Crystal Structure of Calmodulin Binding Domain of Orai1 in Complex with Ca2+•Calmodulin Displays a Unique Binding Mode*

Background: Calmodulin (CaM) binds the calmodulin binding domain (CMBD) of Orai1 causing Ca2+·CaM-dependent inhibition (CDI). Results: Orai1-CMBD binds the C-lobe of CaM in crystal, whereas in solution Orai1-CMBD interacts with both N- and C-lobes of CaM. Conclusion: One CaM, in an extended conformation, binds to two Orai1-CMBDs. Significance: This expands our knowledge of the interaction of CaM with its targets, and provides important information for understanding CDI. Orai1 is a plasma membrane protein that in its tetrameric form is responsible for calcium influx from the extracellular environment into the cytosol in response to interaction with the Ca2+-depletion sensor STIM1. This is followed by a fast Ca2+·calmodulin (CaM)-dependent inhibition, resulting from CaM binding to an Orai1 region called the calmodulin binding domain (CMBD). The interaction between Orai1 and CaM at the atomic level remains unknown. Here, we report the crystal structure of a CaM·Orai1-CMBD complex showing one CMBD bound to the C-terminal lobe of CaM, differing from other CaM-target protein complexes, in which both N- and C-terminal lobes of CaM (CaM-N and CaM-C) are involved in target binding. Orai1-CMBD binds CaM-C mainly through hydrophobic interactions, primarily involving residue Trp76 of Orai1-CMBD, which interacts with the hydrophobic pocket of CaM-C. However, NMR data, isothermal titration calorimetry data, and pulldown assays indicated that CaM-N and CaM-C both can bind Orai1-CMBD, with CaM-N having ∼4 times weaker affinity than CaM-C. Pulldown assays of a Orai1-CMBD(W76E) mutant, gel filtration chromatography data, and NOE signals indicated that CaM-N and CaM-C can each bind one Orai1-CMBD. Thus our studies support an unusual, extended 1:2 binding mode of CaM to Orai1-CMBDs, and quantify the affinity of Orai1 for CaM. We propose a two-step mechanism for CaM-dependent Orai1 inactivation initiated by binding of the C-lobe of CaM to the CMBD of one Orai1 followed by the binding of the N-lobe of CaM to the CMBD of a neighboring Orai1.

formational changes to expose a hydrophobic pocket in each lobe for the binding of target proteins (14 -16). In the absence of target proteins, the central linker of CaM is flexible in solution, and CaM-N and CaM-C do not interact with each other. In the presence of target proteins, CaM-N and CaM-C of a CaM molecule typically wrap around the CaM binding domains (CMBDs) of target proteins (usually an ␣ helix) in a collapsed 1:1 binding mode. Exceptions, however, have been reported, in which CaM binds its target CMBDs in a 2:2 or 1:2 binding mode. For example, in the case of the small conductance Ca 2ϩactivated potassium channel SKca, three helices are bound between CaM-N and CaM-C in an extended 2:2 binding mode (17). In the case of glutamate decarboxylase, each lobe of CaM binds one peptide of glutamate decarboxylase, with the two peptides interacting with each other, resulting in a collapsed 1:2 binding mode (18). The different structures of CaM⅐CMBD complexes indicate that CaM is very flexible allowing it to adopt different conformations. This conformational flexibility is consistent with the lack of sequence homology among CMBDs of its target proteins. The conformational flexibility of CaM and the lack of sequence homology among CMBDs make it difficult to use existing structural information to accurately predict the interaction between CaM and the CMBD of Orai1. We therefore carried out structural studies on the CaM⅐Orai1-CMBD complex, hoping to 1) understand how CaM facilitates CDI of Orai1; 2) add to our knowledge of the interaction modes of CaM with its target proteins.
Here, we report the crystal structure of Ca 2ϩ ⅐CaM in complex with Orai1-CMBD at 1.9-Å resolution. The structure shows that in the crystal Orai1-CMBD is bound only to CaM-C, with residue Trp 76 of Orai1-CMBD fitting into the hydrophobic pocket of CaM-C and making extensive interactions with hydrophobic residues of CaM. Further biochemical analyses and mutational studies indicate that in solution CaM interacts with Orai1-CMBDs in a 1:2 binding mode, with CaM-N and CaM-C each binding one Orai1-CMBD.

EXPERIMENTAL PROCEDURES
Materials-All chemicals were purchased from Sigma unless indicated elsewhere. All restriction enzymes were purchased from New England Biolabs. Crystallization chemicals were purchased from Hampton Research. The Orai1-CMBD peptide, (sequence HSMQALSWRKLYLSRAKLKA), corresponding to residues His 69 -Ala 88 of Orai1, was obtained from GenScript USA Inc.
