Research Article

Crystal Structure of the Calcium Release–Activated Calcium Channel Orai

See allHide authors and affiliations

Science  07 Dec 2012:
Vol. 338, Issue 6112, pp. 1308-1313
DOI: 10.1126/science.1228757


The plasma membrane protein Orai forms the pore of the calcium release–activated calcium (CRAC) channel and generates sustained cytosolic calcium signals when triggered by depletion of calcium from the endoplasmic reticulum. The crystal structure of Orai from Drosophila melanogaster, determined at 3.35 angstrom resolution, reveals that the calcium channel is composed of a hexameric assembly of Orai subunits arranged around a central ion pore. The pore traverses the membrane and extends into the cytosol. A ring of glutamate residues on its extracellular side forms the selectivity filter. A basic region near the intracellular side can bind anions that may stabilize the closed state. The architecture of the channel differs markedly from other ion channels and gives insight into the principles of selective calcium permeation and gating.

In many cell types, the depletion of calcium ions (Ca2+) stored within the endoplasmic reticulum (ER) causes the opening of Ca2+-selective channels in the plasma membrane (1). The resulting influx of Ca2+ through these calcium release–activated calcium (CRAC) channels generates sustained intracellular Ca2+ signals, the roles of which are well appreciated in cells of the immune system, where they mediate signaling through B cell, T cell, and Fc receptors (1). Although this process of “store-operated Ca2+ entry” has been recognized for decades, Orai and stromal interaction molecule (STIM), the CRAC channel pore and its regulator, were identified fairly recently (27). Orai, located in the plasma membrane, forms the Ca2+ pore (810). Humans have three Orai proteins (Orai1 to -3). A mutation in Orai1 that renders the channel nonconductive [Arg91→Trp91 (R91W) (11)] causes a form of immune deficiency, underscoring its importance in the immune system (5). STIM is a single-pass membrane protein in the ER that senses the luminal Ca2+ concentration. It opens the pore of Orai following Ca2+ store depletion, probably via interaction between the cytosolic portions of STIM and Orai, which accumulate at ER-plasma membrane junctions (1, 12).

Orai proteins constitute a highly conserved ion-channel family and have no discernible sequence homology with other ion channels (57), suggesting distinctive architecture and mechanisms for selective ion permeation and gating. From sequence analysis, Orai is predicted to have four transmembrane helices (M1 to M4). The chemical accessibility and intersubunit cross-linking of cysteine residues introduced into the channel by mutagenesis predict that residues on the M1 helix contribute to the ion-conduction pore (13, 14). Several lines of evidence indicate that Orai channels are multimers, but studies have led to differing conclusions about the oligomeric state. Some studies suggest a tetrameric assembly (1518), whereas others hint at a higher-order oligomer (14) or raise the possibility that the oligomeric state of the channel might change when it is activated by STIM (15, 18).

Under physiological conditions, in which the extracellular Ca2+ concentration is 1 to 2 mM, CRAC channels select for Ca2+ over monovalent cations such as sodium (Na+) or potassium (K+) by more than 1000:1, placing them among the most highly-selective Ca2+ channels known (19). However, as is the case for many Ca2+-selective channels, when divalent cations are removed from the extracellular solution, CRAC channels become permeant to monovalent cations, including K+ and Na+ (19, 20). Both the Ca2+ and monovalent cation currents through CRAC channels require activation by STIM and are blocked by gadolinium (Gd3+) and lanthanum (La3+) ions from the extracellular side (inhibition constant Ki < 100 nM) (8, 21, 22).

To better understand ion permeation, ion selectivity, and gating in CRAC channels, we have determined the crystal structure of Orai from Drosophila melanogaster, which shares 73% sequence identity with human Orai1 within its transmembrane region. We have also determined the structure of the nonconductive K163W mutant, corresponding to the R91W mutant in Orai1 that causes immune deficiency in humans. The structures define the architecture of a Ca2+ channel, establish principles of its Ca2+ selectivity, and give insight into the mechanism of activation by STIM.

