Modulation of the slow/common gating of CLC channels by intracellular cadmium

Members of the CLC Cl⁻ channel/transporter family participate in a wide spectrum of biological processes from extreme acid resistance in enteric bacteria to lysosomal acidification and membrane voltage regulation in mammalian cells (Jentsch, 2008; Jentsch et al., 2005; Stauber et al., 2012). These CLC proteins contain two identical subunits (Middleton et al., 1994, 1996; Ludewig et al., 1996), each forming a Cl⁻-transport pathway (Dutzler et al., 2002, 2003). In eukaryotic CLC Cl⁻ channels, such as CLC-0 and CLC-1, the individual anion-transport pathways (or pores) are gated by the negatively charged side chain of a glutamate residue (Dutzler et al., 2003). The swing of this glutamate side chain is believed to be responsible for the “fast gating” operating in the time range of milliseconds. There is a second gating mechanism called “slow gating” (for CLC-0) or “common gating” (for CLC-1), which controls the opening and closing of both pores simultaneously with a slower kinetics (Chen, 2005; Miller, 1982; Miller and White, 1984). Although recent work has identified certain molecular determinants for the slow/common gating of CLC channels (Bennetts and Parker, 2013), the exact mechanism of this gating process is not well understood.

It has been suggested that interactions between the two CLC subunits are involved in the slow/common gating processes (Estévez and Jentsch, 2002; Bykova et al., 2006). Crystal structures of the Escherichia coli CLC (CLC-ec1) and a eukaryotic CLC transporter CmCLC show extensive helix packing in the subunit interface (Dutzler et al., 2002; Feng et al., 2010; Robertson et al., 2010). At this flat, nonpolar interface, hydrophobic residues such as valine, leucine, and isoleucine form most of the cross-subunit contacts (Dutzler et al., 2002; Robertson et al., 2010). Replacing two isoleucine residues, I201 and I422, at the dimer interface of CLC-ec1 (see Fig. 1 for the locations of these two residues) with tryptophan disrupts the dimeric architecture, and the I201W/I422W double mutant of CLC-ec1 exists as a monomer (Robertson et al., 2010). Coincidentally, previous work has also demonstrated that mutations of I290 and I556 of CLC-1 (which correspond to I201 and I422, respectively, in CLC-ec1) alter the common gating (Pusch et al., 1995; Kubisch et al., 1998; Saviane et al., 1999). Collectively, these results suggest that molecular events occurring in the CLC dimer interface contribute to slow/common gating.

The research in our laboratory has revealed previously that extracellular zinc ions (Zn²⁺) inhibit CLC-0 by facilitating the closure of slow gate, although the Zn²⁺-binding site for this effect is not known (Chen, 1998). In the present study, we find that intracellular transition metal cations (such as Cd²⁺) also inhibit the slow gating in WT CLC-0. We have identified that the Cd²⁺-binding site responsible for this inhibition effect in CLC-0 is formed by C229 and H231 located at the dimer interface near the intracellular end of the transmembrane domain (Fig. 1). Interestingly, the Cd²⁺ inhibition potency in CLC-0 is drastically increased by I225W/V490W mutations, corresponding to the dimer interface mutations of I201W/I422W in CLC-ec1. In addition, formation of a disulfide bond between introduced cysteine residues in this binding site is also affected by mutations of I225W and V490W. These results show that a structural change of the Cd²⁺-binding site, as reported by the alterations of Cd²⁺ modification and disulfide-bond formation, is induced by the dimer interface mutations, which also alter CLC channel’s slow/common gating. These results thus suggest that a structural rearrangement at the dimer interface near the intracellular Cd²⁺-binding site is likely involved in the slow/common gating of CLC channels.

MTS reagents were purchased from Toronto Research Chemicals. Stock solutions of 0.3 M MTS reagents were made in distilled water and stored at −80°C. The working solutions with required concentrations of MTS reagents were made immediately before use from the stock solution placed on ice, and were discarded if not used within 5 min. Cadmium chloride (CdCl₂), HEPES, EGTA, and various salts for electrophysiological recordings were from Sigma-Aldrich and Thermo Fisher Scientific.

