Solution Structure of the Cytoplasmic Region of Na+/H+ Exchanger 1 Complexed with Essential Cofactor Calcineurin B Homologous Protein 1*

Na+/H+ exchanger 1 (NHE1) regulates intracellular pH, Na+ content, and cell volume. Calcineurin B homologous protein 1 (CHP1) serves as an essential cofactor that facilitates NHE1 exchange activity under physiological conditions by direct binding to the cytoplasmic juxtamembrane region of NHE1. Here we describe the solution structure of the cytoplasmic juxtamembrane region of NHE1 complexed with CHP1. The region of NHE1 forms an amphipathic helix, which is induced by CHP1 binding, and CHP1 possesses a large hydrophobic cleft formed by EF-hand helices. The apolar side of the NHE1 helix participates in extensive hydrophobic interactions with the cleft of CHP1. We suggest that helix formation of the cytoplasmic region of NHE1 by CHP1 is a prerequisite for generating the active form of NHE1. The molecular recognition detailed in this study also provides novel insight into the target binding mechanism of EF-hand proteins.

Na ϩ /H ϩ exchangers comprise a family of countertransport proteins that catalyze the electroneutral exchange of Na ϩ and H ϩ . Nine isoforms of the Na ϩ /H ϩ exchanger have been isolated and shown to possess similar membrane topologies consisting of 12 N-terminal membrane-spanning helices and a large C-terminal cytoplasmic region (Fig. 1A). The exchanger isoforms exhibit tissue-specific expression, membrane localization, and kinetic and pharmacological properties (1). They participate in a broad range of physiological processes including the regulation of cell volume, transepithelial transport of electrolytes, cell proliferation, apoptosis, and differentiation.
Isoforms NHE1-5, localized at the plasma membrane, are primarily involved in the regulation of intracellular pH (pH i ) 2 and Na ϩ concentration (1).
Among these factors, CHP1 can serve as an essential cofactor and is required by at least three NHE isoforms (NHE1-3) to express high physiological levels of exchange activity (3). It was shown that CHP1 bound directly to the juxtamembrane region of the C-terminal cytoplasmic domain. When GFP-CHP1 and NHE1-3 were co-expressed, it was found that GFP-CHP1 was mostly localized at the cell surface, whereas co-expression of CHP1 and a CHP1 binding-defective NHE1 mutant failed to show co-localization, implying that NHE1 is a principal target of CHP1 (3). In addition to reduced activity in the neutral pH range, the CHP1 binding-defective NHE1 mutant showed a marked reduction in pH i sensitivity (ϳ0.7 pH unit acidic shift) that subsequently abolished various NHE1 regulatory responses. Furthermore CHP1 deprivation resulted in marked reduction (Ͼ90%) of NHE1 activity (3). These observations suggest that the association of NHE1 with CHP1 is critical for activity and the maintenance of NHE1 pH i sensitivity (14).
CHP1 consists of four EF-hands, the primary sequence of which is homologous to calmodulin (CaM) and calcineurin B (CNB), possessing 31 and 41% sequence identity, respectively (Fig. 1B). It is well known that all CaM and CNB EF-hands can bind Ca 2ϩ . However, CHP1 EF-1 and EF-2 are ancestral and do not bind Ca 2ϩ under physiological conditions, whereas EF-3 and EF-4 bind two Ca 2ϩ ions with high affinity (ϳ90 nM) based on the 45 Ca 2ϩ binding experiments for several CHP1 mutants (14). Complex formation between CHP1 and the CHP1 binding domain of NHE1 resulted in a marked increase in Ca 2ϩ binding affinity (K d ϭ ϳ2 nM) (14). This suggests that CHP1 constitutively contains two Ca 2ϩ ions when associated with NHE1 in cells (14).
A second CHP isoform, CHP2, with 61% sequence identity was also identified and found to be involved in the maintenance of abnormally high pH i in malignantly transformed cells. CHP2 is expressed at a relatively high level in malignantly transformed cells and in rat small intestine, suggesting that it plays a specific role in this tissue (20). In addition, tescalcin, an EF-hand protein closely related to CNB, that interacts with the cytoplasmic region of NHE1 has been identified (21,22).
The crystal structure of NHE1-unbound rat CHP1 was recently determined and revealed that the overall structure is similar to CNB where Ca 2ϩ ions are coordinated within EF-3 and EF-4. However, the interaction mechanism between NHE1 and CHP1 remains unknown (23).
