Separation and Characterization of Currents through Store-operated CRAC Channels and Mg²⁺-inhibited Cation (MIC) Channels

The depletion of Ca²⁺ from the ER in nonexcitable cells triggers Ca²⁺ influx across the plasma membrane, a process termed store-operated Ca²⁺ entry. Store-operated channels (SOCs)^* are widely, if not ubiquitously, expressed in nonexcitable cells, where they provide a major route for Ca²⁺ entry (for reviews see Parekh and Penner, 1997; Lewis, 1999; Putney and McKay, 1999). One of the best described SOCs, the calcium release–activated Ca²⁺ (CRAC) channel, is present in mast cells, T lymphocytes, and several related cell lines. CRAC channels influence a variety of important physiological processes, including the release of inflammatory mediators from mast cells during allergic reactions, and the generation of [Ca²⁺]_i oscillations leading to gene expression and the differentiation and activation of T cells (for reviews see Parekh and Penner, 1997; Lewis, 2001). The critical role of CRAC channels in human physiology is shown clearly by the devastating immunodeficiencies that arise from the absence of CRAC channel activity in T cells from human patients (Partiseti et al., 1994; Feske et al., 2001).

In view of their critical physiological functions, considerable effort has been focused on isolating the gene(s) encoding the CRAC channel and on understanding how the channel's gating is regulated. Although several genes of the TRP family have been proposed to encode the CRAC channel, it is unclear as to whether the currents resulting from heterologous expression of these genes are identical to native I_CRAC (Putney and McKay, 1999; Clapham et al., 2001; Prakriya and Lewis, 2002). Several classes of activation mechanisms are currently being studied, including a diffusible messenger released from the ER, depletion-triggered insertion of CRAC channels into the plasma membrane, and physical coupling between the channels and IP₃ receptors in the ER. However, there is as yet no consensus on which if any of these mechanisms may be correct (Putney et al., 2001). Progress in resolving these issues has been hampered by several factors that are unique to this class of channels, including a lack of selective inhibitors, very small whole cell currents (on the order of a few pA per cell), and an extremely small single-channel Ca²⁺ conductance (2–26 fS estimated from noise analysis) (Zweifach and Lewis, 1993), which has precluded single-channel recording in membrane patches.

The study of currents carried by monovalent cations through CRAC channels could, in principle, bypass some of these problems. Like voltage-gated Ca²⁺ channels (Almers and McCleskey, 1984; Hess et al., 1986), CRAC channels under physiological conditions are exquisitely selective for Ca²⁺, conducting Ca²⁺ ∼1,000 times better than the more prevalent Na⁺ (Hoth and Penner, 1993; Hoth, 1995). However, as with voltage-gated Ca²⁺ channels (Almers and McCleskey, 1984), the removal of extracellular divalent cations renders CRAC channels permeable to Na⁺ (Hoth and Penner, 1993; Lepple-Wienhues and Cahalan, 1996). Initially, upon removal of extracellular divalents, this Na⁺ current is approximately six times larger than the preceding Ca²⁺ current, but it declines by >90% over tens of seconds. The slow loss of channel activity is reversed following readdition of extracellular Ca²⁺, a process referred to as Ca²⁺-dependent potentiation (CDP) (Christian et al., 1996b; Zweifach and Lewis, 1996). Kerschbaum and Cahalan (1998) found that removal of intracellular Mg²⁺ made the Na⁺ current larger and sustained, leading to the idea that intracellular Mg²⁺ is required for CRAC channel depotentiation. Furthermore, in the absence of intracellular Mg²⁺ the whole-cell current was seen to arise from the progressive, all-or-none activation of single 40-pS channels having a high open probability (P_o > 0.9) (Kerschbaum and Cahalan, 1999). Similar results were later found in human T cells (Fomina et al., 2000) and RBL cells (Braun et al., 2001). The Na⁺ currents were considered to arise from CRAC channels based on their slow time course of activation and inhibition by extracellular Ca²⁺, Mg²⁺, Ni²⁺, and Gd³⁺ (Kerschbaum and Cahalan, 1999; Fomina et al., 2000; Braun et al., 2001). These results are significant because resolution of CRAC currents at the single-channel level is expected to greatly facilitate studies of the molecular mechanism of store-operated Ca²⁺ entry. In fact, the conductance, selectivity, and high open probability of the Na⁺-conducting channels was exploited in studies of the CRAC channel's activation mechanism (Braun et al., 2001; Rychkov et al., 2001), changes in CRAC channel expression during T cell activation (Fomina et al., 2000), and for the identification of genes that may encode the CRAC channel pore region (Yue et al., 2001).

Although the discovery of the large, sustained mono-valent current in the absence of intracellular Mg²⁺ has offered new opportunities for molecular characterization of CRAC channels, we noted several discrepancies in our own studies which led us to examine its identity in greater detail. We have found that the large sustained monovalent current seen with Mg²⁺-free intracellular solutions arises from a store-independent channel that differs from I_CRAC in its ion selectivity, pharmacology, and regulation. Because its activity is suppressed by intracellular Mg²⁺, we refer to this channel as the Mg²⁺-inhibited cation (MIC) channel. By using conditions that prevent the activation of I_MIC, we have been able to characterize monovalent fluxes through CRAC channels. Several key properties of monovalent CRAC channels, including ion selectivity, unitary conductance, and regulation by intracellular Mg²⁺, differ significantly from those described previously. These new results reveal similarities and differences in the ionic selectivity mechanisms of store-operated and voltage-gated Ca²⁺ channels, and force a revision of the biophysical fingerprint of CRAC channels that will have important implications for the identification of CRAC channel genes.

Jurkat E6-1 human leukemic T cells (American Type Culture Collection) were grown in a medium consisting of RPMI 1640 supplemented with 10% fetal calf serum, 2 mM glutamine, 50 U/ml penicillin and 50 μg/ml streptomycin. The cells were maintained in log-phase growth at 37°C in 6% CO₂.

The standard extracellular Ringer's solution contained (in mM): 155 NaCl, 4.5 KCl, 2 or 20 CaCl₂, 1 MgCl₂, 10 D-glucose, and 5 Na-HEPES (pH 7.4). Ca²⁺-free Ringer's was prepared by substituting 1 mM EGTA + 2 mM MgCl₂ for CaCl₂. The divalent-free (DVF) Ringer's solutions contained (in mM): 155 Na, Cs, or NMDG methanesulfonate, and 10 HEDTA, 1 EDTA and 10 Hepes (pH 7.4 with NaOH, CsOH, or HCl, respectively). The standard internal solution contained (in mM): 150 Cs methanesulfonate, 3–10 mM MgCl₂, 10 BAPTA, and 10 Cs-Hepes (pH 7.2). The Mg²⁺-free (MGF) intracellular solution contained (in mM): 150 Cs methanesulfonate, 10 HEDTA, 0.5 CaCl₂ (calculated [Ca²⁺]_i = 10 nM) and 10 Cs-Hepes (pH 7.2). Where noted, 10 mM BAPTA was substituted for HEDTA. In the experiments on I_CRAC deactivation, the intracellular solution contained (in mM): 150 Cs methanesulfonate, 2 mM CsCl, 1.2 EGTA or 1 BAPTA, and 10 Cs-HEPES (pH 7.2). In the excised patch experiments, the pipette solution contained an Na-based DVF solution (composition listed above), and the cytoplasmic face of the patch was exposed to the MGF solution (listed above, but with 0 CaCl₂) to which an appropriate quantity of MgCl₂ added to yield the free [Mg²⁺] indicated in the figure legends. MgATP and Na₂ATP (Sigma-Aldrich) were added to the intracellular solution in some experiments, and free [Mg²⁺]_i and [Ca²⁺]_i were calculated using MaxChelator software (WEBMAXC 2.10, available at http://www.stanford.edu/~cpatton/webmaxc2.htm).

