Structural determinants of ion permeation in CRAC channels Beth A. McNally1, Megumi Yamashita1, Anita Engh, and Murali Prakriya2 Department of Molecular Pharmacology and Biological Chemistry, Northwestern University School of Medicine, Chicago, IL 60611 Edited by David E. Clapham, Children’s Hospital Boston, HHMI, Boston, MA, and approved November 3, 2009 (received for review August 25, 2009)
CRAC channels generate Ca2ⴙ signals critical for the activation of immune cells and exhibit an intriguing pore profile distinguished by extremely high Ca2ⴙ selectivity, low Csⴙ permeability, and small unitary conductance. To identify the ion conduction pathway and gain insight into the structural bases of these permeation characteristics, we introduced cysteine residues in the CRAC channel pore subunit, Orai1, and probed their accessibility to various thiolreactive reagents. Our results indicate that the architecture of the ion conduction pathway is characterized by a flexible outer vestibule formed by the TM1-TM2 loop, which leads to a narrow pore flanked by residues of a helical TM1 segment. Residues in TM3, and specifically, E190, a residue considered important for ion selectivity, are not close to the pore. Moreover, the outer vestibule does not significantly contribute to ion selectivity, implying that Ca2ⴙ selectivity is conferred mainly by E106. The ion conduction pathway is sufficiently narrow along much of its length to permit stable coordination of Cd2ⴙ by several TM1 residues, which likely explains the slow flux of ions within the restrained geometry of the pore. These results provide a structural framework to understand the unique permeation properties of CRAC channels. Orai1 兩 STIM1 兩 store-operated channels
C
a2⫹ release-activated Ca2⫹ (CRAC) channels are the principal route of Ca2⫹ entry in immune cells and orchestrate functions such as gene expression, motility, and the release of inflammatory mediators (1). Mutations in CRAC channels give rise to devastating immunodeficiencies and abnormalities in muscle, skin, and teeth, highlighting their importance for various organ systems (1). The recent discoveries of STIM1 (the ER Ca2⫹ sensor), and Orai1 (the CRAC channel pore subunit) have provided major breakthroughs to illuminate the molecular basis of CRAC channel function (2). However, while the identification of these proteins has produced rapid progress in our understanding of the cellular events underlying channel activation, the molecular mechanisms of ion selectivity and permeation remain unclear. CRAC channels are distinguished by an extraordinarily high selectivity for Ca2⫹ over monovalent ions (PCa/PNa ⬎ 1,000), a very low unitary conductance (⬍1 pS), and low permeability to Cs⫹ and larger monovalent cations (3). The structural underpinnings of these characteristics have been the focus of much debate but are largely unknown. As with most ion channels, the pore properties of CRAC channels are likely shaped by the arrangement and chemistry of pore-lining residues. Thus, to understand the basis of the unique permeation properties of CRAC channels, the residues lining the ion transport pathway need to be elucidated. Orai1 bears little sequence homology to other ion channel proteins, and consequently, there are few clues regarding the contribution of the different parts of the molecule for pore formation. Electrophysiological studies indicate that the exquisite Ca2⫹ selectivity of CRAC channels arises from high affinity Ca2⫹ binding within the channel, resulting in the occlusion of monovalent cation permeation (4–6). Additionally, recent mutagenesis studies have implicated conserved acidic residues, including E106 and E190 in the first and third predicted transmembrane (TM) segments, and D110, D112, and D114 in the TM1-TM2 linker region of human Orai1 in shaping ion selectivity (7–10). These studies have led to the 22516 –22521 兩 PNAS 兩 December 29, 2009 兩 vol. 106 兩 no. 52
notion that the TM1 and TM3 segments of Orai1 flank the ion conduction pathway of the CRAC channel with the acidic residues within these segments forming coordinating sites for the conducting ions (9, 11). While the mutagenesis studies provided a glimpse into the identity of candidate residues that might regulate Ca2⫹ selectivity, the roles of the various acidic residues for Ca2⫹ coordination and ion selectivity remain uncertain. For example, although a Gln substitution of E190 altered ion selectivity, substitutions to Ala or Asp had no detectable effects (7), raising questions about the true energetic contributions of the identified residues for ion binding. In addition, there is no structural framework to understand the basis of the CRAC channel’s low unitary conductance or its unusual selectivity profile. To address these issues, we have used the substituted cysteine accessibility method to identify critical porelining residues of the CRAC channel and used this information to deduce the architecture of the pore and the contribution of pore-lining residues for Ca2⫹ coordination and ion permeation. Results Mutagenesis studies have implicated acidic residues in TM1, TM3, and the TM1-TM2 loop segments of Orai1 as important determinants of ion permeation and pore diameter of CRAC channels (7–10). To uncover the precise contribution of these regions for pore formation, we systematically substituted cysteines into these segments (Fig. 1) with the rationale that they could serve as targets for thiol-specific reagents applied from the extracellular side of the channel (12). If the engineered Cys side-chain projects into the water-filled pore of the ion channel, its interaction with small, hydrophilic thiol-specific probes would be predicted to produce rapid channel blockade, thus revealing its proximity to the pore (12). We expressed mutant Orai1 constructs together with STIM1 in HEK293 cells and studied the effects of thiol reagents on activated CRAC channels in cells treated with 1 M thapsigargin. Effects of Cysteine Mutations. Barring a few exceptions (H113C,
E106C, and M101C), Orai1 Cys substitutions were well tolerated and produced currents ⬎5 pA/pF in amplitude and with permeation properties characteristic of ICRAC (Table S1). Currents in the E106C and H113C mutants (⬍0.5 pA/pF) were indistinguishable from leak and were not further tested. Currents arising from M101C averaged only 1 pA/pF, but despite their small size, exhibited properties such as inward rectification and positive reversal potentials characteristic of ICRAC. Inactivity of E106C channels is consistent with previous reports showing that Ala and Gln substitutions of this critical residue abrogate channel activity (7–10). The Author contributions: B.A.M., M.Y., A.E., and M.P. designed research; B.A.M., M.Y., A.E., and M.P. performed research; B.A.M., M.Y., A.E., and M.P. analyzed data; and B.A.M., A.E., and M.P. wrote the paper. The authors declare no conflict of interest. This article is a PNAS Direct Submission. Freely available online through the PNAS open access option. 1B.A.M. 2To
and M.Y. contributed equally to this work.
