, inset). The geometry of the coordinate covalent bonds formed between chelating amino acid residues in CaBP4 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 was modeled with structural constraints derived from the x-ray crystal structure of Ca 2ϩ -bound CaM (17), which closely resembles the binding site geometry conserved in other EF-hand proteins (40). Structure calculations performed without the metal binding restraints produced an overall similar fold, and the metal binding restraints decrease the main chain RMSD by 25%.
The N-lobe of CaBP4 with one Ca 2ϩ bound at EF1 (EF2 unoccupied) adopts a "closed" conformation (41); the interhelical angles of EF1 and EF2 are 134.7 and 141.1°, respectively. That EF1 and EF2 both remain in a closed conformation even when Ca 2ϩ is bound at EF1 suggests that the binding energy of one Ca 2ϩ at EF1 (EF2 unoccupied) does not suffice to drive a closed-to-open transition. This implies that the binding energy of two Ca 2ϩ are needed to drive formation of the open state and could explain why two EF-hands are paired together to form a lobe (Fig. 4A). The Ca 2ϩ -bound closed conformation for the N-lobe in CaBP4 is similar to that of CaBP1 (42) with an RMSD of 0.9 Å.
The Ca 2ϩ -bound C-lobe of CaBP4 (two Ca 2ϩ bound) forms the familiar "open" conformation seen for other Ca 2ϩ -bound EF-hand proteins (16). For Ca 2ϩ -bound CaBP4, the interhelical angles of EF3 and EF4 are 104.1 and 88.3°and quite different from the interhelical angles of EF1 and EF2 in the closed conformation ( Fig. 4B and Table 3).
Exposed Hydrophobic Patch in CaBP4 -A surface representation of Ca 2ϩ -bound CaBP4 is illustrated in Fig. 5. The N-lobe surface contains many negatively charged residues (Glu 121 , Glu 125 , Glu 129 , Glu 136 , Glu 163 , and Glu 168 , highlighted in red in Fig. 5A). Only a few hydrophobic residues (Leu 122 , Leu 150 , and Met 164 ) and basic residues (Arg 120 , Arg 155 , and Arg 174 ) are exposed on the N-lobe surface. By contrast, the C-lobe of Ca 2ϩbound CaBP4 has an extensive array of solvent-exposed hydrophobic residues (Leu 207 , Phe 214 , Ile 222 , Leu 235 , Leu 239 , Met 251 , Phe 264 , and Leu 268 , highlighted in yellow in Fig. 5B). This exposed hydrophobic patch on the C-lobe makes important contacts with the Cav1.4 IQ motif (see below). The hydrophobic patch is surrounded peripherally by charged residues

FIGURE 5. Space-filling representations of Ca 2؉ -saturated CaBP4 illustrate exposed residues on the surface of the N-lobe (A) and C-lobe (B).
Acidic residues (Asp and Glu), basic residues (Arg, His, and Lys), and hydrophobic residues (Ile, Leu, Phe, Met, and Val) are colored red, blue, and yellow, respectively. C-lobe residues in the exposed hydrophobic patch make critical contacts with Cav1.4 (see below). The solvent-accessible hydrophobic surface area is 50 Å 2 for the N-lobe versus 92 Å 2 for the C-lobe.  (38), whereas the physiological Mg 2ϩ concentration remains steady at ϳ1 mM (43). Mg 2ϩ -bound CaBP4(100 -271) was not soluble enough for NMR structural analysis. Instead, NMR experiments were performed on separate constructs of the N-lobe (residues 100 -200) and C-lobe (residues 198 -271). 1 H-15 N HSQC spectra of Mg 2ϩ -bound N-lobe and C-lobe fragments are shown in Fig. 6, A and B. These spectra contain downfield shifted peaks assigned to Gly-143 (EF1) and Gly-220 (EF3), indicating that Mg 2ϩ is bound at EF1 and EF3, whereas EF2 and EF4 are unoccupied at saturating Mg 2ϩ levels. Mg 2ϩ binding at EF1 and EF3 was modeled based on the Mg 2ϩ -bound structure of CaBP1 (32). Residues at the 1-, 3-, and 5-positions of the EF-hand loop (in EF1 and EF3) were selected to chelate the bound Mg 2ϩ (30,44). Separate protein structures of the Mg 2ϩ -bound CaBP4 N-lobe and C-lobe were generated by CS-ROSETTA with HN, C␣, C␤, and CO chemical shift values used as structural restraints (45). The five lowest energy structures were selected from 1000 trial structures and refined against the residual D NH restraints as described (31). Initial residual dipolar coupling magnitude and rhombicity were calculated by fitting the measured residual dipolar couplings to the calculated structure using the PALES program (46). The residual dipolar coupling-refined structures of Mg 2ϩ -bound N-lobe and C-lobe are presented in Fig. 6, C and D. The N-lobe structures have a quality Q-factor of 0.28 and an R-factor of 0.95, and the C-lobe structures have a Q-factor of 0.17 and an R-factor of 0.93. Both structures adopt a "closed" conformation in which the helices are nearly antiparallel with interhelical angles defined in Table 3. The overall secondary structure and topology of Mg 2ϩ -bound CaBP1 are very similar to those described above for Ca 2ϩ -bound CaBP4. Ca 2ϩ -induced Conformational Changes-Ca 2ϩ -induced conformational changes in CaBP4 are illustrated by superimposing its Mg 2ϩ -bound and Ca 2ϩ -bound structures (Fig. 6E). For the CaBP4 N-lobe, the Ca 2ϩ -bound closed conformation is similar to that of the Mg 2ϩ -bound conformation (RMSD ϭ 1.1 Å), indicating that Ca 2ϩ binding to the N-lobe does not induce a significant conformational change. However, Ca 2ϩ binding to the C-lobe at EF3 and EF4 caused large changes in the interhelical angles for both EF3 and EF4 (Table 3). This Ca 2ϩ -induced . Resonance assignments are indicated by the residue labels. Downfield peaks at ϳ10.5 ppm are assigned to Gly 143 (N-lobe) and Gly 220 (C-lobe) and indicate that Mg 2ϩ is bound at EF1 and EF3. Main chain structures of Ca 2ϩ -free/Mg 2ϩ -bound CaBP4 N-lobe (C) and C-lobe (D) were calculated using NMR residual dipolar couplings (31) and CS-ROSETTA (45). Bound Mg 2ϩ ions at EF1 and EF3 are not shown. EF-hands are highlighted in color as defined in Fig. 1. E, structures of Ca 2ϩ -free/Mg 2ϩ -bound CaBP4 N-lobe (left) and C-lobe (right) highlighted in gray are overlaid onto structures of the Ca 2ϩ -bound N-lobe (left) and C-lobe (right) shown in color. The N-lobe structure is not affected by Ca 2ϩ , whereas the C-lobe exhibits a Ca 2ϩ -induced decrease in interhelical angles for EF3 and EF4 (Table 3). decrease in interhelical angles for both EF3 and EF4 is consistent with the familiar closed-to-open transition seen previously in CaM and other Ca 2ϩ sensor proteins (16).

Structure of CaBP4 and
CaBP4 Binds to the IQ Motif in Cav1.4 -CaBP4 was suggested previously to interact with the IQ motif (residues 1579 -1605) in Cav1.4 (4). NMR and ITC experiments were performed to quantify CaBP4 binding to a peptide fragment that represents the IQ motif. NMR spectral changes in 15 N-1 H HSQC spectra of 15 N-labeled Ca 2ϩ -bound CaBP4(100 -271) were observed upon adding unlabeled IQ peptide (Fig. 7A). These spectral changes saturated after adding 1 eq of peptide, consistent with a 1:1 stochiometry. Injection of the IQ peptide into Ca 2ϩ -bound CaBP4(100 -271) showed exothermic binding (⌬H ϭ Ϫ2.5 kcal/mol) with 1:1 stoichiometry and a dissociation constant (K d ) of 0.8 M (Fig. 7B (left) and Table 2). By contrast, the Ca 2ϩ -free/Mg 2ϩ -bound CaBP4 did not show any detectable binding to the IQ peptide as judged by a lack of NMR spectral changes in the 15 N-1 H HSQC spectra of CaBP4 after adding a 10-fold excess peptide (Fig. 7A, top right). ITC binding experiments showed that the IQ peptide binds to Ca 2ϩ -bound CaBP4(100 -271) with at least 10-fold higher affinity than it binds to the isolated C-lobe fragment (Fig. 7B (right) and Table  2). No ITC heat signals could be detected upon adding IQ pep-tide to the N-lobe, consistent with a lack of binding, and the Ca 2ϩ -bound N-lobe did not show any detectable binding to the IQ peptide as judged by a lack of NMR spectral changes after adding a 10-fold excess of this peptide (Fig. 7A, bottom right). The much higher affinity binding of the IQ peptide to CaBP4(100 -271) compared with the isolated lobes suggested that both lobes of Ca 2ϩ -bound CaBP4 bind cooperatively to the IQ peptide in contrast to apo-CaM (47,48), in which only the C-lobe forms contacts with the IQ motif.
