CLC Cl− channels are homodimers in which each subunit has a proper pore and a (fast) gate. An additional slow gate acts on both pores. A conserved glutamate (E166 in CLC-0) is a major determinant of gating in CLC-0 and is crucially involved in Cl−/H+ antiport of CLC-ec1, a CLC of known structure. We constructed tandem dimers with one wild-type (WT) and one mutant subunit (E166A or E166D) to show that these mutations of E166 specifically alter the fast gate of the pore to which they belong without effect on the fast gate of the neighboring pore. In addition both mutations activate the common slow gate. E166A pores have a large, voltage-independent open probability of the fast gate (popen), whereas popen of E166D pores is dramatically reduced. Similar to WT, popen of E166D was increased by lowering pHint. At negative voltages, E166D presents a persistent inward current that is blocked by p-chlorophenoxy-acetic acid (CPA) and increased at low pHext. The pHext dependence of the persistent current is analogous to a similar steady inward current in WT CLC-0. Surprisingly, however, the underlying unitary conductance of the persistent current in E166D is about an order of magnitude smaller than that of the transient deactivating inward Cl− current. Collectively, our data support the possibility that the mutated CLC-0 channel E166D can assume two distinct open states. Voltage-independent protonation of D166 from the outside favors a low conductance state, whereas protonation from the inside favors the high conductance state.
CLC proteins are a multigene family present in all phyla (Jentsch et al., 2005). The founding member, called CLC-0, was identified as a voltage-dependent Cl− channel in the electric organ of Torpedo californica (White and Miller, 1979; Miller and White, 1980). After the cloning of CLC-0 (Jentsch et al., 1990), homologues were found in many prokaryotes and all eukaryotes. The physiological importance of CLC proteins is underscored by their involvement in several human genetic diseases and the surprising phenotypes of knockout models (Jentsch et al., 2005; Uchida and Sasaki, 2005). Functions of CLC proteins range from the regulation of the skeletal muscle membrane potential to transepithelial transport and pH homeostasis of intracellular organelles (Jentsch et al., 2005; Pusch and Jentsch, 2005). From the recent X-ray structure of bacterial CLCs it is now clear that CLC proteins contain two identical subunits in which each subunit presents an independent pathway for Cl− ions (Dutzler et al., 2002, 2003). Such a “double-barreled” architecture had been conjectured by Miller and colleagues early on from the single channel behavior of the Torpedo channel (Miller and White, 1984) and it was also predicted from later mutagenesis work and a lower resolution structure (Ludewig et al., 1996; Middleton et al., 1996; Mindell et al., 2001). The homodimeric structure enables two mechanisms of gating: a fast gate acting on each pore independently and a slow gate acting on both pores simultaneously (Miller and White, 1984). The fast gate of each pore of CLC-0 opens and closes in the time range of milliseconds, whereas the slow gate opens and closes both pores simultaneously in the time range of seconds to minutes (Pusch et al., 1997). While the mechanism of the slow gate is still obscure, the X-ray structure of CLC-ec1 allowed important insights into the mechanism of the fast gate of CLC-0. Each subunit of CLC-ec1 contains 18 α helices that wrap around a central part, the putative pore in which three ion binding sites are present (Sint, Scent, and Sext) (Dutzler et al., 2002, 2003). In the bacterial protein CLC-ec1, Sext is occupied by the side chain of the glutamate carboxylate group of residue E148, which is probably negatively charged (Dutzler et al., 2002). This side chain appeared to block the movement of Cl− ions toward the extracellular side, and the structure of the WT protein was therefore assumed to represent a closed pore (Dutzler et al., 2002). In accordance with this hypothesis, in the structure of CLC-ec1 in which the glutamate was mutated to an alanine (E148A), a chloride ion was resolved where the E148 side chain is localized in the WT, whereas the remaining structure was almost unaltered (Dutzler et al., 2003). The interpretation of this mutated structure as an open state was supported by the functional results performed in parallel on CLC-0, which showed that the equivalent mutant in CLC-0 (E166A) had an “open,” voltage-independent phenotype (Dutzler et al., 2003). This straightforward correlation of the structure of the bacterial CLC-ec1 and the function of the vertebrate CLC-0 is also in agreement with other studies that suggest a structural conservation of CLC proteins (Estévez et al., 2003; Lin and Chen, 2003). However, other studies suggested that the fast gating of CLC-0 involves an additional rearrangement of the intracellular part that is not revealed by the crystal structure of E148A (Accardi and Pusch, 2003; Traverso et al., 2003). In fact, the functional equivalence of CLC-0 and CLC-ec1 was challenged when Accardi and Miller demonstrated that CLC-ec1 is actually not a Cl− ion channel but an electrogenic Cl−/H+ antiporter with an apparent stoichiometry of 2 Cl−:1 H+ (Accardi and Miller, 2004). Moreover it has been recently shown that also mammalian proteins of the same family, CLC-3, CLC-4, and CLC-5, exhibit Cl−/H+ antiporter activity (Picollo and Pusch, 2005; Scheel et al., 2005). In contrast, CLC-0 is clearly an ion channel with well-defined, relatively large single channel events (Hanke and Miller, 1983; Bauer et al., 1991) and Nernstian dependence of the reversal potential on the chloride concentration.