Structure determination and channel reconstitution. To obtain well-diffracting crystals of Orai, we evaluated the biochemical behavior of ~50 orthologs and engineered a variety of protein forms, including truncations and fusions. Crystals that diffracted to 3.35 Å resolution were obtained from a construct (termed Oraicryst) spanning residues 132 to 341 of Orai from Drosophila melanogaster in which two nonconserved cysteine residues and two residues in the hypervariable M3-M4 loop were mutated (see the supplementary materials and methods). The construct corresponds to the conserved region of Orai and contains all regions necessary for activation by STIM (fig. S1) (2325).

We tested the function of the purified channel protein by reconstituting it into liposomes and using a fluorescence-based assay to monitor ion flux under divalent ion-free conditions (Fig. 1). As would be expected for a closed channel before activation by STIM, ion flux was not detected for Oraicryst or for the K163W mutant. Introduction of the mutation V174A (V102A in Orai1), which creates channels that are open in the absence of STIM (26), yielded a robust decrease in fluorescence, indicative of ion flux (Fig. 1B). We confirmed that the fluorescence decrease was due to ion flux through Orai by adding the pore-blocker Gd3+, which eliminated the observed flux through the pore but had no effect on the baseline fluorescence induced by the Na+-ionophore monensin (Fig. 1C). Therefore, the reconstituted protein recapitulated known properties of Orai.

Fig. 1

Channel reconstitution in liposomes. (A) Schematic of the fluorescence-based flux assay. Vesicles containing Orai or those prepared without protein (empty vesicles) were loaded with 150 mM NaCl and diluted 100-fold into flux buffer containing a fluorescent pH indicator (ACMA) and 150 mM N-methyl-d-glucamine (NMDG) to establish a Na+ gradient. After stabilization of the fluorescence signal (150 s), a proton ionophore (CCCP) was added to the sample, and an electrical potential arising from Na+ efflux was used to drive the uptake of protons into the vesicles, which quenches the fluorescence of ACMA. The “X” indicates that ACMA is no longer membrane-permeable in the protonated form. (B) Fluorescence measurements for the indicated protein constructs of Orai. Monensin, a Na+ ionophore, was added after 990 s to render all vesicles permeable to Na+ and establish a minimum fluorescence baseline. Fluorescence was normalized by dividing by the initial value. (C) Fluorescence trace observed for V174A Oraicryst in the absence and presence of Gd3+.

The crystals form in the space group P213 and contain two Orai subunits (referred to as subunits A and B of the channel) within the asymmetric unit. Experimental phases were derived from diffraction data collected from crystals containing mercury, iridium, and gadolinium heavy atoms. The phases, which were improved by noncrystallographic symmetry averaging, yielded an electron-density map that allowed placement of the majority of the amino acid side chains (fig. S2). The atomic model contains amino acids 144 to 334 of Orai, except for the M1-M2 loop (amino acids 181 to 190), the M2-M3 loop (amino acids 220 to 235), and amino acids 330 to 334 of subunit B, which are disordered. Anomalous-difference electron-density peaks for sulfur atoms of methionine and cysteine residues confirm the correct assignment of amino acids in the atomic model (fig. S2E), which is refined to a free residual (Rfree) of 28% (table S1).

Architecture of Orai. The crystal structure shows that the channel is formed from a hexamer of six Orai subunits arranged around a central axis (Fig. 2). Cross-linking and light-scattering data are consistent with a hexameric assembly for both purified Orai in detergent and Orai expressed in cell membranes (fig. S3). Each subunit contains four transmembrane α helices (M1 to M4) and a helix following M4 that extends into the cytosol (termed the M4 extension helix). The M4 extension helices of neighboring subunits adopt one of two alternating conformations and pack with each other in pairs, giving this region of the channel threefold symmetry (Fig. 2). Within the channel’s transmembrane region, there is sixfold symmetry.

