The Crystal Structure of the Primary Ca2+ Sensor of the Na+/Ca2+ Exchanger Reveals a Novel Ca2+ Binding Motif*

The Na+/Ca2+ exchanger is a plasma membrane protein that regulates intracellular Ca2+ levels in cardiac myocytes. Transport activity is governed by Ca2+, and the primary Ca2+ sensor (CBD1) is located in a large cytoplasmic loop connecting two transmembrane helices. The binding of Ca2+ to the CBD1 sensory domain results in conformational changes that stimulate the exchanger to extrude Ca2+. Here, we present a crystal structure of CBD1 at 2.5Å resolution, which reveals a novel Ca2+ binding site consisting of four Ca2+ ions arranged in a tight planar cluster. This intricate coordination pattern for a Ca2+ binding cluster is indicative of a highly sensitive Ca2+ sensor and may represent a general platform for Ca2+ sensing.

The Na ؉ /Ca 2؉ exchanger is a plasma membrane protein that regulates intracellular Ca 2؉ levels in cardiac myocytes. Transport activity is governed by Ca 2؉ , and the primary Ca 2؉ sensor (CBD1) is located in a large cytoplasmic loop connecting two transmembrane helices. The binding of Ca 2؉ to the CBD1 sensory domain results in conformational changes that stimulate the exchanger to extrude Ca 2؉ . Here, we present a crystal structure of CBD1 at 2.5 Å resolution, which reveals a novel Ca 2؉ binding site consisting of four Ca 2؉ ions arranged in a tight planar cluster. This intricate coordination pattern for a Ca 2؉ binding cluster is indicative of a highly sensitive Ca 2؉ sensor and may represent a general platform for Ca 2؉ sensing.
Rapid fluxes of Ca 2ϩ across the sarcolemmal membrane are an important component of cardiac excitation-contraction coupling. Ca 2ϩ influx mediated by voltage-dependent Ca 2ϩ channels initiates contractions, while Ca 2ϩ efflux is dominated by the Na ϩ /Ca 2ϩ exchanger (1). Thus, the Na ϩ /Ca 2ϩ exchanger is an important component of regulation of cardiac contractility. Under most physiological conditions, the exchanger uses the energy stored in the inwardly directed Na ϩ gradient to catalyze the extrusion of Ca 2ϩ from the cell with a stoichiometry of 3 Na ϩ for 1 Ca 2ϩ .
Activity of the Na ϩ /Ca 2ϩ exchanger is modulated by the binding of Ca 2ϩ to a high affinity regulatory site on an intracellular portion of the protein. Regulatory Ca 2ϩ is not transported but potently activates exchange activity. Recent evidence suggests that Ca 2ϩ may bind to and dissociate from its regulatory site during the rapid Ca 2ϩ fluctuations that occur during a cardiac contraction cycle (2). The Na ϩ /Ca 2ϩ exchanger protein is predicted to consist of nine transmembrane segments and a large intracellular loop (3,4). The transmembrane segments translocate ions across the membrane, and the intracellular loop is largely responsible for regulation of activity. We have previously identified a region of the intracellular loop of the exchanger (amino acids 371-508) that binds Ca 2ϩ with high affinity and mediates activation of exchange activity by Ca 2ϩ (5,6). This segment comprises the first of two tandem Calx-␤ domains (7). Mutational analysis identified two groups of three aspartate residues within the first Calx-␤ domain that were associated with the binding of Ca 2ϩ (5,6). The binding of Ca 2ϩ to the regulatory site induces substantial conformational changes that presumably mediate regulatory function (2,6,8,9).
A recent major development in the understanding of Ca 2ϩ regulation has been the determination of the structure of the Ca 2ϩ binding region of the large intracellular loop using NMR techniques (9). Two Ca 2ϩ binding domains (CBD1 and CBD2) were identified that correspond to Calx-␤1 and -␤2. CBD1 encompasses the same region that we had identified as being responsible for Ca 2ϩ regulation. Binding of Ca 2ϩ to CBD1 induces a substantial conformational change consistent with earlier studies. In the presence of Ca 2ϩ , both CBD1 and CBD2 have an immunoglobulin fold. CBD2, in the adjoining Calx-␤ repeat region, binds Ca 2ϩ with substantially lower affinity and its functional role is unclear. Unlike CBD1, the removal of Ca 2ϩ from CBD2 does not induce protein unfolding.