whom correspondence should be addressed. E-mail:
[email protected].
This article contains supporting information online at www.pnas.org/cgi/content/full/ 0909574106/DCSupplemental.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0909574106
N
Cytoplasm
C
reduced currents in the H113C and M101C mutants suggest that the native residues at these positions similarly serve important functions and need to be further studied. La3ⴙ and Ca2ⴙ Sensitivities Reside at Distinct Sites. Although Cys
substitutions in the TM1-TM2 loop did not significantly affect Ca2⫹ selectivity (Table S1), several loop mutants (Q108C, D110C, and D112C) nonetheless exhibited lowered sensitivity to blockade by the trivalent ion, La3⫹ (Fig. S1 A), consistent with findings in Drosophila Orai (8). An intriguing aspect of La3⫹ block was that it was unaffected by the E106D Orai1 mutation, which produces the well known loss of Ca2⫹ selectivity and Ca2⫹ binding to the pore (7–10) (Fig. S1B). In contrast to the effects of Ca2⫹, La3⫹ block was also largely voltage-independent (Fig. S1C). These results indicate that the molecular determinants of high affinity La3⫹ and Ca2⫹ binding are distinct: La3⫹ binds tightly to the TM1-TM2 loop residues, but not deeper in the pore, whereas Ca2⫹ binds tightly deeper in the pore but not significantly to the TM1-TM2 loop residues. Notably, this feature is distinct from voltage-gated Ca2⫹ (Cav) channels, where both La3⫹ and Ca2⫹ binding are believed to occur in the selectivity filter (13). Cys Substitutions at Some Sites Produce Inter-Subunit Disulfide Bonds. In several Orai1 Cys mutants localized to the TM1-TM2
loop, including Q108C, D110C, and D112C Orai1, ICRAC detected at whole-cell break-in could be enhanced by the reducing agent, bis(2-mercaptoethylsulfone) (BMS) (Fig. 2A). We suspected that this effect was mediated by the cleavage of preexisting disulfide bonds between the introduced cysteines, resulting in the relief of channel inhibition. Indeed, examination of Orai1 protein by SDS/ PAGE and Western blot analysis revealed that several Orai1 cysteine mutants exhibited bands corresponding to dimers (Fig. 2C). The dimer/monomer ratio declined when the protein lysate was exposed to a reducing environment, consistent with the presence of an inter-subunit disulfide bond (Fig. 2B). Exposing cells to the mild redox catalyst, copper:phenanthroline (1.5:5 M), also rapidly blocked Q108C current (Fig. S2 A), suggesting that the formation of the disulfide bond can occur in the native folded and functional channel in the plasma membrane. The cysteine mutants exhibiting homodimer formation were Q108C, D110C, A111C, and D112C in the TM1-TM2 loop and E106C in TM1 (Fig. 2C and Fig. S3). This pattern of disulfide bond formation was not altered by substituting the endogenous cysteines in Orai1 (at positions 126, 143, and 195) with alanines (Fig. 2 C and D), indicating that the endogenous cysteines are not involved in the formation of inter-subunit disulfide linkages. Collectively, these results argue that the engineered cysteines at some positions in the TM1-TM2 loop and at E106 spontaneously cross-link with the corresponding cysteines in neighboring subunits to inhibit ion conduction. A practical consequence of this effect was that availability of the introduced cysteines for modification was greatly McNally et al.
1 2
-1000 0
TM3 Orai1 177 AFSTVI GTLLFLAEVV 192 GFSTVLGI LLFLAEVV Orai2 GFSTALGTFLFLAEVV Orai3
Fig. 1. Primary structure and predicted topology of Orai1. (A) Important residues identified from structure-function and human linkage-analysis studies are highlighted. The red boxes indicate the regions of the protein targeted for Cys scanning in this study. (B) Amino acid sequence alignment of the three human Orai isoforms showing conserved residues (shaded) in TM1, the TM1TM2 loop, and TM3.
mV -100 1 2
0
3 30 60 Time (s)
Q108C 250 150 100 75 50 37
BMS 100 pA
3 -600
% BME 0.05 1.25 2.5
C
WT Orai1 background
D
1.0 0.5 0.0 0.05 1.25 2.5 % BME
Cys-less Orai1 background
130 100 70 55
150 100 75 50 37
35 25
dimer monomer
Fig. 2. Cysteine substitutions at some positions form inter-subunit disulfide bonds. (A) Application of BMS (5 mM) immediately following whole-cell break-in enhances ICRAC in Q108C Orai1. (B) Western blot of lysates from cells expressing STIM1⫹Q108C Orai1 reveals Orai1 dimers. Exposing the lysate to increasing concentrations -mercaptoethanol dose-dependently decreases the dimer:monomer protein ratio, confirming disulfide cross-linking between Q108C Orai1 subunits. (C) Western blots of several Orai1 Cys mutants reveal bands corresponding to dimers in E106C, Q108C, D110C, A111C, and D112C Orai1 mutants. (D) The pattern of Orai1 mutants exhibiting disulfide crosslinking is similar in the Cys-less Orai1 background. All cells were treated with tunicamycin to eliminate protein glycosylation (Fig. S3A).