Cloning, Expression, and Purification of CaM and CaM-(Ca 2ϩ -N mut )-The rat CaM cDNA was amplified by PCR using Supermix solution (Invitrogen) and cloned into pET21b vector using NdeI and BamHI restriction enzymes. CaM(Ca 2ϩ -N mut ), a CaM mutant with abolished Ca 2ϩ binding capacity at its N-lobe, was constructed by mutating Asp 20 , Asp 22 , Asp 24 , Glu 31 , Asp 56 , Asp 58 , Asn 60 , and Glu 67 to histidines. Primers containing the mutations were synthesized by Sigma. The mutant gene was amplified using multistep PCR, and cloned into pET21b as wild type CaM. The plasmids CaM-pET21 and CaM(Ca 2ϩ -N mut )-pET21 were sequenced to confirm that the gene was correctly inserted. The plasmids were transformed into BL21(DE3) Gold (Agilent Technologies) for protein expression. The bacteria were grown in LB medium at 37°C, 220 rpm until cell density reached A 600 nm ϭ 0.6. The bacteria were then cultured at 23°C for 30 min before induction with 0.2 mM isopropyl 1-thio-␤-D-galactopyranoside (Invitrogen). The culture was continued at 23°C, 170 rpm for another 4 h. The bacteria were harvested by centrifugation at 4,000 ϫ g for 10 min. Cell pellets were kept at Ϫ80°C until further use. CaM and CaM(Ca 2ϩ -N mut ) were purified according to a reported method (19). Briefly, 1-g portions of frozen cell pellets were suspended in 20 ml of lysis buffer (50 mM Tris, pH 7.5, 2 mM EDTA, 0.2 mM PMSF). The bacteria were lysed by sonication. The lysate was centrifuged at 16,000 ϫ g for 20 min to remove cell debris. The supernatant was supplemented with 5 mM CaCl 2 before it was loaded onto a 5-ml HiTrap Phenyl HP column (GE Healthcare). The column was then washed with 50 ml of phenyl wash buffer 1 (50 mM Tris, pH 7.5, 0.1 mM CaCl 2 , 100 mM NaCl), followed by a wash with 50 ml of phenyl wash buffer 2 (50 mM Tris, pH 7.5, 5 mM CaCl 2 , 500 mM NaCl). The column was eluted with phenyl elute buffer (50 mM Tris, pH 7.5, 1 mM EGTA). The eluate was concentrated and further purified on a Superdex 200 column (GE Healthcare) equilibrated with sizing buffer (10 mM Tris, pH 7.5, 2 mM CaCl 2 , 100 mM NaCl).
Expression and Purification of CaM-N, CaM-C, Orai1-CMBD, and Orai1-CMBD(W76E)-CaM was also cloned into pGEX-4T1 vector using BamHI and XhoI restriction enzymes. To obtain CaM-N (aa 1-75), a stop codon was added after Lys 75 in the CaM-pGEX plasmid using the QuikChange mutagenesis kit (Agilent Technologies). The same strategy was used to make the CaM-N-pET21 plasmid from CaM-pET21. The gene of CaM-C (aa 79 -148) was amplified by PCR from CaM-pET21 and cloned into pET28b (Novagen) with a His tag at its N terminus using NheI and BamHI restriction enzymes. Orai1-CMBD (aa 69 -91) was cloned into pET32b vector (Novagen) using NcoI and EcoRI restriction enzymes as a thioredoxin fusion protein (Orai1-CMBD-Trx) with a His tag. Orai1-CMBD(W76E) was made from Orai1-CMBD-pET32 plasmid using the QuikChange mutagenesis kit. All constructs were expressed in Escherichia coli BL21(DE3) Gold (Agilent Technologies) in the same way as described above for the expression of CaM. CaM-N-GST fusion protein was loaded onto a GSTrap HP column (GE Healthcare), which was then washed with 10 column volumes of GSH wash buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 2 mM CaCl 2 ), and eluted with GSH elution buffer (100 mM Tris-HCl, pH 7.5, 50 mM NaCl, 2 mM CaCl 2 , 10 mM GSH). CaM-C and Orai1-CMBD-Trx were purified on a His-Trap FF column (GE Healthcare), which was washed with Ni wash buffer (20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 2 mM CaCl 2 , 20 mM imidazole), and eluted with Ni elution buffer (20 mM Tris-HCl, pH 7.5, 500 mM NaCl, 100 mM EDTA). All eluates were concentrated using Amicon concentrators and further purified on a Superdex 200 column equilibrated with sizing buffer.