The CLC-0 and CLC-1 cDNAs used for electrophysiological recordings were constructed in the pcDNA3 vector without any genetic tags of fluorescent proteins. For the constructs used in biochemical experiments, a “Flag” epitope was added to the N terminus (after the first methionine) of different CLC-0 constructs subcloned in the pcDNA3 vector. Mutations were made using the QuikChange site-directed mutagenesis kit (Agilent Technologies) and were confirmed by commercial DNA sequencing services. In the fluorescent resonant energy transfer (FRET) study used to evaluate the interaction of two channel subunits (Fig. 3), the C-terminal cytoplasmic domain of CLC-0 was removed after Ser530 to make a fair comparison with CLC-ec1 because the latter lacks a cytoplasmic domain. This C terminus–truncated mutant of CLC-0, referred to as CLC0-TM, has been described previously (Bykova et al., 2006). It was constructed in the pEGFP-N3 vector, and the C terminus of the CLC0-TM sequence was tagged by a Gly-Ser linker followed by the sequence of cerulean or eYFP. The I225W/V490W mutations of CLC-0 (corresponding to I201W/I422W mutations in CLC-ec1) were made in these two CLC0-TM constructs, and were used in the FRET study. The CNG channel used as the negative control in the FRET experiment was provided by W.N. Zagotta (University of Washington, Seattle, WA), and fluorescent proteins were also tagged to the C terminus of the CNG channel.

All experiments on WT CLC-0 and its mutants were conducted in human embryonic kidney (HEK)293 cells, whereas experiments on WT and mutant CLC-1 were conducted in tsA201 cells. Transfections of cDNA into HEK293 cells or tsA201 cells followed a procedure described previously (Lin and Chen, 2003; Zhang et al., 2006, 2010; Zhang and Chen, 2009), using the lipofectamine transfection kit (Life Technologies). Experiments were performed 1–3 d after transfections.

Constructs of WT or mutant CLC channels were cotransfected with the cDNA of the green fluorescent protein (GFP). The green fluorescence was visualized via an inverted microscope (DM IRB; Leica) equipped with a GFP filter (Chroma Technology Corp.). Macroscopic currents from excised inside-out membrane patches of the transfected cells were recorded according to methods described previously (Lin and Chen, 2003; Zhang et al., 2006, 2010; Zhang and Chen, 2009). Both Cd²⁺ and Zn²⁺ (not depicted) can inhibit WT CLC-0 from the intracellular side with roughly similar potency. However, Zn²⁺ inhibition on the I225W/V490W mutant has a slower off-rate compared with that of the Cd²⁺ inhibition. Thus, for technical convenience, we focused on the experiments using Cd²⁺. All electrophysiological experiments were conducted using an amplifier (Axopatch 200B; Molecular Devices), and the recording traces were digitally filtered at 5 kHz, and digitized at 10 kHz using a Digidata 1440 digitizing board and pClamp10 software (Molecular Devices). The recording pipettes were fabricated from borosilicate glass capillaries (World Precision Instruments) using the pp830 pipette puller (Narishige International, Inc.). The electrodes normally had a resistance of ∼1–3 MΩ when filled with the pipette (extracellular) solution containing (mM): 130 NaCl, 5 MgCl₂, 10 HEPES, and 1 EGTA, pH 7.4. The bath (intracellular) solution was the same as the pipette solution except that EGTA was not included in the solution containing Cd²⁺. Because the two gating mechanisms in WT CLC-0 operate in very different time scales (Miller, 1982; Miller and White, 1984; Maduke et al., 2000; Chen, 2005), a voltage protocol containing multiple short voltage steps was used to construct the voltage-dependent activation curve (which reflects the fast-gate P_o-V curve): a prepulse at 60 mV was given before the test voltage was clamped from 60 to −160 mV for 100 ms in −20-mV voltage steps, followed by a tail voltage pulse at −100 mV. The voltage-dependent fast-gate activation curve was constructed by normalizing the initial tail current after various test voltages to that after the most depolarized test voltage. For the hyperpolarization-activated mutants such as the I225W/V490W double mutant and the mutants constructed in the I225W/V490W background, the voltage protocol included a test voltage pulse from 60 to −160 mV for 40 ms in −10-mV steps followed by a −100-mV tail pulse. The P/4 leak subtraction routine was applied to the recordings of the hyperpolarization-activated mutants because no current was detected from these mutants at the membrane voltage >0 mV.

The kinetics of the two gating mechanisms of CLC-1 (fast gating and common gating) only differs by severalfold. Separating these two gating mechanisms normally relies on two voltage protocols (referred to as protocol A and protocol B; Accardi and Pusch, 2000; Duffield et al., 2003; Bennetts et al., 2005; Tseng et al., 2007). Protocol A was used to assess the overall open probability (P_o) of CLC-1: the product of the open probability of the fast gate (P_o^f) and that of the common gate (P_o^c). In brief, the membrane voltage was clamped at 0 mV and was first stepped to various test voltages from 120 to −120 mV in −20-mV steps for 200 ms before the tail voltage step at −100 mV for 100 ms. The tail current at −100 mV was fitted to a double exponential function, and the P_o was calculated by normalizing the value of the initial tail current to the maximal initial tail current obtained after the most positive test voltage. Protocol B is basically the same as protocol A except that a short (400-µs) but very positive (170-mV) pulse was inserted before the tail voltage step to completely open the fast gate. In this case, the normalized tail current represents P_o^c at the preceding test voltage, and P_o^f can be calculated by dividing P_o by P_o^c.

