|Home | About | Journals | Submit | Contact Us | Français|
Many physiological and synthetic agents act by occluding the ion conduction pore of ion channels. A hallmark of charged blockers is that their apparent affinity for the pore usually varies with membrane voltage. Two models have been proposed to explain this voltage sensitivity. One model assumes that the charged blocker itself directly senses the transmembrane electric field, i.e., that blocker binding is intrinsically voltage dependent. In the alternative model, the blocker does not directly interact with the electric field; instead, blocker binding acquires voltage dependence solely through the concurrent movement of permeant ions across the field. This latter model may better explain voltage dependence of channel block by large organic compounds that are too bulky to fit into the narrow (usually ion-selective) part of the pore where the electric field is steep. To date, no systematic investigation has been performed to distinguish between these voltage-dependent mechanisms of channel block. The most fundamental characteristic of the extrinsic mechanism, i.e., that block can be rendered voltage independent, remains to be established and formally analyzed for the case of organic blockers. Here, we observe that the voltage dependence of block of a cyclic nucleotide–gated channel by a series of intracellular quaternary ammonium blockers, which are too bulky to traverse the narrow ion selectivity filter, gradually vanishes with extreme depolarization, a predicted feature of the extrinsic voltage dependence model. In contrast, the voltage dependence of block by an amine blocker, which has a smaller “diameter” and can therefore penetrate into the selectivity filter, follows a Boltzmann function, a predicted feature of the intrinsic voltage dependence model. Additionally, a blocker generates (at least) two blocked states, which, if related serially, may preclude meaningful application of a commonly used approach for investigating channel gating, namely, inferring the properties of the activation gate from the kinetics of channel block.
The conduction pore of ion channels can be physically blocked by natural or synthetic agents. Occlusion of the pore by natural blockers underlies important physiological processes, including visual signal transduction, neurotransmission, or shaping of the cardiac action potential (Hille, 2001). Several pharmacological agents produce therapeutic effects by blocking ion channels (Macdonald and Kelly, 1995; Carmeliet and Mubagwa, 1998; Bräu et al., 2001). In some cases, channel block produces severe adverse effects, such as the lethal acquired long-QT syndrome (Keating and Sanguinetti, 2001). Given that ion channel pores possess substantial longitudinal depth, blockers may travel along and interact with parts of the pore before reaching their deepest site. Binding schemes for such interactions will likely comprise multiple steps (Shin and Lu, 2005; Shin et al., 2005; Xu et al., 2009). In cases where blocker affinity for all but the final binding site is negligible, a multistep block may be mistaken for a single-step reaction. As discussed later, the existence of multiple, sequential blocked states may preclude deduction of channel-gating mechanisms from the dependence of blocking kinetics on channel open probability.
Most pore blockers are charged, and the apparent affinity of channels for these blockers often varies significantly with membrane voltage. Two general model types have been proposed to explain this voltage dependence. In one model, the voltage dependence is intrinsic to the binding of the blocking ion within the transmembrane electric field (Woodhull, 1973), and if the blocker is not permeant, the extent of block is expected to be a Boltzmann function of membrane voltage. During the past three decades, this model has been invoked over a thousand times to account for voltage dependence of channel block (e.g., Hagiwara et al., 1978; Neher and Steinbach, 1978; Coronado and Miller, 1979; Blatz and Magleby, 1984; Mayer and Westbrook, 1987; Blaustein and Finkelstein, 1990). The alternative model posits that the blocker does not bind within the electric field, but that the apparent voltage dependence of block reflects the concurrent movement of permeant ions displaced across the electric field by the blocker (Armstrong, 1971; see also Spassova and Lu, 1998, 1999). This model has mostly been used to explain how raising the permeant ion concentration on the opposite side of the membrane lowers the apparent blocker affinity, a phenomenon sometimes dubbed the trans knock-off effect (Armstrong and Binstock, 1965; Hille and Schwarz, 1978; Yellen, 1984; Neyton and Miller, 1988). However, it has relatively infrequently been used to interpret the voltage dependence of channel block itself (e.g., MacKinnon and Miller, 1988; Park and Miller, 1992; Ruppersberg et al., 1994; Spassova and Lu, 1998; Thompson and Begenisich, 2001, 2003; Nimigean and Miller, 2002; Guo et al., 2003; Kutluay et al., 2005; Shin and Lu, 2005; Oseguera et al., 2007; Xu et al., 2009), even though it is the more plausible mechanism for block by organic compounds too bulky to fit into the narrow, high-field part of the pore.
It is noteworthy that in typical voltage-gated channels, the apparent rate of an intracellular blocker entering or exiting the pore may depend on voltage-regulated channel open probability. Such dependence of channel block on voltage gating would complicate studies of the voltage sensitivity of channel block that arises solely from the movement of a blocker and/or permeant ions within the transmembrane electric field along the pore. To avoid this complication, we have used a retinal CNG channel as a model to illustrate the electrophysiological hallmarks and analytical characteristics of both types of mechanism underlying voltage-dependent channel block.