reduced (Fig. S2 B and C). Therefore, in all experiments, cells were first treated with BMS immediately following whole-cell break-in to reduce the disulfide bonds and permit modification by Cys-reactive reagents. Effects of MTS Reagents on ICRAC. To identify the pore-lining cysteines, we applied methanethiosulfonate (MTS) derived reagents differing in size and charge and measured the inhibition of ICRAC persisting following washout of the reagent. We first examined the action of MTS reagents on WT Orai1 channels to test for effects on endogenous cysteines. As exemplified by the response to 2(trimethylammonium)ethyl methanethiosulfonate (MTSET), WT Orai1 currents were inhibited by all MTS reagents examined, but this inhibition fully reversed following washout of the reagent (Fig. 3A). Similar reversible inhibition of ICRAC by MTS reagents was also observed in the Cys-less Orai1 mutant (C126/143/195A), confirming that the inhibition of WT Orai1 by MTSET arises from noncovalent effects on ion conduction (Fig. 3B). Currents in the Cys-less Orai1 mutant were similar to WT Orai1 with respect to high Ca2⫹ selectivity, low permeability to Cs⫹, depotentiation in divalent-free (DVF) solution, and fast inactivation (Fig. S4), indicating that the elimination of the endogenous cysteines does not significantly alter the channel structure. In contrast to WT and Cys-less Orai1, several Cys mutants displayed a large degree of ICRAC inhibition that persisted following washout of the reagent. As illustrated in the response of D110C Orai1 channels to the positively charged MTSET (headgroup size: 5.8 Å) (Fig. 3C), inhibition of the current in the cysteine mutants typically occurred over two phases: a rapid initial phase of inhibition (⬍2 s) was followed by a slow additional decline occurring over seconds, similar to the noncovalent block seen in WT Orai1. Following washout of MTSET, a large portion of the current remained inhibited. This persistent inhibition is indicative of covalent modification of D110C resulting in channel blockade. The ion selectivity of the residual current was unaltered. Exposure to the reducing agent, BMS, reversed the current inhibition (Fig. 3C). Both smaller and larger positively charged MTS reagents (2PNAS 兩 December 29, 2009 兩 vol. 106 兩 no. 52 兩 22517
PHYSIOLOGY
R91 R91
B
Q108C
dimer/monomer
TM1- TM2 Loop Orai1 108 QLDADHDYPPGL 119 Orai2 QLETQYQYPRPL Orai3 QLESDHEYPPGL
M4 T TM4
TM3 TM3
E190 E190
TM2 TM2
TM1 TM1
E106 E106
A
multimer/monome
D110
TM1 90 SRTSALLSGFAMVAMVEV 107 SRTSALLSGFAMVAMVEV SRTSALLSGFAMVAMVEV
E1 V106C Q 07C 1 L1 08C D109C A110C D111C H112C D113C 14 C
Orai1 Orai2 Orai3
I (pA)
B
D112 D114
W E1 T V106C Q 07C 10 L1 8C 0 A1 9 C D 11C 1 D 12C 11 4C
A
B
WT Orai1
I (pA)
MTSET 20 Ca BMS
-500
1 0
C
mV -100
0
Cys-less Orai1 MTSET
100 pA
I (pA)
A
mV -100
100 pA
2,3
3 1-250
2
BMS
0
-100
2
1
3
2 1,3
-50
100 200 Time (s)
100 200 300 Time (s)
D110C Cys-less Orai1 0
I (pA) -200
2 1 0
MTS-TEAE BMS
MTSEA BMS
MTSET BMS 0
0
3 4
100 200 Time (s)
-200 100 200 Time (s)
100 pA
-150 1,4
-200 0
mV -100 2 3
0
50 100 150 Time (s)
Fig. 3. Covalent modification of Orai1 Cys mutants by MTS reagents. (A and B) WT and Cys-less Orai1 channels are insensitive to modification by MTSET. In both cases, ICRAC (measured at ⫺100 mV) was partially inhibited by application of MTSET (100 M), but this inhibition reversed completely following washout of the reagent. BMS (5 mM) treatment transiently inhibited ICRAC by an unknown mechanism. (C) Modification of D110C Orai1 (in the Cys-less background) by MTSET, MTSEA, and MTS-TEAE (100 M each) recorded from the same cell. In each case, significant persistent inhibition is seen following washout of the reagent that is reversed by BMS (5 mM). Voltage ramps in the right graph show the I-Vs at the time points indicated by arrowheads.