In experiments where continuous monitoring of the channel current was required, the membrane potential was clamped at 0 mV. For the depolarization-activated channels (such as WT CLC-0, C212S, and CLC-1), a voltage pulse of 60 mV for 100–200 ms followed by a negative voltage of −100 mV for 100–300 ms was given every 2 s. For the hyperpolarization-activated mutants, the 60-mV voltage step of this pulse protocol was omitted. The steady-state current at the end of the negative voltage step was measured to monitor the various effects from Cd²⁺ modulation or from modification by MTS reagents. Solution exchange was achieved using a solution exchanger (SF-77; Warner Instruments). Various concentrations of Cd²⁺ or MTS reagents were applied to the intracellular side of the excised inside-out membrane patches.

Examining the kinetics of Cd²⁺ inhibition on the I225W/V490W mutant is challenging because monitoring the channel current requires negative voltage pulses that relieve the binding of the Cd²⁺. We therefore designed a two-pulse protocol for studying the kinetics of Cd²⁺ inhibition. To evaluate the inhibition rate of intracellular Cd²⁺, a −100-mV voltage pulse for 50 ms was first given to obtain the initial current in the absence of Cd²⁺, followed by exposing the patch to various concentrations of Cd²⁺ at 0, −40, or −80 mV for various periods of time before the same pulse was given again to measure the remaining current (Fig. 12 A). Combining multiple traces from experiments using this two-pulse protocol generates the time course of the Cd²⁺ inhibition (Fig. 12 B). To measure the recovery rate of Cd²⁺ inhibition, a 50-ms voltage pulse to −100 mV was given to obtain the control current, followed by exposing the patch to 300 µM Cd²⁺ for 2 s. Cd²⁺ was then washed out for various amounts of time at 0, −40, or −80 mV before a second voltage pulse was given to assess the current. The current recorded from the second pulse was normalized to the control current before Cd²⁺ inhibition.

All solution exchanges were achieved by using a solution exchanger (SF-77; Warner Instruments). The time for the electrode tip to cross the border of two laminar flows from the solution delivery pipe of the exchanger is only 2 ms (Zhang and Chen, 2009). Including the dead time of triggering the motor that moves the solution delivery pipes, the solution exchange is usually completed within 20–50 ms.

Experimental methods for the FRET experiments have been described previously (Bykova et al., 2006). In brief, an inverted fluorescence microscope (IX-81; Olympus) equipped with a 100-W mercury lamp was used for imaging. The intensity of the excitation light was controlled by a set of neutral density filters installed in a filter wheel, and the duration of light exposure was controlled by a computer-driven mechanical shutter (IX2-SHA; Olympus) using Metamorph software (Molecular Devices). The cell images and the images of their spectra were recorded by a CCD camera (128B; Roper Scientific). For FRET measurements, two filter cubes (Chroma Technology Corp.) were used: Cube I (excitation filter D436/20 and dichroic mirror 455dclp) was used to excite cerulean; Cube II (excitation filter HQ500/20 and dichroic mirror Q515lp) was used to excite eYFP. No emission filter was used in these cubes. Autofluorescence from untransfected cells was negligible in all experiments.

Fluorescence imaging and analysis were performed using the Metamorph software (Molecular Devices). The experiments were achieved by user-designed journals that initiated automatic collection of a bright field cell image, the fluorescence cell images (taken with 50 ms of light exposure), and finally the spectral images (800 ms of light exposure). Spectral images were generated from the fluorescence signal on the cell by passing the imaging through a slit of a spectrograph (Acton SpectraPro 2150i; Princeton Instruments; Fig. 3 A); the same slit position was applied to both the spectrum from cerulean excitation and the spectrum from eYFP excitation. Fluorescence signals from the truncated WT CLC-0 and the truncated I225W/V490W mutant can be clearly identified at the membrane (Fig. 3 A). Therefore, the edge of the spectra corresponding to the membrane fluorescent signal was used for FRET analysis. Background light spectra, collected from the blank region of the same image, were subtracted from the cerulean and the eYFP spectral signals.

