Abstract
Ethanol modulation of calcium- and voltage-gated potassium (slo1) channels alters neuronal excitability, cerebrovascular tone, brain function, and behavior, yet the mechanism of this modulation remains unknown. Using patch-clamp electrophysiology on recombinant BKCa channels cloned from mouse brain and expressed in Xenopus laevis oocytes, we demonstrate that ethanol, even at concentrations maximally effective to modulate BKCa channel function (100 mM), fails to gate the channel in absence of activating calcium. Moreover, ethanol does not modify intrinsic, voltage- or physiological magnesium-driven gating. The alcohol works as an adjuvant of calcium by selectively facilitating calcium-driven gating. This facilitation, however, renders differential ethanol effects on channel activity: potentiation at low (<10 μM) and inhibition at high (>10 μM) calcium, this dual pattern remaining largely unmodified by coexpression of brain slo1 channels with the neuronally abundant BKCa channel β4 subunit. Calcium recognition by either of the slo1 high-affinity sensors (calcium bowl and RCK1 Asp362/Asp367) is required for ethanol to amplify channel activation by calcium. The Asp362/Asp367 site, however, is necessary and sufficient to sustain ethanol inhibition. This inhibition also results from ethanol facilitation of calcium action; in this case, ethanol favors channel dwelling in a calcium-driven, low-activity mode. The agonist-adjuvant mechanism that we advance from the calcium-ethanol interaction on slo1 might be applicable to data of ethanol action on a wide variety of ligand-gated channels.
Large conductance calcium- and voltage-gated potassium (BKCa) channels, encoded by the Slo1 (KCNMA1) gene are ubiquitous in the nervous system. Increases in channel activity in response to membrane depolarization and/or increase in internal calcium allow BKCa channels to play an important role in action potential (AP) repolarization, after-hyperpolarizations that follow the AP or, in particular, trains of APs, and in controlling the release of neurotransmitters and neurohormones (Weiger et al., 2002; Salkoff et al., 2006).
Given the key role of BKCa channels in controlling neuronal excitability and presynaptic secretion, it is not surprising that this channel type is functionally targeted by drugs that alter nervous system physiology and, thus, behavior. Indeed, several small amphiphiles, including local and general anesthetics and alcohols, have all been reported to modulate BKCa channel activity (Weiger et al., 2002). In particular, BKCa channel activity is increased by short-term exposure to ethanol concentrations obtained in circulation after alcohol consumption [i.e., ethanol <100 mM (Brodie et al., 2007)]. This ethanol action has been demonstrated to 1) accelerate AP repolarization in rat nucleus accumbens neurons (Martin et al., 2004), 2) decrease neuronal excitability in rat dorsal root ganglia, and thus is linked to alcohol-related analgesia (Gruss et al., 2001), and 3) inhibit the release of oxytocin and vasopressin from neurohypophysial nerve endings, the latter effect being linked to alcohol-induced diuresis (reviewed in Brodie et al., 2007). Finally, ethanol inhibition of cerebral artery BKCa channel activity contributes to alcohol-induced cerebrovascular constriction, a drug effect associated with moderate to heavy episodic drinking (Liu et al., 2004).
Ethanol modification of BKCa channel activity is modulated by a variety of factors, including post-translational modification of the channel-forming (slo1) subunit (Liu et al., 2006), coexpression of channel accessory subunits (Martin et al., 2004), and the lipid environment of the slo1 protein (Brodie et al., 2007). However, it is ethanol modification of slo1 channel gating that ultimately determines changes in BKCa activity and, thus, current (Dopico et al., 1998; Dopico, 2003; Martin et al., 2004). The central role of slo in alcohol actions in the body was underscored by demonstrating that BKCa channel activation in dopamine neurons is the major mechanism underlying ethanol-induced motor intoxication in Caenorhabditis elegans. In addition, mutations introduced to Slo by neuronal-specific promoters in Drosophila melanogaster prevent the acquisition of tolerance induced by ethanol (reviewed in Brodie et al., 2007). In synthesis, neuronal slo channel activity is modulated by ethanol, which contributes to major behavioral effects of the drug. Remarkably, the mechanisms and structural basis that determine ethanol modulation of slo1 channel gating, and thus activity, remain unknown.
Four basic processes define slo1 channel gating: intrinsic gating (channel constitutive activity), voltage-, -, and -driven gating (Cox and Aldrich, 2000). Moreover, gating by these biological signals is determined by distinct domains in the slo1 protein: the voltage sensor, the regulation of conductance for K+ domain 1 (RCK1), which includes both a low-affinity Ca2+/Mg2+ recognition site involved in gating by physiological Mg2+ and a high-affinity Ca2+-recognition site, and the “calcium bowl” region. The latter and the RCK1 sense low over (Shi et al., 2002; Xia et al., 2002). Our study demonstrates that ethanol itself, at concentrations that modify slo1 currents and thus neuronal excitability and behavior, does not gate the slo1 channel or modify voltage-driven gating. Instead, ethanol requires physiological , but not , to alter channel activity. Ethanol is merely an adjuvant of activating , which leads to differential ethanol actions on BKCa Po and thus current as a function of activating ligand. Finally, pinpoint mutagenesis results identified the channel structural domains that determine ethanol facilitation of Ca2+-driven gating and the differential contribution of each domain to alcohol modulation of Ca2+ actions.
Materials and Methods
Mutagenesis and Expression. cDNAs coding for mouse brain slo1 (mslo; mbr5) inserted into the pBluescript vector were cut with ClaI and NotI and reinserted into the pBscMXT vector for expression in Xenopus laevis oocytes. Mslo mutants were constructed using QuikChange (Stratagene). Desired mutations and lack of unwanted mutations were confirmed by sequencing at the University of Tennessee Molecular Research Center. Mslo cDNAs were linearized with SalI and transcribed in vitro using T3 polymerase (Ambion, Austin, TX). BKCa β4 cDNA inserted into the pOx vector was linearized by NotI and transcribed using T3 polymerase. BKCa β4 cDNA was a generous gift from Dr. Ligia Toro (UCLA).