2-aminoethyldiphenyl borate (2-APB) was provided by Dr. K. Mikoshiba (Tokyo University, Japan). In some experiments, 2-APB obtained from Sigma-Aldrich was used; no difference was found between the drugs from the different sources. Stock solutions of 2-APB and thapsigargin (Sigma-Aldrich) were prepared in DMSO at concentrations of 20 mM and 1 mM, respectively; SKF 96365 (Sigma-Aldrich) was dissolved in deionized water at a concentration of 10 mM. The drugs were diluted to the concentrations indicated in the legends and applied to the cells using a multi-barrel local perfusion pipette with a common delivery port. The time for 90% solution exchange was measured to be <1 s, based on the rate at which the K⁺ current reversal potential changed when the external [K⁺] was switched from 2 to 150 mM.

Patch-clamp experiments were conducted in the standard whole-cell recording configuration at 22–25°C using an Axopatch 200 amplifier (Axon Instruments, Inc.), an ITC-16 interface (Instrutech) and a Macintosh G3 computer. Recording electrodes were pulled from 100-μl pipettes coated with Sylgard and fire-polished to a final resistance of 2–5 MΩ. Stimulation and data acquisition and analysis were performed using in-house routines developed on the Igor Pro platform (Wavemetrics). The holding potential was 20 mV unless otherwise indicated. Voltage stimuli usually consisted of a 100-ms step to −110 mV, immediately followed by a 100-ms ramp from −110 to 90 mV applied every 1–2 s. Currents were filtered at 1 kHz with a 4-pole Bessel filter and sampled at 5 kHz without series resistance compensation. Data are corrected for the liquid junction potential of the pipette solution relative to Ringer's in the bath (−10 mV) and of the bath DVF solution relative to Ringer's in the bath-ground agar bridge (5 mV). The averaged results are presented as the mean value ± SEM. Curve fitting was done by least-squares methods using built-in functions in Igor Pro 4.0. For the analysis of MIC channel kinetics, excised patches showing the activity of only one MIC channel were selected and processed using TAC (Bruxton Corporation). Channel transitions were idealized by setting a discriminator at 50% of the current between the open and closed levels, and the channel kinetics were obtained from the idealized traces.

The activity of MIC channels when [Mg²⁺]_i is ≤3 mM (conditions of the majority of published papers on I_CRAC) can pose special problems for the isolation of I_CRAC in Jurkat cells. In our experience, MIC activity often changes significantly during the course of whole-cell recording (depending, among other things, on the free [Mg²⁺] in the pipette). Thus, the commonly employed practice of subtracting the current present at the start of whole-cell recording (before I_CRAC induction) from later currents will not necessarily isolate I_CRAC cleanly. A more consistent leak subtraction method is to expose the cell to a Ca²⁺-free extracellular solution (0 Ca²⁺/3 Mg²⁺) shortly before or after measurement of the Ca²⁺ current, and to use this current as the “leak,” because these ionic conditions eliminate I_CRAC (Zweifach and Lewis, 1993), but not I_MIC (see Fig. 7 C). Therefore, we used the zero-Ca²⁺ leak subtraction method to isolate I_CRAC in the experiments reported here. For measurements of I_MIC, leak current was collected in 20 mM Ca²⁺ immediately after whole-cell break-in.

Current noise data were recorded using two methods. In the first, current at a holding potential of −110 mV was recorded continuously at 32-kHz bandwidth onto digital audio tape (Sony DTC-700 modified for DC coupling). For mean-variance analysis, data from tape were then lowpass-filtered at 2 kHz (8-pole Bessel filter; Frequency Devices) and digitized at 5 kHz for analysis. In the second method, current during 500-ms voltage steps to −110 mV applied once per second was lowpass-filtered at 1 kHz, digitized at 5 kHz, and recorded directly to hard disk. The series resistance was typically ∼5 MΩ, which in conjunction with the average membrane capacitance of 10 pF creates a 1-pole rolloff characteristic at 3.2 kHz; this filtering was ignored as it occurred well above the 1–2 kHz 8-pole Bessel cutoff we applied. The mean and variance were calculated from 200-ms segments of the digitized current. This short duration was chosen to minimize changes due to depotentiation of Na⁺-I_CRAC during each sweep (<2% change in current). For spectral analysis, data from tape were replayed through a 20-Hz highpass filter (Krohn-Hite) and a 2-kHz lowpass Bessel filter (Frequency Devices) in series and were digitized at 5 kHz. Power spectra were computed from 512-point sweeps using a Hanning window (Igor Pro; Wavemetrics) and were averaged.

As described in previous studies of Jurkat and human T cells (Kerschbaum and Cahalan, 1998, 1999; Fomina et al., 2000) and RBL cells (Braun et al., 2001), intracellular dialysis with Mg²⁺-free internal solution in the absence of extracellular divalent cations triggers the slow development of a large monovalent conductance. In the experiment shown in Fig. 1, a Jurkat cell was dialysed with a Mg²⁺-free intracellular solution containing HEDTA, and continuous exposure to a DVF Ringer's solution revealed the induction of a large inward current over a period of ∼400 s (Fig. 1 A). With equimolar extracellular Na⁺ and intracellular Cs⁺, the current reversed at −5 mV (Fig. 1 B), indicating approximately equal permeabilities for Na⁺ and Cs⁺. Substitution of NMDG for Na⁺ selectively suppressed the inward current, demonstrating that the larger cation cannot permeate (Fig. 1, A and B). Early in the induction phase of the whole-cell current, the progressive opening of single channels could be resolved (Fig. 1 C). These channels generally opened in an all-or-none fashion to a very high open probability (P_o = 0.97 at −110 mV; n = 4 cells), such that openings of more than five channels could be clearly resolved under whole-cell conditions as described previously (Kerschbaum and Cahalan, 1999; Fomina et al., 2000). Based on the amplitude and reversal potential of the single-channel current, the average chord conductance was 44 ± 3 pS. We observed similar single-channel and whole-cell currents in human T cells freshly isolated from blood, although their activation following break-in was slower (unpublished data).

With DVF Ringer's in the recording pipette, these channels could also be observed in cell-attached patches and after patch excision into a Mg²⁺-free intracellular solution (Fig. 1, D and E). These channels were similar to those seen in whole-cell recordings (Fig. 1 C) in terms of their conductance, lack of selectivity, high open probability, and brief closures. The channels were weakly voltage-dependent, with the average P_o changing from 0.97 at −135 mV to 0.84 at −55 mV (n = 4 patches). Kinetic analysis indicated that the mean closed time (τ_c) varied from 1.1 ± 0.1 ms at −135 mV to 7.9 ± 1.5 ms at −55 mV (n = 4 patches), whereas the mean open time was relatively constant (32.2 ± 3.5 ms at −135 mV and 36.3 ± 4.8 ms at −55 mV). In terms of unitary conductance, kinetics, open probability and reversal potential, these channels closely resemble the 40-pS channels described previously in Jurkat, human T cells, and RBL cells (Kerschbaum and Cahalan, 1998, 1999; Fomina et al., 2000; Braun et al., 2001).