CNG channels mediate visual signal transduction in vertebrate retinal rods and are activated by cGMP (Yau and Baylor, 1989), opening in darkness when the intracellular cGMP concentration is elevated and closing when it is lowered in light. These channels exhibit practically no (or modest) voltage gating in saturating (or subsaturating) concentrations of cGMP (Benndorf et al., 1999; Martínez-François et al., 2009). They can be blocked by divalent cations and polyamines from either side of the membrane in a voltage-dependent manner (Stern et al., 1987; Colamartino et al., 1991; Zimmerman and Baylor, 1992; Root and MacKinnon, 1993; Eismann et al., 1994; Guo and Lu, 2000). Physiological block of the channels reduces the relative contribution of individual channel currents to the overall macroscopic current, a property that allows accurate phototransduction of light intensity (Yau and Baylor, 1989; Rieke and Baylor, 1998).
Complementary DNA for CNGA1 (Kaupp et al., 1989) was cloned into the pGEM-HE plasmid (Liman et al., 1992), which was provided by S. Siegelbaum (Columbia University, New York, NY). The cRNA was synthesized with T7 polymerase (Promega) using linearized cDNA as a template. Oocytes harvested from Xenopus laevis were incubated in a solution containing (in mM): 82.5 NaCl, 2.5 KCl, 1.0 MgCl2, and 5.0 HEPES, pH 7.6, and 2–4 mg/ml collagenase (Worthington Biochemical Corp.). The oocyte preparation was agitated at 100 rpm. It was then rinsed thoroughly and stored in a solution containing (in mM): 96 NaCl, 2.5 KCl, 1.8 CaCl2, 1.0 MgCl2, and 5 HEPES, pH 7.6, and 50 µg/ml gentamicin. Defolliculated oocytes were selected and injected with RNA at least 2 and 16 h, respectively, after collagenase treatment. All oocytes were stored at 18°C.
Macroscopic currents were recorded from inside-out membrane patches of Xenopus oocytes heterologously expressing CNGA1 channels with an amplifier (Axopatch 200B; MDS Analytical Technologies), filtered at 5 kHz, and sampled at 25 kHz using an analogue-to-digital converter (Digidata 1322A; MDS Analytical Technologies) interfaced to a personal computer. pClamp8 software (MDS Analytical Technologies) was used for amplifier control and data acquisition. To elicit macroscopic currents, the voltage across the membrane patch was first hyperpolarized from the 0-mV holding potential to the most hyperpolarizing or depolarizing potential indicated, and then stepped to various test voltages in 10-mV increments and back to 0 mV. The currents in the absence of cGMP were used as templates for subsequent offline background current corrections. Both the internal and external solutions contained (in mM): 0.1 EDTA, 5 HEPES, pH 7.6, and, unless otherwise specified, 130 NaCl. To activate the channel, a saturating (2-mM) concentration of cGMP was included in the internal solution. All chemicals were purchased from Sigma-Aldrich, except for C10-TEA, which was provided by C. Armstrong (University of Pennsylvania, Philadelphia, PA).
Data analysis and curve fitting were performed with Origin 8.0 (OriginLab Corp.). The figures were made using Origin 8.0 and CorelDRAW X14 (Corel Corp.).
We first constructed a dose–response curve for cGMP activation of retinal CNGA1 channels in our recording system. Fig. 1 A shows currents recorded from an inside-out membrane patch in the presence of four concentrations of intracellular cGMP. Confirming previous reports (Kaupp et al., 1989; Goulding et al., 1994; Zagotta and Siegelbaum, 1996; Benndorf et al., 1999), cGMP stimulates current with an EC50 of ~80 µM, approaching saturation at 2 mM cGMP. The dose–response curve differs little between −100 and 100 mV (Fig. 1 B).
The polyamine spermine not only blocks (from either side) but also robustly permeates CNGA1 channels (Lu and Ding, 1999; Lynch, 1999; Guo and Lu, 2000; Nevin et al., 2000). As the latter property complicates the study of the mechanism of block, we used PhTx instead, which can be thought of as a spermine molecule with a bulky group at one end (Fig. 2 A). The spermine portion of the PhTx molecule, like spermine itself, is expected to bind in the narrow selectivity filter, but the bulky group will prevent PhTx from sliding through (Guo and Lu, 2000; see also Bähring et al., 1997). Fig. 2 B shows CNGA1 currents in the absence or presence of three concentrations of extracellular PhTx. The channels were activated with 2 mM of intracellular cGMP, and currents were elicited with the voltage protocol shown. The corresponding steady-state I-V curves are shown in Fig. 2 C. In the absence of PhTx, the I-V curve is approximately linear. Extracellular PhTx blocks the channels in a voltage-dependent manner, rendering the I-V curve outwardly rectifying. To better illustrate the voltage dependence of channel block, we made a plot of the fraction of current not blocked (I/Io) against voltage, which we will call the (voltage-dependent) blocking curve (Fig. 2 D). As expected for a mechanism where a positively charged extracellular blocker enters the transmembrane electric field, the fraction of current not blocked obeys a Boltzmann function that approaches 0 at extreme negative potentials. A fit of the data yields an apparent dissociation constant (appKd) at 0 mV of 2.71 ± 0.20 × 10−5 M and an effective valence (Z) of 1.76 ± 0.02.