FRET involves a nonradiative transfer of light-induced excitation from one fluorophore (donor) to another (acceptor) when the donor and acceptor are located within a distance of <100 Å (Miyawaki and Tsien, 2000; Erickson et al., 2001; Zheng et al., 2002, 2003). The efficiency of the energy transfer (E) depends on the distance, R, between the two fluorophores following the Forster equation:

where R₀ is a spectroscopically determined distance when E = 0.5 (or 50% of FRET efficiency). Our FRET experiments involved passing fluorescence signal of the cell through a slit of a spectrograph and obtaining the spectrum of the fluorescence (Fig. 3 A). The spectra obtained by using the cerulean excitation wavelength (wavelength centered at 436 nm) and that obtained by using the eYFP excitation wavelength (wavelength centered at 500 nm) were from cells transfected with constructs containing cerulean and eYFP, or from cells transfected with constructs containing eYFP only. FRET efficiency was calculated using the method and equations demonstrated in Bykova et al. (2006). The FRET analysis is illustrated in Fig. 3 B. In brief, from cells transfected with cDNA constructs containing eYFP only, ratio A₀ is defined as the ratio between the fluorescence excited by light filtered by Cube I (F^Y(Direct)₄₃₆) and the fluorescence excited by the light passing filter Cube II (F^Y₅₀₀). Namely:

Another ratio (Ratio A) was determined from cells cotransfected with constructs containing cerulean and eYFP. In this case, the cerulean fluorescence by direct excitation was obtained from cells expressing cerulean only, and was subtracted from the cerulean–excitation spectrum obtained from cells coexpressing cerulean and eYFP. The resulted spectrum only contains emission from the acceptor (eYFP), which includes two components: the eYFP emission caused by FRET (F^Y(FRET)₄₃₆) and the emission caused by direct excitation of eYFP (F^Y(Direct)₄₃₆). Thus,

We defined FRET ratio (FR) as the ratio of A and A₀:

The apparent FRET efficiency (E_app) is calculated according to:

where ε_C and ε_Y are molar extinction coefficients of cerulean and eYFP, respectively.

The Flag-tagged CLC-0 constructs were transfected into HEK293 cells, and the cells were harvested 2 d after transfection. The harvested cells were lysed in Dulbecco’s phosphate-buffered saline (Life Technologies) containing 25 mM 3-[(3-cholamidopropyl) dimethylammonio]-1-propanesulfonate (CHAPS) and a protease inhibitor cocktail (Roche). The supernatant of the lysate after centrifugation was treated with or without 10 µM copper phenanthroline before being loaded into a 7.5% SDS-PAGE gel for electrophoresis. The proteins were then transferred to nitrocellulose papers and detected by anti-Flag antibody (M2 antibody; Sigma-Aldrich).

Similar to the extracellular effect of Zn²⁺ on CLC-0 (Chen, 1998), intracellular Cd²⁺ can also inhibit the channel, although much higher concentrations, at least tens of micromolars, are required to generate inhibition (Fig. 2, A and B). The P_o-V curve and the kinetics of the fast gating of CLC-0 appear not to be affected by intracellular Cd²⁺ at a concentration range that significantly inhibits the channel (Fig. 2 C). Furthermore, the inhibition of the channel by Cd²⁺ is largely suppressed by the C212S mutation (Fig. 2 D), which prevents the slow gate of CLC-0 from closing (Lin et al., 1999). Collectively, these results suggest that intracellularly applied Cd²⁺ inhibits CLC-0 by modulating the slow gating.

The hypothesis that the intracellular Cd²⁺ inhibition of CLC-0 is mediated by modulating the channel’s slow gating is consistent with the findings that the inhibition and recovery processes are slow (see Fig. 2 D, top). The slow kinetics of the inhibition rendered a precise characterization of the Cd²⁺ effect technically challenging because of relatively short durations of stable excised inside-out patch recordings of WT CLC-0. The problem was circumvented by an accidental finding that the mutations of I225W and V490W accelerated the inhibition kinetics. Mutations of the corresponding residues in CLC-ec1 (I201W/I422W) have been shown previously to disrupt the subunit interaction, resulting in functional monomeric CLC-ec1 proteins (Robertson et al., 2010). However, the I225W/V490W mutant of CLC-0 remains as a dimer when expressed in HEK293 cells, as shown by the FRET experiment in which the FRET efficiency between a FRET donor and a FRET acceptor attached to the I225W/V490W mutant is no less than that observed in the WT construct (Fig. 3, C and D). Nevertheless, the voltage dependence of the opening of the I225W/V490W mutant channel is opposite to that of the WT channel: the mutant channel’s open probability (P_o) increases as the membrane voltage hyperpolarizes (Fig. 4 A). Moreover, Cd²⁺ inhibition is now faster and more potent: even 1 µM Cd²⁺ significantly inhibits the current (Fig. 4 B).