(aminoethyl)-methanethiosulfonate (MTSEA) and 2-(triethylammonium)ethyl methanethiosulfonate (MTS-TEAE), headgroup sizes: 3.8 and 8 Å, respectively) also produced pronounced inhibition of D110C Orai1 currents (Fig. 3C). Collectively, these results indicate that MTS reagents covalently modify and block D110C Orai1 channels, suggesting that this residue lines the ion conduction pathway. Pattern of Blockade by MTS Reagents. In the predicted TM1-TM2
loop, MTSET blocked currents at several positions with the most significant block evident in a stretch of residues between V107 and D114 (Fig. 4A). By contrast, little block was seen in the TM1 residues below E106. With the exception of the Cys at position V107, blockade by the smaller thiol reagent, MTSEA, was similar to that elicited by MTSET (Fig. 4A). The reactive residues between V107 and D110 in WT background also exhibited strong reactivity in the Cys-less background, indicating that the endogenous cysteines do not significantly contribute to the MTSET effects (Fig. 4 A and B). MTS-TEAE, a larger analog of MTSET with a predicted head group size equivalent to tetraethylammonium (⬇8 Å), also showed pronounced block at positions V107–D110 (Fig. 4B), indicating that these residues flank a vestibule large enough to accommodate bulky MTS compounds. Blockade of currents by MTSET in Q108C and D110C Orai1 occurred with second-order rate constants in the range of 104 M⫺1s⫺1 (Fig. 4C), values within an order of magnitude for the reaction of MTSET with free thiols in solution (⬇90,000 M⫺1s⫺1) (14), and very similar to the rate constants observed in the internal gate region of the shaker Kv channels (15). This finding indicates that these residues are readily exposed to the water-filled ion conduction pathway. Modification rate constants of MTSEA at these positions were also in the same range (Fig. S5A), indicating that the difference in size between these reagents has little impact on their accessibility. By contrast, MTSET modification rate constants were slower at V107 and L109 (⬇103 M⫺1s⫺1; Fig. 4C). Because macroscopic currents were large and similar in amplitude for all of these mutants, dramatic variations in channel Po are not likely to account for these differences in modification rates. The most straightforward inter22518 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0909574106
pretation is that modification of cysteines at V107 and L109 occurs due to less frequently exposed configurations of the Cys side-chains to the water filled pore. However, other factors could also contribute to the slow rates, including steric hindrance of the pathway leading to the substituted Cys or unfavorable local chemistry. The potentiation of currents by MTSEA and MTSES in V107C Orai1 (Fig. 4A and Fig. S5B) also raises the possibility that changes in gating rather than pore occlusion could be responsible for the slow MTSET inhibition in this mutant (Fig. S5C). In contrast to the positively charged MTS reagents, little persistent block was evoked at most positions by the negatively charged reagent, 2-sulfonatoethyl methanethiosulfonate (MTSES), which is similar in size to MTSET (5.8 Å) (Fig. 4 A and D). However, prior exposure to MTSES markedly attenuated subsequent block by MTSET in several TM1-TM2 loop mutants, including D110C (Fig. 4D), indicating that covalent modification of these residues by MTSES renders them unavailable for subsequent modification by other reagents. These findings indicate that Ca2⫹ ions can flow into the pore despite the apparent presence of bound MTSES to this region of Orai1. Thus, inhibition of ICRAC in the TM1-TM2 loop mutants by the positively charged reagents, MTSET and MTSEA, likely stems from electrostatic rather than steric blockade of ion conduction. CRAC Channel Pore Narrows below E106. A puzzling feature of the
pattern of reactivity of MTSEA and MTSET was the dearth of responses in the TM1 segment (Fig. 4A). We suspected that the paucity of MTS responses was due to a constriction in the ion conduction pathway that prevents access of the bulky MTS reagents to deep pore regions. To directly test this idea, we studied the accessibility of a small Cys-reactive probe, Cd2⫹. Cd2⫹ has the advantage that it is a divalent ion like Ca2⫹ and is approximately the same size (ionic diameters of Cd2⫹ and Ca2⫹ are 1.94 and 1.98 Å, respectively). Unlike MTS reagents, however, Cd2⫹ does not form a covalent bond. Instead, high affinity Cd2⫹ binding requires coordination by multiple (ⱖ2) sulfhydryl groups, which is most likely to occur at the central symmetry axis of an ion channel (15–17). Given the tetrameric nature of the CRAC channel (18– 20), Cd2⫹ block is hence predicted to arise from binding to cysteines in two or more subunits in the pore. As a first test to examine the utility of Cd2⫹, we examined the reactivity of this probe to Cys substitutions in the TM1-TM2 loop. Cd2⫹ blocks native CRAC channels with a weak affinity [IC50 of ⬇200 M (21)], and in our tests, we found that a relatively low Cd2⫹ dose (5-M) blocked WT Orai1 currents by ⬍10% (Fig. 5A). By contrast, several TM1-TM2 loop Cys mutants exhibited strong sensitivity to Cd2⫹ blockade (Fig. 5 A and B). Significantly, unlike the MTS reagents, Cd2⫹ strongly blocked currents of several TM1 mutants, including L95C, G98C, and V102C (Fig. 5 A and B), suggesting that TM1 lines the pore. The ability of Cd2⫹ to be stably coordinated by several TM1 residues contrasts starkly with the lack of responses with the bulky MTS reagents and suggests that the pore narrows below E106, preventing MTS reagents from accessing the deeper pore residues of the TM1 segment. Moreover, the rates of Cd2⫹ blockade were considerably faster for cysteines in the TM1TM2 loop than those in the TM1 segment (Fig. 5E), indicating reduced accessibility of Cd2⫹ to the deeper regions of the pore compared to the outer mouth of the pore. With a few exceptions, the overall pattern of Cd2⫹ block was similar in the WT and Cys-less Orai1 backgrounds (Fig. 5B), indicating that tight Cd2⫹ binding does not involve native cysteines. However, reactivity at several TM1-TM2 loop positions: V107, L109, and A111 was reduced, suggesting that the TM1-TM2 loop segments adopt a different conformation in the Cys-less background. Together with the relatively slow rates of MTSET effects described in the previous section (Fig. 4C), these results suggest that modification of the Cys residues at these positions in the WT McNally et al.