The molecular mechanism of transport of CLC-ec1 and CLC-3–5 is not understood and it is currently unclear how the gating and permeation properties of CLC-0 are correlated to it. It seems, however, likely that the proton dependence of the fast gate of CLC-0 is somehow related to the Cl−/H+ antiport of CLC-ec1. For example, extracellular acidification opens CLC-0 (Chen and Chen, 2001; Dutzler et al., 2003), probably by protonating E166, mimicking the effect of mutating E166 to a neutral amino acid (Dutzler et al., 2003). The analogous mutations of CLC-ec1 (E148A) and CLC-5 (E211A) abolish H+ transport (Accardi and Miller, 2004; Picollo and Pusch, 2005; Scheel et al., 2005). Also intracellular acidification strongly activates CLC-0 (Hanke and Miller, 1983) and similarly CLC-1 (Rychkov et al., 1996) by apparently shifting the activation curve to negative voltages.
To better understand the role of protons in the mechanism of protopore gating of CLC-0 we mutated the critical glutamate 166 to the similar amino acid aspartate. Despite the small difference between these two amino acids (one extra CH2 group in the side chain of glutamate) the kinetics of the WT and of the mutant are very different (Traverso et al., 2003). In the present work, we show that indeed both mutants, E166A and E166D, have a drastically altered protopore gate. However, both mutations, despite their disparate effect on the fast gate, appear to eliminate slow gating processes, locking the slow gate open. We then varied pHint and pHext and compared the effects on E166D with those described for WT CLC-0. In contrast with what would be expected from the behavior of WT CLC-0, extracellular pH had no effect on E166D outward currents but affected only a persistent inward current present at negative voltages, a current that appears to represent a different conductance state of the channel. Decreasing intracellular pH drastically increased the popen of the channel in a manner that suggests that the protonation from the inside represents one of the major voltage-dependent steps in the opening of the fast gate.
Materials And Methods
Molecular Biology and Heterologous Expression
Mutations were introduced by recombinant PCR as previously described (Accardi et al., 2001). Tandem dimers (WT-ED, ED-WT, WT-EA, and EA-WT) were generated as previously described (Ludewig et al., 1996). In brief, the stop codon of the NH2-terminal subunit was replaced with an SpeI and a KpnI restriction site. The same sites were introduced before the start codon of the COOH-terminal subunit. The Spe I site was used to link both subunits. The linker sequence consisted of four amino acids (G-T-T-S). All constructs were in the PTLN vector (Lorenz et al., 1996). cRNA was transcribed and injected in Xenopus oocytes as previously described (Accardi and Pusch, 2000).
Currents were recorded using the two-electrode voltage-clamp method and excised patch-clamp recording as previously described (Pusch et al., 2000). For whole oocyte voltage clamp measurements, the bath solution contained (in mM) 100 NaCl (or 100 NaI), 4 MgSO4, 5 HEPES, pH 7.3, and the holding potential was chosen close to the resting membrane potential (−30 to −50 mV). For patch clamping, the intracellular solution contained (in mM): 100 NMDG-Cl, 2 MgCl2, 10 HEPES, 2 EGTA, pH 7.2, whereas the standard extracellular solution contained 100 NMDG-Cl, 5 MgCl2, 10 HEPES, pH 7.2. In both solutions HEPES was substituted by MES for solutions with 5 < pH < 6.5, bis-Tris-propane for solutions with pH > 7.8, and glutamate for solutions with pH < 5. Experiments shown in the figures were performed using regular pH of 7.2 for the extracellular solution (for patch-clamp and voltage-clamp recordings) and the intracellular solution (for patch-clamp recordings) unless otherwise noted.
All substances were purchased from Sigma-Aldrich. Solutions in patch-clamp experiments were changed by inserting the patch pipette into perfusion tubes of ∼0.5 mm diameter. Patch-clamp data were recorded using an EPC-7 amplifier (HEKA) and the Pulse acquisition program (HEKA) or a custom acquisition program (GePulse) (the acquisition program is available at http://www.ge.cnr.it/ICB/conti_moran_pusch/programs-pusch/software-mik.htm).
The holding potential in patch clamp measurements was 0 mV. Voltage clamp measurements employed an NPI-TEC 05 amplifier (NPI Electronics), and were performed as described previously (Traverso et al., 2003).
For all (macroscopic) patch-clamp experiments, the capacitive transients were subtracted offline in the following manner. All pulse protocols contained a final pulse to 0 mV with a duration equal or superior to the longest segment of the voltage-clamp pulse protocol. Since at 0 mV (very close to the reversal potential) no ionic current is expected to flow, the current response reflects only the capacitive transient. This response, after appropriate scaling, was used for the subtraction of the capacitive transients at the other test potentials.
Single Channel Analysis
Single channel recordings were generally filtered at 1 kHz or 500 Hz. For the construction of amplitude histograms a bin width of 5 fA was used. Amplitude histograms were fitted with the sum of Gaussian functions, Gi, one for each peak, i,
where each Gaussian function is characterized by a mean μi, a width σi, and amplitude ai. The relative area, Ai, that is occupied by each Gaussian component was calculated by
and was taken as a measure of the probability to dwell in the conductance state associated with mean μi.
Nonstationary Noise Analysis
Nonstationary noise analysis was performed essentially as described earlier (Pusch et al., 1994). In brief, a test pulse was repeatedly applied, and the variance was calculated from the squared difference of consecutive current responses. For the variance-mean analysis, the current range was subdivided into a certain number of bins for a clearer graphical representation. Background variance was measured at the baseline (at 0 mV) and subtracted. The variance-mean plots were fitted with Eq. 1 (see below) with two free parameters, the single channel current i, and the number of channels, N. All primary data analysis was performed with the custom analysis program downloadable at http://www.ge.cnr.it/ICB/conti_moran_pusch/programs-pusch/software-mik.htm. Figures were prepared with SigmaPlot (SPSS Inc.).