The NMR structure of CBD1 shows a classical immunoglobulin fold with two Ca 2ϩ ions bound in the distal loops (9). However, the heteronuclear single quantum correlation spectra employed by Hilge and colleagues does not directly visualize the presence of Ca 2ϩ but rather infers positions from Yb 3ϩinduced shifts. Here, we describe the crystal structure of CBD1 using x-ray techniques. Like the NMR structure, we find an immunoglobulin fold, and the two structures superimpose well. Strikingly, the x-ray structure reveals the presence of four Ca 2ϩ ions bound in a unique cluster with important physiological consequences.
Crystal Growth and Structure Determination-Purified CBD1 protein was maintained in a solution of 10 mM Tris-HCl, pH ϭ 7.4, ϩ 0.2 mM EGTA at a concentration of 25 mg/ml. This solution was screened against 480 commercially available crystallization conditions with the mosquito crystallization robot (TTP Labtech) using the hanging drop vapor diffusion technique. Crystals were obtained at 20°C in condition number 35 of Hampton Research's Crystal Screen 2 (100 mM HEPES, pH ϭ 7.5, ϩ 70% 2-methylpentane-2,3-diol). These crystals were then optimized by addition of 100 mM guanidine discovered through additive screening using Hampton Research's Additive Screen in conjunction with the Mosquito robot. The resulting crystals diffracted to 2.5 Å resolution (see Table 1 of supplemental material).
Data were collected from a cryo-cooled crystal at beamline 8.2.2 of the Advance Light Source (Berkeley, CA). The crystal belongs to the space group P2 1 2 1 2 with cell dimensions of a ϭ 59.6 Å, b ϭ 45.5 Å, and c ϭ 57.3 Å. Image data were processed using the programs DENZO and SCALEPACK (10). The structure of CBD1 was phased by molecular replacement using the program PHASER (11). The coordinates of the recent NMR structure of CBD1 (PDB accession code 2FWS) were used for the search model. The structure was built using the program COOT (12) and refined using CNS (13) and REFMAC (14) with a final R and R free of 22.2 and 28.4%, respectively.

RESULTS AND DISCUSSION
We sought to uncover the principles underlying Ca 2ϩ regulation of the NCX by resolving the crystal structures of the primary Ca 2ϩ binding domain (CBD1) in the Ca 2ϩ -bound and Ca 2ϩ -free conformations. Initial crystallization trials in the presence of 2 mM CaCl 2 (Ca 2ϩ -bound) and 2 mM EGTA (Ca 2ϩfree) failed. To minimize the impact of these reagents on crystallization, we reduced their concentrations to 0.2 mM. An EGTA-containing sample yielded crystals diffracting to 2.5 Å. The crystal structure had a strong resemblance to the NMR structure (9) maintaining the overall immunoglobulin fold. In addition, the positions of four tightly clustered Ca 2ϩ ions were revealed. Further analysis confirmed a contamination of 0.12 mM Ca 2ϩ in condition number 35 of Hampton Research's Crys-tal Screen 2, which inadvertently led to the Ca 2ϩ -bound structure.
Structure Overview-The NMR and crystal structures were superimposed with a root mean square difference of 1.8 for 128 C ␣ atoms (Fig. 1). The overall positional alignment between the two structures coincides well including the notable ␤-bulge and cis-proline that disrupt the A and G ␤-strands, respectively. The striking new feature of the crystal structure is the presence of a novel Ca 2ϩ binding site situated in the distal loops of the ␤-sandwich containing four Ca 2ϩ ions coordinated by an extensive network of amino acids residues. The previously reported NMR structure showed two Ca 2ϩ ions, which approximately represent a positional average of those observed in the crystal structure (Fig. 1). This newly observed Ca 2ϩ binding motif was only revealed by x-ray crystallography and will provide a framework for further biochemical and mutational analysis.
There had not previously been any indication that four Ca 2ϩ ions were present in the Ca 2ϩ regulatory domain. Ca 2ϩ binding data had suggested the binding of two Ca 2ϩ ions per regulatory domain (8). Hill coefficients have been variable for binding and functional effects of Ca 2ϩ . Values include 0.9 (5), 1.4 (15), and 2.9 (2) consistent with the involvement of multiple Ca 2ϩ ions, although the source of the variability is unclear.