A full characterization of the Cd²⁺ sensitivity of I225W/V490W is shown in Fig. 5 A, in which the currents of the mutant channel in excised membrane patches were recorded in various Cd²⁺ concentrations ([Cd²⁺]). We noticed that Cd²⁺ appears to have two distinct effects. First, the voltage-dependent steady-state activation curve of this mutant shifts to more hyperpolarized voltages in low [Cd²⁺] (<30 µM). Second, higher [Cd²⁺] (such as 1 mM) reduces the maximal level of the channel activation (Fig. 5 B). Because of the dual effects of Cd²⁺, the dose–response curve of the Cd²⁺ inhibition throughout the entire Cd²⁺ concentration range cannot be fitted to a single Langmuir function. We thus focused on the inhibition by low [Cd²⁺] (up to 300 µM), and normalized the inhibition effect to that induced by 1 mM Cd²⁺. The values of IC₅₀ so obtained were 1, 7, and 204 µM at −80, −120, and −160 mV, respectively (Fig. 5 C). Therefore, the inhibition of I225W/V490W by Cd²⁺ is voltage dependent; the apparent affinity is decreased with more hyperpolarized membrane potentials.

To understand the structural basis of the Cd²⁺ inhibition, we searched for the amino acid residues involved in forming the Cd²⁺-binding site. We first tested if cysteine-modifying reagents, such as MTS compounds, could affect Cd²⁺ inhibition because binding of transition metal ions in proteins frequently involves cysteine (Rulíšek and Vondrásek, 1998). Fig. 6 A shows that the Cd²⁺ inhibition of the I225W/V490W current, monitored by a 50-ms voltage pulse from 0 to −100 mV every 2 s, was indeed abolished after the patch was treated with intracellularly applied 2-(trimethylammonium)ethyl MTS (MTSET; Fig. 6 A, left) or 2-sulfonatoethyl MTS (MTSES) (Fig. 6 A, right). Consistently, the presence of intracellular Cd²⁺ reduced the effect of MTSET or MTSES (Fig. 6 B), indicating that at least one cysteine forms part of the binding site, so that Cd²⁺ binding would protect the cysteine from MTS modifications.

Among the 12 endogenous cysteines in CLC-0, we have shown previously that C229, which is located at the dimer interface near the intracellular end of the transmembrane domain, can be modified by intracellular MTS reagents (Zhang et al., 2010). Interestingly, a histidine residue, H231, is located nearby, raising the possibility that it can act with C229 to coordinate Cd²⁺ as a bidentate ligand (Puljung and Zagotta, 2011). We thus tested whether C229 and H231 could both be involved in the formation of the Cd²⁺-binding site. Mutating C229 (C229S), H231 (H231Y), or both (C229S/H231Y) in the background of I225W/V490W indeed suppressed the ability of Cd²⁺ to shift the steady-state activation curve (Fig. 7, A–C, respectively). When the C229S mutation was made in the WT CLC-0, the Cd²⁺ inhibition effect was also significantly reduced (Fig. 7 D). Likewise, the mutation H231Y also reduced the sensitivity of Cd²⁺ inhibition on WT CLC-0 (not depicted). However, the reduction of maximal channel activity by high [Cd²⁺] still existed in these mutants (Fig. 7, A–D), suggesting that it is mediated by a different mechanism. Our results thus demonstrate that C229 and H231 form the high affinity Cd²⁺-binding site, where binding of Cd²⁺ shifts the steady-state activation curve of I225W/V490W. It is likely that C229 and H231 from both CLC-0 subunits together coordinate Cd²⁺ binding, as these residues are located near the dimer interface and therefore are close to each other (see Fig. 1). The conclusion that C229 and H231 form the Cd²⁺-binding site is further supported by experiments from CLC-1. These two residues are not conserved in CLC-1, which is not sensitive to intracellular Cd²⁺ modulation (Figs. 8 A and 9 A). When cysteine and histidine were introduced to the corresponding positions in CLC-1 (S294C/Y296H), a Cd²⁺-binding site was created, and the activation of CLC-1 could be modulated by intracellular Cd²⁺ (Figs. 8 B and 9, A and B). Like the MTS modification of the endogenous cysteine (C229) in CLC-0, modification of the introduced cysteine (S294C) abolished Cd²⁺ inhibition on CLC-1 containing the engineered Cd²⁺-binding site (Fig. 9 C).