Oocytes were removed and defolliculated as described previously (Dopico et al., 1998). Defolliculated oocytes were transferred to ND-96 solution (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl2, and 5 mM HEPES, pH 7.4, containing 2 mg/ml gentamicin and 2.5 mM Na+ pyruvate). Mslo mbr5 cRNA was injected alone (0.1-1 ng/μl) or with BKCa β4 (7.5 ng/μl) cRNA, giving molar ratios ≥6:1 (β:α). cRNA injection (23 nl/oocyte) was conducted using a modified micropipette (Drummond Scientific, Broomall, PA).
Electrophysiology Recordings. Immediately before recordings, oocytes were devitellinized as described previously (Dopico et al., 1998). Recordings were obtained from I/O patches 48 to 72 h after cRNA injection. The electrode solution contained 130 mM K+ gluconate, 5.22 mM CaCl2, 2.28 mM MgCl2, 5 mM EGTA, 1.6 mM HEDTA, and 15 mM HEPES, pH 7.35, free Ca2+ = 11 ± 0.6 μM. Micromolar levels of external Ca2+ are widely known to improve Gigaseal formation and stability, while not modifying BKCa channel function (McManus, 1991; Priel et al., 2007). Bath solutions had varied composition, as follows. In the experiments in which the free was set to <1 μM, the solution contained 130 mM K+ gluconate, 1 mM MgCl2, 5 mM EGTA, and 15 mM HEPES, pH 7.35. In the experiments where the free was set to ≥1 μM, 1.6 mM HEDTA was added. In both cases, varying amounts of CaCl2 (Dopico, 2003) were used to set the free Ca2+ at the desired level, keeping free Mg2+ constant at 1 mM. For the experiments conducted in zero , the bath solution contained 130 mM K+ gluconate, 2.9 mM MgCl2, 5 mM EGTA, 1.6 mM HEDTA, 15 mM HEPES, and 10 mM glucose, pH 7.35. For the experiments conducted in zero , the bath solution contained 130 mM K+ gluconate, 5 mM EGTA, 1.6 mM HEDTA, 15 mM HEPES, 10 mM glucose, pH 7.35, with 5, 5.36, 6.1, 6.42, 6.63, and 6.88 mM CaCl2 to achieve 1, 3, 10, 30, 100, and 300 μM free Ca2+. In making the free Mg2+ solution having 0.3 μM free Ca2+, HEDTA was omitted and 4.1 mM CaCl2 was added. For the experiments with the combined 5D5N, D362A/D367A mutant at 1 mM free , the bath solution contained 130 mM K+ gluconate, 7.5 mM CaCl2, 1 mM MgCl2, 5 mM EGTA, 1.6 mM HEDTA, 15 mM HEPES, pH 7.35. Free Ca2+ and Mg2+ were calculated using Max Chelator (Bers et al., 1994; http://www.stanford.edu/~cpatton/maxc.html) and experimentally validated using Ca2+-sensitive/reference electrodes (Corning Life Sciences, Acton, MA) as described in detail somewhere else (Dopico, 2003).
Patch electrodes were pulled from glass capillaries (Drummond Scientific) as described previously (Dopico et al., 1998). The procedure gave tip resistances of 2 to 5 MΩ (for macropatch recordings) or 5 to 10 MΩ (for conventional I/O single-channel recordings) when filled with electrode solution. An Ag/AgCl electrode was used as ground electrode. After excision from the oocyte, the inner side of the membrane patch was exposed to bath solution containing the desired ethanol concentration and/or free flowing from a computer-controlled, pressurized system (ALA Scientific Instruments, Westbury, NY). Deionized, 100% pure ethanol (American Bioanalytical, Natick, MA) was freshly diluted in bath solution immediately before experiments. Perfusion with urea isosmotically replacing ethanol was used as the control perfusion. Even when applied at maximally effective concentration (100 mM), ethanol failed to modify the bath solution pH, as predicted from the very weak acid properties of the alcohol: pH = 7.376 ± 0.004, 7.375 ± 0.003, and 7.364 ± 0.004 for bath solution, bath solution plus 100 mM urea, and bath solution plus 100 mM ethanol, respectively (n = 35). Isosmotic control solution had no effect on mslo channel Po (n = 11). Iberiotoxin (Alomone) was applied to the extracellular side of outside-out (O/O) patches. The electrode and bath solutions used in O/O recordings corresponded respectively to the bath and electrode solutions described above with I/O recordings. Experiments were carried out at room temperature.
Both macroscopic and unitary currents were acquired using an EPC8 (HEKA, Lambrecht/Pfalz, Germany) amplifier and digitized using a 1320 interface and pClamp8 or pClamp9 software (Molecular Devices, Sunnyvale, CA). Macroscopic currents were evoked from I/O macropatches held at -80 mV by 200-ms, 10-mV depolarizing steps from -100 to 100 (or 160) mV. Currents were low-pass-filtered at 1 kHz with an eight-pole Bessel filter (Frequency Devices, Haverhill, MA) and sampled at 5 kHz. Average current amplitude was determined 175 to 200 ms after the start of the depolarizing step. Unitary currents were low-pass-filtered at 7 to 10 kHz with an eight-pole Bessel filter and sampled at 35 to 50 kHz.
Kinetic Modeling and Analysis. Macroscopic conductance (G)-voltage plots were fitted to a Boltzmann function of the type G(V) = Gmax/1 + exp[(-V + V½)/k]. The effective valence (z) was calculated from: 1/slope = RT/zF, where F = 96,485 C/mol, R = 8.31 J/(mole × K), and T = absolute temperature. Data were fitted to a voltage-dependent Monod-Wyman-Changeux (VD-MWC) for allosteric proteins, using the equation , where G is the macroscopic conductance, Q is the gating charge, L0 is the closed/open equilibrium constant in the absence of ligand, KC is the Ca2+ dissociation constant in the closed-channel conformation, and KO is the Ca2+ dissociation constant in the open-channel conformation. In this VD-MWC model, four parameters of channel gating are obtained: two are Ca2+-independent (Q, L0) and two are Ca2+-dependent (KC and KO) (Cox and Aldrich, 2000). An ad hoc routine (Boltzmann-type of fitting), written and generously provided by Dr. Daniel Cox (Tufts University School of Medicine), was run through Igor-Pro 5 (Wave-Metrics, Lake Oswego, OR), and KO, KC, L0, and Q were directly derived by fitting G/Gmax data to the equation given above. Boltzmann fitting routines were run using the Levenberg-Marquardt algorithm to perform nonlinear least-squares fittings.