At low blocker concentrations, the ON rates of current block are sufficiently slow to allow kinetic analysis. Fig. 3 A shows current block by 0.1 µM PhTx initiated by a voltage step from 100 to −150 mV; it follows a single-exponential time course. Similar records were made for three PhTx concentrations (0.03, 0.1, and 1 µM) at eight voltages between −110 and −180 mV. In Fig. 3 B, we plot the reciprocals of the corresponding averaged time constants against PhTx concentration. At each PhTx concentration, the data points for all eight voltages are superimposed; thus, there is no discernible effect of voltage on the value of the individual slopes, i.e., on the second-order rate constant for blocker binding (Fig. 3 C). Linear extrapolation of the fit in Fig. 3 C to 0 mV yields a second-order rate constant kon (0 mV) of 1.04 ± 0.06 × 109 M−1s−1 for extracellular blocker binding to the pore, with valence (zon, estimated from the extrapolating slope) indistinguishable from 0 (0.01 ± 0.01).
Unblocking kinetics were inferred from the time course of depolarization-induced current recovery. Fig. 4 A shows an example of current recovery from block by 0.1 µM PhTx, initiated by a voltage step from −200 to −50 mV; recovery again follows a single-exponential time course. The reciprocal of the time constant (1/τoff or koff) is independent of PhTx concentration (as expected) but varies with voltage (Fig. 4 B). The averages (over the three concentrations tested) of the natural logarithm of koff are plotted against voltage in Fig. 4 C. A linear fit of the plot yields koff (0 mV) = 9.65 ± 0.35 × 103 s−1 and valence (zoff) = 1.31 ± 0.02. The calculated ratio koff/kon = 9.28 × 10−6 M (at 0 mV) is 34% of the apparent dissociation constant appKd (0 mV) = 2.71 × 10−5 M determined from equilibrium measurements, and zon + zoff = 1.32 is 75% of the latter’s apparent valence Z = 1.76 (Fig. 2). The implication of these differences will be discussed below.
In similar fashion, we examined CNGA1 block by intracellular PhTx. Fig. 5 A shows currents recorded in the absence or presence of 0.3 µM PhTx. Intracellular PhTx, like extracellular PhTx, blocks the channels in a voltage-dependent manner. To examine the voltage dependence of channel block by intracellular PhTx, we plotted the blocking curves for four concentrations of intracellular PhTx in Fig. 5 B. As expected, if PhTx were binding in the electric field, the fraction of current not blocked decreases with membrane depolarization and approaches 0 at extreme positive potentials. A Boltzmann function fits the data well and yields appKd (0 mV) = 8.58 ± 0.38 × 10−7 M and valence Z = 2.67 ± 0.04.
We also performed kinetic analyses of the development of, and recovery from, channel block by intracellular PhTx, initiated with voltage jumps. Fig. 5 (C and D) plots the natural logarithm of the second-order blocking rate constant (kon) and of the unblocking rate constant (koff), respectively, against membrane voltage. Voltage dependence is weak for kon and strong for koff. A linear fit of the data yields kon (0 mV) = 4.23 ± 0.34 × 107 M−1s−1 with low valence zon = 0.23 ± 0.01, and koff (0 mV) = 33 ± 1 s−1 with high valence zoff = 1.60 ± 0.01. The calculated ratio koff/kon = 7.80 × 10−7 M is comparable to appKd = 8.58 × 10−7 M determined from equilibrium measurements (Fig. 5 B), but zon + zoff = 1.83 is 68% of the latter’s Z = 2.67. As demonstrated later, this discrepancy signifies that a two-step (at least) model is required to account for block by intracellular PhTx.
The simplest way to reveal a suspected low affinity blocking state would be to apply much higher PhTx concentrations, but this was cost prohibitive. We therefore increased the apparent affinity of the toxin by lowering the Na+ concentration on both sides of the membrane from 130 mM (used in all experiments thus far) to 30 mM. As shown in Fig. 6 (A and B), under low Na+ conditions, intracellular PhTx, as expected, produces voltage-dependent block with markedly higher affinity. Lowering Na+ now also reveals the existence of dose-dependent but voltage-independent channel block at highly hyperpolarized voltages (−100 to −150 mV). Thus, intracellular PhTx apparently produces both voltage-dependent and -independent blocked states, and the blocking curve is no longer described by a simple Boltzmann function.
We also analyzed the kinetics of block and unblock under low Na+ conditions. The logarithms of kon and koff are plotted against voltage in Fig. 6 (D and E, respectively). As was the case under high Na+ conditions, kon (8.42 ± 0.68 × 108 M−1s−1 at 0 mV) exhibits little voltage dependence (zon = 0.14 ± 0.01), whereas koff (27 ± 2 s−1 at 0 mV) exhibits considerable dependence (zoff = 1.10 ± 0.02).