Fig. 2

Architecture of Orai. (A) Ribbon representation showing the tertiary structure of the channel from the side. The helices are colored: M1 (blue), M2 (red), M3 (green), M4 (brown), M4 extension (yellow in subunit A and gray in subunit B). Also shown are a Ca2+ ion (magenta sphere) and the nearby Glu178 residues (yellow sticks). Based on the hydrophobic region of the channel’s surface, horizontal lines (~30 Å apart) suggest approximate boundaries of the inner (In) and outer (Out) leaflets of the membrane. (B) Orthogonal view of the channel from the extracellular side. (C) Close-up view showing the interaction between the M4 extension helices.

The ion pore is located at the center of the channel along the sixfold axis, and the transmembrane helices are arranged in three concentric rings (Fig. 2B). Six M1 helices, one from each subunit, make up an inner ring of helices and line the ion pore. The M2 and M3 helices constitute a middle ring that surrounds the transmembrane portion of the M1 helices and separates them from M4 helices, which are arranged in an outer ring and located at the periphery of the channel. The channel extends just beyond the extracellular side of the membrane, and the M1 helix and the M4 extension helix protrude 20 to 25 Å into the cytosol.

The M1 helices extend from Thr144 to Gln180 and are ~55 Å long. The region of M1 that extends into the cytosol was not expected to be part of M1 from sequence analysis and includes four helical turns at its N terminus (residues 144 to 157). As this region does not make contacts with other parts of the channel in the structure, it could conceivably interact with cellular factors such as STIM.

The M2 helix traverses just beyond the thickness of a lipid bilayer, and the M3 helix extends about two helical turns further into the cytosol. The arrangement of the M2 and M3 helices in a ring surrounding the transmembrane portion of the M1 helices, while leaving the cytosolic portion of M1 free, suggests that M2 and M3 help provide structural integrity to the transmembrane region of the pore but would permit widening of the intracellular region of the pore.

The M4 helices are located furthest away from the ion pore. In comparison to the M1 to M3 helices, the M4 helices have fewer contacts with other parts of the channel, and their peripheral location may allow mobility (Fig. 2A). A bend near the middle of the transmembrane region (Pro288) means that approximately half of M4 (within the outer leaflet) is roughly perpendicular to the plane of the membrane, whereas the other half is oriented diagonally (Fig. 2A). Connected to M4 by a highly conserved hinge region (residues 305 to 308), the M4 extensions of subunit A and subunit B pair with one another through an antiparallel coiled-coil helix packing arrangement (Fig. 2C). The interaction involves two residues from each subunit (Ile316 and Leu319) that form a hydrophobic patch on the M4 extension, which is otherwise predominantly decorated with hydrophilic amino acids. Mutation of either residue (L273S or L276D mutants of Orai1) disrupts interaction with STIM and its ability to open the channel (2729).

Chemical composition of the pore. The chemical environment along the length of the ion pore is determined by side chains of the amino acids of M1 that line the pore. This is in marked contrast to the pore of K+ channels, where a substantial part of it is formed by an extended conformation of the polypeptide backbone (30). The pore of Orai has four distinct sections (Fig. 3, A and B): (i) a ring of glutamates at its extracellular end (termed the glutamate ring and composed of six Glu178 side chains), (ii) a hydrophobic section spanning three helical turns of M1 within the transmembrane region, (iii) a basic section spanning three helical turns near the intracellular side, and (iv) a wider section that extends into the cytosol.

Fig. 3

Ion pore. (A) Two M1 helices are drawn (four are omitted for clarity), showing the amino acids lining the pore in yellow. The approximate boundaries of the membrane-spanning region are shown as horizontal lines. The corresponding amino acids in human Orai1 are shown in parentheses. Ser161, Ser162, Thr164, Ser165, Ser169, and Gly170 are drawn as gray sticks. (B) View of the pore. Within a ribbon representation of four M1 helices (two in the foreground are removed for clarity) is a representation (teal) of the minimal radial distance from the center to the nearest van der Waals protein contact. The sections of the pore discussed in the text are labeled. (C and D) Molecular surface of Orai viewed from the extracellular (C) and intracellular (D) sides and colored according to the electrostatic potential contoured from –10 kT (red) to +10 kT (blue) (dielectric constant: 80).