Single-channel analysis was initially performed using pClamp9 (Molecular Devices). The product of the number of channels in the patch (N) and the probability that a channel is open (Po) was used as an index of channel steady-state activity. NPo was calculated from the area under the curve of the Gaussian fit of all-points amplitude histograms. From a Poisson distribution of histogram data resulting from the independence and identical behavior of channel gating, NPo = xi, with i = 1... n, where n is the maximum number of simultaneous conducting channels during the observation period, and x is the area under the curve corresponding to each opening, as explained in our first study of ethanol action on mslo channels (Dopico et al., 1996, 1998). This method of NPo calculation allowed us to easily identify 1) BKCa channels from contaminant ion channels that could be present in the cell membrane, avoiding the use of blockers, and 2) possible subconductance states through which the BKCa channel might sojourn. If present, the method would have determined the contribution of these subconductances to the total channel activity and their possible modulation by calcium and/or ethanol. NPo values were obtained from gap-free recording of single channel activity for 1 to 3 min under each condition (control bath, ethanol, washout for any given ionic gradient).
From patches where N = 1, dwell-time histograms were constructed using the half-amplitude threshold criterion from data low-pass-filtered at 5 to 10 kHz, rendering an effective dead time for event analysis that ranged from 18 to 28 μs. A maximum-likelihood minimization routine was used to fit exponential curves to the distribution of open and closed-times. An F table (P < 0.01) was used to determine the minimum number of exponential components to appropriately fit dwell-time histogram data. The number of components in the exponential fit to the open (closed) time distribution provided a minimum estimate of the number of open (closed) states in which the channel population sojourns (Colquhoun and Hawkes, 1983). Kinetic modeling and derivation of individual rate constants were obtained using the QuB program (http://www.qub.buffalo.edu).
For the construction of a simple kinetic model of channel behavior, digitized data were first idealized using the segmental k-means algorithm, which uses hidden-Markov models to find the most likely sequence of events in the data set and estimate model parameters. A maximum likelihood interval analysis method was used to compute the likelihood of the experimental series of open and closed times for a given set of trial rate constants and to search for the rate constants maximizing the likelihood algorithms.
Both macroscopic and microscopic data are expressed as mean ± S.E.M., where n = number of patches/cell. Analysis of variance and Bonferroni's test were conducted using Instat 3.05 (GraphPad Software), and further data plotting and fitting were performed using Origin 7.0 (OriginLab Corp., Northampton, MA).
Computational Modeling. The sequences of calmodulin (PDB entry 1CLL) and the two high-affinity Ca2+-binding sites of slo1 were aligned to match the first Ca2+-interacting acidic residue (Asp56 in calmodulin, Asp362 in the RCK1 domain, and Glu912 in the calcium bowl), and models of both slo1 calcium binding sites were constructed using the homology modeling feature in MOE 2006.05 (Chemical Computing Group, Montreal, QC, Canada). Calcium and ethanol positions were transferred directly from the calmodulin crystal structure. The RCK1 domain model was modified to coordinate Asp367 to the calcium ion by applying distance constraints during geometry optimization of residues 366 to 368, with other atoms in the model held fixed using the MMFF94 force field.
Results
Ethanol Modulation of Channel Activity Depends on Ca2+i. We first studied ethanol action on brain BKCa (mslo) channel steady-state activity (NPo) in the presence of highly buffered, physiological levels of free (1 mM) and (1 μM), two metal ligands that gate the BKCa channel and thus increase its NPo (Shi et al., 2002; Xia et al., 2002). Short-term exposure of the cytosolic side of I/O patches expressing mslo to 100 mM ethanol, a maximally effective alcohol concentration on native and recombinant (slo1) BKCa channels (Dopico et al., 1998; Brodie et al., 2007), robustly increased NPo (Fig. 1A, middle trace) in seven of seven cells; average NPo reached 260 ± 22% of control (p < 0.01) (Fig. 1B) and returned to pre-ethanol values (99 ± 18% of control; n = 7) within 3 to 4 min after washing in an alcohol-free solution. Activating seemed sufficient for ethanol to increase NPo, as ethanol action was observed in Ca2+-containing, Mg2+-free solutions (Fig. 1A, bottom trace). In contrast, ethanol action was lost in internal medium containing physiological levels of Mg2+ but no Ca2+ (Fig. 1A, top trace; averages given in Fig. 1B). Therefore, physiological levels of activating , but not physiological , are necessary for ethanol to increase channel activity.
To begin to determine the mechanisms underlying ligand dependence of ethanol action, we next probed ethanol on channel function across a wide voltage range, and at obtained in neurons under physiological or pathological conditions (Verkhratsky, 2005). Confirming an early observation (Dopico et al., 1998), ethanol potentiated BKCa currents at submicromolar levels of (Fig. 2A, top), with the drug effect diminishing as increased. Remarkably, ethanol consistently inhibited currents when ranged from 10 to 100 μM (Fig. 2A, bottom). Ethanol “dual” (activation versus inhibition) effects on mslo current were also obtained when the drug was probed at 50 mM (Fig. 3); these concentrations are submaximal (∼EC75) to activate neuronal BKCa channels (Dopico et al., 1998; Brodie et al., 2007) and are reached in circulation in humans after moderate to heavy episodic alcohol consumption (Thombs et al., 2003).