We measured the extracellular pH close to the oocyte surface with a pH-sensitive microelectrode as described by Picollo and Pusch (2005). In brief, silanized (with dichlorodimethylsilane; Sigma-Aldrich) microelectrodes were filled with proton ionophore B (Fluka) and then with a solution containing (in mM) 150 NaCl, 23 NaOH, 40 KH2PO4, pH 6.8. Pipettes were connected to a custom built high impedance (>1015 Ohm) amplifier and responded with a slope of 57–63 mV per pH unit. The pH-sensitive microelectrodes were gently pushed against the vitelline membrane without penetrating the oocyte. The pH-related signal was recorded as a second input channel in parallel to the membrane current. The extracellular solution for pH measurements contained 100 mM NaCl, 4 mM MgCl2, 0.1 mM HEPES, pH 6.4.
Noise Analysis of Mutant E166D Reveals a Slightly Reduced Conductance
We analyzed mutations of the glutamate E166 of CLC-0 to aspartate (ED) and alanine (EA). Mutating E166 in CLC-0 to alanine leads to an almost constitutively open channel (Dutzler et al., 2003; Traverso et al., 2003), whereas E166D slows down opening at positive potentials and increases the rate of closure at negative potentials (Traverso et al., 2003). In addition, mutant E166D consistently expressed much less current than WT CLC-0 or mutant E166A. In fact, in order to keep currents at a manageable magnitude that could be measured with the two electrode voltage clamp, <1 ng of cRNA was injected for CLC-0, and currents were measured after 1–2 d. In contrast, ∼20 ng had to be injected for mutant E166D, and oocytes had to be incubated for at least 3 d to achieve a sizable current (unpublished data). To find out whether a reduced single channel conductance of E166D was responsible for the smaller currents we performed nonstationary noise analysis (Fig. 1). This method provides a good order-of-magnitude estimate of the single channel current and may allow the determination of the absolute open probability (Heinemann and Conti, 1992). Fig. 1 A shows the mean of 80 traces obtained each by stepping the voltage from 0 to 120 mV in an inside-out patch. Fig. 1 B shows the corresponding variance trace, whereas in Fig. 1 C the variance is plotted versus the mean (symbols) together with a fit (line) of the equation
where σ2 is the variance, I the macroscopic current, i the single channel current, and N the number of channels. The average single-channel current obtained from noise analysis at 120 mV was 0.55 ± 0.14 pA (mean ± SD, n = 6). Assuming a linear single-channel i-V, this corresponds to a conductance of 4.6 pS. This value is only ∼35% smaller than the conductance of WT CLC-0 (∼7 pS) (Accardi and Pusch, 2003) and this reduction can thus not explain the small macroscopic currents of mutant E166D. From the nonstationary noise analysis it was not possible to obtain a reliable estimate for the number of channels, and consequently of the absolute open probability, because the variance-mean plots showed little curvature. This indicates that the maximal open probability is significantly smaller than 0.5, as is indeed the case (see below).
Tandem Mutant-WT and WT Mutant Dimers Differentiate Effects on Fast and Slow Gate
The next question we asked was: In what specific manner do mutations E166A and E166D affect the fast (protopore) gate and the slow (common) gate? The slow gate of CLC-0 is voltage dependent with activation favored at negative voltages (Miller and Richard, 1990; Pusch et al., 1997) (Fig. 2 A). No hyperpolarization activation could be detected for either E166A (Fig. 2 B) or E166D (Fig. 2 C). A further hint that the slow gate is locked open in mutant E166D was obtained from the study of the double mutant E166D/C212S that includes the mutation C212S, that locks open the slow gate of CLC-0 (Lin et al., 1999). The double mutant E166D/C212S had properties indistinguishable from the single mutant E166D (unpublished data). These results indicate that the slow gate is either locked open or rendered voltage independent by both mutations E166A and E166D.
We next made four tandem dimeric constructs consisting of one WT and one mutant subunit: WT-ED, WT-EA, ED-WT, and EA-WT. For example, the notation WT-ED indicates that the first subunit is WT and the second bears the E166D mutation. No difference could be observed between heterodimers with opposite order (i.e., ED-WT versus WT-ED and EA-WT versus WT-EA) (unpublished data). All dimeric constructs, including a control WT-WT dimer, expressed similar levels of current (conductance ranged between 50 and 200 μS after 1 d of incubation), suggesting that the E166D mutation has no (dominant) effect on protein processing or stability. However, we cannot completely rule out that mutation E166D alters protein trafficking.
Interestingly, all dimeric constructs partially recovered a voltage dependence of the slow gate (Fig. 2, D and E). The fact that the slow gate was only partially recovered demonstrates that for all dimeric constructs, both subunits contribute to channel gating. Macroscopic currents of EA-WT dimers on a faster time scale from inside-out patches are shown in Fig. 3 A. The voltage dependence of the fast gate, plotted in Fig. 3 B, differs from that of WT CLC-0 mainly by an increased offset at negative voltages, reflecting probably the voltage-independent contribution of the E166A pores. This interpretation was further confirmed by single channel analysis.