In previous studies, the large, sustained monovalent current and the underlying 40-pS single-channel currents were ascribed to the activity of CRAC channels (Kerschbaum and Cahalan, 1998, 1999; Fomina et al., 2000; Braun et al., 2001). However, we noted several discrepancies in our own experiments that led us to question this conclusion. First, monovalent inward current under Mg²⁺_i-free conditions activated much more slowly than the Ca²⁺ current in the presence of 20 mM extracellular Ca²⁺, with half times of 210 ± 27 s (n = 5) and 109 ± 16 s (n = 6), respectively. Second, the ratio of I_Ca to I_Na varied widely among cells (from 13 in Fig. 1 A to more than 100), even to the point where monovalent currents in the nA range occurred in some cells lacking a measurable Ca²⁺ current. Finally, using a DVF pipette solution we observed 40-pS channels in the majority of cell-attached patches from resting cells in which stores were not deliberately depleted, suggesting that the large monovalent current may be store-independent. To determine whether it does in fact reflect CRAC channel activity, we compared its dependence on store depletion and its pharmacological profile with that of I_CRAC.

If the large monovalent current flows through CRAC channels, then it should be active immediately upon establishing the whole-cell recording configuration in cells with empty Ca²⁺ stores. To test this, 1 μM thapsigargin (TG) was applied for 5–10 min before seal formation to fully deplete stores and maximally activate CRAC channels. Soon after break-in to the whole-cell configuration, application of 20 mM Ca²⁺ to the cell causes an inward current to develop over ∼10 s (Fig. 2 A). This inward current is I_CRAC as judged from its current-voltage relationship (Fig. 2 D, top graph), Ca²⁺_o- and store-dependence, sensitivity to various pharmacological agents (unpublished data), and characteristic delayed appearance following each application of Ca²⁺. This latter process, termed CDP, has been described previously (Zweifach and Lewis, 1996). Following Ca²⁺ readdition, the first application of DVF solution revealed only a small, transient inward Na⁺ current at −110 mV that declined >60% within 20 s (arrow, Fig. 2 A). Subsequent short applications of DVF solution revealed the slow development of a large Na⁺ current over the next 400 s, shown more clearly at lower gain (Fig. 2 A, right graph). Two observations suggest that the large Na⁺ current and I_CRAC are not related. First, the amplitudes of the inward Ca²⁺ and Na⁺ currents did not increase in parallel (Fig. 2 B), causing the ratio of the Na⁺ to the Ca²⁺ current to change from ∼2:1 at the beginning of the experiment (t = 37 s) to ∼50:1 at later times (t = 570 s). Thus, the amplitudes of the Ca²⁺ current and the large Na⁺ current are not well correlated in time. Second, during each application of DVF, the large Na⁺ current was roughly constant, even though CRAC channels appeared to be closing. This is shown by the small initial Ca²⁺ current seen immediately after each Ca²⁺ readdition, followed by a prominent increase due to CDP. Similar results were obtained in five cells.

Importantly, several key properties of the monovalent currents changed during the experiment. Whereas the Na⁺ current seen at early times in DVF solution decayed rapidly, at later times (t > 200 s) it was more sustained (Fig. 2, A and B). Second, the shape of the current-voltage relationship changed dramatically. At early times the monovalent current showed no clear reversal potential, even up to 90 mV, suggesting that intracellular Cs⁺ permeates poorly if at all compared with Na⁺ (Fig. 2 C, top), whereas at later times the current did not rectify and reversed near 0 mV, consistent with equal permeabilities for the two ions (Fig. 2 C, bottom). Time-dependent changes in current selectivity and rectification were also observed in the presence of extracellular Ca²⁺. At the beginning of the experiment (t < 50 s), the whole-cell current in 20 mM Ca²⁺_o displayed prominent inward rectification, characteristic of I_CRAC (Fig. 2 D, top). However, at later time points (>300 s) the whole-cell current displayed strong outward rectification, generating a large outward current at potentials above 50 mV (Fig. 2 D, bottom). A likely explanation for these results is that an additional type of channel, unrelated to CRAC, activates slowly and produces a large, sustained and nonselective monovalent current that eventually dominates the whole-cell current under DVF conditions, and produces the outwardly rectifying current seen in the presence of Ca²⁺. An alternative explanation (Kerschbaum and Cahalan, 1998) is that gradual removal of intracellular Mg²⁺ changes the properties of the CRAC channel, keeping it from inactivating and allowing it to conduct Cs⁺ more efficiently.

To distinguish between these possibilities, we asked whether the large monovalent conductance deactivates in parallel with I_CRAC in response to store refilling. We have shown previously that in the presence of weak intracellular Ca²⁺ buffering (1 mM EGTA in the pipette solution), prolonged Ca²⁺ influx through CRAC channels can overwhelm the buffer, causing a global rise of [Ca²⁺]_i, store refilling, and deactivation of I_CRAC (Zweifach and Lewis, 1995). This behavior is illustrated in the experiment of Fig. 3 A (left graph), in which I_CRAC measured in the presence of 20 mM Ca²⁺ activated slowly in response to passive store depletion, rose to a peak, then declined over ∼150 s back to baseline. DVF solution was applied periodically in this experiment to monitor the large monovalent current. The monovalent current did not decline in parallel with I_CRAC, but instead grew progressively larger over 400 s (Fig. 3 A, right). A plot of the time courses of the Ca²⁺ current and the Na⁺ current in this cell shows clearly that the amplitude of the large Na⁺ current increases even as CRAC channels close (Fig. 3 B). Similar results were observed in 5/5 cells. Together with the changes in ion selectivity and kinetic behavior noted above, these results argue strongly that the large monovalent current arises from store-independent channels distinct from the CRAC channel.

To further establish differences between the large monovalent current and CRAC channels and find conditions that could be used to isolate each current, we compared their sensitivities to SKF 96365 and 2-APB. SKF 96365 is an imidazole antimycotic compound that inhibits CRAC channels and several other SOCs with IC₅₀ values of 0.6–16 μM (Franzius et al., 1994; Christian et al., 1996a). Application of 20 μM SKF 96365 caused robust and partially reversible inhibition of I_CRAC (Fig. 4 A). On average, the peak current amplitude was diminished by 87 ± 4% with a time constant of 17 ± 7 s (n = 6). In contrast, the same concentration of SKF 96365 had very little effect on the large monovalent current (Fig. 4 B), inhibiting by only 10 ± 2% after 120 s of exposure (n = 4). Two results argue that the resistance of the monovalent current to block by SKF 96365 is not due to DVF conditions per se. First, with 20 mM Ca²⁺ present, the compound also failed to inhibit the outwardly rectifying current seen in Fig. 2 D, which we believe is mediated by the same channels that conduct the large monovalent current under DVF conditions (see below). Furthermore, complete inhibition of I_CRAC by drug application in the presence of Ca²⁺ did not affect or delay the appearance of the large monovalent current upon removal of extracellular divalents (Fig. 4 C). Thus, CRAC channels and the large monovalent current are distinctly different in their sensitivities to inhibition by SKF 96365.

A comparison of the effects of 2-APB provides further evidence that the two currents are distinct. 2-APB, originally described as a noncompetitive antagonist of IP₃ receptors, also inhibits store-operated Ca²⁺ entry in several cell types (Ma et al., 2000; Broad et al., 2001; Prakriya and Lewis, 2001). We have shown previously that I_CRAC in Jurkat cells is enhanced several-fold by low concentrations of 2-APB (≤5 μM) and that this potentiation is followed by nearly irreversible inhibition at higher concentrations (IC₅₀ of ∼10 μM) (Prakriya and Lewis, 2001). The dual effects of 2-APB are illustrated in Fig. 5, which shows potentiation of I_CRAC by 5 μM 2-APB (Fig. 5 A), and potentiation and subsequent inhibition by 50 μM 2-APB (Fig. 5 B). The large monovalent current differed dramatically in its response to 2-APB. First, low concentrations of 2-APB (≤5 μM) failed to enhance the large monovalent current. Second, 50 μM 2-APB inhibited the large monovalent current by <50% (Fig. 5 C). In six cells, 40 μM 2-APB inhibited I_CRAC by 97 ± 2% (Prakriya and Lewis, 2001), but inhibited the large Na⁺ current by only 48 ± 6%. Inhibition by 50 μM 2-APB of the large monovalent current was also much slower (τ = 90 ± 8 s; n = 6) than inhibition of I_CRAC (τ = 10 ± 1 s, n = 6; Prakriya and Lewis, 2001). Finally, inhibition of the large Na⁺ current was rapidly and completely reversible, whereas block of I_CRAC was essentially irreversible (see Fig. 5 B; Prakriya and Lewis, 2001). Thus, the large monovalent current and I_CRAC exhibit clearly different sensitivities and responses to 2-APB.