Ethanol caused a leftward (current potentiation) or rightward (current reduction) shift in the normalized macroscopic current conductance (G/Gmax)-voltage (V) relationship (Fig. 2B). The resulting half-maximal voltage (V½)-Ca2+ plots show that the “crossover” from ethanol activation (decrease in V½) to inhibition (increase in V½) occurs at ∼10 μM free (Fig. 2C), which is also shown in G/G versus plots at any given voltage (Fig. 2D). As a first approximation to understand ethanol modulation of channel gating, we fitted macroscopic current data to a VD-MWC model (Cox and Aldrich, 2000). Ethanol significantly modified the two Ca2+-dependent parameters of the model: the open channel-Ca2+ dissociation constant (KO) and the closed channel-Ca2+ dissociation constant (KC). In contrast, ethanol failed to alter the other two parameters of the model: the closed-to-open equilibrium constant in the absence of Ca2+ (L0), which reflects constitutive gating, and the gating charge (Q) (Supplemental Fig. S1), both of which are Ca2+-independent (Cox and Aldrich, 2000). Ethanol's lack of action on the channel gating charge is evident from identical aspects of the V½ versus Ca2+ plots before and after factoring Q (Figs. 2E versus 2C). The unmodified gating charge is consistent with the parallel shifts along the voltage axis in the G/Gmax-V plots regardless of ethanol potentiated or inhibited current (Fig. 2B); at any given , the channel effective valence (z) obtained from these plots (Materials and Methods) was not different between control and ethanol (e.g., at 1 μM , z = 0.34 ± 0.01 versus 0.33 ± 0.02; n = 9, p > 0.5). Therefore, whether activating or inhibiting current, ethanol does not alter the amount of effective charge that gates the channel.
The ethanol actions on macroscopic currents in 1 mM free described above were identical to those in Mg2+-free solution (Fig. 4), buttressing the idea that physiological Mg2+ is not necessary for ethanol to modulate BKCa channel gating. Collectively, our data indicate that modulation of BKCa currents by clinically relevant concentrations of ethanol is dependent on selective targeting of Ca2+-driven gating.
From G/G versus plots (e.g., Fig. 2D), we determined the maximal effect (Emax; G/Gmax = 1), the apparent dissociation constant (KD(apparent), in μM), and the equivalent to the Hill coefficient (nH) for Ca2+ in both the absence and presence of ethanol: Emax, 0.98 ± 0.01 versus 0.93 ± 0.01 (p < 0.01); KD(apparent), 6.32 ± 0.17 versus 4.95 ± 0.18 (p < 0.0001); nH, 3.84 ± 0.21 versus 2.55 ± 0.15 (p < 0.0001) (n = 9). Data demonstrate that the “efficacy” of the channel natural ligand (i.e., Ca2+) is reduced in the presence of ethanol. In the absence of changes in the number of channels present in the membrane (Dopico et al., 1998), the decreased nH and Emax for Ca2+ caused by short-term ethanol exposure can be explained by 1) ethanol reduction in unitary current amplitude (this would be more evident as ethanol modulation of Ca2+-driven gating that leads to increased Po reaches a “ceiling effect” at high Ca2+) and/or 2) ethanol favoring ligand (Ca2+)-driven desensitization. To address these possibilities, we next evaluated the actions of ethanol at the single-channel level.
Ethanol Effects on Current Result from Modulation of Ca2+-Driven Kinetics. Single-channel data obtained at different and constant voltage (60 mV) and free (1 mM) clearly show that ethanol failed to modify unitary current (i) amplitude (Figs. 1 and 5, A-D). The lack of change in i was observed within a wide voltage range (-40 to +120 mV), rendering slope conductances (γ) for the ohmic section of the i/V relationship in symmetric 130 mM K+ of 219 ± 21 and 221 ± 22 pS in control and ethanol, respectively (Supplmental Fig. S2A). Ethanol also failed to introduce subconductances during channel openings (Fig. 5, B and D). In addition, ethanol did not alter the high selectivity of the BKCa pore for K+ over Na+, as determined from the lack of change in Nernst potential shift when Na+ substituted for K+ (Supplemental Fig. S2A). This ethanol failure to modify slope conductance and Nernst shift were consistently observed under a wide variety of divalent conditions: 0.3 μMCa2+ (Supplemental Fig. S2A), at which ethanol potentiates activity; 30 μMCa2+ (Supplemental Fig. S2B), at which ethanol reduces activity; physiological (Supplemental Fig. S2, A and B) or 0 Mg2+ (Supplemental Fig. S2C). The lack of ethanol action on channel K+ permeability and selectivity over Na+ strongly suggests that the drug does not modify the conformation of the channel protein region involved in ion conduction. Moreover, the lack of ethanol action on channel conduction, together with ethanol failure to modify channel intrinsic gating (Fig. S1), indicates that ethanol is not acting as a solvent that alters the overall conformation of the mslo channel protein.
The dual ethanol effects on macroscopic current (Fig. 2) were paralleled by dual actions on Po: e.g., potentiation at 0.3 μM (Fig. 5, A versus B) and inhibition at 30 μM (Fig. 5, C versus D). Thus, in the absence of changes in N and γ, ethanol effects on current are caused by drug-induced modification of Po. Dwell-time analysis (Fig. 6) and empiric kinetic modeling (Fig. 7) explain ethanol's dual actions on Po. At low (0.3 μM), both open- and closed-time distributions could be well fit to triple exponentials, suggesting the existence of at least three open and three closed channel states. At 30 μM , however, the channel enters a low-activity mode of gating, which can be seen in single-channel recordings low-pass-filtered at 1 kHz as interburst periods of very low Po lasting hundreds of milliseconds (Fig. 5, C and D). At the bottom of each panel, time-expanded records low-pass-filtered at 7 kHz demonstrate that this low-activity mode includes flickery openings (arrows), resulting in an additional closed time life of ∼3 ms. Thus, at 30 μM , the closed time distribution could be satisfactorily fit to four exponentials (Fig. 6). These data with mslo channels support a previous study with native BKCa channels reporting that > 10 μM drives channel entry into a complex, low-activity mode (Rothberg et al., 1996).
At submicromolar levels, ethanol drastically decreased channel long-closed events (τC2 and τC3). In addition, ethanol increased the probability of occurrence of long open events (τO3) (Fig. 6). These ethanol actions on channel time-lives are similar to those of Ca2+ (Rothberg et al., 1996) and explain the resulting increase in Po at < 10 μM. On the other hand, ethanol reduction of P at 30 μM (Fig. 5, D versus C) is determined by a robust reduction in the average duration of long-open events (τO3) and, more significantly, a major increase in both duration and probability of occurrence of long-closed events (τC3 and τC4) (Fig. 6), as the channel spends more time in the low-activity mode (see time-expanded traces in Fig. 5, D versus C). Thus, ethanol facilitates channel dwelling in a low-activity mode, resembling the actions of high micromolar Ca2+ on channel behavior (Rothberg et al., 1996).