WT Orai1 background
TM3
E190
E190
E190
D114 *H113 D112 A111 D110 L109 Q108 V107 *E106 V105 M104 A103 V102 G98 L95 WTO1
D114 *H113 D112 A111 D110 L109 Q108 V107 *E106 V105 M104 A103 V102 G98 L95 WTO1
D114 *H113 D112 A111 D110 L109 Q108 V107 *E106 V105 M104 A103 V102 G98 L95 WTO1
0
B
40 80 MTSET Block (%)
-80 0 40 80 MTSEA Block (%)
C
D
D114
D114
D112
D112
D110 L109 Q108 V107 *E106 V105
D110 L109 Q108 V107 *E106
V102
CLO1 -20 0 40 80 MTSES Block (%)
MTSES
D110C
MTSET-1
BMS
V107C 10
3
2
3 4 56
4
2
3 4 5
-400
10 -1 -1 Rates (M s )
40 80 MTSET Block (%)
250
500 Time (s)
0 40 80 MTS-TEAE Block (%)
MTSET-2 BMS
I (pA)
Q108C
CLO1 0
0
L109C
Cys-less Orai1 background
Covalent Block (%)
TM1
TM1-TM2 Loop
A
80 60 40 20 0
MTSET-1 MTSET-2
Fig. 4. MTS reagents block Cys mutants localized exclusively in the TM1-TM2 loop. (A) Summary of persistent inhibition [100 * (1 ⫺ Ipost/Ipre), where Ipre and Ipost are the currents prior and post washout of the MTS reagent, respectively] of single-Cys mutants in the WT Orai1 background given as mean (⫾SEM; n ⫽ 4 –12 cells). WTO1, wild-type Orai1. Asterisks indicate mutations that produced nonfunctional channels. (B) Pattern of block of the mutants in the Cys-less background by MTSET and MTS-TEAE. Positions V107 to D110 are highly reactive to MTSET both in the WT and Cys-less backgrounds, indicating minimal contributions of native cysteines to the observed reactivity. CLO1, Cys-less Orai1. (C) MTSET reaction rate constants of the most susceptible mutants (mean ⫾ SEM of four cells). (D) MTSES does not produce persistent blockade of D110C currents, but eliminates subsequent persistent blockade by MTSET. Persistent MTSET block is restored following treatment with BMS, which presumably removes MTSES. The right graph shows the mean ⫾ SEM (n ⫽ 4) persistent block by MTSET applied at the time points indicated in the left trace.
Secondary Structure. The pattern of Cd2⫹ blockade supports the
predicted secondary structure for the TM1-TM2 loop and the TM1 segment (Fig. 5 B and F). In the TM1-TM2 loop segment, several closely spaced residues spanning positions V107–D112 were susceptible to Cd2⫹ block, resembling the pattern seen with MTS reagents, and reaffirming that this segment forms an extended loop region. However, in the TM1 region, strongly reactive residues such as L95C, G98C, and V102C were well separated by stretches of less reactive residues. Together with E106, this pattern of reactivity is consistent with TM1 being a pore-lining helix (Fig. 5F). R91C also exhibited modest reactivity (Fig. 5B). Its location at the bottom of the predicted TM1 segment suggests that it borders the cytoplasmic rim of the pore. The proximity of R91 to the pore implies that mutations that introduce hydrophobic bulky side-chains at this position [such as the human SCID mutation, R91W (22)] could interfere with ion conduction due to steric occlusion of the inner pore. TM1 Lines a Narrow Pore. Estimates of the distances between the
backbone carbons of cysteines involved in high affinity Cd2⫹ binding are in the range of 5–9 Å (23, 24). Thus, the ability of several McNally et al.
TM1 residues to coordinate Cd2⫹ suggests that the TM1 helices must be close to one another and therefore line a narrow pore. In particular, the blockade of V102C Orai1 currents was nearly complete, largely irreversible, and occurred with high affinity (IC50 ⬇300 nM; Fig. S6A), suggesting that V102 must be particularly close to the central symmetry axis of the channel. The degree of Cd2⫹ blockade depended on the concentration of the permeant ion (Ca2⫹) (Fig. S6B), signifying that subsaturating block observed in several TM1 mutants arises from the displacement of Cd2⫹ by Ca2⫹, as expected for pore-lining residues. Recovery from Cd2⫹ block in the V102C and L95C mutants required perfusion of a DVF solution containing EDTA to chelate Cd2⫹ out of the channel, or in some cases, BMS (Fig. 5A and Fig. S6C), consistent with an extremely high Cd2⫹ affinity for these sites. E190 Is Not a Pore-Facing Residue. On the basis of the loss of Ca2⫹
selectivity caused by the E190Q mutation, E190 in TM3 has been implicated in the coordination of Ca2⫹ in the pore (7, 9, 10). Yet, the E190D mutation fails to affect ion selectivity, and Ca2⫹ imaging experiments indicate that the E190A substitution has no effect on store-operated Ca2⫹ entry (7). To more directly investigate whether E190 is exposed to the pore, we studied the reactivity of E190C to Cys-modifying reagents. E190C Orai1 protein failed to exibit disulfide bond formation, and currents in this mutant were indisPNAS 兩 December 29, 2009 兩 vol. 106 兩 no. 52 兩 22519
PHYSIOLOGY
background likely occurs in rarely exposed configurations, or due to gating effects.