Here we report on the solution structure of the cytoplasmic region of NHE1 bound to CHP1 as determined by NMR. Details of the NHE1-CHP1 interaction are described. We present mutational binding data to delineate the significance of the interactions observed in the complex. Based on the structure, we suggest a role for CHP1 in terms of NHE1 activation. Comparisons of the binding mode between NHE1-CHP1 and calcineurin A (CNA), CNB, and other related four-EF-hand proteins are also discussed.

EXPERIMENTAL PROCEDURES
Sample Preparation-The NHE1-CHP1 complex was co-expressed and co-purified. DNA encoding NHE1 was cloned into the pET24a vector (Novagen), and CHP1 was subcloned into pET11a (Novagen), which produces recombinant protein with a hexahistidine (His 6 ) sequence at the C terminus. The proteins were co-overexpressed in Escherichia coli BL21(DE3) cells. Uniformly 15 N-and 15 N/ 13 C-labeled proteins were prepared by B, alignment of CHP1, CHP2, and human calcineurin B. In A and B, sequence alignment was performed using ClustalW. Secondary structure elements of the proteins are shown schematically at the top of the alignments. N-and C-terminal domains of CHP1 are colored in blue and magenta, respectively, for clarity. Conserved and semiconserved residues are colored in yellow and green, respectively. The 12-residue motif involved in the EF-hand is underlined, and key residues are indicated as X, Y, Z, -Y, -X, and -Z. growing bacteria in minimal medium containing 15 NH 4 Cl with or without [ 13 C 6 ]glucose. Uniformly 15 N/ 13 C-labeled and fractionally deuterated protein sample was prepared using medium containing 60% 2 H 2 O. The NHE1-CHP1 complex was purified using a standard Ni-NTA affinity column protocol (Qiagen). Further purification was performed by gel filtration using Superdex 200 (GE Healthcare). NMR samples contained 0.5-0.9 mM protein in 50 mM Tris-d 7 buffer (pH 6.9), 1 mM dithiothreitol-d 10  NMR Spectroscopy-NMR data were recorded at 37°C on Bruker AVANCE 500 and DRX 800 NMR spectrometers. Resonance assignments for 1 HN, 15 N, 13 C␣, 13 C␤, and 13 CЈ nuclei for the CHP1-NHE1 complex were obtained through the following 2 H-decoupled, triple resonance spectra applied to a fractionally deuterated 15 N/ 13 C-labeled sample: three-dimensional HNCACB/HN(CO)CACB (24) and three-dimensional HN(CA)CO/HNCO experiments (25,26). Side-chain 1 H and 13 C resonances were assigned on a fully protonated sample or a fractionally deuterated 15 N/ 13 C-labeled sample using three-dimensional C(CO)NH, three-dimensional H(CCO)NH, four-dimensional HC(CO)NH, and three-dimensional HCCH TOCSY (27)(28)(29). Stereospecific assignment of leucine and valine methyl groups were obtained from a Constant Time-1 H/ 13 C HSQC spectrum of a 15% 13 C-labeled sample (30). 1 H␣ and 1 H␤ resonance assignments were supplemented with three-dimensional H(CACO)NH (31), 15 N-edited TOCSY-HSQC (32), and threedimensional HBHA(CBCACO)NH (33). Aromatic resonances of both NHE1 and CHP1 were mainly assigned using threedimensional 13 C-aromatic-edited/ 15 N-separated NOESY-HSQC, three-dimensional 13 C-edited NOESY, 1 H-1 H TOCSY, and NOESY experiments (32). The assignment was verified using the Constant Time-1 H/ 13 C HSQC spectrum of a 15% 13 C-labeled sample (30). The following NOESY spectra were recorded for the protonated sample and used to generate distance restraints for structure calculations: three-dimensional 15 N-edited NOESY (80-ms mixing time), three-dimensional 13 C-edited NOESY (80-ms mixing time), and two-dimensional 1 H NOESY (80-ms mixing time). Slowly exchanging amide protons were identified from a series of two-dimensional 15 N HSQC spectra recorded after the H 2 O buffer was replaced with 2 H 2 O buffer. All NMR spectra were processed using NMRPipe/NMRDraw (34) analyzed using SPARKY (35).