Structural Model of CaBP4 Bound to the IQ Motif-The relatively low solubility of the CaBP4-IQ complex has thus far hampered all efforts to directly solve the complex structure by NMR or x-ray crystallography. Instead, an experimentally guided computational approach was used to dock the NMR structures of Ca 2ϩ -bound CaBP4 (Fig. 4) onto a modeled helical structure of the Cav1.4 IQ motif, derived from the helical Cav1.2 IQ motif seen in previous crystal structures (49,50). The helical structure of the IQ peptide bound to CaBP4 was confirmed by circular dichroism (Fig. 7C). The docking calculation performed with HADDOCK (34 -36) was experimentally constrained with chemical shift perturbation data obtained by comparing NMR spectra of CaBP4 alone and in the presence of IQ peptide (Fig. 7A). The NMR resonances of CaBP4 that showed largest changes in chemical shift were assigned to exposed residues (Glu 121 , Leu 122 , Gly 123 , Phe 137 , Met 154 , Phe 186 , Thr 198 , Ala 199 , Leu 207 , Phe 214 , Leu 235 , Gly 236 , Met 251 , Glu 263 , Phe 264 , and Leu 268 ). Residues that affected this binding (see "Experimental Procedures") were selected as "active restraints" within HADDOCK to calculate an ensemble of docked structures (interface RMSD is 1.7 Å). A representative docked structure of CaBP4 bound to the IQ motif is shown in Fig. 8A.
Both lobes of Ca 2ϩ -bound CaBP4 formed separate contacts with Cav1.4 on opposite sides of the IQ helix (C-lobe (cyan) and N-lobe (light gray) in Fig. 8A). Exposed hydrophobic residues in the CaBP4 C-lobe (Leu 207 , Phe 214 , Met 251 , Phe 264 , Leu 268 ) interacted primarily with Ile 1592 and Tyr 1595 on the same face of the IQ helix. This is consistent with the I1592A/Y1595A mutation in Cav1.4 that significantly weakened CaBP4 binding (4). In the current study, the Cav1.4 mutation, Y1595E, nearly abolished binding of CaBP4 to Cav1.4 in pull-down assays (Fig. 7D). Also, 15 N-1 H HSQC spectra of Ca 2ϩ -bound CaBP4 remained unaffected by the addition of the Y1595E IQ mutant in contrast to large spectral changes caused by wild type IQ (Fig. 7A). Furthermore, mutating CaBP4 residues Met 251 , Phe 264 , and Leu 268 that contact Tyr 1595 (Fig. 8A) also weakened the binding to the IQ peptide. The CaBP4 mutants M251A, F264E, and L268A showed at least 2-fold weaker binding with a significant loss in binding entropy, consistent with the removal of a hydrophobic interaction for each mutant ( Table 2). The CaBP4 C-lobe bound to the Cav1.4 IQ helix had the same relative orientation as the CaM C-lobe bound to the Cav1.2 IQ helix (49,50) (Fig.  8B). The RMSD between main chain atoms of CaBP4 C-lobe versus CaM C-lobe in both complexes was 1.5 Å, suggesting that their overall main chain structures are fairly similar.
The CaBP4 N-lobe formed hydrophobic contacts and a salt bridge with the opposite face of the IQ helix (Fig. 8A). CaBP4 residue, Phe 137 contacted Phe 1586 on the IQ helix. The IQ residue, Arg 1597 formed electrostatic contacts with CaBP4 residue, Glu 168 . The IQ helix bound to the CaBP4 N-lobe had an opposite orientation compared with the Cav1.2 IQ helix bound to the CaM N-lobe (49,50) (Fig. 8B). We suggest that the opposite orientation for the IQ helix bound to the CaBP4 N-lobe versus the CaM N-lobe might explain why L-type channels bind more tightly to CaBP4 than CaM (6).