2
1
-100
3
100 pA
-300 20 40 Time (s)
2 -150 1,3
Q108C DVF mV
0
-100 2
2
1
1
-800 0
100 pA
-150
TM1
40 80 120 Time (s)
V102C I (pA)
Cd2+
DVF mV 2 -100
0 1
2
100 pA
-400
1 -75
0
E
L95C Cd2+
DVF BMS
I (pA)
0
-100
1 0
mV -100 2
2 3
D114 *H113 D112 A111 D110 L109 Q108 V107 *E106 V105 M104 A103 V102 *M101 A100 F99 G98 S97 L96 L95 A94 S93 T92 R91 S90 CysO1 0
40 80 Time (s)
100 pA
3 -80 1
100 200 Time (s)
Cys-less background
C
E190C Cd2+
L119 G118 P117 P116 Y115 D114 *H113 D112 A111 D110 L109 Q108 V107 *E106 V105 M104 A103 V102 M101 A100 F99 G98 S97 L96 L95 A94 S93 T92 R91 S90 WTO1
TM1-TM2 Loop
I (pA)
Cd 2+
WT background
50 100 Cd Block (%)
D
50 Time (s)
50 100 Cd Block (%)
97 104
100 107
2 -100 1
50 100 Cd Block (%)
90 S
S
101 M
94 A
M
V
105 G 98
S
TM1
A
91
R V
V
96
6
10 10 -1 -1 Rates (M s )
100
V192 V191 E190 A189 L188 F187 L186 L185 T184 G183 I182 V181 T180 S179 F178 A177 WTO1 0
93
5
2
0
0
100 pA
1
D110C Q108C V102C G98C L95C 4
mV -100
-200
F
10
20 Ca
0
I (pA)
mV
0
TM1-TM2 Loop
I (pA)
20 Ca Cd2+
TM3
B
WT Orai1
TM1
A
102
L
L A
103
T
F
92
99
95
E
106
Fig. 5. block reveals pore-lining residues in TM1 (A) Blockade of ICRAC in various Cys mutants by 5 M (WT background). Although Cd2⫹ block in V102C and L95C was poorly reversed by washout, application of a chelator-containing DVF solution (V102C) or the reducing agent BMS (L95C) reversed Cd2⫹ block. (B) Pattern of Cd2⫹ blockade (5 M) of TM1 and TM1-TM2 loop Cys mutants (⫾SEM; n ⫽ 4 –12 cells). In the TM1 segment, three residues (L95C, G98C, and V102C) exhibited strong reactivity to Cd2⫹. The dotted line depicts a sine function with a periodicity of 3.5. Asterisks indicate mutations that produced nonfunctional channels. (C and D) E190C and the other TM3 positions (WT background) do not exhibit significant sensitivity to Cd2⫹ (5 M). (E) Rate constants of Cd2⫹ blockade of selected TM1 and TM1-TM2 loop Cys mutants in WT background (⫾SEM; n ⫽ 4 cells). (F) The most Cd2⫹ sensitive sites in TM1 (green boxes) localize on one face of a helical wheel, suggesting that this segment forms a ␣ helical structure. E106C (red) is nonfunctional. R91C (yellow box) exhibited only modest reactivity. Cd2⫹
tinguishable from WT Orai1 in reversal potentials, inward rectification, and fast inactivation (Table S1 and Fig. S7), indicating that removal of the negative charge at this position does not alter the key functional properties of CRAC channels. Responses of the E190C mutant to MTSEA, MTSET, and MTSES were indistinguishable from that of WT Orai1, revealing little or no block (Fig. 4A). Cd2⫹ also failed to significantly inhibit E190C currents, indicating that the lack of MTS effects is not due to E190C lining a region of the pore too narrow for bulky MTS reagents to penetrate (Fig. 5 C and D). Likewise, no Cd2⫹ block was seen at the other TM3 positions (Fig. 5D). These results indicate either that E190C cannot be modified by Cys-reactive probes or that its modification has no effect on ion conduction. Either way, these results strongly suggest that E190 and the TM3 segment do not flank the ion-conduction pathway. Discussion By examining the distinct effects of various thiol-reactive reagents differing in size and charge with Cys residues introduced into Orai1, we show that the CRAC channel pore exhibits many structural characteristics that could explain its intriguing permeation properties. Our results 22520 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0909574106
Cd2⫹
indicate that the CRAC channel pore has an outer vestibule formed by the TM1-TM2 loops that sharply narrows at the beginning of the TM1 segment near E106, a critical residue that controls Ca2⫹ selectivity. The remainder of the pore is also sufficiently narrow along much of its length to permit stable Cd2⫹ coordination by pore-lining cysteines; this long narrow pore is likely responsible for the low unitary conductance of CRAC channels. The other conserved acidic residues: E190 in TM3 and D110/112/114 in the TM1-TM2 loop are not involved in regulating ion selectivity, strongly suggesting that E106 forms the main Ca2⫹ binding site that controls Ca2⫹ selectivity. Importantly, these results also confirm the essential role of Orai1 as the pore-forming subunit of the CRAC channel and refute recent proposals that Orai1 mediates a regulatory function as a non-pore forming -subunit to TRPC channels (25, 26). The outer region of the pore exhibits several notable features. First, it can accommodate the negatively charged 6 Å MTSES reagent without sterically obstructing ion flow, as well as the bulky MTS-TEAE (headgroup diameter ⬎8 Å), indicating that these residues flank a wide outer vestibule. Second, the vestibule is nonselective; none of the single Cys mutations in the TM1-TM2 loop overtly affected ion selectivity, and more surprisingly, the negatively charged MTSES could approach McNally et al.