Fig. 4 A shows a single channel trace of the dimer EA-WT at −100 mV. For most of the time the channel behaves as the superposition of a WT pore and a pore, of similar conductance, that is almost always open. Occasionally, currents reach the baseline (indicated by arrows in Fig. 4 A). These events probably represent longer closures of the “EA pore.” A similar behavior has been also previously observed in homomeric E166A channels (Traverso et al., 2003). These closures are not closures of the slow gate because the neighboring “WT pore” continues to open and close. Whether these slow transitions of the EA pore reflect “residual” fast gate transitions remains to be studied. An amplitude histogram over a stretch without such long closures (see bracket with * in Fig. 4 A) yielded a popen of 0.64 of the fluctuating pore (Fig. 4 C), relatively close to the WT popen at −100 mV (popen[WT] ∼ 0.75; Accardi and Pusch, 2003). This strongly suggests that indeed the fast fluctuating conductance level represents the WT pore whereas the long open times correspond to the EA pore, demonstrating that the mutation strongly affects the fast protopore gate, without affecting the fast gate of the neighboring WT pore.
The fast gate properties of the ED-WT resemble those of homodimeric WT CLC-0 as illustrated in Fig. 3 (C and D). The same holds true for WT-ED dimers (unpublished data). In particular, at positive voltage, no slowly activating component similar to that seen in Fig. 1 is visible in patch or voltage-clamp recordings from dimers containing E166D. Thus, there seems to be no contribution of the ED subunit to the “heterodimeric” ED/WT currents, even though we know from the slow-gate recovery, that the ED subunit is present in the dimer. This might be caused by a very small open probability of the fast gate of mutant E166D as suggested by the noise analysis described above. To directly test this hypothesis we measured single ED-WT channels. The advantage of using the dimer instead of measuring directly single E166D channels is that the presence of only one channel in the patch can be assured by the presence of a single WT-like pore. Indeed, single channel events of WT-ED dimers almost behaved as if they consisted of an isolated single WT pore. This is illustrated in Fig. 5 A, which shows a recording of a patch subjected to voltage pulses to first −100 mV and then +80 mV. At −100 mV, a single pore switches between open and closed states, with an apparent open probability (p ∼ 0.77) close to that of the fast gate of CLC-0 at this voltage (popen[WT] ∼ 0.75; Accardi and Pusch, 2003). At +80 mV, the pore is almost permanently open, which is consistent with the high open probability of the fast gate of CLC-0 at this voltage. However, closer inspection of several traces recorded at 80 mV (Fig. 5 B) reveals rare short openings to a higher conductance level (see arrow in Fig. 5 B). Also many brief interruptions to the baseline can be seen, representing probably brief closures of the WT fast gate. In fact, the expected mean closed time of the WT fast gate is given by 1/α (where α is the opening rate) being ∼0.3 ms at 80 mV, which we calculated using the known parameters for WT (Chen and Miller, 1996, Eq. 9 and Table II) together with the external chloride concentration used in our experiments. To quantify the contribution of the openings to the second open conductance level (arrow in Fig. 5 B) to the overall current, we constructed an amplitude histogram on a logarithmic scale with I > 0.7 pA, excluding thus the brief closure events. Fitting the histogram with the sum of two Gaussian functions (Fig. 5 C) results in an estimate of a contribution of ∼0.5% of the small peak at ∼1.3 pA to the overall current. Given that the open probability of the fast gate of WT is practically one at 80 mV, this contribution directly reflects the open probability of the fast gate of E166D if the openings are interpreted as arising from an E166D pore. The conductance of the putative ED pore estimated from the histogram analysis at 80 mV (∼4.7 pS) is very similar to that obtained from the nonstationary noise analysis at 120 mV (∼4.6 pS), consistent with this interpretation of the single-channel results. Thus, the almost complete lack of an obvious contribution of E166D pores to macroscopic and single channel currents in WT/ED heteromers probably arises from the very small open probability of E166D pores even at 80 mV.
Collectively, results from macroscopic and single channel current measurements show that both E166 mutations (E166A and E166D) strongly affect the fast gate, without influence on the fast gate of the neighbor subunit, and that the conservative mutation E166D leads to a drastic reduction of the open probability of the fast gate.
Dependence of E166D on Extracellular pH
Having established that E166D strongly alters the fast gate, our next purpose was to analyze the effects of different extracellular and intracellular pH on this mutant and compare it to the published effects on WT CLC-0. In two-electrode voltage clamp measurements we varied pHext from 5.8 to 8.3. WT CLC-0 is strongly activated by a reduction of pHext (Chen and Chen, 2001; Dutzler et al., 2003). Given that the open probability of E166D is very small (see above and below), we expected that its popen would be strongly augmented at low extracellular pH. Surprisingly, almost no change of E166D currents were observed at positive voltages, and only the inward currents at negative voltages were significantly enhanced (Fig. 6, A and B). The latter phenomenon was studied in further detail using outside-out patches. Changing pHext from 7.2 to 5.8 in outside-out patches outward currents were almost unchanged, whereas we observed a significant increase of the steady-state inward current present at −140 mV (Fig. 6, C and D). This current is not a leak current as it is blocked by p-chlorophenoxy-acetic acid (CPA) (see below, Fig. 9 A). A more detailed investigation of its properties is presented below.