Crystallization of CaM⅐Orai1-CMBD Complex-Peak fractions of CaM from the Superdex 200 column were pooled and concentrated to 10 -20 mg/ml using an Amicon concentrator. The Orai1-CMBD peptide powder was dissolved in sizing buffer, and mixed with CaM at a 2:1 molar ratio. Initial crystallization trials were screened around the reported conditions for Calmodulin and Orai1 Form an Unusual 1:2 Complex DECEMBER 14, 2012 • VOLUME 287 • NUMBER 51 CaM and CaM⅐target peptide complexes. Clusters of long needles were initially obtained in 0.1 M BisTris-HCl (pH 6.0), 45% polypropylene glycol P400 (PPG P400) at 4°C. Optimization of crystallization was carried out using an Additive Screen (Hampton Research) with a Mosquito liquid handling system (TTP Labtech). Microseeding significantly facilitated the crystal growth of CaM⅐Orai1-CMBD. Single needle crystals were obtained using 1 l of calmodulin (20 mg/ml) ϩ synthetic Orai1-CMBD at a 1:4 CaM:peptide molar ratio in 10 mM Tris-HCl (pH 7.5), 100 mM NaCl, 5 mM CaCl 2 , mixed with 1 l of reservior solution (0.1 M BisTris-HCl (pH 6.0), 40% PPG P400, 14% ␥-butyrolactone, 1% n-octyl-␤-D-glucopyranoside), and equilibrated against 500 l of reservoir solution at 20°C by the hanging drop vapor diffusion method.
Data Collection and Processing-Crystals of CaM⅐Orai1-CMBD were flash frozen in liquid nitrogen without additional cryo-protectant. X-ray diffraction data were collected using inhouse x-ray generator MicroMax-007 HF (Rigaku). The crystals diffracted to 1.9 Å in space group C2, with cell dimensions of a ϭ 100.07 Å, b ϭ 24.71 Å, c ϭ 60.90Å, and ␤ ϭ 100.1°. The data were processed and scaled with DENZO/SCALEPACK (20). The statistics of the diffraction data are summarized in Table 1.
Structure Determination and Refinement-The structure was solved by molecular replacement using PHASER molecular replacement from the PHENIX program (21). Initial search with CaM from the CaM⅐MARCKS structure (PDB code 1IWQ) as a search model did not yield a solution. When CaM-N (aa 4 -75) and CaM-C (aa 85-146) were used as individual search models, we were able to find rotation and translation solutions for molecular replacement. After several rounds of refinement using PHENIX and manual model building with COOT (22), the central linker of CaM and Orai1-CMBD were successfully built into the electron density map. Four ␥-butyrolactone molecules were also built and refined. The statistics of refinement are shown in Table 1. The interaction interfaces were analyzed with PISA program in the CCP4 Molecular Graphics Documentation (23). All the graphic presentations of structures were made with PyMOL.
Pulldown Assays-In pulldown assays for synthetic Orai1-CMBD, 50 l of 50 M CaM-N-GST or CaM-C-His 6 were mixed with 50 l of 50 M Orai1-CMBD. All samples were in sizing buffer. 50 l of GSH resin or nickel resin was added to mixtures of CaM-N-GST ϩ Orai1-CMBD or CaM-C-His 6 ϩ Orai1-CMBD, respectively. The mixtures were incubated on ice for 20 min, and centrifuged at 2,000 ϫ g for 3 min. The supernatants were removed and the resins were washed three times with 1 ml of GSH wash buffer or Ni wash buffer. 100 l of 2ϫ SDS sample buffer was added to each resin. The resins were spun at 16,000 ϫ g for 1 min and the supernatants were loaded onto a SDS-PAGE gel.
The pulldown assays testing for interaction of CaM and CaM-N with recombinantly expressed wild type (WT) Orai1-CMBD-Trx or Orai1-CMBD(W76E)-Trx mutants were carried out in a similar manner as described above. Orai1-CMBDs were expressed as His 6 -tagged thioredoxin fusion proteins. WT and W76E Orai1-CMBD were incubated with GSH and nickel resins as a control to test the nonspecific binding.
Isothermal Titration Calorimetry (ITC) Experiments-ITC measurements were carried out in sizing buffer using an iTC 200 MicroCalorimeter (GE Healthcare) at 25°C. Substrate solutions containing synthetic or recombinantly expressed Orai1-CMBD at 0.3-1 mM were injected into a reaction cell containing 30 -60 M CaM, CaM-N, or CaM-C. Fifty injections of 0.4 -0.7 l at 120-s intervals were performed. Data acquisition and analysis were performed using the Origin Scientific Graphing and Analysis software package (OriginLab). Data analysis was performed by generating a binding isotherm and best fit using the following fitting parameters: N (number of sites), ⌬H (cal/mol), ⌬S (cal/mol/deg), and K (binding constant in M Ϫ1 ) and the standard Levenberg-Marquardt methods (24). Following data analysis, Determination of Protein Concentrations-CaM has a low extinction coefficient due to the lack of tryptophan, the main contributor for UV absorbance of a protein at 280 nm. Therefore, the CaM concentration calculated with the extinction coefficient is susceptible to influence of contaminating proteins or DNA. Accuracy of the protein concentration is critical for ITC data analysis. Therefore, we determined the CaM concentration using a bicinchoninic acid protein assay kit (Thermo-Scientific). BSA at different concentrations determined by weight over volume was used as standard. The concentrations of His-tagged CaM-C, CaM-N-GST, and synthetic Orai1-CMBD were also determined using the same bicinchoninic acid method.