We show that CaBP4 binds to the Cav1.4 IQ motif with a collapsed structure of both lobes that appears similar to the collapsed structure of Ca 2ϩ -CaM bound to the Cav1.2 IQ (49, 50) (Fig. 8B). This similarity is consistent with the suggestion that CaBP4 binding to Cav1.4 can block CaM binding to prevent Ca 2ϩ -dependent inactivation (5) and allow Cav1.4 channels to remain open at high Ca 2ϩ levels in dark-adapted photoreceptor cells.

DISCUSSION
In this study, we determined the NMR solution structures of CaBP4 in both Mg 2ϩ -bound and Ca 2ϩ -bound conformational states and characterized the CaBP4 structural interaction with Cav1.4. The overall main chain structures of Mg 2ϩ -bound and Ca 2ϩ -bound CaBP4 (Figs. 4 and 6) are similar to those seen previously for CaBP1 (32). In the presence of physiological Mg 2ϩ levels (and in the absence of Ca 2ϩ ), CaBP4 has Mg 2ϩ bound at EF1 and EF3 ( Fig. 2A, inset). At saturating Ca 2ϩ levels, the N-lobe of CaBP4 adopts a closed conformation with Ca 2ϩ bound at EF1 (and no metal bound at EF2), in contrast to the Ca 2ϩ -bound open conformation seen in CaM (17) and TnC (51). The CaBP4 C-lobe binds Ca 2ϩ at EF3 and EF4 (Fig. 2B, inset) and adopts the familiar Ca 2ϩ -bound open conformation ( Fig. 4B) with an exposed hydrophobic patch (Fig. 5). Many of the exposed hydrophobic residues in CaBP4 (Leu 207 , Phe 214 , Met 251 , Phe 264 , and Leu 268 ) make contact with the IQ motif (Fig. 8A). The corresponding residues in CaM make similar contacts with IQ, and this may explain why CaBPs and CaM compete for binding to IQ motifs in voltage-gated Ca 2ϩ channels (52).
The Ca 2ϩ -bound closed conformation for the CaBP4 N-lobe with Ca 2ϩ -bound at EF1 (EF2 unoccupied; Fig. 4A) could prevent unwanted binding of low affinity molecules and play a role in enhancing target specificity in photoreceptors. The closed conformation of the Ca 2ϩ -bound CaBP4 N-lobe conceals hydrophic residues (Leu 130 , Leu 150 , and Phe 186 in Fig. 4A) that would otherwise be exposed in an open conformation. When CaBP4 binds to Cav1.4, we suggest that the target binding free energy may drive the closed Ca 2ϩ -bound N-lobe into a semiopen conformation that can more readily recognize the IQ helix (53). Indeed, the modeled structure of CaBP4 bound to IQ peptide shows that the N-lobe (with Ca 2ϩ bound at EF1) adopts a more opened conformation that exposes Phe 137 and Phe 186 to FIGURE 8. Structural comparison of Ca 2؉ -bound CaBP4 and Ca 2؉ -bound CaM bound to the IQ motif. A, structural model of the docked complex for CaBP4 (C-lobe (cyan) and N-lobe (light gray)) bound to Cav1.4 IQ motif (red). The docking calculation was performed using HADDOCK as described under "Experimental Procedures." B, structure of Ca 2ϩ -bound CaM (C-lobe (cyan) and N-lobe (light gray)) bound to Cav1.2 IQ motif (red) from (49). Side chain atoms in the exposed hydrophobic patch of CaBP4 and CaM are highlighted in yellow.
interact with Cav1.4 residue Phe 1586 (Fig. 8A). This target-induced opening of the N-lobe (which lacks Ca 2ϩ -binding at EF2) suggests that the target binding free energy can compensate for the lack of Ca 2ϩ binding at EF2. Thus, disabling Ca 2ϩ binding at EF2 in CaBP4 helps to ensure that the N-lobe remains closed in the absence of Cav1.4 to prevent binding of lower affinity molecules. A similar enhancement of target specificity is also seen for CaBP1 (42) and cardiac TnC (54), both of which have only one Ca 2ϩ bound to the N-lobe.