We also tested for the presence of the large monovalent current in Jurkat cell mutants that express low levels of I_CRAC. One of these lines, CJ-2, expresses ∼14% of the normal level of I_CRAC (Fanger et al., 1995). Despite this defect in store-operated Ca²⁺ entry, CJ-2 cells express normal levels of I_MIC (−114 ± 8 pA/pF [n = 5] in CJ-2 vs. −127 ± 23 pA/pF [n = 7] in wild-type cells; measured at −110 mV). These results provide genetic evidence that CRAC channels and the large monovalent current represent distinct conduction pathways.

In sum, multiple independent lines of evidence demonstrate that the large monovalent current evoked in the absence of extracellular divalent ions and intracellular Mg²⁺ does not flow through CRAC channels. The large monovalent current is not store-dependent, its time course is independent of changes in CRAC channel activity, it is pharmacologically distinct from I_CRAC and its expression is regulated independently of CRAC in mutant cells. In the following section we examine the properties of the large monovalent conductance and describe methods for inhibiting it so that monovalent currents through CRAC channels can be measured in isolation.

The induction of the large monovalent current by intracellular pipette solutions lacking Mg²⁺ suggests that it may be activated as intracellular Mg²⁺ is dialysed out of the cell. We obtained direct support for this idea at both the whole-cell and single-channel levels. Addition of 8 mM MgCl₂ to the whole-cell pipette solution completely suppressed the activation of the large monovalent current under DVF conditions (Fig. 6 A) with an IC₅₀ of ∼0.6 mM (Fig. 6 B). Intracellular Mg²⁺ also inhibited the development of the outwardly rectifying current seen in 20 mM Ca²⁺_o (Fig. 2 D) with about the same efficacy (Fig. 6 B), suggesting that both currents arise from the same channel.

Mg²⁺ applied to the cytoplasmic side also inhibited single-channel monovalent currents in excised inside-out patches. An intriguing feature of Mg²⁺-inhibition of these channels in excised patches was that full inhibition was observed even at Mg²⁺ concentrations as low as 100 μM (Fig. 6 D; 3/3 patches). At present, we do not understand why the sensitivity to Mg²⁺ increases after patch excision. Unlike the effect of 100 μM Mg²⁺, application of 2 mM Mg²⁺ to the cytoplasmic face of the membrane caused irreversible channel closure (Fig. 6 E; 8/8 patches). Similar effects have been reported for the 40-pS channels in RBL cells (Braun et al., 2001). The slow time course and variable reversibility of inhibition by Mg²⁺ suggests that Mg²⁺ causes channel closure rather than acting as a pore blocker. To denote their Mg²⁺ sensitivity and nonselective cation permeability, we refer to these channels as Mg²⁺-inhibited cation channels (MICs), and to the whole-cell current as I_MIC. We apply this name for simplicity, and not to imply that intracellular Mg²⁺ is a physiological regulator of channel activity (see discussion).

To isolate I_MIC for further characterization, I_CRAC was blocked irreversibly by pretreatment with 100 μM 2-APB for 15 min. Following break-in with a Mg²⁺-free internal solution, ramp currents were recorded while switching between 20 mM Ca²⁺ and DVF Ringer's solutions. Under these conditions, I_MIC (observed during brief applications of DVF Ringer's) developed in parallel with an outwardly rectifying current (observed during applications of 20 Ca²⁺ Ringer's) (Fig. 7, A and B). Similar behavior was seen in 5/5 cells. As with I_MIC, the outwardly rectifying current was not observed when 8–10 mM Mg²⁺ was included in the pipette solution (unpublished data), supporting the idea that this current is conducted by MIC channels.

Under DVF conditions, MIC currents reverse at −6 ± 2 mV (n = 5), indicating a P_Cs/P_Na ratio of 1.2. In the presence of extracellular divalent ions, however, the permeability of MIC channels is complex. Addition of divalent ions transforms the linear I-V relation in DVF conditions to a outwardly rectifying one, suggesting that external divalent ions block monovalent current flow in a voltage-dependent manner. I_MIC reverses around 0 mV under these conditions also (Fig. 7), consistent with a lack of selectivity among cations. Our preliminary results from ion substitution experiments indicate that Na⁺, Ca²⁺, and Mg²⁺ all carry current through MIC channels at negative potentials; however, Mg²⁺ and Ca²⁺ also block the passage of Na⁺, making it difficult on the basis of ion substitution experiments to determine the exact proportion of current carried by each ionic species. For this reason, we did not characterize the permeation properties of MIC channels further, but instead focused on ways to prevent contamination of I_CRAC measurements with I_MIC. Importantly, the inward MIC current is unaffected by switching from 20 mM Ca²⁺_o to a Ca²⁺-free Ringer's solution containing 2 mM Mg²⁺ (Fig. 7 C). Therefore, for purposes of isolating I_CRAC at negative potentials, leak subtraction using currents obtained in a Ca²⁺-free solution effectively removes the contaminating current arising from MIC channels (see materials and methods).

A recently cloned member of the TRP family of ion channels, called TRP-PLIK (Runnels et al., 2001) or LTRPC7 (Nadler et al., 2001), has an ionic selectivity and current-voltage relation quite similar to that of MIC. LTRPC7 is expressed in Jurkat cells and is inhibited by intracellular Mg²⁺ as well as by intracellular MgATP (Nadler et al., 2001). Likewise, we found that MgATP at concentrations of 4 mM and higher in the intracellular pipette solution completely inhibited the development of I_MIC (Fig. 6 C), whereas recordings with 3 mM MgATP exhibited sporadic MIC channel activity. Our combined observations suggest that I_MIC is mediated by TRP-PLIK/LTRPC7 channels, and is probably identical to an outwardly rectifying MgATP-sensitive current (called MagNuM, for Mg²⁺ nucleotide–regulated metal) that has been reported in Jurkat, RBL, and HEK293 cells (Nadler et al., 2001).

Inhibition of I_MIC by intracellular Mg²⁺ offers a convenient strategy for isolating the monovalent current through CRAC channels for characterization. With 8 mM intracellular Mg²⁺ to suppress I_MIC, passive depletion of Ca²⁺ stores by intracellular BAPTA (10 mM) slowly evoked I_CRAC in the presence of 20 mM Ca²⁺ (Fig. 8 A). Periodic exposure to DVF solution revealed a monovalent current which developed with the same time course as I_CRAC (Fig. 8 B). Similar results were seen in 7/7 cells. The monovalent current was transient during each DVF episode, peaking immediately after solution exchange and subsequently declining with a time constant of ∼10 s. The ratio of the peak amplitude of the Na⁺ current to I_CRAC measured just before solution exchange was 7.8 ± 0.4 (n = 17).