1. Feske S (2009) ORAI1 and STIM1 deficiency in human and mice: Roles of store-operated Ca2⫹ entry in the immune system and beyond. Immunol Rev 231:189 –209. 2. Lewis RS (2007) The molecular choreography of a store-operated calcium channel. Nature 446:284 –287. 3. Prakriya M (2009) The molecular physiology of CRAC channels. Immunol Rev 231:88 –98. 4. Lepple-Wienhues A, Cahalan MD (1996) Conductance and permeation of monovalent cations through depletion-activated Ca2⫹ channels (ICRAC) in Jurkat T cells. Biophys J 71:787–794. 5. Prakriya M, Lewis RS (2006) Regulation of CRAC channel activity by recruitment of silent channels to a high open-probability gating mode. J Gen Physiol 128:373–386. 6. Bakowski D, Parekh AB (2002) Monovalent cation permeability and Ca2⫹ block of the store-operated Ca2⫹ current ICRAC in rat basophilic leukemia cells. Pflugers Arch 443:892–902. 7. Prakriya M, et al. (2006) Orai1 is an essential pore subunit of the CRAC channel. Nature 443:230 –233. 8. Yeromin AV, et al. (2006) Molecular identification of the CRAC channel by altered ion selectivity in a mutant of Orai. Nature 443:226 –229. 9. Vig M, et al. (2006) CRACM1 multimers form the ion-selective pore of the CRAC channel. Curr Biol 16:2073–2079. 10. Yamashita M, Navarro-Borelly L, McNally BA, Prakriya M (2007) Orai1 mutations alter ion permeation and Ca2⫹-dependent inactivation of CRAC channels: Evidence for coupling of permeation and gating. J Gen Physiol 130:525–540. 11. Fahrner M, et al. (2009) Mechanistic view on domains mediating STIM1-Orai coupling. Immunol Rev 231:99 –112. 12. Karlin A, Akabas MH (1998) Substituted-cysteine accessibility method. Methods Enzymol 293:123–145. 13. Babich O, Reeves J, Shirokov R (2007) Block of CaV1.2 channels by Gd3⫹ reveals preopening transitions in the selectivity filter. J Gen Physiol 129:461– 475. 14. Stauffer DA, Karlin A (1994) Electrostatic potential of the acetylcholine binding sites in the nicotinic receptor probed by reactions of binding-site cysteines with charged methanethiosulfonates. Biochemistry 33:6840 – 6849. 15. Liu Y, Holmgren M, Jurman ME, Yellen G (1997) Gated access to the pore of a voltage-dependent K⫹ channel. Neuron 19:175–184. 16. Yellen G, Sodickson D, Chen TY, Jurman ME (1994) An engineered cysteine in the external mouth of a K⫹ channel allows inactivation to be modulated by metal binding. Biophys J 66:1068 –1075.
McNally et al.
Our results also suggest that the prevailing view of Ca2⫹ permeation in CRAC channels must be modified. Based on changes in ion selectivity caused by specific mutations (7–10), current models of the CRAC channel pore propose that several acidic residues: E190, E106, D110, D112, and D114 are involved in Ca2⫹ selectivity. However, in the broad screening strategy used here, we found no evidence for involvement of E190 in Ca2⫹ coordination and no significant changes in ion selectivity either by mutating D110, D112, D114 individually to Cys (Table S1), or by covalently modifying these sites by positively charged reagents. Thus, our results effectively rule out scenarios in which E190 and D110/112/114 form high affinity Ca2⫹ binding sites and suggest that the effects of E190Q and the double/triple mutations of D110/112/114 are likely due to allosteric effects. Together with evidence suggesting that occupancy of the pore by a single Ca2⫹ ion is sufficient to block monovalent CRAC currents (5, 10), these results suggest that only a single high-affinity Ca2⫹ binding site encoded by E106 controls Ca2⫹ selectivity in CRAC channels. Such a “single-locus” model is qualitatively reminiscent of L-type Cav channels and TRPV6 channels (30, 31). However, the strikingly wider selectivity filters of Cav and TRPV6 channels (31, 32) and the unusually narrow dimensions of the CRAC channel ion conduction pathway highlight important differences that are likely responsible for the slow flux rates in the CRAC channel and its unique ion permeation profile. Materials and Methods ICRAC was recorded by stepping the membrane voltage from a ⫹30 mV (holding) to ⫺100 mV followed by a ramp from ⫺100 to ⫹100 mV (100 ms each) applied at 1 s intervals in a Ringer’s solution containing 20 mM Ca2⫹. MTS reagents or Cd2⫹ were added to this solution at the indicated concentrations. Cells were pretreated with 1 M TG before seal formation. Rate constants were determined at a constant holding potential of ⫺80 mV. Cells were exposed to BMS for 30 – 60 s immediately following whole-cell break-in to remove preexisting disulfide bonds, if any. Repeated BMS applications indicated that after the initial application, no further enhancement of ICRAC could be induced by subsequent BMS exposure, suggesting that this duration of BMS treatment is sufficient to remove preexisting disulfide bonds. Detailed standard methods are provided in SI Text. ACKNOWLEDGMENTS. We thank S. Feske (New York University) for gift of the polyclonal Orai1 antibody, and A. Gross, K. Swartz, C. Lingle, R. Lewis, T. Hornell, A. Somasundaram, and S. Feske for helpful discussions and comments.This work was supported by the American Heart Association Grant 0630401Z and the National Institutes of Health Grant NS057499 (to M.P.).