Structure Calculations-Intramolecular and intermolecular distance constraints were identified in the three-dimensional 15 N/ 13 C-separated NOESY spectra using a 15 N/ 13 C-labeled NHE1-CHP sample with mixing times of 80 ms. Backbone hydrogen bond restraints within regular secondary structure elements that were consistent with backbone amide hydrogen/ deuterium exchange data were included in the structure calculations. The initial structure calculations were performed by iterative automated assignment of the NOE spectra using CANDID (36) in addition to manually assigned NOE-derived distance restraints. The restraints, leading to converged structures, were subsequently utilized for the iterative automated assignment of all spectra including aromatic residues using CANDID. Finally refinement of the structures (including two Ca 2ϩ ions) using XPLOR-NIH (version 2.96) was performed (37). The final structure calculations used a total of 4022 NOE-derived distance restraints obtained from the manual and the CANDID-assisted assignments from the 15 N-or 13 C-edited NOE data. A total of 100 simulated annealing structures were calculated, and 20 structures were selected that possessed no NOE violations greater than 0.5 Å and no dihedral violations greater than 5°. Final structures were evaluated using the program ProcheckNMR (38). Structures and figures were drawn using MOLMOL (39), GRASP (40), and Chimera (41).
Mutagenesis and GST Pulldown Assay-For the binding analyses, CHP1 was expressed as a fusion protein with GST in E. coli, which was then subcloned into modified pGEX6P-3 (Novagen) (42). Site-directed mutant proteins were prepared using the QuikChange kit (Stratagene). DNA sequencing confirmed the mutations. Vectors were transformed into BL21(DE3)star (Invitrogen). Cells were grown at 37°C and then induced with 1 mM isopropyl 1-thio-␤-D-galactopyranoside for 12 h at 20°C. Harvested cells were disrupted via sonication in HEPES (pH 8.0) containing 10% glycerol, 10% sucrose, 1 mM dithiothreitol, and 30 mM KCl. The GST-CHP1 fusion protein was purified using glutathione-Sepharose (GE Healthcare) and a standard protocol except that the equilibrium buffer was changed to 50 mM HEPES buffer (pH 8.0) containing 10% glycerol, 1 mM dithiothreitol, and 30 mM KCl, and the column was washed extensively with 20 mM CHAPS. The effect of mutation on the binding of CHP1 to NHE1 peptides was characterized using a GST pulldown assay. Synthetic NHE1 peptide (residues 514 -545 including a hexahistidine sequence at the C terminus) was purchased from Greiner Japan (Tokyo, Japan). Briefly wildtype and CHP1 mutant GST fusion proteins, in addition to GST alone, were incubated for 30 min at 20°C with NHE1 peptide in binding buffer containing 50 mM HEPES (pH 7.5), 20 mM CHAPS, 10% glycerol, 1 mM Pefabloc, and 1 mM dithiothreitol. GST protein-bound Sepharose beads were washed extensively with binding buffer. Proteins were resolved by 12% NuPAGE (Invitrogen) and blotted onto membranes, and then His-NHE1 was analyzed with Ni-NTA-conjugated alkaline phosphatase (Promega) and Western Blue substrate (Promega). Quantification was represented as the average value of experiments performed in triplicate.

RESULTS AND DISCUSSION
Structure Determination-To better understand the mechanism pertaining to CHP1-regulated NHE1 activity, the solution structure of unmyristoylated CHP1 complexed with the cytoplasmic region (503-545) of NHE1 was determined by NMR spectroscopy. Our structural studies were initially hampered by the fact that NHE1-free CHP1 tended to aggregate during NMR measurements, and the CHP1-unbound cytoplasmic region (503-545) of NHE1 readily degraded during the expression and purification steps. Consequently NMR structural analysis of coexpressed and co-purified samples was undertaken. Co-expression of CHP1 and NHE1-(503-545) produced a stable complex for structural studies and showed no significant degradation and aggregation for several weeks.
The 1 H-15 N HSQC spectrum of NHE1-unbound CHP1 displays many broadened peaks presumably due to formation of a dimer, multimer, or an equilibrium between these states in solution. In contrast, the HSQC spectrum of 15 N-labeled CHP1 complexed with NHE1 is well dispersed with favorable line shapes (Fig. 2), suggesting that the complex essentially adopts an ordered monomeric structure in solution.