Dependence of E166D on Intracellular pH
In contrast to pHext, pHint had a drastic effect on the currents carried by mutant E166D. In Fig. 7 A, currents measured from the same inside-out patch are shown at different pHint (pH 7.2, pH 6.8, pH 6.3, and pH 5.8). These currents were elicited by pulses up to 220 mV and channel activation was monitored by a tail pulse to −140 mV. Currents are clearly activated by low pHint. This can be seen directly from the current measured at the variable test voltages. But also the initial current at the constant tail pulse to −140 mV is clearly increased at low pH up to 5.3 (Fig. 7 A, insets). A qualitatively different behavior was seen when the pH was further lowered down to 4.3. Fig. 7 B shows current traces from a different inside-out patch at pHint 7.2, 5.8, 4.8, and 4.3. Surprisingly, currents at pHint 4.8 were smaller than currents at pHint 5.8 but increased again at pHint 4.3. Most importantly, the tail currents at the fixed tail voltage of −140 mV were significantly smaller at pHint 4.8 and even smaller at pHint 4.3 compared with pHint 5.8 (Fig. 7 B, insets). Thus, reducing pHint has a biphasic effect on tail current amplitude of mutant E166D: tail currents increase up to pH ∼5.3 and then decrease again. In fact, the tail currents at pHint < 5.3 were too small to allow a quantitative analysis. We noted also that it took several minutes to reach a steady-state current level after perfusing patches with pHint < 5.3, and also recovery after washing with the control solution was much slower than the speed of the solution exchange. Because of these complications we restricted the quantitative analysis of the effect of pHint on the open probability to values pHint > 5. For the quantitative analysis we assumed that the initial current at the constant final “tail” pulse is proportional to the popen at the end of the prepulse to Vp. In Fig. 8 A the normalized initial tail currents are plotted versus Vp. Currents were normalized to the maximum response seen in the same patch for voltages ≥200 mV and pH ≤5.8. This normalization is justified because curves saturated at high voltages at low pHint. The normalization implicitly assumes that the single-channel current at −140 mV is not influenced by pHint. In principle, this is a reasonable assumption because intracellular H+ are expected to be pushed away from the pore at these negative voltages. For pHint > 5, the assumption seems to be justified also by the data, whereas for more acidic pH, tail currents decrease in an apparently paradoxical manner (Fig. 7 B). For pHint > 5 the I-V curves were fitted by
where the parameter pmin accounts for possible leak current and the persistent inward current carried by E166D. The fits resulted in estimates of the voltage of half-maximal activation, V1/2, as a function of pHint (Fig. 8 B). Overall it can be concluded that E166D is dramatically activated by intracellular protons. The popen(V) curve is shifted by ∼81 mV per pH unit (see solid line in Fig. 8 B). Assuming that the maximal popen achieved at high voltages and low pHint is close to unity, allows to estimate the absolute popen at all other voltages and pH values. For example at pHint 7.2 and V = 80 mV a popen of 0.95% is estimated (neglecting pmin that accounts for possible leak current and the persistent inward current). This is relatively close to the value estimated from the single channel analysis of ED-WT heterodimers (0.5%, see above). However, it has to be kept in mind that the assumption that the open probability approaches one at saturating voltages is probably not fully justified (see Discussion). The low popen at physiological pH values inferred from these measurements and from the single channel analysis most likely explains the small macroscopic current amplitude in voltage clamp experiments. As can be seen in the insets in Fig. 7 A, pHint affects also the kinetics of deactivation: deactivation is slightly faster at high pHint.
The Persistent Inward Current Is Possibly Mediated by a Different Open State
We characterized the properties of the persistent inward currents at −140 mV in more detail. First, to assure that it is not an unspecific leak current, we applied 5 and 20 mM intracellular CPA, a known blocker of CLC-0 (Accardi and Pusch, 2003; Estévez et al., 2003). Most of the persistent inward current is indeed blocked by 5 mM CPA and practically all inward current is blocked by 20 mM CPA, in a reversible manner (Fig. 9 A). Since the size of the inward current strictly correlated with the size of the outward current (unpublished data), this result shows that the current is carried by E166D proteins and is not an artifact. A further proof of the specificity of the persistent inward current was obtained by reducing intracellular chloride: reducing [Cl−]int to 14 mM abolished the persistent inward current almost completely (Fig. 9 B).
In noise analysis experiments such as that shown in Fig. 1, we noted that the persistent inward current was associated with a surprisingly small variance. To study this in detail we applied high-resolution nonstationary noise analysis to the deactivating and steady-state part of the current at −140 mV under various conditions. Fig. 10 shows examples performed at various values of pHext. Data are from different inside-out patches at the indicated pH values. In each case at least 50% of the persistent current at −140 mV was blocked by the application of 5 mM CPA to the intracellular side of the patch showing that the current is not caused by leak (unpublished data). Clearly, at pHext 6.4, and even more so at pHext 5.4, the persistent current is much larger relative to the transient inward current (Fig. 10, A–C, a). Actually, at pHext 5.4, the inward current at −140 mV is practically exclusively composed of an activating component without a transient deactivating component. In each case, the variance associated with the persistent current was very small (Fig. 10, A–C, b). In the subpanels c (Fig. 10, A–C) the variance of the current during the pulse to −140 mV is plotted versus the absolute value of the corresponding mean current, after appropriate binning (circles). For pHext 7.2, the data are reasonably fitted by Eq. 1 (Fig. 10 A, c, solid line). However, the data are better fitted by a straight line with slope 0.8 pA, that does not cross the origin (Fig. 10 A, c, dashed line; overlaps with data points). At pHext 6.4 (Fig. 10 B) it is obvious that the data cannot be well fitted by Eq. 1 because the variance is close to zero at steady state even though the current is substantial (Fig. 10 B, c, solid line). The dashed line in Fig. 10 B, c, that fits the data well, has a slope of ∼1 pA. This shows that the transient, deactivating response of E166D at −140 mV is associated with an elementary current of ∼1 pA, corresponding to a single-channel conductance of ∼7 pS. Assuming that the persistent current is carried by ion channel activity, the small variance of the steady-state current could be due to a small unitary conductance or a larger unitary conductance associated with a close-to unity open probability. However, since the persistent current is mediated by E166D proteins, very many of which must be present in the patch illustrated in Fig. 10 to generate the sizeable outward current at 120 mV, the latter possibility can be excluded: a close-to unity open probability together with a conductance similar to that estimated for the transient inward current would generate a much larger inward current than observed. We can therefore conclude that the elementary conductance of the persistent inward current is small. To estimate the elementary current of the steady-state component we use the equation
and assuming for simplicity a small open probability i = σ2/I. From the measured values of the persistent current, I, and the associated variance, σ2, an elementary current of ∼0.11 pA at pHext 6.4 can be calculated, about 10-fold smaller than the elementary current associated with the transient component (similar results were obtained in more than five patches).