The parallel activation kinetics of I_CRAC and the transient monovalent current suggests that the latter current represents flux through CRAC channels. As a further test, we examined the transient monovalent current's dependence on store refilling under conditions that inhibit I_MIC activation (8–10 mM Mg²⁺_i). In the presence of 1 mM intracellular BAPTA and 20 mM Ca²⁺_o, activation of I_CRAC was followed by characteristic slow deactivation, presumably as Ca²⁺ entering the cell overwhelmed the buffer and refilled the stores (Zweifach and Lewis, 1995) (Fig. 8 C). Periodic removal of external divalents showed that the transient Na⁺ current decreased in parallel with I_CRAC (Fig. 8, C and D). Similar behavior was seen in 6/6 cells. Together, these results show that the transient Na⁺ current is store-dependent, similar to I_CRAC but unlike I_MIC.

The pharmacological profile of the transient Na⁺ current also closely matched that of I_CRAC. As shown in Fig. 9 A, 20 μM SKF 96365 inhibited both currents to similar extents. On average, I_CRAC was reduced by 84 ± 3% (n = 4), whereas the inactivating Na⁺ current was reduced by 74 ± 3% in the same cells. Moreover, 5 μM 2-APB potentiated both currents, enhancing I_CRAC by 241 ± 41% (n = 7) and the peak amplitude of the transient Na⁺ current in the same cells by 185 ± 31% (Fig. 9 B). 2-APB also enhanced the steady-state component of the Na⁺ current by a similar amount (224 ± 26%; n = 6). These results suggest that the residual current remaining after the Na⁺ current has declined is also due to CRAC channels, a conclusion that is consistent with the observation that the peak and the residual currents reverse at the same potential (see below).

High concentrations of 2-APB inhibited the transient Na⁺ current as they did I_CRAC. As illustrated in Fig. 9 C, 40 μM 2-APB was applied to irreversibly inhibit I_CRAC, and the transient Na⁺ current was measured under DVF conditions following washout of the drug. In three cells, 50 μM 2-APB reduced I_CRAC in the presence of 20 mM Ca²⁺ by 83 ± 3% and the monovalent current by 80 ± 6%. These results also show that block of the transient monovalent current by 2-APB outlasts bath application of the drug, as is the case for I_CRAC but not I_MIC (Fig. 5). Thus, the pharmacological properties of the transient Na⁺ current are consistent with those of I_CRAC.

Based on these results as well as the evidence given above that I_CRAC and the transient Na⁺ current activate and deactivate in parallel and are store-dependent, we conclude that 8–10 mM intracellular Mg²⁺ suffices to isolate the monovalent current through CRAC channels in the absence of extracellular divalent ions. For convenience, the inward Na⁺ and Ca²⁺ currents through CRAC channels will hereafter be referred to as Na⁺-I_CRAC and Ca²⁺-I_CRAC, respectively.

The identification of a Mg²⁺-sensitive monovalent current raises new questions about the possible role of Mg²⁺ in shaping the properties of CRAC channels. Previous studies have concluded that intracellular Mg²⁺ blocks permeation of Cs⁺ through CRAC channels under DVF conditions (Kerschbaum and Cahalan, 1998). Using the methods described above for separating I_MIC from I_CRAC, we reassessed the role of Mg²⁺ in CRAC channel selectivity.

Fig. 10 A shows the transient monovalent CRAC current with 8 mM Mg²⁺_i and with equimolar Cs⁺ and Na⁺ as the principal intracellular and extracellular cations, respectively. The I-V relations recorded as the current depotentiated show pronounced inward rectification and intersect at a single potential. This supports the idea that the whole-cell current decreases due to the closure of a single channel type, i.e., CRAC channels, and that the residual steady-state current is also due to CRAC channels as discussed above. The crossover point in eight cells was 52 ± 2 mV, implying that CRAC channels under DVF conditions are only weakly permeable to Cs⁺ (P_Cs/P_Na = 0.13 based on the Goldman-Hodgkin-Katz equation). A similar reversal potential was obtained from single ramp currents after subtraction of leak current recorded in 0-Ca Ringer's (unpublished data). Thus, unlike MIC channels, CRAC channels in the absence of divalent cations select strongly against Cs⁺ (compare Fig. 7 B). Consistent with this conclusion, replacement of extracellular Na⁺ with Cs⁺ eliminates nearly all of the inward monovalent current through CRAC channels (Fig. 10 B).

These experiments do not address the role of Mg²⁺ in determining the selectivity against Cs⁺, because 8–10 mM intracellular Mg²⁺ was present to block I_MIC. Unfortunately, if intracellular Mg²⁺ is removed, I_MIC develops to such an extent that it readily swamps the small monovalent current through CRAC channels. Therefore, to measure Na⁺-I_CRAC in the absence of Mg²⁺, we exploited the ability of 2-APB to enhance I_CRAC selectively. After the full development of I_MIC with 0 Mg²⁺_i, 5 μM 2-APB was applied to elevate I_CRAC. The current immediately before 2-APB application was subtracted from the enhanced current to remove the contribution of I_MIC (Fig. 10 C). The I-V relation of the additional monovalent current is similar to that of the current recorded in the presence of 8 mM Mg²⁺_i (Fig. 10 A). The 2-APB-enhanced current reverses at potentials >40 mV and shows inward rectification, though to a slightly lesser extent than in the presence of Mg²⁺_i (compare Fig. 10 C with 10 A). Thus, the low permeability of CRAC channels to Cs⁺ does not require intracellular Mg²⁺, but is instead likely to be an intrinsic property of the CRAC channel pore.

The activity of CRAC channels slowly declines by up to 80% after removal of extracellular Ca²⁺, and recovers over 10–20 s after its reapplication by a process referred to as Ca²⁺-dependent potentiation, or CDP (Christian et al., 1996b; Zweifach and Lewis, 1996). Thus, the current's decline following the removal of extracellular divalents reflects depotentiation. Previous findings that transient monovalent currents (thought to be through CRAC channels) became large and sustained in the absence of intracellular Mg²⁺ led to the suggestion that internal Mg²⁺ is required for the depotentiation process, and that CDP may arise from the expulsion of Mg²⁺ ions from the pore of CRAC channels by permeant Ca²⁺ ions (Kerschbaum and Cahalan, 1998). Because I_MIC was likely mistaken for I_CRAC in the absence of intracellular Mg²⁺, we reexamined the role of Mg²⁺ in CRAC channel depotentiation under conditions that prevent activation of I_MIC.

To effectively inhibit I_MIC under conditions of low intracellular Mg²⁺, we applied MgATP in combination with Na₂ATP to reduce free [Mg²⁺]_i. 6–8 mM MgATP + 2–6 mM Na₂ATP in the pipette solution inhibited I_MIC while reducing free [Mg²⁺]_i to calculated values of ∼92–185 μM (see materials and methods). Under these conditions, Na⁺-I_CRAC still depotentiated after removal of extracellular divalents (Fig. 11 A), and the I-V relationship of the current was similar to that recorded with 8 mM Mg²⁺_i. Neither the time course (Fig. 11 C) nor the amplitude of depotentiation were altered relative to cells with 8 mM Mg²⁺_i; the fraction of current remaining after 50 s of DVF application was 20 ± 4% (n = 7) with 131 μM Mg²⁺_i vs. 18 ± 2% (n = 8) with 8 mM Mg²⁺_i. Finally, if depotentiation were to arise from Mg²⁺ binding in the pore of the CRAC channel and within the membrane field, hyperpolarization would be expected to inhibit it by reducing Mg²⁺ binding. However, shifting the holding potential from 40 mV to −110 mV did not significantly alter the rate of depotentiation at either high (6–10 mM) or low (131 μM) levels of Mg²⁺_i (Fig. 11, B and C), even though the maximum amplitude of Na⁺-I_CRAC was increased, consistent with the voltage dependence of CDP (Zweifach and Lewis, 1996). Together, these results argue against a necessary role for intracellular Mg²⁺ in the depotentiation of CRAC channels.