Calculation of Molecular Weights by Size Exclusion Chromatography-Bio-Rad gel filtration standard containing bovine thyroglobulin (670 kDa), bovine ␥-globulin (158 kDa), chicken ovalbumin (44 kDa), horse myoglobin (17 kDa), and vitamin B 12 (1350 Da) were run on a Superdex 200 column eluted with sizing buffer. The log 10 (MW) were plotted against the elution volumes to obtain a standard curve. The molecular weights of protein samples were calculated by interpolation based on their elution volumes.
NMR Assignments-Calmodulin and CaM-N in pET21 vector were expressed in BL21(DE3) Gold in minimal media containing [U-13 C]glucose and 15 NH 4 Cl, and purified as described above. Prior to the NMR experiments the protein was concentrated to 60 M. NMR spectra were acquired with Orai1-CMBD at 0, 30, 60, 90, and 120 M. The experiments to assign the protein resonances were acquired in the presence of Orai1-CMBD at 120 M. The buffer was: 10 mM NaP i (pH 6.5), 2 mM CaCl 2 , 5% D 2 O, 50 M 5,5-dimethylsilapentanesulfonate to serve as a chemical shift reference. NMR spectroscopy was performed at 25°C on Varian UNITY INOVA 500 MHz spectrometer equipped with a 5-mM triple resonance cold probe with actively shielded z axis gradients. The NMR data were processed with NMRPipe (25) and the spectra were analyzed with NMRView (26). The backbone resonance assignments were made from analysis of the HNCACB and CBCA(CO)NH experiments acquired using the Varian BioPack pulse sequences.

Orai1-CMBD Interacts Only with CaM-C in Crystal-We
have determined the structure of CaM⅐Orai1-CMBD in the presence of Ca 2ϩ in space group C2 at 1.9 Å. There is one CaM⅐Orai1-CMBD complex in one asymmetric unit. The final R/R free factors were 0.18/0.22 (Table 1). There are four calcium ions bound to CaM, two to each lobe. Residues 3-146 of CaM are well defined by the electron density map, and all residues from Orai1-CMBD fit well to the electron density map. The overall quality of the electron density is good (supplemental Fig.  S1).
Surprisingly, CaM adopts an extended conformation in the complex, with its central linker forming an ␣ helix (Fig. 1, A and B). The structure of CaM here is almost identical to Ca 2ϩ ⅐CaM that is free of ligand (PDB code 3CLN) (28) (Fig. 1C). When our structure and that of ligand-free Ca 2ϩ ⅐CaM were superimposed along their C␣, the superposition gave a root mean square deviation of 0.44 Å over 127 C␣. In fact, upon structural superposition, the surface of the C-lobe of Ca 2ϩ ⅐CaM accommodates Orai-CMBD well, indicating that their side chains also adopt similar conformations. The major change in the hydrophobic pocket of CaM-C is the side chain of Phe 92 that rotates about 90°along the C␤-C␥ bond to accommodate the side chain of Leu 79 of Orai1-CMBD (Fig. 1D). In one asymmetric unit, only CaM-C interacts with Orai1-CMBD; CaM-N has no interaction with either Orai1-CMBD or CaM-C. Orai1-CMBD, however, interacts with CaM-N of a symmetry-related CaM molecule (supplemental Fig. S2). This interaction consists of a hydrophobic interaction (between Leu 81 of Orai1-CMBD and Leu 4 and Leu 70 of CaM) and two hydrogen bonds (CaM Lys 13 NE-Orai1-CMBD Lys 87 O, and CaM Asp 64 OG1-Orai1-CMBD Lys 78 NE) (supplemental Fig. S2A). The interface of CaM-C and Orai1-CMBD has a higher shape complementarity than that of Orai1-CMBD and symmetry-related CaM-N, resulting in a surface area of 780 Å 2 for Orai1-CMBD/CaM-C interaction, and 420 Å 2 for the Orai1-CMBD/symmetry-related CaM-N interaction (supplemental Fig. S2, B and C). Shape complementarity analysis using the PISA server from CCP4 indicates that Orai1-CMBD interaction with the N-lobe of symmetry-related CaM is unlikely to exist in solution. In addition, size exclusion chromatography of the CaM⅐Orai1-CMBD complex does not show the existence of a dimeric complex, which would have been formed by this interaction. Thus the interaction of Orai1-CMBD with CaM-N observed in the crystal structure is simply a consequence of crystal packing.