At pHext 5.4, data are reasonably well fitted with Eq. 1, resulting in an estimate of i = 0.13 pA (Fig. 10 C, c), very similar to the above estimate for the elementary current associated with the persistent current at pHext 6.4.
Collectively, the noise analysis demonstrates that the persistent inward current is carried by a different mechanism, of lower elementary conductance, than the transient inward current or the slowly activating outward current. It likely represents either a different open state or a transport mode instead of a channel mode.
The Persistent Inward Current Is Not Associated with H+ Transport
It might be hypothesized that the persistent inward current reflects Cl−/H+ antiport. This would explain the marked dependence on pHext and the small variance. To test for this we measured the extracellular pH close to the oocyte surface using a pH-sensitive microelectrode in a solution of low buffer capacity (0.1 mM HEPES) and a slightly acidic pH of 6.4 to increase the magnitude of the inward current. To assay for H+ transport we applied a train of pulses to negative voltages (between −100 and −140 mV). In CLC-5–expressing oocytes, tested at positive voltages in the same solution and at similar current levels (Fig. 11 A), a robust change of the extracellular pH could be recorded (Fig. 11 B; Picollo and Pusch, 2005). Since inward currents generated by mutant E166D are relatively small (e.g., Fig. 6), we used only oocytes that expressed currents >12 μA at 80 mV for this analysis (Fig. 11 C). However, even in such highly expressing oocytes, no significant increase of the extracellular pH close to the oocyte surface could be detected (Fig. 11 D). This result suggests that the persistent inward current is not mediated to a large extent by protons.
The recent discovery that CLC-ec1 is not a Cl− channel but a Cl−/H+ antiporter induced us to investigate better the role of protons in gating of CLC-0. Dutzler et al. (2003) have shown that a major element that is responsible for CLC gating is the glutamate residue E166. Mutating it to alanine, glutamine, or serine leads to a complete loss of voltage and chloride dependence of gating of CLC-0 (Dutzler et al., 2003; Traverso et al., 2003) with channels appearing almost permanently open. X-ray crystallographical analysis of the equivalent mutations of the bacterial CLC-ec1 (E148A, E148Q) beautifully revealed that the location of the presumably negatively charged side chain of E148 in the WT structure was occupied by a Cl− ion in the mutant structures (Dutzler et al., 2003). Accardi and Miller found that the Cl−/H+ antiport activity of CLC-ec1 was abolished by the mutation E148A and that it behaves like a passive Cl−-selective channel or uniporter (Accardi and Miller, 2004). These results indicate that there is an intimate relationship between gating of CLC-0 and Cl−/H+ antiport of CLC-ec1. A Cl−/H+ antiport function was recently described also for the mammalian proteins CLC-4 and CLC-5 (Scheel et al., 2005; Picollo and Pusch, 2005), demonstrating that the function of the bacterial proteins is of far greater relevance to human physiology than previously thought.
In the present work we sought to obtain more insight into the relationship between gating and H+ transport by analyzing in detail the properties of an interesting mutation of the glutamate E166 in CLC-0; a mutation into the very similar amino acid aspartate. Aspartate differs from glutamate only by the lack of one CH2 group while the acidic properties of the two side chains are quite similar (pKAsp ∼3.9, pKGlut ∼4.1; Fersht, 1998). Despite this small structural difference, the mutant E166D has drastic effects on gating. E166D slows down opening at positive potentials (Traverso et al., 2003), increases the rate of closure to negative potentials, and has a very low open probability at physiological pH and voltages <100 mV. Our first aim was to understand whether this drastically different gating of E166D (and E166A) was due to a specific alteration of the fast gate or of the slow gate. From the published single channel data of the mutant E166A (Dutzler et al., 2003; Traverso et al., 2003) it was clear that the fast gating transitions were abolished in the mutant E166A. However, no single channel data were available for E166D. Interestingly, we found here that both mutations (E166A and E166D) appear to lock the slow gate in an open state, similar to the C212S mutation (Lin et al., 1999). Slow-gate closing was partially recovered in tandem heterodimers containing one mutant and one WT subunit. The mechanism of the slow gate and its relationship with the fast gate, however, remain an enigma.
Based on single-channel analysis of dimeric constructs containing one WT and one mutant subunit, we could definitely conclude that in addition to their effect on the slow gate, both mutants strongly alter the fast gate. E166A strongly increases the open probability and renders it insensitive to voltage and chloride concentration, whereas E166D strongly reduces popen. An effect of these mutations on the fast gate might seem almost trivial since the glutamate side chain is directly “occluding” the individual “protopores” in the WT CLC-ec1 structure (Dutzler et al., 2002). However, mutating only one amino acid upstream of E166 in CLC-0, namely K165 into arginine, a mutation that leads to a strong inwardly rectifying phenotype (Ludewig et al., 1997), drastically affects mainly the slow gate, rendering it much faster in a manner that the overall macroscopic gating relaxations are reflecting mostly slow gate transitions (unpublished data), with little direct effect on the fast gate.