The pore is ~55 Å long, and its diameter varies along its distance (Fig. 3B). In the glutamate ring, the Glu178 side chains point toward the center of the pore with their side chain oxygen atoms separated by ~6 Å across the pore, giving the extracellular entrance to the pore negative electrostatic potential (Fig. 3, B and C). Glu178, which is conserved in Orai channels, is the only acidic residue lining the pore (31).

The side chains of Val174, Phe171, and Leu167 are separated by 8 to 10 Å across the pore and line the walls of the hydrophobic section, which is roughly tube-shaped and ~18 Å long (Fig. 3B). The hydrophobic side chains are well packed in the pore, make extensive van der Waals interactions with one another, and are strictly conserved among Orai channels. This region also has well-defined electron density (fig. S2) and the lowest temperature factors of the structure, suggesting that it is relatively rigid. In addition to lining the pore, the hydrophobic residues form a core of the protein that likely contributes to the channel’s structural integrity. Mutations of amino acids in the hydrophobic region have yielded channels with dramatically altered ion-conduction properties. For instance, mutations of Val174 (Orai1 mutants V102C, V102A, V102T, or V102S) or Gly170 (Orai1 mutants G98P and G98D), which closely packs its Cα carbon with the side chain of Phe171, result in channels that are constitutively conductive and demonstrate altered ion selectivity (26, 32). The close packing in the hydrophobic region suggests that the functional abnormalities of these mutants may be due to reduced structural integrity and/or slightly altered packing in the pore.

Below the hydrophobic region in Fig. 3B and near the intracellular side of the membrane, the pore is lined by three basic residues from each subunit (Lys163, Lys159, and Arg155) that are conserved as lysine or arginine among Orai orthologs (fig. S1). The 18 lysine or arginine residues that line this portion of the pore create an unexpected environment for the pore of a channel that conducts cations (Fig. 3D). The M1 helix is slightly bent at Ser162 and is enriched for residues that are often associated with bends in helices (serine, threonine, and glycine) in the region that spans the junction between the hydrophobic and basic portions of the pore (Fig. 3A and fig. S4). This raises the possibility that the conformation of the basic region could change, perhaps as part of the gating process, while the hydrophobic section and glutamate ring remain fairly fixed.

External Ca2+ binding and ion selectivity. In crystals soaked in Ca2+, electron density consistent with a Ca2+ ion is observed on the extracellular side of the glutamate ring (in what we term the external site) (Fig. 4A). Because it is difficult to distinguish Ca2+ from other ions or water molecules in electron-density maps, we collected diffraction data from a crystal soaked in barium (Ba2+), which permeates through the CRAC channel (22, 33) and can be identified in electron-density maps from its anomalous diffraction signal. The anomalous-difference electron-density map contained a 10σ peak for Ba2+ in the external site, consistent with the assessment that it is a binding site for Ca2+ (Fig. 4A). Ba2+ is positioned slightly farther away from the glutamate ring than Ca2+, and this correlates with the ionic radii of the ions (1.35 Å for Ba2+ and 0.99 Å for Ca2+). It is possible that the ion coordination, which could be either direct or water-mediated, is not sixfold symmetric due to the flexibility of glutamate side chains. The backbone carbonyl oxygen of Glu178 at the C-terminal end of M1 might also contribute to Ca2+ binding (Fig. 4A). To investigate if Gd3+ blocks the channel by binding to the same site, we collected diffraction data from crystals soaked in Gd3+, which also has a strong anomalous diffraction signal. The anomalous-difference electron-density map showed a 14σ peak for Gd3+ in the external site (Fig. 4B). The density is positioned closer to the glutamate ring than for Ca2+; this may be indicative of the high affinity of Gd3+ and its +3 charge. Notably, the single Ca2+ binding site identified is located just outside the pore rather than within it. This configuration of ion binding distinguishes Orai from K+ channels, which have multiple ions present within the pore in their structures (34), and this probably reflects differences in the mechanisms for ion permeation and selectivity.