Our target complex was ϳ27 kDa in size, assuming a 1:1 complex of CHP1 (22 kDa) and NHE1 (5 kDa). This represented a relatively large molecular weight in terms of conventional NMR studies. Consequently utilization of triple labeling ( 2 H, 13 C, and 15 N) and the recently developed computational methodology, CANDID, was extremely helpful in the structure determination. Sequential backbone assignments and most side-chain assignments were obtained from a 60% 2 H/ 15 N/ 13 C-labeled sample using standard triple resonance experiments. Missing 1 H resonances were supplemented using heteronuclear three-dimensional NOESY experiments with 15 N/ 13 C-labeled samples. Resonance assignments of methyl groups were carefully confirmed in a stereospecific manner using two-dimensional Constant Time HSQC spectra recorded for a 15% randomly enriched 13 C sample. Methyl groups for 18 of 24 leucines and nine of 11 valines were stereospecifically assigned. Aromatic ring proton assignments, essential for delineating hydrophobic core and protein-protein interactions, were obtained using twodimensional TOCSY, two-dimensional NOESY, two-dimensional HCCH(arom) TOCSY, and three-dimensional 13 C(arom)-edited 15 N-separated NOESY experiments. NMR spectra including the three-dimensional 13 C-edited NOESY spectrum used to monitor inter/intramolecular 1 H-1 H NOEs were of adequate quality to pursue a structural determination of the NHE1-CHP1 complex. Use of partial deuteration and the almost complete resonance assignment of methyl groups forming the hydrophobic core facilitated an initial determination of the overall protein fold. A high resolution structure was subsequently obtained using CANDID for automated assignments, which included the use of ambiguous NOEs from 15 N-and 13 C-edited NOESY experiments recorded for 15 N-or 15 N/ 13 C-labeled protein samples. An iterative approach was used for assigning NOEs in addition to the manually assigned unambiguous NOEs. The solution structure of the CHP1-NHE1 complex was determined from a total of over 4000 NMR-derived restraints, including 134 intermolecular distance restraints ( Table 1). The ensemble of 20 structures in excellent agreement with a large body of experimental data were well defined (Fig. 3A). The r.m.s. deviations of backbone and heavy atoms over residues 518 -537 of NHE1 and residues 10 -92 and 108 -193 of CHP1 were 0.53 and 1.15 Å, respectively. Of the NMR structures determined, the one with the smallest total energy was selected as representative for further discussion. The complex is predominantly ␣-helical, and the CHP1 helices constitute a cleft. A helix of the cytoplasmic region of NHE1 associates with CHP1 in 1:1 stoichiometry via the cleft (Fig. 3, A and B).

PROCHECK Ramachandran plot (185-254)
Residues in most favored regions (%) 78.6 Residues in additional allowed regions (%) 18 Fig. 3A). The N-terminal half of the helix (residues 518 -530) binds to the C-terminal domain of CHP1, and the C-terminal half of the helix (residues 531-537) binds to the N-terminal domain of CHP1 (Fig. 4A). Side-chain conformations of the helix are also well defined particularly for apolar residues that make extensive contacts with CHP1. For example, NMR spin-echo difference 3 J NC␥ and 3 J CЈC␥ experiments, which bring about 1 rotamer information of aromatic side chain, showed that the His-523 and Phe-526 adopted gϩ and t conformations, respectively. The helix exhibits amphipathic character in which the bulky hydrophobic residues Ile-518, Ile-522, His-523, Phe-526, Leu-527, Leu-530, Leu-531, Ile-534, and Ile-537 are clearly confined to one side, and hydrophilic residues are exposed at the other side (Fig. 4B). The hydrophobic residues form a continuous apolar surface (Fig. 4B). The main-chain and side-chain conformations of residues preceding and following the helix, residues 503-517 and 538 -545, respectively, are poorly defined in the NMR structure because of the absence of medium and long range NOEs involving these regions. The narrow resonance linewidths, chemical shift index, and steady state { 1 H}-15 N heteronuclear NOE suggest that these regions are unstructured in the complex.
The four CHP1 EF-hands form a deep hydrophobic pocket, which constitutes the interaction surface for the NHE1 amphipathic ␣-helix. CHP1 binds to the apolar side of NHE1 with the four EF-hands through a side-by-side manner (Fig.  4A). This contrasts with the well known canonical CaM-target binding mode that represents a wrap-around manner in which two pairs of EF-hands bind to the target IQ motif helix on opposite sides to each other (43,44).