In addition, the dimeric ED-WT (and WT-ED) constructs allowed us to determine the absolute open probability of the fast gate of the ED pore. This was not possible using nonstationary noise analysis, because this method fails if the popen is significantly smaller than 0.5. It turned out that at 80 mV the fast gate of E166D pores has a popen of <0.01, whereas WT CLC-0 has a maximal popen of 1 at this voltage. It would have been practically almost impossible to determine such a low popen using direct single channel recording of ED homomers because of the difficulty in determining the number of channels if each has a low popen (Colquhoun and Hawkes, 1990).
Having established that the E166A and E166D mutations affect the fast gate, we next investigated the fast gate modification of E166D by pHint and pHext. The fast gate of WT CLC-0 is known to be affected by pH but in a qualitatively different manner by pHint and pHext. Changing pHint mainly “shifts” the activation curve (Hanke and Miller, 1983), whereas lowering pHext mainly increases the “residual” popen at negative voltages (Chen and Chen, 2001). The differential effect of internal and external pH on the open probability is related to the two possible routes by which the fast gate of CLC-0 can open (Chen and Miller, 1996). The opening rate constant shows a biphasic voltage dependence, rising both at negative as well as at positive voltages (Chen and Miller, 1996; Chen and Chen, 2001), however with different voltage dependencies. Extracellular protons increase the rate of opening favored at negative voltages and have no effect on the closing rate, whereas intracellular protons mostly seem to affect the closing rate (Hanke and Miller, 1983; Chen and Chen, 2001).
The molecular acceptor for intracellular protons is unknown. In contrast, it is believed that the increase of popen of WT CLC-0 at low pHext is caused by a protonation of E166 and a consequent “unblocking” of the external Cl− site, Sext (Dutzler et al., 2003). Therefore, since popen of the mutant E166D is so low we expected that decreasing pHext would dramatically increase currents over the whole voltage range by protonation of Asp166. Surprisingly, practically the only effect of lowering pHext was to increase the persistent current at negative voltages. Qualitatively, the effect of reducing pHext in WT CLC-0 and mutant E166D seems to be nevertheless very similar: steady inward currents are increased, whereas the “voltage-dependent part” of the popen–voltage relationship is only slightly affected. There is, however, a significant difference. In WT CLC-0, single channel events that reflect the “persistent” conductance at negative voltages have the same conductance as the single channel events seen at more positive voltages. In contrast, in mutant E166D the persistent current is carried by a different, lower-conductance, open state. Several possibilities exist to explain this behavior. First, the persistent current of E166D could be mechanistically different from that in WT CLC-0. Even though we are not able to exclude this possibility, the similar pHext dependence and the relative voltage insensitivity suggest that these phenomena reflect the same molecular mechanism, i.e., opening of the channel by protonation of E166/D166 from the extracellular side. The major difficulty with this interpretation is that the conductance of this state in mutant E166D is ∼10-fold smaller than the depolarization-induced open state.
Lowering pHint dramatically increased currents carried by E166D. For pHint values >5 we could describe the activation by a shift of the popen(V) curve along the voltage axis. At more acidic pHint (pH < 5), a qualitatively new behavior appeared that was not followed further in detail. In particular, at very acidic pHint inward “tail” currents became very small, despite the large outward currents. Future studies are needed to characterize this interesting phenotype in more detail.
For pHint > 5, the mechanism of action of intracellular protons seems to be similar to that of WT CLC-0, i.e., lowering pHint “shifts” the popen(V) curve to more negative voltages. In contrast to WT CLC-0 the voltage of half-maximal activation was very positive (+74 mV) even at the lowest pH for which we were able to perform a Boltzmann analysis (5.3). Activation saturated at voltages ≥200 mV and pHint ≤5.8, even if it is not entirely clear if the open probability approaches unity under these conditions. Assuming a saturating popen of 1 allows estimation the absolute popen for the conditions under which the single channel measurements were performed (pH 7.2, V = 80 mV). The value from these macroscopic measurements (0.95%) is in qualitative agreement with that obtained from the single-channel measurements (0.5%). However, the noisy appearance of the current traces at these saturating voltages suggests that the open probability was actually below unity, meaning that the value of 0.95% is an overestimation of the true open probability.