The Cav1.4 IQ motif (residues 1579 -1605) binds tightly to Ca 2ϩ -bound CaBP4 but not to Ca 2ϩ -free/Mg 2ϩ -bound CaBP4. Ca 2ϩ -dependent CaBP4 binding to IQ differs from the Ca 2ϩindependent binding of IQ motifs to CaM (47,48) and other EF-hand proteins (55). The Cav1.4 IQ peptide binds to fulllength CaBP4 with at least 10-fold higher affinity than IQ binding to the individual lobes. This positive cooperativity between the two lobes in CaBP4 is consistent with the collapsed structure of the two lobes that surround the IQ helix (Fig. 8). Similar cooperative lobe interactions are also seen in the crystal structures of Ca 2ϩ -bound CaM bound to the Cav1.2 IQ (49, 50) and other protein targets (56).
CaBP4 makes primarily hydrophobic contacts with the IQ motif (Phe 1586 , Tyr 1587 , Ile 1592 , and Tyr 1595 ). CaBP4 interacts most extensively with the Cav1.4 residue Tyr 1595 , which we show is essential for binding (Fig. 7, A and D). CaBP4 residues (Leu 207 , Met 251 , and Leu 268 ) that contact Tyr 1595 (Fig. 8A) are not conserved in CaM. These non-conserved hydrophobic contacts in the CaBP4-Cav1.4 complex could help explain why L-type channels bind more tightly to CaBP4 than to CaM (6).
Modulation of Cav1.4 by CaBP4 has been shown to be physiologically relevant by studies on knock-out mice and patients with retinal disease (1,14). CaBP4 knock-out mice show defects in photoreceptor synaptic function and organization similar to those in mice that lack Cav1.4. In addition, homozygous mutations in both CaBP4 and Cav1.4 are associated with congenital stationary night blindness 2 (11,57). In patients with pArg216X mutations, most of the CaBP4 C-lobe is absent, and this mutant form of CaBP4 cannot activate Cav1.4 (57). This is consistent with our finding that the CaBP4 N-lobe alone does not bind to the IQ motif (Fig. 7A). Patients with the c.800_801delAG deletion (lacking CaBP4 residues Glu 263 -Gly 271 ) show an impaired modulation of Cav1.4 by CaBP4, leading to deficient neurotransmitter release from photoreceptors (11). This effect of deleting the C-terminal residues (Glu 263 -Gly 271 ) is consistent with our structural model that indicates that Phe 264 and Leu 268 are important for contacting the IQ motif (Fig. 8A).
Based on our structure of the CaBP4-IQ complex (Fig. 8), we propose a schematic mechanism for Ca 2ϩ -dependent regulation of Cav1.4 channel activity in photoreceptor cells (Fig. 9). In dark-adapted rods, the cytosolic Ca 2ϩ concentration is maintained at high levels (38), which allows Ca 2ϩ -bound CaBP4 to bind the IQ motif in Cav1.4. We suggest that CaBP4 binding to IQ could prevent IQ association with ICDI. The Cav1.4 IQ motif was suggested to interact with ICDI and promote channel closure, because deletion of the ICDI causes channel opening (4). The ICDI was also suggested to interact with the EF-hand region (residues 1445-1493, highlighted in green in Fig. 9A) (5). We suggest that Ca 2ϩ -induced CaBP4 binding to IQ could disrupt the ICDI domain from interacting with both the IQ and EF-hand region and thus destabilize the Cav1.4 closed state at high Ca 2ϩ levels. Therefore, Ca 2ϩ -induced binding of CaBP4 to IQ is proposed to stabilize the Cav1.4 open state at high Ca 2ϩ levels so that the channels remain open in dark-adapted photoreceptors. Light activation of these photoreceptors produces a drop in cytosolic Ca 2ϩ (38) that in turn causes the Ca 2ϩ -free CaBP4 to dissociate from Cav1.4. We propose that light-induced dissociation of Ca 2ϩ -free CaBP4 would allow the IQ motif to interact with ICDI and promote channel closure upon light activation.