Empiric, simple kinetic models explain in more detail how ethanol favors channel dwelling into the low-activity mode (Fig. 7). Based on the number of exponentials used to properly fit the dwell-times distribution without over-parametization (see Materials and Methods), we started our kinetic channel modeling by considering three open and three closed states, this initial input model (control, 0.3 μMCa2+) having 12 rate constants. Optimization of data by QuB rendered the models shown in Fig. 7, A and B. It is noteworthy that we could only model the channel behavior at high satisfactorily by introducing an additional “state” (low-activity mode; Fig. 7, C versus A). Therefore, final (optimized) models for control and ethanol in 30 μMCa2+ contain 7 “states” and 14 rate constants (Fig. 7, C and D) (other formalism such as the differential equations used and their corresponding matrices are given as Supplemental Information, online). Comparison of these optimized models shows that ethanol mildly diminishes the C3→C2 transition within the normal gating mode and drastically shifts the equilibrium between C3 and the low-activity mode toward the latter (×∼5 times) (Fig. 7, D versus C). These drug actions effectively favored the probability of the channel dwelling in the low-activity mode, from <5% to >10% in the absence and presence of ethanol. This change, together with a decrease in the probability to dwell in open states of intermediate duration (O2) results in an overall decrease in Po (for Po derivation from the kinetic rate constants used in the optimized models, please also see Supplemental Data; online). Ethanol actions on rate constants within the “normal” gating mode, which explain the ethanol increase in Po at submicromolar levels of , are described in the Fig. 7 legend. In brief, ethanol increase in Po results from the drug increasing the probability of the channel dwelling in O2 and decreasing the probability of the channel dwelling in C3 (Supplemental Information; online).
A fact that contributes to the simplicity of the models shown in Fig. 7 is that they include only variant in the presence or absence of ethanol. Others determinants of gating, such as constitutive activity, , and gating charge, have not been considered because they do not interfere with ethanol action (Figs. 2 and 4 and Supplemental Fig. S1). The models, while simple, appropriately describe the ethanol-Ca2+ interaction and its overall effect on Po. First, at all conditions (low and high , absence or presence of ethanol), Po values calculated from the model rate constants are similar to those determined experimentally from all-points amplitude histograms (compare Po values in the caption of Fig. 7 versus those in Fig. 5). Second, we ran our models and obtained simulated single-channel records (Fig. S3) that look practically identical to those obtained experimentally at all conditions (Fig. 5). Collectively, dwell-time histogram analysis and kinetic modeling seem to indicate that, whether the overall effect is increased P (at < 10 μM) or decreased P (at > 10 μM), ethanol's final effect on mslo activity results primarily from facilitation of Ca2+-driven events and kinetic transitions. Our single-channel data and analysis, together with the lack of ethanol action on constitutive activity and voltage-driven gating (Fig. 2), and the necessity of physiologically activating (but not ) for ethanol effects on current led us to conclude that the drug, at concentrations that maximally modify slo1 channel function and current (Dopico et al., 1998; Brodie et al., 2007), cannot gate the channel in the absence of . Ethanol (a “coagonist”) works as a selective adjuvant of activating (the agonist), which results in current potentiation or inhibition at low and high agonist concentrations, respectively.
Our results underscore a fundamental interaction among two ligands ( and ethanol) and a simple receptor system: the BKCa-forming mslo subunit and its immediate proteolipid environment. Although homotetrameric slo1 channels seem to exist in some tissues (Papassotiriou et al., 2000), most brain BKCa channels consist of the association of slo1 and accessory subunits of the β4 subtype (Brenner et al., 2000; Salkoff et al., 2006). Thus, we probed next whether the fundamental interaction among , ethanol and mslo subunits that is reflected by a -dependent dual modulation of current by the alcohol is modified by the presence of neuronally abundant BKCa β4 subunits.
The functional expression of a BKCa β4 subunits was determined by the refractoriness of the nslo+β4 heteromeric channel to iberiotoxin block in O/O patches, in contrast to the sensitivity of homomeric slo1 channels to this peptidyl blocker (data not shown) (Bukiya et al., 2007, 2008). At 0.3 μM , exposure to 100 mM ethanol consistently increased mslo+β4 channel NPo (Supplemental Fig. S4, B versus A, and E). At 30 μM , however, 100 mM ethanol caused a mild but significant decrease in mslo+β4 channel NPo (Supplemental Fig. S4, D versus C, and E). This alcohol dual action on channel NPo occurred in absence of changes in unitary current amplitude (Supplemental Fig. S4, A-D). Collectively, these data clearly indicate that within physiological calcium levels (0.3-30 μM ), the presence of functional BKCa β4 subunits does not drastically modify ethanol pattern of modulation of slo1 channel activity (but see Discussion).
Structural Domains in slo1 That Determine Ethanol Facilitation of Ca2+-Driven Gating. After determining that ethanol action on BKCa currents results from alcohol-specific facilitation of Ca2+-driven gating of the slo channel, we set to determine which functional domains in the slo1 subunit are involved in this fundamental alcohol action. Slo1 channels sense activating Ca2+ through at least three recognition sites, which can be distinguished based on their differential selectivity for divalents and Ca2+ affinities: 1) the calcium-bowl region includes residues Glu912 and Asp923, thought to contribute to divalent coordination, and a penta-aspartate sequence (the 5D5N mutation drastically diminishes the Ca2+ sensitivity of the slo1 channel (Bian et al., 2001; Bao et al., 2004; Sheng et al., 2005; Salkoff et al., 2006); 2) the high-affinity “site” in the RCK1 domain corresponds to Asp362 and Asp367, which are thought to coordinate divalents (Xia et al., 2002); and 3) the low-affinity “site” in the RCK1 domain, determined by Glu374 and Glu399 (Shi et al., 2002). The first two sites selectively discriminate Ca2+ over Mg2+, and nonconserved mutations in these sites drastically reduce the channel activation by low micromolar levels of . In contrast, the third site responds to activation by hundreds of micromolar Ca2+ and millimolar Mg2+ (Zeng et al., 2005).