Fig. 4

Cation binding in the external site. (A) Ba2+ and Ca2+ binding. The glutamate ring and hydrophobic section of the pore are depicted as in Fig. 3A. Anomalous-difference electron density for Ba2+ is shown in green mesh (calculated from 20 to 5 Å resolution using phases derived from the final model and contoured at 6σ from a diffraction data set collected from a crystal soaked in 50 mM BaCl2 using x-rays with λ = 1.70 Å. Electron density corresponding to Ca2+ is shown from a simulated annealing Fo-Fc omit map (blue mesh, calculated from 20 to 3.8 Å resolution and contoured at 4σ) from a crystal soaked in 50 mM CaCl2 (table S1). The backbone carbonyl oxygen of Glu178 is also shown in stick representation. (B) Gd3+ binding. Anomalous-difference electron density is shown in purple mesh from a crystal soaked in 1 mM GdCl3 (calculated from 20 to 5 Å resolution using model phases and contoured at 11σ from a diffraction data set collected using λ = 1.70 Å x-rays).

CRAC channels may achieve their exquisite Ca2+ selectivity by high-affinity binding of Ca2+ to the glutamate ring that probably constitutes the ion selectivity filter. Accordingly, mutation of the glutamate (Glu178 in Orai or Glu106 in Orai1) to aspartic acid dramatically increases the channel’s permeability to monovalent cations (810, 13). Because the ionic radii of Ca2+ and Na+ are nearly identical (0.99 and 0.95 Å, respectively), selectivity for Ca2+ is most likely due to the ion’s greater positive charge. Consistent with this trend, we observe that high-affinity blockage of Orai by the trivalent lanthanide Gd3+ is due to binding in the external site.

Identification of the external site raises the question of whether there is only one Ca2+ binding site or if there are other site(s) that are occupied during permeation. In Orai, the Ca2+ permeation rate [up to ~104 ions per second (35, 36)] is much slower than the rate at which Ca2+ ions would diffuse to the extracellular mouth of the pore [~106 ions per second at physiological Ca2+ concentration (37)], so a single high-affinity Ca2+ binding site with a fast on-rate (from the extracellular solution) and a slow off-rate (into the pore) might be able to account for Ca2+-selective permeation. However, some evidence, including an anomalous mole fraction effect (33), suggests that there may be more than one Ca2+ binding site. A second binding site might account for the property that, although Na+ currents through CRAC channels are blocked by micromolar concentrations of Ca2+, Ca2+ does not appreciably flow until the external solution contains millimolar concentrations of it (19, 20). Additionally, the extent of blockage of Na+ current at micromolar levels of Ca2+ is dependent on the voltage applied across the membrane (~10% block at –20 mV and ~50% block at –80 mV with 20 μM external Ca2+) (36, 38), which suggests that an ion binding site is located within the voltage gradient, whereas the external site would probably be located outside of it.

If a second binding site for Ca2+ in the selectivity filter exists, we expect that it would be located just below the glutamate ring. This hypothetical internal site, which might only be occupied transiently during permeation, would reside within a voltage field across the membrane and may explain the voltage-dependence for blockage of Na+ current by Ca2+. The binding of a Ca2+ ion in the external site could displace a Ca2+ ion from the internal site due to electrostatic repulsion between the ions, thereby allowing selective Ca2+ permeation.