Hanke and Miller (1983) described the effect of pHint on CLC-0 with a four-state model
in which unprotonated channels open and close with rate constants α and β, respectively, whereas protonated channels open and close with rate constants αH and βH, respectively. In the model, protonation of closed and open states occurs with binding constants KC and KO, respectively. Protonation favors opening if αH/βH > α/β. Assuming microscopic reversibility, Model 1
Hanke and Miller (1983) concluded that most of the voltage dependence of opening should be attributed to the rate constants α and β (and αH and βH), whereas protonation and deprotonation, characterized by the constants KO and KC, was deduced to be almost voltage independent. The dependence of the voltage of half-maximal activation, V1/2, on pHint was thus described by the equation
where z is the apparent gating valence that describes the steepness of the voltage dependence of the open probability (Eq. 18 in Hanke and Miller, 1983). For the Torpedo channel, Hanke and Miller obtained a value for z ∼1 (Hanke and Miller, 1983), whereas for E166D we found a slightly smaller value of z ∼0.74 (Fig. 8, legend). K1(0) is the value of α/β at 0 mV. Eq. 5 predicts that V1/2 levels off at low [H] and high [H] at values −RT/zF*ln(K1(0)) and −RT/zF*ln(K1*KC/KO), respectively. In contrast to this prediction of the model of Hanke and Miller (1983), no clear saturation of the V1/2 values is seen for E166D at either low or high pH values (Fig. 8 B). Indeed, the data shown in Fig. 8 B can only be well fitted by Eq. 5 if the parameter z is adjusted to a value of 0.59, significantly smaller than the steepness of the popen(V) curve at pH 5.3 and pH 5.8 (dashed line in Fig. 8 B). Furthermore, the fit predicts that the pK of the hypothetical proton-accepting group must change from 12.2 in the closed state to 5.3 in the open state, a change by 7 log units. Such a huge change of pK is rather unlikely. A more consistent description of our data can be obtained if, in disagreement with the conclusion of Hanke and Miller (1983) it is assumed that the protonation/deprotonation reaction carries most of the voltage dependence, whereas the “conformational” rates α and β are less voltage dependent. Our data are not detailed enough to allow a precise determination of an exact model. We will thus consider only a simplified model in which opening occurs only for protonated channels. This simplification appears to be justified by the finding that the open probability of E166D in the absence of protons is extremely small.
Assuming that αH and βH are voltage independent, and that instead the protonation constant KC depends exponentially on voltage
and if βH/αH << 1, the open probability follows practically a Boltzmann distribution. Most importantly, Eq. 8 predicts that the voltage of half maximal activation depends linearly on pH. This is indeed found for the mutant E166D (Fig. 8 B), and seems also to be an appropriate description of the data reported for CLC-0 (see Fig. 5 in Hanke and Miller, 1983). The argument Hanke and Miller used to conclude that protonation/deprotonation is voltage independent was that the apparent gating valence of the Boltzmann distribution describing the voltage dependence of popen was independent of pH. However, also Eq. 7 predicts a constant gating valence even though the voltage dependence arises exclusively from protonation/deprotonation. Thus, also the data of Hanke and Miller (1983) may allow the alternative interpretation proposed here that at least part of the voltage dependence lies in the protonation step.
We found a change of V1/2 of ∼81 mV per pH unit (see solid line in Fig. 8 B) corresponding to a voltage dependence of ∼0.71 elementary charges in Eq. 6, in good agreement with the observed steepness of the voltage dependence of the popen(V) curves (z ∼ 0.74; Fig. 8, legend). In this simple model the steepness of the voltage dependence of the activation curve reflects mainly the voltage dependence of the protonation from the intracellular side.
The quantitative conclusions from the above considerations are limited by (at least two) factors. First, the data obtained for E166D are not ideal because the popen(V) curve does not saturate for the two more alkaline pH values shown in Fig. 8 B. Thus, the V1/2 values for pH 7.2 and pH 6.8 shown in Fig. 8 B are relatively rough estimates. Unfortunately at more acidic pH values, for which better defined V1/2 values might be expected to be definable, a qualitative new behavior of the mutant emerged, rendering impossible an analysis in terms of the usual open probability. Second, the Model 2 is certainly an oversimplification because the contribution of chloride ions to the voltage dependence of gating is completely neglected. Furthermore, the model does not consider the effect of extracellular protons on the open probability. Despite all these uncertainties, we believe that the idea that one of the major sources of voltage dependence of the gating of CLC-0 arises from a protonation step from the intracellular side is a viable hypothesis. The simple Model 2 assumes protonation directly from the intracellular solution. This seems to be unrealistic, however, if the target of protonation is the residue at position 166. In fact, it is more likely that protonation involves several intermediate proton acceptors that remain to be identified. It is also not clear whether protons from the intracellular side use the same pathway as permeating chloride ions.
Collectively, from our detailed analysis of the conservative mutant E166D, two principal speculations are suggested. First, a major voltage-dependent step of the activation of the fast gate results from a voltage-dependent protonation from the intracellular side. Second, and even more speculatively, protonation of the acidic residue at position 166 (E or D) from the outside and protonation from the inside leads to different open states that have a different single-channel conductance in mutant E166D. In WT CLC-0 these two open states have either a similar conductance or they immediately collapse into a common open state. In contrast, in the mutant E166D the two open states are clearly distinct and not simply interconvertable. In this respect it is noteworthy that a recent theoretical study proposed that the side chain of E166 can adopt an inwardly directed conformation, one that may be susceptible for accepting intracellular protons (Bisset et al., 2005). It will also be highly interesting to study the functional and structural effect of mutating the corresponding glutamate residue to aspartate in the bacterial CLC-ec1. The continuous feedback between functional and structural data will undoubtedly provide an ever better understanding of this unique class of ion channels and transporters.
We thank Laura Elia for expert technical assistance, Alessandra Picollo, and Elena Babini for suggestions on the manuscript, Giacomo Gaggero for help in constructing the perfusion system, and Enrico and Giacomo Gaggero for constructing the high impedance amplifier.
The financial support by Telethon Italy (grant GGP04018) and the Italian Research Ministry (FIRB RBAU01PJMS) is gratefully acknowledged. S. Traverso received a Consiglio Nazionale delle Ricerche doctoral fellowship.
David C. Gadsby served as editor.
Abbreviations used in this paper: CPA, p-chlorophenoxy-acetic acid; WT, wild type.