Ca2+ bowl 5D5N channel mutants were characterized by a significant decrease in apparent sensitivity, which is evident from the right-shift in the V½- plot compared with that of wt mslo (Fig. 2B). Ethanol, however, still caused channel activation and inhibition at low and high levels (Fig. 8A), as it did in wt mslo, indicating that a functional Ca2+ bowl is not necessary for drug action on slo1 channels. In contrast, the D362A/D367A mutation (Xia et al., 2002) abolished ethanol inhibition (n = 8) yet did not modify ethanol potentiation (Fig. 8B). This finding indicates the involvement of the RCK1 high-affinity site in the Ca2+-ethanol interaction that leads to decreased Po. Because this decrease was due to facilitation by ethanol of channel dwelling in a low-activity mode (Figs. 5 and 7), ethanol results with the D362A/D367A mutant led us to hypothesize that 1) at ≫ 10 μM, the D362A/D367A mutant fails to enter a low-activity mode and 2) ethanol fails to reduce (N)Po in this mutant. Single-channel data correctly prove both hypotheses (Supplemental Fig. S5). Thus, the RCK1 high-affinity site is sufficient to mediate the ethanol-calcium interaction that results in decreased Po. Because ethanol inhibition remained in both the calcium bowl and the RCK low-affinity site mutant (Fig. 8C), it seems that the RCK1 Asp362/Asp367 site is not only sufficient but also necessary for this ethanol action.
On the other hand, mutations in each of the sites that participate in sensing micromolar physiological levels of failed to modify ethanol-induced potentiation (Fig. 8, A and B). However, combining the 5D5N with the D362A/D367A mutations not only suppressed ethanol-induced inhibition (as found with the D362A/D367A mutant itself), but also consistently abolished (eight of eight cells) ethanol activation (Fig. 8D). These data were obtained in 1 mM free and 100 μM free , a divalent concentration that seems sufficient to gate the channel (Xia et al., 2002; Zeng et al., 2005). Even when was raised to 1 mM, the combined 5D5N, D362A/D367A mutant failed to be activated by ethanol (Supplemental Fig. S6). Together, data from Fig. 8, A-C, indicate that neither a functional Ca2+ bowl nor a functional RCK1 high-affinity site is necessary for ethanol-induced activation, yet each domain is sufficient for supporting this drug action.
Finally, consistent with the failure of millimolar to modulate ethanol actions (Figs. 1 and 4), the mutations E374A/E399A failed to modify BKCa activation or inhibition by ethanol (Fig. 8C). In conclusion, as far as submicromolar-low micromolar , the channel-specific and natural ligand, is sensed by one of its two high-affinity sites in the slo1 protein, ethanol modulates channel gating. In absence of slo1 crystallographic data, structural insights into the -ethanol interactions on these slo1 high-affinity sites remain largely speculative. Crystal structures of apocalmodulin and Ca2+-bound calmodulin in the presence and absence of ethanol (Chattopadhyaya et al., 1992) provide a precedent for an ethanol binding site created upon Ca2+ binding. This model of Ca2+-dependent ethanol binding may be applicable to slo1 and thus explain the requirement of Ca2+ presence for ethanol to modulate slo1 Po, as explored under Discussion.
Discussion
Our study demonstrates that ethanol at concentrations that are maximally effective to modify neuronal BKCa currents and, thus, excitability, brain function, and behavior (Gruss et al., 2001; Martin et al., 2004; Brodie et al., 2007) can modulate slo1 channel activity only in the presence of activating . Furthermore, ethanol actions depend on the concentrations of this activating ligand. As summarized in Fig. 9, at submicromolar to low micromolar levels of , the equilibria from nonconducting to conducting states as a result of Ca2+ binding to the calcium bowl or the RCK1 high-affinity site present in the slo1 subunit are shifted by ethanol, rendering increased Po and thus macroscopic current. The equilibrium from the channel “normal” gating mode to a low Po mode is favored by higher and involves the RCK1 high-affinity binding site. This transition is favored by ethanol, diminishing overall Po and thus current. In contrast, ethanol seems not to modify gating transitions involving divalent recognition by the RCK1 low-affinity site or movement of gating charge. In addition, ethanol fails to modify the channel intrinsic gating and ion conduction properties. Therefore, ethanol is not changing channel function by altering the overall conformation of the slo1 protein or the arrangement of the slo1 subunits in a functional tetrameric channel. Instead, ethanol action is that of a facilitator of the specific natural ligand that gates the channel (i.e., ).
A distinct feature of BKCa channel gating is that both independent and synergistic activation by transmembrane voltage and can occur (Cox and Aldrich, 2000; Niu et al., 2004). Synergism and independence in BKCa gating are determined by the summation of forces at the channel S6 gate, resulting from the coupling of both voltage sensors (S4) and the several Ca2+ sensors to the gates via peptidic “springs” in the channel structure (Niu et al., 2004). In the absence of Ca2+, a linear relationship between Po and spring distance is consistent with the linker-gating ring acting as a passive spring attached to the S6 gate (Niu et al., 2004). Evidence that ethanol 1) does not modify voltage-driven (Fig. 2 and Supplemental Fig. S1) or 2) Mg2+-driven (Fig. 4) gating and 3) cannot gate the channel in the absence of Ca2+ (Fig. 1) all seem to indicate that ethanol fails to modify the behavior of the passive spring attached to the S6 gate. On the other hand, we show that ethanol action is unaffected by the absence of (Figs. 1 and 4) or the Glu374/Glu399 mutations (Fig. 8C), which define the Mg2+ recognition site (Shi et al., 2002). The lack of ethanol interactions with voltage- and Mg2+-driven gating is consistent with the idea that slo1 channel gating driven by these two biological signals may be coupled (Hu et al., 2003).