Although there is modest sequence similarity between CHP1 and CaM, it should be noted that the latter interacts with a large number of proteins with various interaction modes including canonical 1:1 binding and non-canonical 1:1, 1:2, and 2:2 binding (43,44). It has been suggested that the observed binding versatility of CaM could be derived from the variable positioning of the two domains, linked by a flexible linker, that can accommodate different targets. Although target multiplicity has been reported in the case of CHP1, it seems to possess more limited binding modes than CaM. One notable feature of CHP1 that illustrates its difference to CaM has to do with the interdomain interaction. This interaction restricts domain orientation, a phenomenon absent in CaM. In our CHP1 structure, residue Leu-62 of the N-terminal domain participates in a hydrophobic interaction with residues Val-138 and Met-141 of the C-terminal domain, and residue Ala-69 of the N-terminal domain interacts with residue Leu-122 of the C-terminal domain. A side-byside interaction mediated by four EF-hands has also been reported to take place with voltage-gated potassium channel (Kv)-interacting protein (KChIP) (45) and with CNA-CNB (46, 47) (Fig. 5A).
The atomic r.m.s. difference of well fitted parts between CHP1 complexed with NHE1 and CNB complexed with CNA (Protein Data Bank code 1AUI) is 2.7 Å, indicating that the fit is not very good, although the topology is identical with a high Z-score of 11.8 from a distance matrix alignment (DALI) search. The r.m.s. deviation value improves to 1.7 Å when only the N-terminal domains are superimposed, and it is 1.8 Å when only the C-terminal domains are superimposed. This indicates that the higher r.m.s. deviation for both domains originates from an interdomain swiveling between NHE1-CHP1 and CNA-CNB, although both proteins bind cognate targets in a side-by-side manner. Similarly domain swiveling was observed between NHE1-bound and NHE-free CHP1. Swiveling of the N-and C-terminal domains could create a binding surface for cognate targets.
Comparison with Other EF-hands-It is interesting to note that both EF-1 and EF-2 adopt an open conformation in NHE1-bound CHP1 without Ca 2ϩ . This is especially evident when comparing the angles between the EF-hand helices. In Fig.  5B, a graphical view of the interhelical angles between the first and second helices of EF-1 and EF-2 of apoCaM (a typical closed conformation), Ca 2ϩ -CaM (a typical open conformation, CNB), KChIP1, NHE1-bound CHP1, and NHE1free CHP1 is displayed using a vector geometry mapping method (48,49). The first helices of the EFhands are superimposed along the z axis, and the spatial localization of the second helices are shown as a cylinder. This indicates that both EF-1 and EF-2 of NHE1-bound CHP1 adopt an open conformation, whereas EF-1 and EF-2 of NHE1free CHP1 adopt an open and semiopen conformation, respectively ( Fig. 5B and supporting information S1). This implies that EF-1 and EF-2 adopt a constitutively open conformation. However, it should be noted that the hydrophobic cleft of the N-terminal domain of free CHP1, as revealed by the crystal structure, is plugged by additional linker residues (Leu-Ala-Ala-Ala-Leu-Glu-His) (23) derived from the expression vector, partly mimicking the NHE1 helix (Fig. 5A). Binding of the vector-derived linker might have facilitated adoption of the open and semiopen conformations of EF-1 and EF-2, respectively. Therefore, the possibility of a "closed to open" conformational transition of EF-1 and EF-2 remains to be evaluated.
The binding of EF-hands in an open conformation without Ca 2ϩ to target molecules has been found following crystal structure investigations of a Kv-KChIP1 in which the Kv fragment is covalently linked to the C terminus of KChIP1 (45). In this case, the Kv fragment binds to EF-1 and EF-2 of the N-terminal domain of KChIP1 through hydrophobic interactions (Fig. 5A). Of particular note, KChIP1 forms a dimer utilizing the surface formed by the Kv peptides and helix 10 of the C-terminal domain of KChIP1 in contrast to NHE1-CHP1, which exists as a monomer (Fig. 5C). The interhelix angles of EF-1 and EF-2 are 82°and 72°, respectively (supporting information S1), which are slightly larger than those of CHP1. A dimeric interaction in addition to binding of the Kv fragment may contribute to broadening of the cleft constituted by EF-1 and EF-2.
CHP1 shares 18% sequence identity with KChIP1, and the folding topology is almost identical (Fig. 5A). However, the target recognition mechanism differs from that of KChIP1 as judged from the determined structures. KChIP1 belongs to the extensively studied NCS1 family, which act as important regulators of various functions among certain higher eukaryotes (50). We propose that the NHE1-CHP1 interaction represents a novel binding mode utilized throughout the four-EFhand proteins, which constitute a distinct subfamily to the NCS1 family. Furthermore detailed comparison of the binding mode of CHP1 and CNB will be presented below.