The low affinity, voltage-independent block was unaffected, i.e., the plateau of the blocking curve seen at hyperpolarized voltages in low Na+ conditions remained unchanged by lowering the cGMP concentration from 2 mM (saturating) to 20 µM (Fig. 6 C), where only 14.1 ± 1.9% (n = 8) of maximal current was activated. However, lowering the cGMP concentration shifted the blocking curve to the right. As explained in the Discussion, this shift may reflect a cGMP-sensitive gating transition that is not directly probed by the following kinetic measurements. Fig. 7 (A and B) compares the ON and OFF kinetics of intracellular PhTx block of channels activated by either 2 mM or 20 µM cGMP. Depolarization-induced block of current activated by 2 mM cGMP (Fig. 7 A; lower black trace) follows a single-exponential time course. At low (20 µM) cGMP, the time course of current block (Fig. 7 A; lower blue trace) initially tracks that in 2 mM cGMP and then becomes slower. A slow component is expected because at low cGMP concentrations, outward current elicited by a depolarizing voltage pulse also develops in two phases: a nearly instantaneous step, followed by a slow phase that reflects modest further enhancement of channel activation by voltage (Benndorf et al., 1999; Martínez-François et al., 2009), which is seen only at subsaturating cGMP concentrations (Fig. 7 A; upper traces). It follows then that only the fast phase of voltage step–induced current block in 20 µM cGMP reflects the actual kinetics of channel block, whereas the slow phase is primarily rate limited by the voltage-dependent additional channel activation that occurs at subsaturating cGMP concentrations. Because the fast phase in 20 µM cGMP tracks faithfully the time course of current block in 2 mM cGMP, we conclude that the kinetics of current block by intracellular PhTx are not significantly influenced by cGMP concentration. The time courses of current recovery at low and high cGMP concentrations are similarly superimposable (Fig. 7 B). As for the kinetics of channel interaction with extracellular PhTx, we found that they, too, are virtually independent of cGMP concentration (Fig. 7, C and D).
CNGA1 channels are blocked by millimolar concentrations of QAs from the intracellular but not extracellular side (Goulding et al., 1993; Stotz and Haynes, 1996; Contreras and Holmgren, 2006). Because QAs are generally too bulky to enter the narrow selectivity filter, the voltage sensitivity of their block of CNGA1 channels must arise indirectly, as it does when they block K+ channels (Armstrong, 1971; French and Shoukimas, 1981; Spassova and Lu, 1998; Shin and Lu, 2005; Xu et al., 2009). We examined the voltage dependence of CNGA1 block by a series of QAs to learn how it differs from block by PhTx. Using decane-bis-trimethylammonium (bis-QAC10; Fig. 8 A) as an example, we will first illustrate the basic properties of QA block.
Fig. 8 B shows CNGA1 current records in the absence or presence of 5 mM of intracellular bis-QAC10, a molecule of the same length as spermine but with wider head groups. Like PhTx, bis-QAC10 blocks the channels in a voltage-dependent manner, rendering the I-V curve inwardly rectifying (Fig. 8 C). The fraction of current not blocked by 5 mM bis-QAC10 is plotted against voltage in Fig. 8 D. This blocking curve is superficially similar to that of PhTx in that it exhibits a voltage-independent blocking component at very negative voltages, followed by a voltage-dependent component. However, unlike that of PhTx, the blocking curve of bis-QAC10 does not proceed to completion at extreme positive voltages but reaches a voltage-independent plateau, i.e., it deviates markedly from a pure Boltzmann function.
We tested six additional intracellular QA derivatives that cause voltage-dependent block (Fig. 9) and render the I-V curve inwardly rectifying (Fig. 10). Given that these QAs block CNGA1 channels with much lower affinity than PhTx, much higher concentrations of QAs must be used. Consequently, the time course of current block is barely or not resolvable (Figs. 8 and and9),9), depending on the concentration of the blockers (as well as their inherent second-order blocking rate constant). To further illustrate the voltage dependence of channel block, we plotted the fraction of current not blocked against voltage for all seven QAs (Figs. 11 and 12 A). These QAs differ in size, shape, and chemistry, exhibiting various affinities and somewhat differing voltage dependence. Despite their differences, the blocking curves for all seven QAs share a characteristic shape: they exhibit two voltage-independent blocking phases, one at extreme positive and one at extreme negative voltages, with an intervening voltage-dependent phase. Thus, QA blockers appear to create more than one blocked state.
Again using bis-QAC10 as an example, we show that altering the permeant Na+ concentration affects only the voltage-dependent phase of the curve, not the two voltage-independent ones, whereas altering blocker concentration affects all three phases (Fig. 12 A).