Anion binding in the basic region of the pore. Electron-density maps have strong density in the basic region of the pore that appears to plug it. The anomalous-difference electron density was also strong in this site (up to 20σ), aiding the identification of the bound atom(s) (Fig. 5A). The density cannot be due to the surrounding protein side chains because they would have negligible anomalous scattering, and the components of the crystallization solution did not contain elements that would give a strong anomalous signal. To identify possible metal ions that might have copurified with Orai, purified protein was analyzed by inductively coupled plasma mass spectrometry (ICP-MS) and found to contain ~0.6 molar equivalents of iron and ~0.1 molar equivalents of zinc (table S2), both of which have anomalous scattering properties. Diffraction data collected using x-rays with a wavelength (λ) of 1.735 Å, slightly shorter than an absorption edge for iron (1.7433 Å) where the anomalous signal from iron would be strong (the imaginary component of the anomalous scattering factor f′′ for iron is approximately 3.9 electrons at λ = 1.735 Å), yielded a 20σ peak in the anomalous-difference map (Table 1). Using λ = 1.900 Å x-rays, where the anomalous signal from iron would be weaker (f′′ ~ 0.5 electrons), the peak was 5σ. Fo-Fc electron-density maps calculated from these data sets had comparable (~10σ) peaks, indicating that the difference in the strength of the anomalous signal was not due to differing occupancies of the bound entity between the crystals. Data sets collected above and below an absorption edge for zinc (λ = 1.284 Å) yielded comparable anomalous-difference density in the pore, but indicated that zinc was bound in a crystal contact (fig. S6). We therefore conclude that a major species bound in the basic region of the pore contains iron, which was present in the media used for cell growth.

Fig. 5

Anion binding and the K163W mutant. (A) Anomalous-difference electron density in the basic region of the WT pore contoured at 7σ (orange mesh) and 14σ (blue mesh) from a native unsoaked crystal. The M1 helices are depicted as in Fig. 3A. The map was calculated using phases derived from the protein model and a diffraction data set collected using λ = 1.735 Å x-rays (resolution range 20 to 5 Å) (Table 1). (B) Anomalous-difference electron density present in the WT pore from a crystal soaked in (IrCl6)3–. The map is contoured at 7σ (green mesh) and 20σ (red mesh) (calculated using model phases from 20 to 5 Å resolution from a diffraction dataset collected using λ = 1.1033 Å x-rays). (C and D) The K163W mutant. Two M1 helices of the K163W mutant are depicted in the same manner as in (A). (C) Electron density (cyan mesh) is shown for the tryptophan side chains at residue 163 (represented as sticks). The density is from a simulated annealing Fo-Fc map in which the tryptophan side chains have been removed from the model (calculated from 20 to 3.35 Å resolution and contoured at 3σ). (D) Anomalous-difference electron density (green mesh) in the pore of a K163W mutant crystal soaked in (IrCl6)3– (calculated from 20 to 5 Å resolution using model phases and contoured at 7σ from a diffraction data set collected using λ = 1.1033 Å x-rays).

Table 1

Electron density in the basic region using different x-ray wavelengths. Peak heights of the Fo-Fc and anomalous-difference electron-density maps within the basic region are indicated for diffraction data sets from crystals (native unsoaked crystals from the same protein preparation and of comparable size and diffraction quality) that were collected at the indicated wavelengths. Fo-Fc maps were calculated using protein model phases (without ions in the pore) after rigid body refinement and with the resolution range 20 to 4 Å. Anomalous-difference electron-density maps were calculated using these phases and the resolution range 20 to 5 Å. Additional statistics are listed in table S3. The theoretical values for f′′ were obtained from the Web site.

View this table:

Because this region of the pore is highly basic, we tested whether it could bind anions by soaking crystals in the complex anion iridium hexachloride, (IrCl6)3–. Diffraction data were collected using λ = 1.1033 Å x-rays to optimize the anomalous diffraction signal from iridium (f′′ ~ 10.1 electrons), where it would be stronger than that for iron (f′′ ~ 1.8 electrons) (Table 1). The anomalous-difference electron-density map contained two peaks in the basic region: (i) a 29σ peak for (IrCl6)3– between Lys163 and Lys159 and (ii) an 8σ peak adjacent to Arg155 (Fig. 5B). Otherwise, the overall structure of Orai complexed with (IrCl6)3– was indistinguishable from the native structure. We conclude that the basic region can bind anion(s). Because we observe iron in this anion binding site, we speculate that it is also bound in the form of a complex anion or mediated by counter ions. In support of this, hexachloroferrate anions, (FeCl6)3–, are known to be stabilized by hydrogen-bond interactions with amines (39). Although possible, it seems unlikely that iron binds within the pore in a physiological setting because of the limited amount of iron freely available within cells. More likely, the basic region of the pore binds available intracellular anion(s), with possible candidates being a sulfate or phosphate species (phosphate, pyrophosphate, etc.).