CBD1 is arranged in a classical immunoglobulin fold, where the ␤-sandwich motif is formed by two antiparallel ␤-sheets consisting of strands A-B-E and strands D-C-F-G (Fig. 2a). The presence of a ␤-bulge in strand A disrupts the antiparallel hydrogen bonding pattern between strands AЈ and B. Following the ␤-bulge, strand AЈ associates with strand GЈ from the opposing sheet, rather than resuming its interactions with strand B (Fig. 2c). Additionally, there is a cis-proline residue that induces an abrupt loop in the middle of strand G, but unlike strand A, strand G resumes a normal hydrogen bonding pattern with strand F. These geometrical distortions are often observed in external strands A and G of immunoglobulin folds (16,17) and have been suggested to be protective in preventing aggregation between multiple immunoglobulin domains by disrupting potential intermolecular hydrogen bonding surfaces (18). This suggestion seems particularly relevant based on the model presented by Hilge et al. (9), predicting that the high affinity Ca 2ϩ sensor (CBD1) and the low affinity Ca 2ϩ sensor (CBD2) form a heterodimer stacked along the A-G interface.
The coordinates for CBD1 were compared against other three-dimensional structures using the distance matrix alignment server (Dali) (19) revealing a number of structural homologues including fibronectins, cadherins, and integrins. Although there appears to be no apparent sequence identity or functional similarities, members of the immunoglobulin fold family share a common core structure (16), which is one of the most prevalent domains encoded by the human genome (20).
Ca 2ϩ Coordination-The striking difference between the crystal and NMR structures is at the Ca 2ϩ binding region. Hilge and colleagues (9) were able to assign the positions for two Ca 2ϩ ions by using a three prong approach, which included the recording of pseudo-contact shift data, obtaining spectra from the sample in the presence of Yb 3ϩ ions and utilizing biochemical and mutagenesis data for distance constraints. However, the crystal structure revealed an extensive coordination scheme connecting four Ca 2ϩ ions clustered in the distal loops of the ␤-sandwich. It appears that the two Ca 2ϩ sites predicted in the NMR structure represent a positional average of those observed in the crystal structure (Fig. 1). The four binding sites are arranged in a parallelogram-like configuration, where the distances between Ca 2ϩ sites 1 and 2, 2 and 3, and 3 and 4 are 4.27, 4.30, and 3.93 Å, respectively (Fig. 2, a and b). These binding sites are primarily coordinated by aspartic and glutamic acid residues forming polydentate interactions, often between two or three Ca 2ϩ ions.
The majority of the residues involved in coordinating the Ca 2ϩ ions are located at the C terminus (Asp 498 , Asp 499 , Asp 500 ) and in loop E-F (Asp 446 , Asp 447 , Ile 449 , Glu 451 , Glu 454 ). Additional interactions occur with Glu 385 in the A-B loop, Asp 421 in the C-D loop, and three water molecules. The overall coordination scheme for each Ca 2ϩ site is summarized in Table 2 of the supplemental material. In short, Ca1 and Ca4 are penta-coordinated, while Ca2 and Ca3 are hexa-and hepta-coordinated, respectively. Glu 451 , Asp 421 , and Asp 500 coordinate multiple Ca 2ϩ ions and appear to be the key residues in forming a tight binding cluster of four Ca 2ϩ ions. Glu 451 is centrally located coordinating Ca1, Ca2, and Ca3. Asp 421 coordinates both Ca1 and Ca2, while Asp 500 coordinates Ca3 and Ca4. These three residues appear to orient the four Ca 2ϩ ions into a tight binding cluster. Although never previously observed, a similar arrangement of a four Ca 2ϩ ion binding cluster has been predicted for another Ca 2ϩ sensor domain, the C 2 domains of synaptotagmin I and phospholipase C (21).
Two acidic segments, each characterized by three consecutive aspartic acid residues (498 -500 and 446 -448), were previously suggested to be Ca 2ϩ binding regions (6); mutations in residues Asp 447 , Asp 448 , Asp 498 , and Asp 500 each result in an apparent 3-fold decrease in Ca 2ϩ affinity (2). Additionally, a recent mutation, E454K, showed an 8-fold decrease in Ca 2ϩ affinity (9). We directly visualize three residues (Asp 446 , Ile 449 , Asp 499 ) and three water molecules that are ligands for Ca 2ϩ , which are not part of the Ca 2ϩ binding structure in the NMR study (9). Conversely, Hilge et al. (9) place Asp 448 as a Ca 2ϩ ligand, but we find that this residue is not directly involved in the binding of Ca 2ϩ . In total, the Ca 2ϩ binding region is tightly regulated through a complex coordination scheme composed mostly of carboxylate moieties.