Two general model types have been proposed to account for the voltage dependence of ion channel block by charged blockers. One model assumes that the blocker itself traverses (a portion of) the transmembrane electric field to reach its binding site in the pore (Woodhull, 1973), and the voltage dependence is thus a property intrinsic to blocker binding. Characteristic of this intrinsic model is that the fraction of current not blocked (I/Io) by a nonpermeant blocker is described over the entire membrane voltage range by a single Boltzmann function. Thus, for a positively charged intracellular blocker (Fig. 13, gray curve):
where appKd is the apparent equilibrium dissociation constant for the blocker-binding reaction in the absence of a membrane potential, [B] is the concentration of blocker, Z is the effective valence (sometimes denoted as zδ), V is the membrane voltage, and F, R, and T have their usual meaning. As shown here, the voltage dependence of CNGA1 channel block by extracellular PhTx can be well accounted for by this type of mechanism, as equilibrium channel block varies with voltage from none to complete block according to single Boltzmann functions (Figs. 2 D and 5 B).
The other model posits that the blocker itself does not bind within the electric field, but that the apparent voltage dependence reflects phenomena extrinsic to blocker binding per se, namely, displacement by the blocker of permeant ions across the field (Armstrong, 1971). We show below that block by intracellular QAs too bulky to enter the channel’s selectivity filter (Goulding et al., 1993) is entirely accounted for by the extrinsic model. A prediction of this ion displacement model is that the voltage dependence of blocker–channel interaction should vanish if permeant ions were removed, a prediction not testable electrophysiologically because removing all permeant ions would abolish current. Fortunately, an alternative approach exists, as argued with the following example. According to the model, an intracellular cationic blocker binds at a site, outside the field, normally occupied by a permeant cation about to enter the electric field. Increasingly, strong depolarizations will lower the permeant ion occupancy of that site until eventually a blocker could occupy it without encountering and displacing a permeant ion. In the limit when blocker binding is no longer coupled to the movement of permeant ions, channel block must lose its extrinsic voltage dependence. As a consequence, the blocking curve characteristically deviates from a pure Boltzmann function at extreme positive voltages to reach a nonzero current plateau (Fig. 13, black curve). We now proceed to a quantitative examination of these two types of voltage-dependent block using PhTx and QAs as examples.
PhTx blocks the channel from either side of the membrane in a strongly voltage-dependent manner (Figs. 2 and and5).5). The voltage-dependent blocking curves for extracellular (Fig. 2) or intracellular PhTx (Fig. 5) are well described by true Boltzmann functions, so that current vanishes completely at sufficiently strong potentials of the polarity that favors blocker binding. Such behavior is expected for a blocker that binds within the electric field.
In kinetic terms, block by extracellular PhTx (Figs. 2–4)) appears at first sight to be a simple one-step bimolecular reaction. Although voltage jump–induced blocking and unblocking transients follow single-exponential time courses, the ratio koff/kon at 0 mV is only 34% of the apparent equilibrium dissociation constant appKd (0 mV), and zon + zoff is 75% of the effective valence of the voltage-sensitive apparent equilibrium appKd. Additionally, the voltage dependence associated with appKd is entirely attributable to koff, whereas kon is essentially voltage independent. As we will discuss below, these phenomena betray more complex kinetics.
The characteristics of CNGA1 block by intracellular PhTx are similar to those of block by extracellular PhTx. As in the case of block by extracellular PhTx, the sum of zoff (1.60) and zon (0.23) from kinetic measurements of intracellular PhTx block is also noticeably smaller than the Z value (2.67) obtained at equilibrium. This discrepancy indicates that the blocking process is not a one-step bimolecular reaction. At least one additional transition must exist, which we have not been able to measure directly. For example, there may be a blocked state with such low affinity that it is insignificantly populated at the blocker concentrations we used. Higher PhTx concentrations might reveal a potential low affinity state, but this is impractical given the high cost of PhTx. An alternative approach is to raise the apparent blocker affinity by lowering the permeant ion concentration to reduce competition and/or trans knock-off by permeant ions (Armstrong and Binstock, 1965). Having lowered the Na+ concentration on both sides of the membrane from 130 to 30 mM, we find that at negative potentials (which minimize intrinsic voltage-sensitive blocker binding), PhTx indeed blocks current in a dose-dependent but voltage-independent manner (Fig. 6, A and B). Therefore, binding of intracellular PhTx to the channel evidently produces at least two blocked states: a voltage-dependent state plus a voltage-independent state.
The two blocked states (ChB1 and ChB2) could in principle be formed via either sequential or parallel transitions (Fig. 14, A or B). For either model, the fraction of current not blocked is given by the ratio of the nonblocked state to the sum of all states:
For the sequential model (Fig. 14 A), the equilibrium constants in the absence of an electric field are defined as K1 = [Ch][B]/[ChB1] and K2 = [ChB1]/[ChB2]. In the presence of an applied electric field, the fraction of current not blocked is then given by:
where Z1 and Z2 are the effective valences associated with K1 and K2, respectively. For the parallel model (Fig. 14 B), the fraction of current not blocked is given by:
where K1 = [Ch][B]/[ChB1], K’2 = [Ch][B]/[ChB2]. Eqs. 3 and 4 are of identical form, but K2 in the sequential model (Eq. 3) and K’2 in the parallel model (Eq. 4) have different physical meanings. The ΔG of the blocked states with respect to the unblocked state is ΔGChB2 = −RT × ln (K1K2) for the sequential model and ΔGChB2 = −RT × ln K’2 in the parallel model; for both models, ΔGChB1 = −RT × ln K1. As discussed in the next section, a sequential model has important practical implications for attempts to deduce channel gate properties from channel block kinetics. For this reason, we will discuss in detail the sequential two-step model as we analyze block by intracellular PhTx (under low Na+ conditions to reveal the low affinity blocker-binding step).