Because binding sites for transition metal cations can be formed by four cysteine residues (Yellen et al., 1994), we tested the effect of Cd²⁺ on CLC-0 with the H231C mutation. Such a CLC-0 construct contains cysteine at all four positions of 229 and 231 from two subunits (and thus will be called “4C” mutant). In the WT background, the 4C mutant becomes very sensitive to Cd²⁺, and the Cd²⁺ inhibition is nearly irreversible (Fig. 10 A, top). Cd²⁺ inhibition of the 4C mutant with I225W/V490W background mutations, however, is very different; the inhibition can be quickly washed out, but with the time of patch excision the channel is less and less sensitive to Cd²⁺ (Fig. 10 A, bottom). We suspected that this time-dependent reduction of Cd²⁺ sensitivity might result from the formation of a disulfide bond among these four cysteine residues after patch excision, which could change the oxidation/reduction condition of the channel (Zhang et al., 2008). Indeed, Western blots shown in Fig. 10 B demonstrate that the monomeric band of the 4C mutant in the I225W/V490W background disappeared after oxidation (Fig. 10 B, right). On the other hand, the majority of the 4C mutant in the WT CLC-0 background still remained as monomers in the SDS gel after copper phenanthroline treatment (Fig. 10 B, left). Similar results were observed in a “2C” mutant in which cysteine was present only at position 231 but not at position 229 (Fig. 10 B), suggesting that it is the two cysteines at position 231 that form a disulfide bond. Thus, through the differences in Cd²⁺ inhibition and disulfide bond formation, we demonstrate that like the WT Cd²⁺-binding site, the structure of the mutant Cd²⁺-binding site with four cysteines is changed by the I225W/V490W mutations.

Because C229 and H231 are located near the intracellular end of the transmembrane domain and interactions between the transmembrane domain and the cytoplasmic domain have been demonstrated in CLC molecules (Feng et al., 2010; Bennetts and Parker, 2013), the cytoplasmic domain of CLC-0 may influence Cd²⁺ inhibition. We thus tested the intracellular Cd²⁺ effect on a CLC-0 mutant in which the C-terminal region of the channel after residue 530 is replaced with a fluorescent protein. As reported previously (Bykova et al., 2006), this CLC-0 mutant (abbreviated as “CLC0-TM”) generates hyperpolarization-activated currents (Fig. 11 A). Interestingly, Cd²⁺ did not inhibit but potentiated this mutant (Fig. 11 B), suggesting that the cytoplasmic domain of the CLC channel may also be involved in the intracellular Cd²⁺ modulation, likely through its interaction with the transmembrane domain.

To gain further insights into the Cd²⁺ modulation mechanism, we studied the kinetics of Cd²⁺ inhibition on I225W/V490W of CLC-0, taking advantage of the faster kinetics of Cd²⁺ inhibition in this mutant. Fig. 12 A illustrates a two-pulse protocol used to examine the rate of Cd²⁺ inhibition by measuring the current of the I225W/V490W mutant elicited by a −100-mV voltage pulse before (first pulse) and after the patch was exposed to Cd²⁺ for a certain amount of time (second pulse). Combining multiple recording traces from the same patch generates the time course of Cd²⁺ inhibition (Fig. 12 B). It is worth noticing that there appears to be a delay of the current inhibition because the current recorded after applying Cd²⁺ for 0.2 s is almost the same as that before the Cd²⁺ application. This delay has little voltage dependence (Fig. 12 C), but as [Cd²⁺] is increased, the delay is shortened (Fig. 12 D). To analyze the kinetics of Cd²⁺ inhibition like that shown in Fig. 12 B, we normalized the current induced by the second voltage pulse to the current before Cd²⁺ application, and plotted the time course of the current inhibition by Cd²⁺ (Fig. 13 A). Fitting the Cd²⁺ inhibition time course with an exponential function gives the inhibition time constant (τ_on), the inverse of which (1/τ_on) is defined as the apparent Cd²⁺ inhibition rate (or apparent on rate), k_on. The observation that k_on is a saturable function of [Cd²⁺] (Fig. 13 B) suggests that Cd²⁺ inhibition is not a simple bimolecular reaction involving only Cd²⁺ binding. These results suggest that the binding of Cd²⁺ to the open-state channel may not immediately close the channel, and the delay in current inhibition upon Cd²⁺ application could result from the conformational change between the open and the closed state of this dimer interface mutant.

We have also measured the rate of current recovery after Cd²⁺ washout (apparent off-rate, or k_off) using a similar two-pulse protocol. The current recovery process was well described by a single-exponential function (recording traces not depicted). These kinetic measurements of Cd²⁺ inhibition and recovery were conducted at multiple holding voltages (Fig. 13, C and D). The values of the apparent τ_on (in 30 µM Cd²⁺) and τ_off at 0, −40, and −80 mV are compared in Fig. 13 E. Although τ_on is insensitive to the change of membrane potential, the value of τ_off decreases as membrane potential becomes more negative. Therefore, the lower apparent affinity of Cd²⁺ inhibition at more negative voltages on the I225W/V490W mutant results mostly from a faster off-rate by membrane hyperpolarization. Collectively, these results suggest a state-dependent binding of Cd²⁺ to the channel; Cd²⁺ binding to the closed state of the channel is energetically more favored than the binding to the open state.