Rather, it is Ca2+ sensing by either the calcium bowl or the RCK1 high-affinity site that allows ethanol to modulate Po and thus current. It is noteworthy that both these sites (but not the Glu374/Glu399 site) can sense and participate in gating driven by low micromolar Ca2+ but not millimolar (Shi et al., 2002; Zeng et al., 2005). To obtain structural insights into the Ca2+ dependence of ethanol action, homology modeling of the calcium bowl and RCK1 high-affinity sites was performed. The most relevant structural template for ethanol interaction with a Ca2+-binding protein is calmodulin. Supplemental Fig. S7A demonstrates that ethanol per se does not have any effect on the apocalmodulin domain folding, which is modified solely upon Ca2+ binding. Ethanol is positioned in a groove near the amino end of an α-helix and not far away from the Ca2+-coordinating residues. Remarkably, ethanol is present only when Ca2+ is bound to the metal coordinating residues.
To determine the relevance of this template to the high-affinity Ca2+-binding sites in slo1, sequences were aligned to match the first Ca2+-interacting acidic residue (Materials and Methods). The slo1 calcium bowl and RCK1 high-affinity site models include ethanol located in a groove between two α-helices and near their amino ends, not far away from the Ca2+-coordinating residues (Supplemental Fig. S7B). The ethanol location is near but not identical to that of activating Ca2+. Thus, although some kinetic actions of both ligands overlap (Fig. 5, 6 and 7), Ca2+ and ethanol should be considered heterotropic ligands of the slo1 channel. Ethanol is near a region that allows hydrogen bonding. This pattern is common to several proteins whose activity is modulated by ethanol (Dwyer and Bradley, 2000). This location predicts that more efficient hydrogen bond donor/acceptors than ethanol will interact more efficiently than ethanol will. Consistent with this, trichloroethanol effects on BKCa channel activity and AP in dorsal root ganglia group-A neurons are more robust than those of ethanol (Gruss et al., 2001).
Ethanol increases Ca2+-binding to calmodulin (Ohashi et al., 2004). It is conceivable that an equivalent ethanol location in slo1 high-affinity binding sites could also increase their affinity for the metal and thus facilitate Ca2+-driven gating, as demonstrated by the changes in KC/KO (Fig. 2 and Supplemental Fig. S1), Po (Fig. 5), and kinetic analysis (Figs. 6 and 7). Application of the apocalmodulin/calmodulin model to slo1 high-affinity site Ca2+-dependent ethanol binding might also explain why ethanol per se cannot activate the channel (i.e., cannot substitute for activating ), because ethanol does not have any effect on domain folding. Finally, the residue responsible for the Ca2+-induced conformational change in calmodulin (Glu67) has a homolog in the slo1 calcium bowl (Asp923). The RCK1 high-affinity site, however, lacks an acidic residue at the top of the C-terminal helix. Thus, the calcium-induced conformational changes reported for calmodulin are likely to be followed more closely by those in the calcium bowl than those in the RCK1 high-affinity site. On the other hand, the RCK1 high-affinity site, but not the calcium bowl, determines ethanol-induced reduction in Po at high micromolar Ca2+ (Fig. 8, A and B). It is noteworthy that the contribution of these two sites to the slo1 gating process is not identical; macroscopic kinetic analysis shows that the Asp362/Asp367 site, but not the calcium bowl, slows slo1 channel deactivation at 10 to 300 μM (Zeng et al., 2005). It might be possible that the RCK1 high-affinity site favors both slow deactivation and entry into a desensitized state(s) or low-affinity mode from a common kinetic state(s) (e.g., C3 in Fig. 7). From our data, it is clear that the D362A/D367A mutant fails to enter the low-activity mode and, accordingly, is not inhibited by ethanol (Supplemental Fig. S5).
Among members of the TM6 K+ channel superfamily, slo1 presents a unique, high sensitivity to ethanol (EC50, ∼25 mM; Emax, ∼100 mM; Dopico et al., 1998; Liu et al., 2006). We show that this sensitivity is secondary to ethanol modulation of Ca2+-driven gating determined by Ca2+ interactions with the calcium bowl and the RCK1 high-affinity site, two structures missing in voltage-gated K+ channels other than slo1 (Salkoff et al., 2006). In addition, slo1 distinctively contains an extra segment (S0), rendering an exofacial N-end (Salkoff et al., 2006). CamKII-induced phosphorylation of Thr107 in the S0-S1 linker of bovine aorta slo1 (bslo) channels (which share all relevant Ca2+ sensing sites with mslo mbr5) can gradually switch ethanol responses from activation to inhibition. These data were obtained at < 10 μM, indicating that phosphorylation of Thr107 in bslo can override ethanol amplification of Ca2+ activation of slo1, and suggest a functional coupling between the S0-S1 linker and the Ca2+-sensing sites involved in the Ca2+-ethanol interaction.
Our study identifies a functional interaction between ethanol and that occurs at the BKCa channel-forming slo1 subunit. The final ethanol effect on BKCa channel gating, however, should be fine-tuned by accessory proteins that control the Ca2+ sensitivity of the channel complex. In particular, BKCa beta1 subunits drastically increase the apparent Ca2+ sensitivity of slo1 channels (Brenner et al., 2000; Cox and Aldrich, 2000; Nimigean and Magleby, 2000). According to our model, an increase in apparent sensitivity should facilitate agonist-mediated channel dwelling into the low-activity mode (which is further favored by ethanol presence; Fig. 7), diminishing ethanol potentiation. Heterologous coexpression of β1 subunits consistently reduces ethanol potentiation of hslo channels (Feinberg-Zadek and Treistman, 2007).
In the brain, most native BKCa channels consist of the association of slo1 and β4 subunits (Brenner et al., 2000; Weiger et al., 2002). Within physiological levels of found in neurons (0.3-30 μM), the presence of β4 subunits did not modify the basic fundamental interaction between ethanol and . However, the degree of ethanol-induced potentiation observed at submicromolar was diminished by β4. In addition, the ethanol-induced inhibition observed at 30 μM was somewhat increased by β4 (Supplemental Fig. S4, C-E versus Fig. 5, C and D). This pattern is consistent with the fact that β4 produces a very modest shift in apparent Ca2+-sensitivity within the low micromolar to 30 μM range (Wang et al., 2006) and likely explains data showing some modulation of alcohol action on hslo channels by β4 subunits (Feinberg-Zadek and Treistman, 2007). In brief, in the presence of at resting levels (≤ 10 μM), our ligand-adjuvant mechanism predicts that ethanol will potentiate BKCa currents in neurons. Indeed, this is a widespread finding (Brodie et al., 2007). On the other hand, overall in neurons can reach several tens of micromolar during pathophysiological processes, including excitotoxicity, seizures, and aging (Tymianski and Tator, 1996). At these levels, ethanol will inhibit BKCa channels, impairing one of the major channel populations that protects a cell from toxic levels (Han et al., 2007).