A blocker is likely to transiently contact parts of the pore with some affinity on its way to the energetically most stable site. That is, a realistic blocking scheme likely consists of sequential interactions (regardless of whether there are additional parallel blocking transitions). In principle, the first blocked state may reflect the engagement of a blocker with a part of the pore as shallow at its innermost (or outermost) end. Indeed, as previously demonstrated with inward-rectifier K+ channels (Shin and Lu, 2005; Shin et al., 2005; Xu et al., 2009), a sequential model is physically plausible given the significant depth of ion channel pores.
We fitted the steady-state blocking curves in Fig. 6 B with Eq. 3 to obtain equilibrium constants K1 = 1.04 ± 0.03 × 10−5 M and K2 = 2.03 ± 0.14 × 10−3, and valence Z2 = 2.08 ± 0.04. Z1 was set to 0 because formation of the first (shallow) blocked state is of little voltage sensitivity (witness the voltage-insensitive but dose-dependent asymptotes at extreme negative voltages). As for the individual rate constants at 0 mV, analysis of the blocking and unblocking kinetics (Fig. 6, D and E) yields kon = 8.42 × 108 M−1s−1 with zon = 0.14, and koff = 27 s−1 with zoff = 1.10. As a first approximation, the experimental blocking rate kon primarily reflects k1 in the model (and thus z1 = zon), and the model’s k–1 can then be calculated as K1k1 = 8.76 × 103 s−1. This value of k–1 is ~300-fold larger than the experimental unblocking rate koff. The latter must therefore approximate the model’s OFF-rate–limiting k–2 (and zoff must approximate z–2). Finally, we can estimate the model’s k2 as k–2/K2 = 1.33 × 104 s−1 with z2 as Z2 − z–2 = 0.98. Our kinetic analysis in terms of a sequential two-step blocking model shows that most of the voltage sensitivity of channel block by intracellular PhTx resides in the transition between the two blocked states, with k2 and k–2 exhibiting comparable voltage dependence.
According to Eq. 3 derived for the sequential two-step blocking model, the apparent equilibrium constant at 0 mV is given by:
Because in the present example k–2 (= 27 s−1) k2 (= 1.3 × 104 s−1), appKd reduces to
and because in this instance the values of k–1 and k2 differ by less than twofold, appKd is approximately equal to k–2/k1. As current is blocked from the instant the first blocked state is reached, k1 must effectively determine the experimental ON rate of channel block. Given the low PhTx concentrations we used, the pseudo first-order rate constants k1[B] were resolvable by our recording system. As for k–2, we saw above that it is vastly slower than k–1 in this example; therefore, it effectively determines the experimental OFF rate for recovery from PhTx block. The finding that in this instance the equilibrium appKd approximates the experimental koff/kon ratio is then merely a coincidence and, by no means, indicative of a one-step process. A clear indication that the blocking reaction involves more than one step is the discrepancy between the equilibrium valence, Z, and the sum of the valences zon and zoff associated with kinetic constants kon and koff. Not only for intracellular, but also in the case of extracellular PhTx block, is experimental koff voltage dependent and kon independent. Within the theoretical framework of a simple one-step blocking model, almost the entire voltage dependence of equilibrium block would have to be attributed to the sole dissociation reaction—an unlikely scheme. In the sequential two-step model, however, blocker binding to and unbinding from the first (shallow) site are voltage insensitive, but the transitions to and from the deep site are sensitive in such a manner that the voltage dependence is nearly evenly allocated between the forward and backward reactions—a physically more plausible scheme. Even though experimental demonstration has thus far succeeded only for intracellular PhTx, it is likely that extracellular PhTx also interacts with the channel in more than one step as it penetrates into the selectivity filter.
Mindful of the possibility that channel-blocking kinetics encompass several binding steps, we now examine how the existence of multiple, sequential blocked states may preclude inferences being drawn regarding channel-gating behavior from the dependence of blocking kinetics on channel open probability. The apparent rate of channel inhibition caused by the binding of an open-channel blocker (or by, for example, a cysteine residue modifier) that travels through the channel gate region is expected to vary with the gating state. Such variation has been successfully exploited to delineate the channel gate, e.g., in voltage-gated K+ channels (Liu et al., 1997). To our surprise, the apparent rate of channel block by either extracellular or intracellular PhTx exhibits little dependence on channel open probability, even though block is strongly voltage dependent in both cases. This unexpected finding may help us stipulate the necessary conditions under which properties of the channel gate may be inferred from the dependence of blocking kinetics on gating status.