Orai1-CMBD Partially Adopts a Helical and Partially an Extended Conformation-The N-terminal part of the peptide (residues 69 -74) adopts an extended conformation. Some residues (His 69 , Gln 72 , and Ala 73 ) on this segment of the peptide interact with neighboring CaM and Orai1-CMBD molecules resulting from crystal packing. The C-terminal segment of Orai1-CMBD adopts ␣ helical conformation. Most of the resi-   dues interacting with CaM are located in this helical segment (Fig. 1, A and B). Orai1-CMBD Interacts with CaM-C Mainly through Hydrophobic Interactions-Residues Trp 76 and Leu 79 of Orai1-CMBD bind to the hydrophobic pocket of CaM-C ( Fig. 2A). The side chain of Trp 76 is deeply buried in this pocket (Fig. 2B) Fig. S3B). Thus, Trp 76 and Leu 79 of Orai1-CMBD make the major contribution to its interaction with CaM. In addition, Leu 74 of Orai1-CMBD makes hydrophobic interactions with residues Met 109 , Glu 114 , Leu 116 , and Met 124 of CaM, and Tyr 80 of Orai1-CMBD interacts with Ile 85 , Ala 88 , Phe 141 , and Met 145 of CaM (supplemental Fig. S3, A and C).
CaM-N Binds Orai1-CMBD in Solution-In the crystal structure, CaM-N interacts with Orai1-CMBD from a symmetry-related complex. To evaluate the interaction of Orai1-CMBD with CaM-N, we expressed and purified CaM-N (aa 1-75) as a GST fusion protein and CaM-C (aa 79 -148) with a His tag. The individual lobes of CaM were used to pull down the synthetic Orai1-CMBD. Orai1-CMBD was pulled down not only by CaM-C, but also by CaM-N (Fig. 3A). It is noteworthy that the controls for the pulldown assay indicate a nonspecific binding of Orai1-CMBD to GSH resin. However, the nonspecifically bound Orai1-CMBD was removed from GSH resin during the wash. Thus, Orai1-CMBD was indeed pulled down by CaM-N GST fusion protein. ITC experiments also showed that both CaM-N and CaM-C alone bound Orai1-CMBD, with a K d of 4.6 M for CaM-N, and 1.1 M for CaM-C (Fig. 4). This indicates that CaM-N alone can bind Orai1-CMBD, but with a slightly weaker affinity than CaM-C. The binding of CaM-N to Orai1-CMBD detected by pulldown assays and ITC suggested that the interaction between CaM-N and Orai1-CMBD  observed in crystal packing might contribute to the interaction in solution. However, when we mutated Trp 76 of Orai1-CMBD to Glu, a mutation reported to abolish the binding of Orai1-CMBD to CaM and thus disrupt CDI (13), the pulldown assay (Fig. 3B) and ITC results (data not shown) showed that the W76E mutant of Orai1-CMBD did not bind to CaM, CaM-N, or CaM-C, indicating that the interaction of CaM-N with Orai1-CMBD observed at the lattice contact, which does not involve Trp 76 of Orai1-CMBD, differs from the interaction observed in solution. Based on these data, we concluded that CaM-N and CaM-C both bind Orai1-CMBD through Trp 76 , with CaM-C having a higher affinity than CaM-N. Because CaM-N and CaM-C share high sequence homology, we superimposed the CaM-N structure to CaM-C, and noticed that Trp 76 of Orai1-CMBD also fits into the hydrophobic pocket of CaM-N (Fig. 2, C and D), making Trp 76 a ligand for both CaM-N and CaM-C.
CaM Binds Orai1-CMBD at a 1:2 Ratio-The binding of CaM-N and CaM-C to Orai1-CMBD described above suggests that full-length CaM binds Orai1-CMBD at a 1:2 ratio in solution, and that the crystal structure reflects an intermediate toward this 1:2 complex. Indeed, ITC data show that one CaM molecule binds two Orai1-CMBDs with a K d of 1.3 M (Fig. 4).