17. Loussouarn G, Phillips LR, Masia R, Rose T, Nichols CG (2001) Flexibility of the Kir6.2 inward rectifier K⫹ channel pore. Proc Natl Acad Sci USA 98:4227– 4232. 18. Ji W, et al. (2008) Functional stoichiometry of the unitary calcium-release-activated calcium channel. Proc Natl Acad Sci USA 105:13668 –13673. 19. Mignen O, Thompson JL, Shuttleworth TJ (2008) Orai1 subunit stoichiometry of the mammalian CRAC channel pore. J Physiol 586:419 – 425. 20. Penna A, et al. (2008) The CRAC channel consists of a tetramer formed by Stim-induced dimerization of Orai dimers. Nature 456:116 –120. 21. Hoth M, Penner R (1993) Calcium release-activated calcium current in rat mast cells. J Physiol 465:359 –386. 22. Feske S, et al. (2006) A mutation in Orai1 causes immune deficiency by abrogating CRAC channel function. Nature 441:179 –185. 23. Swartz KJ (2005) Structure and anticipatory movements of the S6 gate in Kv channels. J Gen Physiol 126:413– 417. 24. Krovetz HS, VanDongen HM, VanDongen AM (1997) Atomic distance estimates from disulfides and high-affinity metal-binding sites in a K⫹ channel pore. Biophys J 72:117– 126. 25. Liao Y, et al. (2007) Orai proteins interact with TRPC channels and confer responsiveness to store depletion. Proc Natl Acad Sci USA 104:4682– 4687. 26. Liao Y, et al. (2008) Functional interactions among Orai1, TRPCs, and STIM1 suggest a STIM-regulated heteromeric Orai/TRPC model for SOCE/Icrac channels. Proc Natl Acad Sci USA 105:2895–2900. 27. Bera AK, Akabas MH (2005) Spontaneous thermal motion of the GABA(A) receptor M2 channel-lining segments. J Biol Chem 280:35506 –35512. 28. Benitah JP, et al. (1997) Molecular motions within the pore of voltage-dependent sodium channels. Biophys J 73:603– 613. 29. Zweifach A, Lewis RS (1993) Mitogen-regulated Ca2⫹ current of T lymphocytes is activated by depletion of intracellular Ca2⫹ stores. Proc Natl Acad Sci USA 90:6295– 6299. 30. Ellinor PT, Yang J, Sather WA, Zhang JF, Tsien RW (1995) Ca2⫹ channel selectivity at a single locus for high-affinity Ca2⫹ interactions. Neuron 15:1121–1132. 31. Voets T, Janssens A, Droogmans G, Nilius B (2004) Outer pore architecture of a Ca2⫹-selective TRP channel. J Biol Chem 279:15223–15230. 32. Wu XS, Edwards HD, Sather WA (2000) Side chain orientation in the selectivity filter of a voltage-gated Ca2⫹ channel. J Biol Chem 275:31778 –31785.
PNAS 兩 December 29, 2009 兩 vol. 106 兩 no. 52 兩 22521
PHYSIOLOGY
and modify residues in the vestibule. A previous report has suggested that the negatively charged Asp residues in the loop region concentrate cations at the mouth of the pore (8). While our results do not directly address this issue, these acidic residues apparently do not prevent the entry of anions into the outer vestibule. Third, several engineered cysteines in the TM1-TM2 loop exhibited spontaneous disulfide bond formation with corresponding cysteines in neighboring subunits. This finding is consistent with the ability of the TM1-TM2 loop Cys mutants to coordinate the small ion, Cd2⫹, but is seemingly at odds with the notion that they form a wide aqueous outer vestibule. These findings could be reconciled if the TM1-TM2 loop segments are highly flexible, permitting Cd2⫹ coordination and even disulfide cross-linking to occur across the pore axis when the loops are in close proximity to each other. Support for this idea comes from the finding that La3⫹ can interact tightly with loop residues, whereas Ca2⫹ does not, suggesting that the loops rearrange to interact with La3⫹, but do not (or cannot) do so for Ca2⫹. Therefore, we favor the idea that TM1-TM2 loop segments are highly flexible and mobile. Flexibility of pore-lining residues is a recurrent theme in many ion channels (17, 27, 28). In the case of the CRAC channel, flexibility of the pore region around TM1-TM2 loop may be important for channel gating, an idea that will have to be tested in future studies. Moving deeper into the pore, the distinct effects of MTS reagents and Cd2⫹ in the TM1 segment indicates that the pore markedly constricts in the vicinity of E106. This constriction restricts permeation of ions as small as 3.8 Å, and instead only allows the passage of very small cations such as Na⫹, Ca2⫹, and Cd2⫹. The pattern of tight binding of Cd2⫹ indicates that TM1 is a helical segment flanking a narrow pore. Together with E106, the periodicity of the most reactive TM1 residues (V102, G98, L95, R91) is consistent with four turns of an ␣-helical structure that is expected to span approximately 24 Å along the vertical axis of the helix. Intriguingly, in contrast to the TM1-TM2 loop region, the rate constants of Cd2⫹ blockade in TM1 are relatively slow, approximately 105 M⫺1s⫺1. This rate constant is similar to the effective rate constant of the overall ion transfer process itself, approximately 2.3 ⫻ 105 M⫺1s⫺1, derived from an estimated unitary flux of approximately 25,000 s⫺1 in 110 mM Ca2⫹ (5, 29). The close similarity in these reaction rates suggests that the rate of Cd2⫹ block in deep pore residues is limited by the low unitary conductance of the CRAC channel.