However, an alternative explanation could be that the putative Mg²⁺ site for depotentiation has a high affinity and does not lie within the membrane field. We examined this possibility by exploring Mg²⁺ effects on Ca²⁺-I_CRAC. Full removal of intracellular Mg²⁺ in the presence of extracellular Ca²⁺ evokes only a small inward I_MIC, and the current is not affected by removal of Ca²⁺_o alone (Fig. 7 C), providing a way of isolating it from I_CRAC through leak subtraction. Therefore, we used these conditions to ask whether Ca²⁺-I_CRAC depotentiates in the absence of intracellular Mg²⁺. Depotentiation of I_CRAC in the absence of Ca²⁺_o can be detected indirectly by the slow recovery of full current amplitude after readdition of Ca²⁺ (Fig. 11 D). Under control conditions (8 mM Mg²⁺_i), Ca²⁺ readdition causes Ca²⁺-I_CRAC to appear in two kinetically distinct stages: a small fraction of the current appears immediately (i.e., within the 2-s exchange time of the perfusion system) that represents the current through active CRAC channels, followed by a several-fold increase in current amplitude over the next 10–20 s due to CDP. Thus, the size of the slowly relaxing current component reflects the extent of depotentiation that occurred before the readdition of Ca²⁺. Significantly, neither the amplitude nor the time course of CDP was affected by intracellular Mg²⁺. At a holding potential of 20 mV, CDP caused a 104 ± 5% increase in Ca²⁺-I_CRAC with an average τ of 2.9 ± 0.3 s (n = 6) in cells with 8 mM Mg²⁺_i, and a 114 ± 7% increase with τ = 3.1 ± 0.4 s (n = 4) in cells in which Mg²⁺_i was chelated by 10 mM HEDTA. These results are consistent with those of Figs. 2 and 3, where I_CRAC depotentiates during applications of DVF solution in the absence of Mg²⁺_i. Thus, intracellular Mg²⁺ appears to have no influence on CRAC-channel depotentiation triggered by removal of extracellular Ca²⁺.

The finding that the 40-pS monovalent channels seen under DVF conditions are MIC channels raises new questions about the unitary monovalent conductance of CRAC channels. Under DVF conditions with 10 mM intracellular Mg²⁺ to inhibit I_MIC, we were unable to detect any clear single-channel transitions of amplitude >0.4 pA during the induction of Na⁺-I_CRAC by ionomycin in 5/5 cells and during the or depotentiation of Na⁺-I_CRAC. This contrasts with the obvious appearance of single-channel MIC currents when intracellular Mg²⁺ is washed out during whole-cell recording (Fig. 1). Therefore, we applied stationary fluctuation analysis to estimate the single-channel monovalent CRAC conductance.

Cells were treated with 1 μM TG to activate CRAC channels, and after establishing whole-cell recording the mean macroscopic current (I) and corresponding variance (σ_I²) were measured at a holding potential of −110 mV. Application of 5 μM 2-APB enhanced I_CRAC in the presence of 20 mM Ca²⁺ and caused a barely detectable increase in current noise (Fig. 12 A). Subsequent removal of divalent cations (DVF conditions) evoked a transient Na⁺ current through CRAC channels in parallel with a more robust increase in noise. Several 200-ms current sweeps used to compute the current mean and variance for the DVF condition are shown in Fig. 12 B, and indicate the lack of any clear single-channel transitions as the current declines. Plots of the current variance against mean current amplitude were well fitted by straight lines (Fig. 12, C and D). The average slope (σ_I²/I) was −3.8 ± 0.6 fA (n = 5) in 20 mM Ca²⁺, and −31 ± 2 fA (n = 10) under DVF conditions.

The relation of the variance to the mean current can be used to estimate the single-channel current given certain assumptions. For a homogeneous population of N independent channels with a single conducting level, i, stochastic gating predicts a parabolic relation between the current variance and the mean current as the open probability (P_o) varies between zero and one (Sigworth, 1980):

where σ₀² is the variance at the zero-current level. Since I = NiP₀, it follows that

Given these assumptions, the linear shape of the σ_I²/I relation implies that the open probability of the channel is very low (<0.5) and that the unitary current amplitude is given by the slope (Eq. 1). Thus, the results would suggest that the unitary amplitude of monovalent I_CRAC at −110 mV is −31 fA; given the reversal potential of 52 mV, this corresponds to a single-channel chord conductance of ∼0.2 pS.

There are several ways in which fluctuation analysis may seriously underestimate the unitary conductance. First, the bandwidth of the recording may not extend to high enough frequencies to capture the fluctuations due to very brief gating events. To test this, we analyzed the power spectrum of the current noise under DVF conditions. Fig. 12 E shows the background spectrum obtained in 0 Ca²⁺ + 2 mM Ni²⁺, together with the spectrum of Na⁺-I_CRAC at its peak and after it depotentiated to a steady level. In both cases, the noise approaches the background level at frequencies above ∼1 kHz, suggesting that no large component of high-frequency noise was missed under the 1–2 kHz filtering conditions of the current variance experiments.

Second, noise measurements may underestimate the size of i if P_o is high and the macroscopic current amplitude reflects changes in N, the number of activatable channels, rather than P_o (Jackson and Strange, 1995). In this case, σ_I² will increase linearly with N:

and the σ_I²/I ratio underestimates i by a factor of (1 − P_o) (Eq. 1). If this situation applies, a small reduction in the P_o resulting from extracellular blockade of the channels would be expected to produce an increase in channel noise due to an increased number of gating events occurring within each 200-ms sampling period. To examine this possibility, we added 1 μM free Ca²⁺ to the extracellular DVF solution to partially block CRAC channels and reduce channel P_o. The peak I_Na/I_Ca ratio was reduced to 3.1 ± 0.3 (n = 4; versus a ratio of 7.8 ± 0.4 without the blocker), indicating >50% inhibition of the Na⁺-I_CRAC. Under these conditions, the σ_I²/I ratio was −19 ± 3 fA (n = 4). This ratio is even smaller than the σ_I²/I ratios obtained in the absence of the blocker.¹ Given these results, it seems unlikely that very high P_o of CRAC channels is causing a gross underestimation of the single channel current. Thus, from the results of noise analysis, we estimate that the monovalent conductance of the CRAC channel is ∼0.2 pS.

In the absence of extracellular divalent cations, removal of intracellular Mg²⁺ during whole-cell recording activates a large monovalent current in T cells and RBL cells. This large Mg²⁺-sensitive current has previously been attributed to monovalent permeation of CRAC channels, and its signature conductance and selectivity have been employed as a biophysical fingerprint to screen candidate CRAC channel genes. In this paper, we show that this large monovalent current, I_MIC, and the 40-pS channels that underlie it, differ in fundamental ways from I_CRAC and therefore represent a distinct ion channel type. These findings necessitate a revision of the biophysical fingerprint of CRAC channels, and have important implications for studies of CRAC channel activation, permeation, and cellular function.

A number of characteristics distinguish MIC channels from CRAC channels. These include the mode of activation, inhibition by Mg²⁺ or MgATP, kinetic properties, selectivity, pharmacology, and unitary conductance. These differences are summarized in Table I and are discussed below.