To further confirm the 1:2 binding ratio of CaM to Orai1-CMBD, we checked complex formation of CaM with the Orai1-CMBD-Trx fusion protein (Orai1-CMBD fused to the C terminus of thioredoxin encoded in the pET32 vector) by size exclusion chromatography (Fig. 5). Purified Orai1-CMBD-Trx, CaM, and a mixture of these two proteins were individually run on Superdex 200 column, and their molecular weights were calculated from the elution volumes using a standard curve. CaM has a theoretical molecular mass of 16.7 kDa, and Orai1-CMBD-Trx of 19.1 kDa. Their molecular weights calculated from size exclusion chromatograms are consistent with their theoretical molecular weights (Fig. 5). The calculated molecular mass of CaM is 20 kDa, slightly higher than the theoretical value. This could be due to the elliptical rather than the globular shape of CaM in the absence of target peptide. The calculated molecular weight of Orai1-CMBD-Trx is 16 kDa, slightly smaller than the theoretical value. At 1:1 molar ratio, the calculated molecular mass of the CaM⅐Orai1-CMBD-Trx mixture is 43 kDa, a number between the molecular masses of 1:1 and 1:2 CaM⅐Orai1-CMBD complexes. Because the affinity of Orai1-CMBD to CaM-C is only 4-fold stronger than that to CaM-N, at 1:1 molar ratio, a major fraction of Orai1-CMBD-Trx will bind to CaM-C, and the other fraction of Orai1-CMBD-Trx will bind to CaM-N, resulting in a mixture of 1:1 and 1:2 complexes. We suggest that there is equilibrium between the 1:1 and 1:2 complexes. This suggestion is supported by the NMR chemical shift perturbation data in Fig. 6, which indicates weak binding of CaM-N to Orai1-CMBD with an intermediate exchange rate. This model is consistent with our failure to observe two separate peaks for 1:1 and 1:2 complexes on the size exclusion chromatogram of the 1:1 mixture. The calculated molecular mass of the CaM⅐Orai1-CMBD-Trx mixture at a 1:2 molar ratio is 48 kDa, indicating a 1:2 CaM⅐Orai1-CMBD-Trx complex. When CaM and Orai1-CMBD-Trx were mixed at a 1:4 molar ratio, an additional elution peak containing Orai1-CMBD-Trx appeared, representing the excess of Orai1-CMBD-Trx. These results indicate that CaM and Orai1-CMBD-Trx form a 1:2 complex, rather than a 1:1 complex. Because Orai1-CMBD does not form dimers or oligomers, as indicated by its size exclusion chromatogram, we propose that upon binding to Orai1-CMBD, CaM remains in its extended conformation.
The property of proteins can make the elution volume on size exclusion chromatography deviate from the regular elution volume, resulting in a wrong calculated molecular weight. To rule out this possibility, we mutated the two calcium binding sites in CaM-N. The mutant, CaM(Ca 2ϩ -N mut ), will have only CaM-C capable of binding to Orai1-CMBD, and therefore can only form a 1:1 complex with Orai1-CMBD. When we subjected the 1:1 molar ratio mixture of CaM(Ca 2ϩ -N mut ):Orai1-CMBD-Trx to size exclusion chromatography, the calculated molecular weight of the complex was 35 kDa, corresponding to the 1:1 complex (Fig. 5B). At 1:2 molar ratio of CaM(Ca 2ϩ -N mut ): Orai1-CMBD-Trx, an additional peak corresponding to free Orai1-CMBD-Trx appeared, and no complex appeared larger than the 1:1 complex formed (Fig. 5B). At the 1:4 molar ratio of CaM(Ca 2ϩ -N mut ):Orai1-CMBD-Trx, the dominant peak on the size exclusion chromatogram was the free Orai1-CMBD-Trx (Fig. 5B). These results confirmed that complexes in the 1:2 and 1:4 CaM⅐Orai1-CMBD-Trx mixtures observed on size exclusion chromatograms were 1:2 CaM⅐Orai1-CMBD-Trx complexes.
NMR data provided confirmation of the binding of two peptides to one CaM, and confirmed that CaM-N has a weaker affinity than CaM-C for Orai1-CMBD. Fig. 6 tracks a representative resonance from the 15 N-1 H heteronuclear single quantum coherence of CaM-C (Ile 130 ) and CaM-N (Aal 57 ) upon titration of Orai1-CMBD using full-length CaM. The full 15 N-1 H heteronuclear single quantum coherence is provided in supplemental Fig. S4. Fig. 6A shows two peaks for Ile 130 at a 1:0.5 molar ratio of the CaM⅐Orai1-CMBD mixture indicative of slow exchange between a bound and unbound conformation for Ile 130 ; this is common for tight binding. At a 1:1 ratio, the C-terminal domain is fully saturated and the peak changes very little upon further addition of peptide to 1:1.5 and 1:2 molar ratios. In comparison, Fig. 6B shows that the Ala 57 peak from the N-terminal domain broadens slightly at 1:0.5 and 1:1 but broadens significantly at 1:1.5, but does recover full intensity when a ratio of 1:2 is reached. This result indicates that binding to the C-terminal domain is tighter and occurs first in the titration. Broadening of the amide resonances of the N-terminal domain upon addition of peptide can also be observed with the CaM-N construct (Fig. 6C). At a 1:0.5 ratio, the Ala 57 peak of CaM-N is broadened due to intermediate exchange, consistent with a weaker binding to this domain. The difference is most striking between the Ile 130 of full-length CaM and Ala 57 of CaM-N at the same molar ratio (Fig. 6, A and C). Based on these comparisons, it appears that Orai1-CMBD interacts independently with the N-and C-terminal domains.