Differential subunit coexpression could contribute to the relative refractoriness of native BKCa channels in nucleus accumbens neuronal dendrites to ethanol activation compared with their counterparts in the somata. The somatic channels express both β1 and β4 subunits, whereas the dendritic channels express primarily β1 (Martin et al., 2004). When evaluated within a range at which β1-subunit coupling to slo1 effectively translates into increased Po (Nimigean and Magleby, 2000) (i.e., increased apparent sensitivity), the dendritic BKCa channels are indeed more -sensitive than their somatic counterparts (Martin et al., 2004). Likewise, native BKCa channels in the somata versus nerve endings of supraoptic neurons display different current phenotypes, including apparent sensitivity, consistent with functional expression of slo1+β1 subunits in the somata, and expression of slo1+β4 in the nerve endings. It is noteworthy that these nerve-ending BKCa channels are sensitive to clinically relevant alcohol concentrations, whereas the somata channels are not (Brodie et al., 2007).
Clinically relevant concentrations of ethanol can modify independently of cell integrity the activity of the vast majority of ligand-gated ion channels (Lima-Landman and Albuquerque, 1989; Lovinger et al., 1990; Wu and Miller, 1994; Parker et al., 1996; Valenzuela et al., 1998; Aistrup et al., 1999; Beckstead et al., 2002; Trevisani et al., 2002; Zhang et al., 2002; Möykkynen et al., 2003; Wallner et al., 2003; Davies et al., 2005). Among inwardly rectifying channels, only G protein-activated K+ channels (Lewohl et al., 1999) are ethanol-sensitive, and among the voltage-gated TM6 K+ channel superfamily, BKCa are highly sensitive (Dopico et al., 1998; Martin et al., 2004; Brodie et al., 2007). Thus, we speculated whether the adjuvant-ligand interpretation that we applied to our results of ethanol-Ca2+ interactions on BKCa could explain some functional ethanol data obtained with ligand-gated channels other than BKCa. First, our interpretation requires ethanol to modulate activity in the presence of efficacious ligand (agonist). This requirement should be overcome in constitutively active channels, because a mutation substituting for agonist-binding diminishes the energy required to drive the channel from the inactivated to the activated state (Galzi et al., 1996). Indeed, although strychnine-sensitive glycine receptors are resistant to ethanol and anesthetics in the absence of glycine, constitutively active glycine receptor mutants are sensitive to these molecules (Beckstead et al., 2002). Likewise, whereas wt 5-HT3 receptors are ethanol-insensitive in absence of serotonin, constitutively active 5-HT3 mutants are ethanol-sensitive (Zhang et al., 2002).
In addition, the adjuvant-ligand interpretation requires that ethanol modulation of channel activity depends on the concentration of the agonist. In channels such as slo1, where the ligand-dependent shift toward a low-activity mode or desensitized state(s) occurs only at high agonist concentrations, adjuvating the agonist with ethanol must result in ethanol-induced activation and inhibition at low and high agonist concentrations, respectively, as shown in the present study. On the other hand, in channels with minor ligand-induced desensitization, ethanol would potentiate ligand-driven activation, the potentiation being diminished as the agonist reaches maximal effect. This also was reported for the Gly receptor (Beckstead et al., 2002). In general, in ligand-gated channels that are potentiated by ethanol, the ethanol effect would diminish with agonist concentration, as found with P2X3 (Davies et al., 2005), bungarotoxin-insensitive nACh (Aistrup et al., 1999), 5-HT3 (Parker et al., 1996), and GABA-A (Wallner et al., 2003) receptors. Finally, in channels with significant desensitization processes, our interpretation predicts that ethanol will primarily reduce activity. Thus, blocking desensitization by pharmacological agents or mutations should reduce ethanol inhibition or even turn it into ethanol potentiation. These patterns were observed with α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (Möykkynen et al., 2003) and N-methyl-d-aspartate channels (Lima-Landman and Albuquerque, 1989), respectively.
In brief, our interpretation of ethanol action on BKCa channels might apply to several results obtained with ethanol on a wide variety of ligand-gated channels. The fundamental requirements of the model are that 1) ethanol cannot gate the channel unless an activating ligand (or a mutation substituting for it) is present and bound to the receptor channel; 2) ethanol, acting as an adjuvant of the activating ligand (agonist), may evoke differential responses in channel activity, which depend on agonist concentration; and 3) the ethanol site is different from that of the channel agonist (heterotropic ligands). Whether nearby or far away in the protein or protein-lipid interface, however, the ethanol binding site must be functionally coupled to the binding site(s) of a channel agonist.
Acknowledgments
We deeply thank Jonathan Jaggar and David Armbruster for critically reading the manuscript, Daniel H. Cox for advice on macroscopic data fitting, David Colquhoun, Fred Sachs, R. Adron Harris, and John J. Woodward for comments, Anna N. Bukiya for discussion, and Maria T. Asuncion-Chin for technical assistance.
Footnotes
- Received May 6, 2008.
- Accepted June 13, 2008.
-
Supported by National Institutes of Health grants AA11560 and HL77424 (to A.M.D.). J.L. is a Postdoctoral Fellow of the American Heart Association Southeast Affiliate.
-
ABBREVIATIONS: BKCa, large conductance calcium- and voltage-gated potassium; AP, action potential; RCK, regulation of conductance for K+; Po, channel open probability; O/O, outside out; I/O, inside out; VD-MWC, voltage-dependent Monod-Wyman-Changeux.
-
↵ The online version of this article (available at http://molpharm.aspetjournals.org) contains supplemental material.
- The American Society for Pharmacology and Experimental Therapeutics