Several observations indicate that unlike I_CRAC, I_MIC is not store-operated (Figs. 2 and 3). For example, I_MIC slowly activates in cells in which I_CRAC is already preactivated by store depletion, and it activates even as I_CRAC deactivates due to store refilling. I_MIC is effectively suppressed by intracellular Mg²⁺ or MgATP, whereas CRAC channel gating is not noticeably affected. I_MIC activates about twice as slowly as I_CRAC in Jurkat cells; thus, the activation kinetics clearly do not reflect the time course of store depletion. Moreover, because introduction of 10 mM HEDTA into the cell would be expected to quickly buffer Mg²⁺_i, it is unlikely that the time course of I_MIC is limited by the rate at which Mg²⁺ or MgATP are removed from the cell. Activation of single 40-pS channels after patch excision from RBL cells into a Mg²⁺- and ATP-free solution can also be slow (Braun et al., 2001), suggesting that the rate-limiting step lies downstream of Mg²⁺ or MgATP removal. Interestingly, Mg²⁺ inhibits MIC channels more effectively in excised patches than in whole-cell recordings (Fig. 6), suggesting that the inhibition mechanism is more complex than simple channel blockade and may involve a diffusible molecule that is lost after patch excision.

Following removal of extracellular divalents, Na⁺-I_CRAC is transient (Fig. 10 A), whereas I_MIC is relatively sustained (Fig. 5 C). The peak phase of the Na⁺ current is linked to CRAC channels, based on its close correlation with Ca²⁺-I_CRAC during store depletion and refilling (Fig. 8), and its sensitivity to SKF 96365 and 2-APB (Fig. 9). Na⁺-I_CRAC depotentiated in DVF solutions to an average level of ∼20%. We believe the steady-state activity is also from CRAC channels because its amplitude varies in parallel with the peak Na⁺-I_CRAC and it has the same reversal potential, relative Cs⁺ permeability and sensitivity to SKF 96365 and 2-APB, as the peak current. Although I_MIC is generally sustained under Mg²⁺_i-free conditions after application of DVF solutions, it declines slowly in the presence of low levels of Mg²⁺_i (0.5–1 mM; unpublished data). Thus, current decline under DVF conditions cannot be used as a specific indicator for I_CRAC.

Finally, the pharmacological signatures of I_CRAC and I_MIC also differ significantly (Figs. 4 and 9). SKF 96365 at micromolar levels is an effective blocker of I_CRAC (Franzius et al., 1994; Christian et al., 1996a), but not of I_MIC. High concentrations of 2-APB (20–40 μM) inhibit I_CRAC completely and irreversibly (Prakriya and Lewis, 2001), whereas they cause modest and rapidly reversible inhibition of I_MIC. Low concentrations of 2-APB (≤5 μM) which strongly enhance I_CRAC (Prakriya and Lewis, 2001) fail to affect I_MIC.

The rather large number of properties that distinguish the two conductances strongly suggest that I_CRAC and I_MIC arise from two different channels. However, an alternative explanation might be that removal of intracellular Mg²⁺ merely alters the properties of CRAC channels, making them acquire the characteristics of what we now call MIC channels (Kerschbaum and Cahalan, 1998). We do not believe that the MIC channel is an altered state of the CRAC channel for several key reasons. First, it seems unlikely that removal of Mg²⁺ alone would be able to alter such a diverse set of properties, including store-dependence, depotentiation, ion selectivity, pharmacological profile, and unitary conductance. Second, we find that removal of intracellular Mg²⁺ does not in fact reduce the selectivity of CRAC for Na⁺ over Cs⁺ (Fig. 10 C), nor does it prevent depotentiation of CRAC channels (i.e., make CRAC activity sustained) under 0-Ca²⁺_o conditions (Figs. 2 A, 3 A, and 11 D). Finally, MIC channel activity is normal in mutant Jurkat cells that have very low levels of I_CRAC. Thus, the most parsimonious explanation for these data is that MIC and CRAC channels are separate and distinct proteins.²

In light of our results, various conclusions made in previous reports regarding the regulation of CRAC channels probably apply more to MIC channels than to CRAC channels. For example, the voltage-dependent block of monovalent CRAC currents by extracellular Ca²⁺ (Kerschbaum and Cahalan, 1998; Fomina et al., 2000), the upregulation of CRAC channels in activated T cells (Fomina et al., 2000), and regulation by cytoplasmic Mg²⁺ and Ca²⁺ (Braun et al., 2001) were all measured under conditions optimized for activation of I_MIC. These issues may have to be revisited using recording conditions explicitly optimized for isolating the monovalent CRAC current. No specific inhibitors of I_MIC are yet known, but high intracellular Mg²⁺ (8–10 mM) or MgATP (>4 mM) effectively blocks I_MIC without affecting I_CRAC. It is important to note that moderate levels of Mg²⁺ (2–3 mM) commonly used in studies of I_CRAC may not be sufficient to completely eliminate MIC activity, as we often found 3–5 active MIC channels in Jurkat cells under these conditions. Although this represents a small fraction of the total MIC conductance in the cell, the unitary current of MIC channels is so much larger than that of CRAC channels that it can seriously contaminate noise measurements under DVF conditions (see below). Activity of MIC channels can also lead to inadvertent contamination of I_CRAC in the presence of divalent ions, although appropriate subtraction of the “leak” current can alleviate this problem (see materials and methods).

Given that MIC channels are distinct from CRAC, what is their molecular identity? Recently, a novel member of the TRP family of ion channels has been cloned which has been named TRP-PLIK (Runnels et al., 2001) or LTRPC7 (Nadler et al., 2001). Several characteristics of TRP-PLIK/LTRPC7 are similar to MIC in Jurkat cells. LTRPC7 is expressed in cell lines commonly employed for studies of I_CRAC, such as Jurkat T cells and RBL cells. Activation of LTRPC7 in transfected cells is elicited by whole-cell dialysis with chelators of Mg²⁺, and introduction of Mg²⁺ and/or Mg-ATP inhibits channel activity (Nadler et al., 2001). TRP-PLIK/LTRPC7 produces a nonselective cation channel with strong outward rectification in the presence of extracellular divalent ions (Nadler et al., 2001; Runnels et al., 2001), much like the outward rectification of I_MIC we observe under similar conditions (Fig. 7). Elimination of extracellular Ca²⁺ or Na⁺ alone does not affect the inward current through LTRPC7 (Nadler et al., 2001) or MIC channels (Fig. 7 C and unpublished data). Together, these similarities to I_MIC in Jurkat cells suggest that TRP-PLIK/LTRPC7 encodes the MIC channel.

The activation mechanism and physiological roles of MIC channels are not well understood at this point. A small number of MIC channels appear to be active in resting Jurkat cells, based on the current seen in DVF conditions immediately after break-in with 0-Mg²⁺ pipette solution (Fig. 6 A). Although depletion of cytoplasmic Mg²⁺ can further activate MIC channels, it seems unlikely that this is the physiological stimulus (see Nadler et al., 2001), as cytosolic Mg²⁺ in many cells is held relatively constant between 0.5 and 1 mM (Romani and Scarpa, 2000). Its sensitivity to inhibition by Mg²⁺-nucleotides has led to the hypothesis that it opens in response to declining ATP levels (Nadler et al., 2001), but there is also evidence that channel activation requires kinase activity of the COOH-terminal PLIK domain (Runnels et al., 2001). The physiological consequences of MIC activation are also unclear. Nadler et al. (2001) have proposed that LTRPC7 provides a conduit for Ca²⁺ entry that regulates mitochondrial activity and ATP homeostasis (Nadler et al., 2001). However, the precise Ca²⁺ permeability of this channel has not been measured. Although it has been stated that the block of Na⁺ flux by external divalent ions renders LTRPC7 solely permeable to divalent ions at negative potentials (Nadler et al., 2001), our preliminary experiments on MIC channels in Jurkat cells suggest that Na⁺ can carry inward current in the presence of Mg²⁺_o (unpublished data). Thus, while MIC channels may provide a pathway for Ca²⁺ entry, they may also depolarize the cell by conducting monovalent ions. From the relative amplitudes of the single-channel and whole-cell currents we estimate that Jurkat cells express ∼250–500 MIC channels per cell.