To model interactions of the N-terminal domain with the peptide, the N-terminal domain was superimposed on the C-terminal domain with the peptide bound to it. Inspection of this model suggests that several methyl groups of the N-terminal domain would potentially interact with Trp 76 of Orai1-CMBD, particularly Ile 27 , Leu 32 , Ile 52 , Val 55 , and Ile 63 (Fig. 2D). Indeed, an examination of the methyl region of the 13 C-1 H HMQC spectra in Fig. 7A showed that the resonances of Ile 52 and Ile 63 shift considerably upon addition of Orai1-CMBD to CaM-N. To confirm this interaction in full-length CaM, Fig. 7B shows similar shifts for resonances of the N-terminal domain, at a ratio of 1:2 CaM:Orai1-CMBD.
To more precisely understand the sites of interaction, an isotope-filtered NOESY study was performed, with the 13 C, 15 N doubly labeled protein and the unlabeled Orai1 peptide. This allows for detection of intermolecular signals that connect the resonances in the (labeled) calmodulin with resonances in the (unlabeled) peptide. There were clear signals from the C-terminal lobe Met 124 , Met 144 , Ile 100 , Ile 125 , and Val 136 to aromatic proton shifts typical of a tryptophan side chain; Trp 76 is the only tryptophan in Orai1-CMBD (Fig. 8A). This is consistent with what we observed in the crystal structure (Fig. 2B). Isotopefiltered NOEs were also detected between CaM-N alone and Orai1-CMBD. Fig. 8B shows the interactions of Trp 76 of Orai1-CMBD with the N-terminal Leu 32 of CaM. This is consistent with the homology model described above (Fig. 2D). However, we did not observe NOE signals between Trp 76 of Orai1-CMBD and other residues in the hydrophobic pocket of CaM-N (see "Discussion" below). In total, these results confirm that CaM interacts with two Orai1-CMBD peptides via the N-and C-terminal domains, and that CaM-N has a weaker affinity for the peptide than CaM-C.
Structure of Orai1-CMBD Reveals a Modified Helical Wheel for Interaction with Camodulin-Most of calmodulin binding domains adopt a helical structure. Mullins et al. (13) predicted a helical wheel model of Orai1-CMBD. Based on the crystal structure (Fig. 9A), we made a helical wheel presentation of Orai1-CMBD (Fig. 9B), and compared it with the helical wheel model predicted by Mullins et al. (13) (Fig. 9C). The helical part of the peptide in the structure is in agreement with the predic-tion. However, the residues N-terminal to Ser 75 of Orai1-CMBD adopt an extended conformation in the crystal structure (Fig. 9B), rather than the predicted helical conformation (Fig. 9C).

DISCUSSION
Unique Binding Mode of Orai1-CMBD to CaM-We show a unique binding mode of CaM to the target peptide Orai1-CMBD: binding of one peptide to each lobe of CaM (Fig. 10A). The binding mode observed here is unique in two aspects: each lobe of CaM binds one target peptide to form a 1:2 complex, and CaM adopts an extended conformation in the complex rather than a collapsed conformation. It is well known that CaM has plasticity to bind different target peptides. Based on the number of residues of CMBDs that are critical for interaction with CaM-N and CaM-C, CaM target peptides can be divided into different groups: 1-7, 1-10, 1-14, 1-16, and 1-18 ( Fig.  10B) (29,30). Because each Orai1-CMBD interacts with only one lobe of CaM, we categorize it as group 1-0. To date, there has been only one structure of 1 CaM:2 target peptides reported that falls into this group, CaM/glutamate decarboxylase (GAD)-CMBD (Fig. 10B). The structure of this CaM⅐GAD complex shows that CaM-N and CaM-C each binds one GAD-CMBD. The two GAD-CMBDs interact with each other to induce CaM to adopt a collapsed conformation (Fig. 10C) (18). CaM-N and CaM-C interact with GAD-CMBD in similar ways. The tryptophan residue on GAD-CMBD, however, is not deeply buried in the hydrophobic pocket of CaM-N or CaM-C (Fig. 10D). CaM⅐Orai1-CMBD may resemble the structure of CaM⅐GAD-CMBD in that Orai1-CMBD interacts with CaM-N and CaM-C in similar ways. The extended conformation in the CaM⅐Orai1-CMBD complex instead of a collapsed conformation (Fig. 10A) may result from the fact that Orai1-CMBD does not form a dimer in solution (Fig. 5) as GAD-CMBD does. Therefore, CaM⅐Orai1-CMBD displays a unique binding mode of CaM to its target peptides, and again shows the flexibility of CaM to bind a large variety of target proteins. In addition to Orai1-CMBD and GAD-CMBD, IQ calmodulin binding peptides from Na V 1.5 (31), Na V 1.2 (32), and PEP-19 (33)(34)(35) bind selectively to the C-lobe of CaM, leaving the N-lobe unoccupied, and thus fall in this group 1-0.