Comparison with Other Ca 2ϩ Binding Proteins-Analysis of sequence and structural data has revealed a number of protein modules that are widespread and repeated throughout nature (22,23). These protein modules facilitate the regulation of numerous proteins that vary dramatically in function and impact multiple cellular processes. Analysis of the human genome revealed a number of Ca 2ϩ binding modules (24). The binding of Ca 2ϩ to proteins has a variety of roles. These include enhancing protein stability (25,26) and inducing conformational changes to facilitate secondary actions as seen with calmodulin (27)(28)(29) and other Ca 2ϩ sensors (30). CBD1 forms a unique binding cluster that may be utilized by other Ca 2ϩ sensor proteins.
We note sequence and structural similarities between the CBD1 domain and the larger family of C 2 domains. C 2 domains FIGURE 1. Structural alignment of the crystal and NMR structures of CBD1. The NMR structure (yellow) was superimposed onto the crystal structure (blue) with a root mean square deviation of 1.8 Å for 128 C ␣ atoms. The Ca 2ϩ ions are represented as spheres maintaining the same color code. The two Ca 2ϩ ions from the NMR structure (yellow) seem to represent a positional average of the four Ca 2ϩ ions seen in the crystal structure (blue).
are the second most abundant Ca 2ϩ binding module present in nature (24). The majority of proteins with C 2 domains are involved in signal transduction or membrane trafficking (31). The two C 2 domains that are most extensively studied on a structural level are those of synaptotagmin (32,33) and phospholipase C (34), both of which form an eight-stranded ␤-sandwich. The ␤-sandwich scaffold permits variable loops that are widely separated in the primary sequence to facilitate the binding of multiple Ca 2ϩ ions in a cluster. Similar to CBD1, the binding sites are comprised primarily of aspartic acid residues forming polydentate interactions between two or three Ca 2ϩ ions. Sequence alignments (Fig. 3) between CBD1 and C 2 domains show a number of similarities around the first acidic segment. However, the existence of a fourth Ca 2ϩ site, as found in CBD1, would require additional acidic coordinating residues not seen in C 2 domain structures. As seen in the current struc- . Sequence alignment between CBD1 and a representative set of C 2 domains positioned around the two acidic segments. The two acidic segments located in CBD1 are underlined in green, and the residues coordinating Ca 2ϩ are shown in yellow background. Residues whose backbone carbonyl groups coordinate Ca 2ϩ ions are shown on a blue background. The blue arrows represent ␤-strands E-F-G (CBD1), 6-7-8 (SYN1) and 5-6-7 (PLC1). CBD1 is the sequence of the primary Ca 2ϩ binding sensor of NCX (canine); SYNI is the sequence of the C 2 A domain of synaptotagmin I (rat); PLCI is the C 2 domain of phospholipase C (rat). FIGURE 2. Structure of CBD1. a, ribbon representation of CBD1. The seven ␤-strands are colored from the N terminus (N) in blue to the C terminus (C) in red in the same orientation as in Fig. 1. The four Ca 2ϩ ions are depicted as green spheres. b, stereo view of the Ca 2ϩ binding sites. The main chain is represented in blue. The four Ca 2ϩ ions and three water molecules are colored as green and red spheres, respectively. The side chain carbons and oxygens are yellow and red, respectively. Coordination to the Ca 2ϩ ions is represented by black dashed lines. c, secondary structure schematic of CBD1. The ␤-strands are depicted as blue arrows labeled from A to G, and the residues involved in Ca 2ϩ binding are shown in red circles. The A and G strands are disrupted by a ␤-bulge and a cis-proline, respectively, resulting in strands AЈ and GЈ. The red box around AЈ indicates a break in hydrogen bonding arrangement, which results in a parallel alignment with strand GЈ. ture, these additional residues are located toward the C terminus in the second acidic segment but there is no structural or sequence similarity for this region in C 2 domains. Although functionally diverse, the CBD1 and the C 2 domains share a common Ca 2ϩ coordination scheme that may be general for Ca 2ϩ sensing.
The crystal structure of CBD1 reveals a new Ca 2ϩ binding motif consisting of four Ca 2ϩ ions arranged in a tight cluster. This coordination scheme utilizes carboxylate moieties from aspartic and glutamic acid residues to form polydendate interactions with multiple Ca 2ϩ ions. This unique cluster facilitates the reversible binding of Ca 2ϩ in an environment where the concentration of free Ca 2ϩ is kept low. Further biochemical and mutational analysis based on the crystal structure and structure determinations of other components of the cytosolic loop will facilitate our understanding of the sensory mechanism of NCX.