The CHP1-NHE1 Interface-The protein-protein interface consists of an extensive hydrophobic concave CHP1 undersurface and an apolar NHE1 surface. The concave undersurface spans both the N-and C-terminal domains of CHP1. The total surface area buried at the interface is 2338 Å 2 , slightly smaller than the value of 2625 Å 2 for the interface between CNA and CNB (Protein Data Bank code 1AUI). Approximately more than 90% of the total buried surface area between NHE1 and CHP1 is hydrophobic, similar to the complexed structure of CNA and CNB.
The interface formed between NHE1 and CHP1 includes methylcontaining and aromatic hydrophobic residues. The residues of the N-terminal domain of CHP1, Ala-69, Phe-90, Ile-66, Thr-86, Leu-87, Phe-35, and Leu-54, comprise a hydrophobic cleft that interacts with the apolar surface formed by the side chains of NHE1 residues Leu-530, Ile-534, and Ile-537. Additionally the side chain of NHE1 Leu-531 expands the hydrophobic area interacting with CHP1 Phe-90 (Fig.  6A). The C-terminal domain residues of CHP1, Ala-163, Thr-159, Leu-139, Leu-135, Tyr-122, Phe-176, Ile-171, Phe-117, Ala-118, and Leu-121, constitute a continuous hydrophobic cleft that interacts with the N-terminal portion of NHE1 helix residues Ile-518, Ile-522, and Phe-526, which protrude at the apolar side. NHE1 residues Leu-527 and His-523 make additional hydrophobic contact with CHP1 residues Tyr-122, Phe-176, and Val-185 (Fig. 6A). These interface-forming hydrophobic residues are well conserved within NHE and CHP isoforms, reflecting the importance of these interactions (Fig.  1B). Furthermore the acidic side chain of Asp-528 at the center of NHE1 helix seems to form a salt bridge with the basic side . Protein Data Bank entries are shown in parentheses. B, vector geometry mapping of EF-1 and EF-2 apocalmodulin, Ca 2ϩ -calmodulin, calcineurin B, Kv-KChIP1, NHE1-bound CHP1, and NHE1-free CHP1 are denoted as "apoCaM," "CaCaM," "CNB," "KChIP," "NHECHP," and "CHP," respectively. Hyphenated numbers 1 and 2 denote EF-1 and EF-2, respectively, for each protein. C, comparison of the KChIP1 dimer and CHP1 structures colored in brown and blue, respectively. Bound Kv and NHE1 are colored in pink and green, respectively. The dimer interface of KChIP1 is shown as a dashed line, and one molecule of the dimer is shown in light colors. KChIP1 and CHP1 are depicted in the same orientation.
In A and C, the N-and C-terminal regions, ␣A, and the CHP loop of CHP1 are not shown in the comparison for reasons of clarity.

Solution Structure of the NHE1-CHP1 Complex
chain of CHP1 Lys-188. The presence of this interaction was observed in 10 of the final 20 NMR structures derived using the MONSTER server (51), which identifies interacting residues and assigns the nature of those interactions based on the structure. These residues are also conserved (Figs. 1, A and B, and 6A).
NHE1 binding orientation relative to CHP1 is probably established by surface complementarity that comprises the hydrophobic surfaces of NHE1 and CHP1. NHE1 side chains of Ile-534 and Ile-537 protrude into the shallow cleft made by the CHP1 N-terminal domain. Meanwhile NHE1 aromatic side chains of His-523 and Phe-526 protrude into the deep cleft of the CHP1 C-terminal domain (Fig. 6A). This deep cleft allows enough space so that they can interact with the bulky side chains of NHE1. The difference between the shapes of the domains is critical for NHE1 binding orientation.
Notably the shape of the interaction surfaces of NHE1-CHP1 and CNA-CNB differs. The spatial arrangement of NHE1 and CNA residues that contribute to form the interaction surfaces differs for each (Fig. 6B). Only Val-349, Phe-350, and Val-357 of CNA are located at equivalent positions to Ile-522, His-523, and Leu-530 of NHE1 unlike other residues. Cleary Trp-352 and Phe-356 of CNA that interact with the roof of the hydrophobic cleft formed by CNB are missing in NHE1. Furthermore the interactions mediated by the C-terminal stretch region of CNA, where Leu-369 and Leu-371 protrude, are absent in NHE1. On the other hand, the corresponding residues to Phe-526 and Ile-518 of NHE1 are absent in CNA.