Transition metal ions have been widely used to study the structure–function relationship of ion channels. In the Shaker K⁺ channel, for example, the state-dependent affinity change of extracellular Cd²⁺ inhibition revealed a structural change on the external mouth of the channel during the slow (C-type) inactivation (Yellen et al., 1994). Intracellularly applied Cd²⁺ was also used to show that the activation gate of the Shaker K⁺ channel can be trapped in the open state when an intersubunit metal bridge is formed (Holmgren et al., 1998). CLC-0 and CLC-1 have been found to be inhibited by Zn²⁺ or Cd²⁺ applied from the extracellular side (Rychkov et al., 1997; Chen, 1998), and extracellular Zn²⁺ is known to inhibit WT CLC-0 by facilitating the closure of the slow gate (Chen, 1998). We found in the present study that the inhibition of WT CLC-0 by intracellular Cd²⁺ is likely mediated by a modulation of the slow gating of CLC-0 as well because intracellular Cd²⁺ has little effect on the fast gating of the channel (Fig. 2 C), and because the inhibition is largely suppressed by the C212S mutation (Fig. 2 D), which prevents the slow gate of CLC-0 from closing (Lin et al., 1999).

The apparent affinity of intracellular Cd²⁺ on WT CLC-0 is quite low—tens to hundreds of micromolars of [Cd²⁺] is required for an observable effect. However, the potency of Cd²⁺ is significantly enhanced by tryptophan substitutions of I225 and V490, which correspond to dimer interface residues I201 and I422, respectively, in CLC-ec1. The I201W/I422W mutation in CLC-ec1 disrupts dimer architecture, leading to the formation of CLC monomers (Robertson et al., 2010). The corresponding mutations in CLC-0, however, do not separate the two channel subunits as indicated by a high FRET efficiency between the FRET donor and acceptor tagged to the C termini of the transmembrane domains (Fig. 3). The formation of a disulfide bond between the two cysteine residues at position 231 further demonstrates a dimeric CLC molecule in the I290W/V490W background mutations (Fig. 10). Despite the fact that CLC-0 with I225W/V490W mutations remains as a dimer, the mutant channel’s voltage-dependent gating properties have been inverted from that of the WT channel (Fig. 4 A). It is conceivable that introducing bulky tryptophan side chains into the dimer interface alters the interaction between the two CLC subunits, thus increasing apparent affinity and kinetics of Cd²⁺ inhibition (Fig. 4 B), and facilitating disulfide bond formation between cysteines at position 231 (Fig. 10).

A closer examination of the Cd²⁺ inhibition on the I225W/V490W mutant reveals that intracellular Cd²⁺ appears to have two effects. At low micromolar [Cd²⁺], the mutant channel’s steady-state activation curve is shifted to more hyperpolarized voltages by Cd²⁺. We observed a competition between intracellular Cd²⁺ inhibition and MTS modification (Fig. 6), a clear indication that like many other transition metal-binding proteins, the amino acid cysteine is directly involved in forming the Cd²⁺-binding site. Further mutagenesis experiments showed that the mutation of C229S, H231Y, or both abolishes the Cd²⁺-induced shift of the steady-state activation curve in WT CLC-0 and in the I225W/V490W mutant (Fig. 7), indicating that the high affinity Cd²⁺ inhibition is mediated through the binding site formed by C229 and H231. On the other hand, the maximal open probability of these mutants was still inhibited by Cd²⁺ at the millimolar concentrations, suggesting that the low affinity Cd²⁺ inhibition may be mediated by a different mechanism.

The residues C229 and H231 of CLC-0 correspond to R205 and Q207, respectively, in CLC-ec1, which are in close proximity to the same residues in the partner subunit (Fig. 1). As seen in many metalloproteins, tetrahedral is a preferred coordination arrangement for Cd²⁺ binding, usually involving four cysteine/histidine residues (Rulíšek and Vondrásek, 1998). Therefore, it is likely that the Cd²⁺ site in CLC-0 is formed by two subunits, each contributing two chelators from C229 and H231 to bind Cd²⁺. These cysteine and histidine residues, however, are not conserved in CLC-1, thus explaining the fact that WT CLC-1 is insensitive to intracellular Cd²⁺, as shown in Figs. 8 A and 9 A, and also reported previously by others (Rychkov et al., 1997; Fahlke et al., 1998). In this study, we are able to construct a Cd²⁺-binding site in CLC-1 by engineering S294C/Y296H mutations to mimic the Cd²⁺ site in CLC-0. This strategy yields a CLC-1 mutant whose activity is modulated by low micromolar intracellular Cd²⁺, although the consequence of the Cd²⁺ modulation is potentiation but not inhibition of the channel current! Such Cd²⁺ effect, again, is abolished by pretreating the channel with intracellular MTS reagents (Fig. 9 C), indicating that a Cd²⁺-binding site is indeed created. Previous studies have indicated an interaction of the transmembrane domain and the C-terminal cytoplasmic domain of CLC proteins (Feng et al., 2010; Bennetts and Parker, 2013). Experiments in Fig. 11 show that removing the C-terminal region of CLC-0 changes the Cd²⁺ effect from inhibition to potentiation. Thus, the interaction of the transmembrane domain with the C-terminal cytoplasmic domain as well as the subunit interaction at the dimer interface are both involved in intracellular Cd²⁺ modulation of CLC-0.