The K163W mutant. The crystal structure of the nonconductive K163W mutant gives further insight into the basic region of the pore. The overall structure of the K163W mutant is indistinguishable (at the current resolution) from the structure with the wild-type (WT) pore (root mean square deviation for Cα atoms is 0.1 Å). The only discernible difference with regard to the protein is that the side chain of the tryptophan residue at position 163 projects into the pore (Fig. 5C). The six tryptophan residues, one from each subunit, are tightly packed and extend the hydrophobic region of the pore an additional helical turn.

Electron-density maps for the K163W mutant did not contain the anomalous-difference density within the pore ascribed to iron. Accordingly, ICP-MS analysis of the K163W mutant protein indicated that it contains only about one third as much iron (table S2). Lys159 and Arg155 have less well-defined electron density in the mutant, presumably due to the absence of an anion. To explore the effect of the K163W mutant on the ability of the pore to bind anions, we calculated an anomalous-difference electron-density map from a crystal soaked in (IrCl6)3– and found that its distribution in the pore is different than for the WT pore. For the K163W mutant, we observed a single peak (12σ) in the pore, located between Lys159 and Arg155 (Fig. 5D), indicating that the mutant eliminates one of the anion binding sites.

Ion permeation and gating. Flux assays indicate that the pore of Oraicryst is closed, as would be expected before activation by STIM (Fig. 1). Because of the narrow width of the basic region of the pore and its extreme positive charge (Fig. 3B), we suspect that this section is not permeable to Ca2+ in its current conformation and forms a closed gate. Hydrophobic amino acids substituted for Lys163 (Arg91 in human Orai1), within the basic region, prevent channel activation (40). The structure of the K163W mutant indicates that the tryptophan residues make extensive hydrophobic interactions with each other and that the mutation eliminates one of the anion binding sites. Anion(s) bound in the basic region of a WT pore would probably occlude the pore and help stabilize the closed conformation. Whereas the hydrophobic packing in the mutants would probably be static, locking the channel in a closed conformation, an anion bound at Lys163 could dissociate, allowing the WT pore to open.

The structure suggests that the glutamate ring and hydrophobic section would be permeable to Ca2+ in the observed conformation. The distance between the oxygen atoms of the glutamate ring and the central axis of the pore is ~3 Å, which is similar to the typical coordination distance between Ca2+ and oxygen in crystal structures of Ca2+-binding proteins (~2.4 Å) (41). This suggests that a Ca2+ ion would pass through the glutamate ring by coordinating the side chain oxygen atoms directly and becoming at least partially dehydrated. After emerging from the glutamate ring, the permeating Ca2+ ion would likely coordinate additional waters within the hydrophobic section. In support of this hypothesis, electron density consistent with water molecules is present within the hydrophobic section (fig. S5).

Experiments suggest that a cytosolic portion of STIM interacts with Orai to activate the channel (24, 4244) and that the interaction involves N- and C-terminal regions of Orai that correspond to the cytosolic region of M1 and the M4 extension in the structure (23, 24, 2729, 45). Deletion of the M4 extension or mutation of the residues that are observed to mediate its coiled-coil packing (Ile316 or Leu319, corresponding to Leu273 and Leu276 in human Orai1) disrupts STIM binding and channel activation (23, 24, 2729). This raises the possibility that STIM, which contains regions predicted to form coiled-coils, binds to the M4 extension by substituting for the coiled-coil interaction observed between the Orai subunits and that the arrangement of the M4 extensions in the structure is a quiescent one prior to the binding of STIM. Although C-terminal modifications of Orai disrupt STIM binding, deletion of the cytosolic portion of M1 or certain mutations within it (including the Orai1 mutants R91W and Lys85, which faces away from the pore, to glutamate) have little effect on STIM binding but prevent channel activation (23, 24, 45). These observations are consistent with a hypothesis proposed by Park et al. that STIM binding provides the energy for channel gating by bridging the N- and C-termini of the channel (24).