In previous studies, the removal of intracellular Mg²⁺ under DVF conditions appeared to change a small transient and Na⁺-selective current through CRAC channels into a much larger, sustained, and nonselective current with equal permeability to Cs⁺ and Na⁺ (Lepple-Wienhues and Cahalan, 1996; Kerschbaum and Cahalan, 1998). These observations were interpreted to mean that Mg²⁺_i removal prevents depotentiation of CRAC and alters their selectivity. We have found that intracellular Mg²⁺ does not affect Ca²⁺-dependent potentiation of CRAC channels, or the reverse process of depotentiation (Fig. 11), nor is it required to achieve selectivity for Na⁺ over Cs⁺ (Fig. 10 C). This discrepancy with the earlier results can now be explained by the fact that removal of Mg²⁺_i activates I_MIC, which is much larger, more sustained, and less selective than I_CRAC.

Our results indicate that the monovalent conductance of CRAC channels is too small to be resolvable in whole-cell recordings. Noise analysis of whole-cell Na⁺-I_CRAC suggests a conductance in the range of ∼0.2 pS. This is dramatically different from recent reports in T cells and RBL cells, where a conductance of ∼40 pS was reported for monovalent currents through CRAC channels (Kerschbaum and Cahalan, 1999; Fomina et al., 2000; Braun et al., 2001). This discrepancy is explained by the fact that the earlier experiments were performed in the absence of cytoplasmic Mg²⁺, which as we have shown here activates store-independent MIC channels. Our estimate of the unitary conductance is also smaller than an earlier estimate of 2.6 pS based on noise analysis (Lepple-Wienhues and Cahalan, 1996). Using the ionic conditions of this earlier study (3 mM intracellular Mg²⁺ + 1 μM extracellular Mg²⁺), we also detect a significantly larger noise, which appears to arise from flickery block of several active MIC channels in the cell. Thus, even slight contamination of Na⁺-I_CRAC with I_MIC may lead to large differences in the estimated unitary currents.

We considered several possible factors that could lead us to underestimate the unitary currents. Current fluctuations could be missed if they occur at frequencies outside the recording bandwidth. However, spectral analysis showed that most of the power of the current noise occurred at frequencies <1 kHz, well within the recording bandwidth (Fig. 12 E). A more difficult problem could arise if activation of I_CRAC occurs through an increase in N, the number of activatable channels, each of which has a high open probability and therefore contributes little noise. However, we were unable to detect discrete jumps in current during the activation or depotentiation of monovalent I_CRAC. In addition, partial inhibition of monovalent I_CRAC with 1 μM extracellular Ca²⁺ did not increase the noise, indicating that whatever the mechanism of activation, the P_o of active CRAC channels conducting Na⁺ is quite small. Together with our estimates of the single channel conductance, these observations set limits on the maximum size of the single-channel conductance to <1 pS.

Our estimate of −3.8 fA at −110 mV for the unitary Ca²⁺ current compares well with a previous estimate in 110 mM Ca²⁺_o at −80 mV (−3.7 fA) (Zweifach and Lewis, 1993). Note that this estimate is roughly eightfold smaller than the single-channel estimates of the Na⁺ current given above (−31 fA). The fact that the whole-cell Na⁺-I_CRAC is also about eightfold larger than Ca²⁺-I_CRAC suggests that the increase in whole-cell current seen upon exchanging the Ca²⁺_o solution for the DVF solution can entirely be accounted for by the increase in conductance of the CRAC channels without a significant change in channel P_o. From the ratio of the unitary and whole-cell peak Na⁺ currents, we estimate that the number of CRAC channels per cell is at least 5,000–10,000. More would be expected if the open probability is low, as is suggested by the linear relation of variance to mean current. Finally, it should be noted that our estimate for the unitary conductance of Na⁺-I_CRAC (0.2 pS) is significantly smaller than the 42-pS conductance reported for the CaT1 channel under similar DVF conditions (Yue et al., 2001), consistent with recent suggestions that CaT1 may not comprise the complete CRAC-channel pore (Voets et al., 2001).

Similarities and differences between the permeation properties of CRAC channels and voltage-gated Ca²⁺ (Ca_V) channels have interesting implications for mechanisms of ion selectivity in CRAC channels. Under physiological conditions, Ca_V channels have an extremely high selectivity for Ca²⁺ over monovalent ions (P_Ca/P_Na > 1,000). The high Ca²⁺ selectivity is thought to arise from the high-affinity binding of Ca²⁺ within the pore, which prevents Na⁺ from conducting. In fact, the removal of extracellular divalent ions allows Na⁺ to conduct freely (Almers and McCleskey, 1984; Hess and Tsien, 1984). CRAC channels show a comparably high selectivity for Ca²⁺ over monovalent ions (Hoth and Penner, 1993), and similarly become freely permeable to Na⁺ in the absence of extracellular divalents. Thus, it is likely that the Ca²⁺ selectivity of CRAC channels also arises from high affinity Ca²⁺ binding within the pore, and this is consistent with the ability of 1 μM Ca²⁺ to block Na⁺-I_CRAC by ∼50% (see results).

The most obvious differences between the two channels relates to their unitary conductances and permeability to Cs⁺. Ca_V channels have single-channel conductances for Ca²⁺ in the range of 5–20 pS and ∼85 pS for Na⁺ under DVF conditions (Hess et al., 1986). By contrast, the unitary conductances of CRAC channels is estimated to be ∼21 fS for Ca²⁺ and 0.2 pS for Na⁺, or ∼500-fold smaller than Ca_V channels. In addition, under DVF conditions Ca_V channels readily pass Cs⁺ (P_Cs/P_Na ∼0.6; Hess et al., 1986), whereas CRAC channels do not (P_Cs/P_Na = 0.13; Fig. 10; see also Lepple-Wienhues and Cahalan, 1996). These observations suggest that although CRAC channels and voltage-gated Ca²⁺ channels achieve selectivity for Ca²⁺ over other ions by high affinity Ca²⁺ binding within the pore, many structural features may be significantly different between these two channel types.

What unique structural features of CRAC channels could give rise to such a low conductance? A popular model proposes that the high throughput of Ca_V channels arises from ion–ion interactions as ions move single-file through the pore (Tsien et al., 1987). Thus, one possibility is that weak ion–ion interactions underlie the low flux rate of Ca²⁺ through CRAC channels. However, this idea cannot adequately explain why CRAC channels would have an equally low conductance for monovalent ions. We hypothesize that CRAC channels possess a nonselective rate-limiting barrier to ion permeation in series with the selectivity filter. Such a barrier may explain not only the uniformly low permeability to various ions, but also the lack of permeability to Cs⁺. Identification of CRAC channel genes will provide much needed tools for elucidating the mechanism of CRAC channel permeation. In the interim, systematic examination of permeation of various ions through native CRAC channels will improve our understanding of the properties of the channel pore.

The authors would like to thank K. Mikoshiba for the gift of 2-APB. We thank members of the Lewis lab for helpful discussions and Dr. Rick Aldrich, Dr. Adam Zweifach, and Diana Bautista for comments on the manuscript.

This work was supported by a postdoctoral fellowship from the Irvington Foundation for Immunological Research to M. Prakriya and National Institutes of Health grant GM45374 to R.S. Lewis.