Kinetics of CNGA1 channel block or unblock are essentially unaltered by lowering the cGMP concentration from 2 mM (saturating) to 20 µM (Fig. 7). This holds for both extracellular and intracellular PhTx. There are two possible explanations for these unexpected results. One possibility is that the gate location, e.g., in the selectivity filter (Sun et al., 1996; Bucossi et al., 1997; Becchetti et al., 1999; Becchetti and Roncaglia, 2000; Liu and Siegelbaum, 2000; Flynn and Zagotta, 2001; Contreras et al., 2008), is inaccessible to PhTx from either side of the membrane. This would mean, in the present example, that the ~20-Å long spermine portion of the PhTx molecule cannot reach deep into the ~12-Å long filter, even though spermine itself readily permeates the channel from either direction. This possibility appears unlikely.
Alternatively, constancy of the apparent channel-blocking rate regardless of ligand concentration may reflect the fact that the rate-limiting transition in the PhTx-blocking sequence is not affected by channel gating. In the initial step of the blocking sequence, PhTx associates with the channel, docking at a site which may, in principle, be as shallow as the innermost part of the pore. It then travels deeper along the pore to form an additional blocked state (or more). If a gate were located between the shallow and the deep sites (or at the deep site), formation of only the deep, not the shallow, blocked state would depend on the gate’s open probability. The second-order rate constant for PhTx associating with the channel to form the first blocked state is very large and arguably near diffusion limited: kon = 8.42 × 108 M−1s−1 in symmetric 30 mM Na+ or 4.23 × 107 M−1s−1 in 130 mM Na+. However, because we used very low PhTx concentrations to achieve experimentally resolvable ON rates, the resulting pseudo first-order rate constant (k1[PhTx]) for formation of the first blocked state in fact became rate limiting, being much lower than the (first-order) rate constant k2 for PhTx transit from the shallow to the deep binding site. As a result, any gating effect on k2 would be obscured. Although k2 might well be sensitive to the gating state, the reverse rate constant k–2 clearly (in the present example) is not because recovery from PhTx block, which is rate limited by k–2 (see above), is insensitive to cGMP concentration. Now, if only k2 is gating sensitive and k–2 is not, this sensitivity will still be reflected in K2 (Fig. 14 A). Indeed, lowering cGMP from 2 mM to 20 µM, which decreases the macroscopic current (that is, Po) by 8.6 ± 1.2–fold (n = 8), causes a 7.5-fold increase in K2 for PhTx block (Fig. 6 C). Therefore, insensitivity of the apparent PhTx-blocking rate to cGMP concentration may simply be a consequence of the multistep character of channel block. In any case, whereas probing the gate by examining the kinetics of channel block (or cysteine modification) is conceptually straightforward for a single-step blocking reaction, this approach is possible for a reaction involving multiple sequential steps only if the observed effect is rate limited by a gating-dependent transition. The mere demonstration that a second-order association rate constant kon is (nearly) diffusion limited does not ensure that the association step is not in fact rate limiting under the actual experimental conditions.
We next discuss extrinsic voltage dependence of channel block by QAs in the framework of an ion displacement model. The most economical version of this model exhibiting the features described in Results (Figs. 8–12)) considers three states: one blocked plus two states of the ion conduction cycle that are blocker free (Fig. 15 A). In this model, the narrow external part of the channel (Ch) can only be occupied by permeant ions, whereas the wide internal part can accommodate either a permeant (e.g., Na+) or a blocking (B) ion. The transmembrane electric field exists exclusively across the narrow part of the pore. The two states of the steady-state Na+ conduction cycle are represented on the left. Extracellular, but not intracellular, Na+ must traverse the narrow part of the pore to reach the internal site in a process that is voltage sensitive. The upper blocking transition is voltage independent as the empty blocker site is outside the field. The left and upper transitions together represent competition between Na+ and intracellular blocker for the internal site. In the limit when the empty species (Ch) vanishes and Na+ and blocker move in a concerted manner, these two processes collapse into a single (electrostatic) displacement reaction as described by the lower transition. This “concerted” blocking transition is voltage sensitive as the blocker site is already occupied by a permeant Na+ ion that must then enter the field.
For this simple three-state model, the fraction of current not blocked is given by:
The apparent equilibrium blocker dissociation constant is
where the equilibrium constants for the blocker-binding transitions without and with concomitant Na+ displacement are KB = [Ch][B]/[ChB] and KB-Na = [ChNa][B]/([ChB][Na+]), respectively (Spassova and Lu, 1998). Eq. 7.1 parses how the two blocking transitions contribute to the overall blocking process, for the experimentally simplest case where the Na+ concentrations on both sides of the membrane are equal.