The NMR studies indicate that the interactions of each lobe of the calmodulin with the Orai1-CMBD are homologous, so that the two binding lobes each compete for the same set of residues. This type of interaction requires that binding be sequential, with the first Orai1-CMBD interacting with the stronger binding C-terminal lobe, and the second binding event involving the N-terminal lobe. This sequential binding effect is particularly apparent from the titration study shown in Fig. 6.
The first structure reported for CaM⅐PMCA-CMBD showed that CaM interacted with PMCA-CMBD through CaM-C only (36). Because the PMCA-CMBD used for their studies was a splice variant of PMCA lacking additional C-terminal residues (Fig. 10B), it was suggested that the structure of the complex was an intermediate toward the collapsed structure (36). As shown in Fig. 10B, when additional residues were added, the Val residue could have served as a binding site for CaM-N, resulting in a collapsed complex, as observed in the structure of CaM⅐smMLCK complex (37) with CaM wrapped around the peptide. In reality, the anchor residue for CaM-N binding was a Phe rather than Val, 17 residues away from the anchoring Trp residue at CaM-C (Fig. 10B), creating a new 1-18 binding mode (38). However, in our case, extension of residues at the C terminus of Orai1-CMBD is not feasible as Orai1-CMBD is imme-   (13). In both B and C, the residues are labeled with their amino acid numbers according to human Orai1 sequence. Ser 75 is colored in red. Note that, N-terminal to Ser 75 , the crystal structure differs from the hypothetical model. The residues that reflect the discrepancy between the structure and the hypothetical model are colored in green. In all three panels, residues Leu 74 , Trp 76 , and Leu 79 involved in hydrophobic interaction with CaM are boxed.
diately N-terminal to the 1st transmembrane helix (Fig. 11A). Therefore, no additional residues can be added to the C terminus of Orai1-CMBD to provide a binding site for CaM-N to allow CaM to adopt a collapsed structure without incurring a clash between CaM-N and the plasma membrane.
Possible Mechanism for Ca 2ϩ ⅐CaM-dependent Inhibition (CDI) of Orai1-Based on this unique feature of the CaM/Orai1 interaction, we propose a mechanism for the CDI of Orai1. Orai1 exists as a dimer in the resting state, and upon activation by STIM1, forms a tetramer (39). STIM1 was reported to interact with the cytosolic N and C termini of Orai1 to activate ion flux (11,12). We therefore propose that CaM-C interacts with one subunit of Orai1 followed by CaM-N interacting with another adjacent subunit in the Orai1 oligomer (Fig. 11B). In so doing CaM displaces STIM1 from interacting with Orai1, and thus inactivates the channel.
Interpretation of Different Binding Affinities of Orai1-CMBD for CaM-N and CaM-C-ITC data showed that the binding affinity of Orai1-CMBD for CaM-C is four times stronger than that for CaM-N. To interpret this difference, we superimposed CaM-N with CaM-C to obtain the model of CaM-N⅐Orai1-CMBD complex (Fig. 2). By doing so, we noticed three significant clashes between CaM-N and the peptide (Fig. 2C). Two of  . Proposed mechanism for the interaction of CaM with Orai1 to carry out CDI. A, schematic presentation of the proposed Orai1 topology. The sequence of Orai1-CMBD from the structure is shown in circles. Leu 74 , Trp 76 , and Leu 79 that are involved in hydrophobic interaction with CaM are colored in red. The three residues between the Orai1-CMBD and the first transmembrane helix (TM1) are colored in green. B, proposed binding of one CaM to two Orai1. Orai1 is shown as a tetramer. Subunit 1 is displayed with four transmembrane helices. The remaining three subunits are displayed as cylinders. Here, one CaM (blue ribbon) binds to two CMBDs (green ribbons) of the two adjacent Orai1 subunits, displacing STIM1 from binding to Orai1 to inactivate the channel. The other CaM for binding to subunits 3 and 4 is not displayed for clarity.