Consequently the interaction mechanism dominated by hydrophobic interactions through amphipathic helices and four EF-hands is common in both complexes, but their shape differs. This surface complementarity in terms of a knobs-into-hole mode of interaction defines the binding specificity of these proteins, although the folding topology is identical.
Correlation with Mutagenesis Studies-The solution structure of the CHP1-NHE1 complex is essentially consistent with previous mutagenesis studies concerning NHE1. Co-immunoprecipitation experiments showed that the 4Q mutant of NHE1, in which Phe-526, Leu-527, Leu-530, and Leu-531 are substituted with glutamine, displayed no binding to CHP1, whereas the I518Q/I522Q NHE1 double mutant displayed some binding albeit with decreased affinity (3). Our structure determination revealed that Phe-526, Leu-527, Leu-530, and Leu-531 form a hydrophobic core that interacts with the center of the hydrophobic cleft of the C-terminal domain of FIGURE 6. Noncovalent interactions at the NHE1-CHP1 interface and contribution to the overall stability of the complex. A, molecular surface of CHP1 and backbone tube representation of NHE1 with yellow and red sticks that show hydrophobic and acidic side chains, respectively. Hydrophobic and basic surface residues of CHP1 are colored in orange and blue, respectively. B, superposition of NHE1-CHP1 and CAN-CNB (Protein Data Bank code 1AUI). NHE1 and CNA are shown in green and magenta, respectively, and the molecular surface of only CHP1 is depicted for clarity. Complex-forming hydrophobic residues of NHE1 and CNA are shown as sticks and are labeled. CNA residues are represented in italics. The left-hand image and the right-hand image are close-up views of the N-and C-terminal domains. C, binding of His-NHE1 peptide to GST-CHP1 mutants in an in vitro pulldown assay. Bound peptide was separated via SDS-PAGE, blotted to a membrane, and then visualized using Ni-NTA-conjugated alkaline phosphatase (top). The amount of GST-CHP1 mutant used was estimated by Coomassie Blue staining (bottom) for calibration. The binding of NHE1 peptide to GST-CHP1 mutants was quantified and expressed as a ratio of binding of wild-type protein at the top of the gel images. CHP1, whereas Ile-518 and Ile-522 interact with the rim of the hydrophobic cleft of the C-terminal domain of CHP1 (Figs. 4B and 6A). Deletion mutagenesis indicated that NHE1 (residues 510 -575) retained binding affinity similar to the wild-type protein, implying that the juxtamembrane region of NHE1 comprising residues 503-509 is unimportant for CHP1 binding (3). These residues were found to be unstructured in the present study and form no direct contact with NHE1.
The binding activity of NHE1 (residues 530 -656) lacking the N-terminal segment was found to be completely impaired (3). Our NHE1-CHP1 complex structure shows that an absence of residues preceding Leu-530 results in almost complete loss of interaction between NHE1 and the C-terminal domain of CHP1, whereas the many hydrophobic interactions mediated by NHE1 residues Leu-530, Leu-531, Ile-534, and Ile-537 are retained (Fig. 6A). This indicates that interaction of the N-terminal residues of NHE1 and the C-terminal domain of CHP1 is indispensable in maintaining the NHE1-CHP1 complex. Namely this part of the interaction plays a dominant role in NHE1-CHP1 complex formation.
Dramatic effects were observed in terms of CHP1 binding to NHE1 with mutations F117Q, Y122Q, and V185Q (Fig. 6C). We expected the floor-forming residues located at the bottom of the cleft to be critical for the interaction, but the rim-forming residue Val-185 showed significant reduction similar to the negative control. Based on the solution structure of the NHE1-CHP1 complex, it was appeared that Phe-117, Tyr-122, and Val-185 were confined to the C-terminal domain of CHP1 and were packed against the apolar side of NHE1. This suggests that the C-terminal hydrophobic cleft represents a mutation hot spot, implying that it plays a key role in the NHE1-CHP1 interaction. This result is consistent with the NHE1 deletion result indicating that interaction between the N-terminal segment of NHE1 and the C-terminal domain of CHP1 is dominant. This represents a unique feature of the NHE1-CHP1 interaction.
Although mutation of residues possibly involved in salt bridge formation such as K188E and K188R resulted in a marked decrease in binding interaction, the effect was not strong, and the charge-reversing effect of K188E was unclear. Thus, it appears that although this salt bridge contributes to NHE1-CHP1 complex formation, it is not the main force possibly because this bond is exposed to solvent (Fig. 6A), and therefore solvation could weaken the strength of the interaction.