These experimental results thus allow us to compare the functional roles of subunit interactions between CLC channels and bacterial CLC proteins. It was demonstrated previously that cross-linking the two CLC-ec1 subunits covalently poses little impact on the Cl⁻ transport function (Nguitragool and Miller, 2007). Similarly, the fast gating of CLC-0 and CLC-1 is not affected by restricting molecular motions at the dimer interface with Cd²⁺ (Figs. 2 and 9). This is consistent with the accepted concept that the Cl⁻/H⁺ exchange cycle of CLC-ec1 and the fast gating of CLC channels are homologous, mediated by molecular motions (for example, swinging of the glutamate side chain) confined in individual subunits. Coordinating opening and closing of the two Cl⁻ pathways requires the slow/common gating mechanism that enables the cross talk of the two subunits through the dimer interface. Cd²⁺ binding or the introduction of bulky residues at the dimer interface would perturb such a cross talk and thus modulate slow/common gating of CLC channels. It is possible that mutation of I225W and V490W speeds up the inactivation of CLC-0 (closure of the slow gate), explaining the faster kinetics of Cd²⁺ inhibition in the I225W/V490W double mutant than in the WT channel. Interestingly, Cd²⁺ binding inhibits the current of WT CLC-0 and I225W/V490W but potentiates the current of CLC0-TM (Fig. 11) and the CLC-1 mutant S294C/Y296H (Figs. 8 and 9). The different consequences of Cd²⁺ binding (potentiation vs. inhibition of the current) may result from the difference between the cross talk of the two subunits in these various channels. Another possibility is that the allosteric interactions between the Cd²⁺-binding site and the C-terminal cytoplasmic domain are different in these channels. It will await further studies to understand the different functional consequences of Cd²⁺ modulations in CLC-0, CLC-1, and the C terminus–truncated CLC channel.

If C229 and H231 of CLC-0 are located at the intracellular end of the transmembrane domain, like their corresponding residues R205 and Q207 in CLC-ec1 (Fig. 1), it is expected that the Cd²⁺ site would be outside the transmembrane electric field, and thus Cd²⁺ binding should be voltage independent. Indeed, the apparent on-rate of Cd²⁺ is voltage independent (Fig. 13 E). However, the apparent Cd²⁺ affinity is lower at more hyperpolarized voltages (Fig. 5 C) because membrane hyperpolarization (which favors the opening of the I225W/V490W mutant) accelerates Cd²⁺ dissociation (Fig. 13, D and E). The voltage dependence of the apparent off-rate thus suggests a state-dependent Cd²⁺ inhibition; namely, Cd²⁺ binding to the open state is energetically less favorable than the binding to the closed state. Furthermore, Cd²⁺ inhibition shows an appreciable delay (Fig. 12) and the apparent on-rate of Cd²⁺ inhibition is a saturable function of [Cd²⁺] (Fig. 13 B), both corroborating the idea that Cd²⁺ binding to the open-state channel cannot immediately shut the pore. Instead, it will require a subsequent molecular event, likely a conformational change associated with slow gating, to close the channel.

In summary, we have characterized the inhibition of CLC-0’s slow gating by intracellular Cd²⁺, and have identified a Cd²⁺-binding site formed by C229 and H231 at the intracellular end of the dimer interface of the transmembrane domain. We have also shown that the residues I225 and V490 of CLC-0, like the corresponding residues in other CLC channels, are important for the channel gating mechanism likely because of their roles in the subunit interaction. Binding of Cd²⁺ to the dimer interface and mutations of dimer interface residues thus may modulate the slow/common gating of CLC channels by altering subunit interactions.

We would like to thank Drs. Robert Fairclough and Tzyh-Chang Hwang for critical readings of the manuscript.

This work was supported by a grant from National Institutes of Health (R01GM065447) and by the Jong L. Chen Family Neuroscience Research Fund.

The authors declare no competing financial interests.

Merritt C. Maduke served as editor.