Based on the structure and the available functional data, we constructed a working model for an open conformation of the channel (Fig. 6). In this model, the M4 extension helices project (>50 Å) into the cytosol and interact with a cytosolic portion of STIM. The basic region of the pore is dilated by the outward bending of the M1 helices (modeled to occur between Ser161 and Gly170), which interact via their N-terminal regions with STIM. The M2-M3 loops (amino acids 220 to 235), while disordered in the current structure, might also make contributions, given their intracellular location and substantial conservation (fig. S1). Whether slight or dramatic, the dilation is expected to reduce the affinity of the basic region for anion(s), allowing them to be displaced.

Fig. 6

Observed structure and hypothetical open state. (A) Observed (apparently closed) structure of Orai. The view is from the side with the M1 helices drawn as blue ribbons and the other helices shown as cylinders (M2, red; M3, green; M4 and M4 extension, yellow). A Ca2+ ion in the external site is depicted as a magenta sphere; an anion in the basic region of the pore is depicted as a gray sphere. Approximate boundaries of the lipid membrane are shown as horizontal lines. (B) Hypothetical model of an open state. The pore is widened by the outward dilation of the M1 helices (left-right arrow). A downward-pointing arrow indicates that Ca2+ is able to move though the pore unobstructed. The intracellular ends of the M1 helices are thought to interact with a cytosolic portion of STIM, as are the M4 extensions, which are modeled to protrude into the cytosol. The depiction of a cytosolic portion of STIM is meant to suggest that it might bridge the cytosolic portions of the M1 helices and the M4/M4 extension helices and is not meant to imply a particular structure, oligomeric state, or stoichiometry with Orai.

Discussion. The crystal structure of Orai reveals an architecture that appears well suited to its physiological role. The ~10,000-fold resting concentration gradient of Ca2+ across the plasma membrane (~1 mM extracellular and ~100 nM intracellular Ca2+ concentrations) must be maintained to prevent aberrant Ca2+ signaling within cells; therefore, Orai must be tightly sealed shut when it is closed. Anion(s) bound in the basic region probably help serve this purpose. In an open conformation, the basic region would tend to prevent the permeation of cations that have modest concentration gradients (such as K+ and Na+, which have ~50-fold gradients) and, thus, would contribute to Ca2+-selective permeation under physiological conditions. To establish a slow permeation rate that prevents overloading of the cell with Ca2+ and allows for fine tuning of sustained Ca2+ signals, the channel imposes energy barriers to ion permeation, which could include dissociation of Ca2+ from the selectivity filter (into the pore), permeation through the hydrophobic section, and electrostatic repulsion by the basic section. Ultimately, the channel’s design takes advantage of the large Ca2+ gradient across the plasma membrane, while tightly regulating the flow of Ca2+ into the cell to mediate one of the most fundamental types of signaling in cell biology.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S6

Tables S1 to S3

References (4669)

References and Notes

  1. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.
  2. Aspartate residues on the disordered M1-M2 loop (D182 and D184, corresponding to D110 and D112 in Orai1) probably also contribute to the acidic nature of the extracellular mouth of the pore.
  3. Acknowledgments: We thank J. Goldberg, M. B. Long, R. MacKinnon, and N. P. Pavletich for discussions and comments on the manuscript; C. D. Lima and members of the laboratory for discussions; E. Gouaux for green fluorescent protein fusion vectors; the staff of beamlines X25 and X29 (National Synchrotron Light Source, Brookhaven National Laboratory); the staff of the monoclonal antibody and proteomics and microchemistry facilities at MSKCC; the Keck Biophysics Resource at Yale University; and the Center for Applied Isotope Studies at the University of Georgia for protein analysis. This work was supported by the NIH (grant GM094273) and by a Burroughs Wellcome Career Award (to S.B.L). Atomic coordinates and structure factors have been deposited in the Protein Data Bank with accession numbers 4HKR (wild type) and 4HKS (K163W mutant).
View Abstract

Stay Connected to Science

Navigate This Article