Role of CHP1-NMR investigations of NHE1 complexed with CHP1 revealed that the juxtamembrane region comprising res-idues 503-517 was unstructured in solution. This region is rich in basic residues with a previous study reporting that NHE1 comprising residues 506 -576 bound to PIP 2 in vitro (13). In addition, residues 513-520 and 556 -564 might represent PIP 2 binding sites (13). Accordingly it is likely that this juxtamembrane region (residues 503-517) following the last transmembrane helix (H12, residues 478 -499) is incapable of forming a continuous straight helical structure into the cytoplasm due to interaction with the membrane. Rather the overall structure of NHE1 presumably turns or bends toward the cytoplasmic membrane following a PIP 2 -mediated interaction. Assuming that flanking regions of the helix, residues 513-520 and 556 -564, attach to the membrane (PIP 2 ) (13), the helix and CHP1 should be located immediately beneath the cytoplasmic membrane.
Furthermore it has been reported that NHE1 acts as a scaffold protein linked to actin filaments via ezrin-radixin-moesin proteins in addition to possessing function as an ion exchanger (52,53). Our NMR studies revealed that the cytoplasmic helixforming residues of NHE1 comprise residues 518 -537, which showed little overlap from the previously reported ezrin-radixin-moesin binding region (512-520 and 550 -565). This ensures simultaneous binding of NHE1 to CHP1 and ezrinradixin-moesin proteins.
The juxtamembrane region of NHE1 forms a particular tertiary or quaternary structure that is mediated by interactions with the membrane (PIP 2 ), CHP1, and ezrin-radixin-moesin proteins. The overall structure of the juxtamembrane region might play an important role in NHE1 activity. The 90% loss in activity following CHP1 depletion might be due to disruption of the structure of the cytoplasmic region of NHE1 around the membrane. In fact, although this represents in vitro evidence, the amphipathic helix is disrupted in the absence of CHP1 based on the CD and NMR data (supporting information S2 and S3).
Possible Mode of Regulation-CHP1 deprivation resulted in impaired regulation of NHE1 following external stimuli, implying that CHP1 acts as regulator of NHE1 by involvement in the processing of intracellular signals derived from external stimuli. However, the regulatory mechanism remains unclear.
Although it was reported that CHP1 is an N-myristoylated protein, CHP1 does not exert myristoyl switching in a Ca 2ϩ -dependent manner under normal physiological conditions because EF-3 and EF-4 constitutively bind Ca 2ϩ ions where N-myristoylation was not required for NHE1 binding, activation, or localization (14).
Of particular note, it was reported that CHP1 is a phosphorylated protein, although the phosphorylation sites were not determined. According to the phosphorylation prediction server NetPhos (54), the CHP1 sequence contains potential phosphorylation sites located at residues Thr-36, Ser-37, Ser-47, Ser-131, and Ser-172 (score, Ͼ0.8). Similarly the phosphorylation server Scansite (55) identified potential phosphorylation sites located at residues Thr-36, Ser-37, and Ser-172 (score, Ͼ0.5). From the determined NHE1-CHP1 structure, residues Thr-36, Ser-37, and Ser-172, predicted by both servers as potential phosphorylation sites, are locate at the terminal part of the EF-hand helix or its flanking loop where the side chains of the aforementioned residues are exposed to the protein surface reinforcing the possibility of phosphorylation. Phosphorylation-induced conformational changes in CHP1 and the subsequent regulation of NHE1 activity are interesting areas that remain to be investigated.
Conclusion-We have determined the solution structure of the cytoplasmic region of NHE1 complexed with CHP1. Although previous biochemical analyses suggested that the hydrophobic residues of NHE1 were likely to interact with CHP1, the present study has delineated the structural basis for this interaction. The solution structure provides concrete evidence that the cytoplasmic region of NHE1 forms an amphipathic helix that interacts directly with the large concave undersurface of CHP1. This helix is disrupted in the absence of CHP1; thus the loss in activity following CHP1 depletion might be due to disruption of the structure of the juxtamembrane region of NHE1. Our structure provides a first step toward understanding the regulation of NHE1 activity. Moreover it revealed a novel target binding mechanism mediated by four EF-hands. These findings should facilitate future studies aimed at understanding the mechanism underlying recognition utilized by EF-hand proteins that are engaged in signal transduction pathways and many other molecular and cellular events. During the initial review of our manuscript, a study appeared that describes a crystal structure of the NHE1 peptide complexed with CHP2 containing Y 3ϩ ions instead of Ca 2ϩ (56).