At sufficiently hyperpolarized voltages where [Na+] >> KB, Eq. 7 reduces to:
an equation dominated by the voltage-dependent blocking transition that involves Na+ displacement. Conversely, at highly depolarized potentials, approaches 0 and Eq. 7 reduces to
which describes voltage-insensitive block not accompanied by Na+ displacement. Eq. 9 accounts for the voltage-independent plateau at depolarized potentials whose amplitude varies with blocker concentration. Indeed, the blocking curve for all seven intracellular QAs tested becomes voltage independent at strongly depolarized potentials. Using bis-QAC10 as an example, we show below that this ion displacement model accounts well for channel block by QAs that are too bulky to enter the selectivity filter (compare Huang et al., 2000).
As shown in Fig. 8 D for bis-QAC10, the blocking curve clearly deviates from a Boltzmann function (dotted curve) at depolarized potentials but is well fitted by the ion displacement model (solid curve). As already discussed for PhTx, the voltage-insensitive block at very negative potentials points to an additional blocked state. A detailed treatment of this additional blocked state will be presented in the next section. To limit the present discussion to the properties of a three-state ion displacement model, we simply scaled the fraction of current not blocked at hyperpolarized potentials (Fig. 12 A) to unity (Fig. 12 B). Eq. 7 derived for our three-state model makes the following experimental prediction: increasing the blocker concentration would not only shift the blocking curve to the left but also lower the plateau seen at positive potentials. This is consistent with our experimental observations (Fig. 12 B; compare black with blue symbols). Mathematically, at higher blocker concentrations, half-maximal block will occur at a lower voltage (Eq. 7) and, consequently, the blocking curve shifts leftward. Thermodynamically, the total energy needed to achieve a given amount of block remains unchanged, and thus the relative energy contribution of electric or chemical origin varies reciprocally. At highly depolarized potentials, Eq. 7 reduces to Eq. 9, meaning that the extent of channel block in this voltage range is independent of voltage and determined solely by the blocker concentration. Therefore, the plateau of the blocking curve at very depolarized potentials decreases with increasing blocker concentration. These predictions make intuitive sense in that strong depolarization reduces the permeant ion occupancy at the shared site, so the blocker will be able to occupy the site without the need to displace permeant Na+. In the limit, binding becomes a standard bimolecular chemical reaction, where the extent of block increases with blocker concentration.
Another prediction is that decreasing the permeant ion concentration will shift the blocking curve leftward without affecting the plateau level at positive potentials, as in fact we observed when we lowered the Na+ concentration from 130 to 65 mM (Fig. 12 B; compare filled with open symbols). Eq. 7 shows that for a given blocker concentration, at lower Na+ concentrations lower voltage suffices to produce 50% block because less electric energy is needed to generate the same probability of vacancy for the blocker. It follows that block of a channel with lower affinity for permeant ions should lose its extrinsic voltage dependence at lower voltages. Eq. 9, on the other hand, indicates that the extent of channel block at extreme positive potentials is independent of Na+ concentration because at sufficiently strong depolarization, Na+ occupancy will vanish.
When we analyzed (in the above section) intracellular QA block in the framework of an ion displacement model, we deferred discussion of voltage-independent block at very negative potentials. To account for that additional feature, we now extend the ion displacement model to include a sequential blocking scheme, analogous to the one we formulated for PhTx (Fig. 15 B). Inserting an extra blocked state in each of the two blocking transitions raises the number of states in the model to five. The additional blocked states ChB1 and ChNaB1 represent a blocker bound at a shallow site without and with a Na+ at the deep site, respectively. For simplicity, we assume that the equilibrium between the shallow blocked state and the unblocked state is characterized by the same equilibrium dissociation constant KB1. KB2-Na and KB2 are the equilibrium constants for the two subsequent transitions to the deep blocked state (ChB2) with and without Na+ displacement, respectively. For this five-state model, the fraction of current not blocked is given by:
The equilibrium constant for blocker binding to the shallow site is:
Blocker movement between the shallow and the deep site is governed by:
for blocker partition to the empty deep site and
for blocker binding to the deep site occupied by Na+. The latter step, which displaces Na+, is the only one assumed to be voltage dependent; this is accounted for by the exponential term in Eq. 10.
At sufficiently hyperpolarized voltages where
approaches 0, Eq. 10 reduces to a simple voltage-independent blocking mechanism:
Eq. 14 accounts for voltage-independent block at hyperpolarized potentials. At these potentials, the block primarily involves the two left horizontal transitions, and the fraction of current not blocked is a function of blocker concentration.
At extreme positive voltages, the term
approaches 0 and Eq. 10 reduces to
where the block again becomes voltage independent. At depolarized potentials, the block primarily involves the two upper transitions, and Eq. 15 thus contains both KB1 and KB2. The curves superimposed on the blocking curves of all seven QAs (Figs. 8 D, ,11,11, and 12 A) are fits of Eq. 10 (all resulting parameters are listed in Table I). Thus, the five-state model fully accounts for QA block.
A brief comment is in order regarding how an additional, shallow blocking site may affect extrinsic voltage dependence of channel block. Recall that a reduction in ion occupancy due to a lowered permeant ion concentration shifts the voltage-dependent blocking curve in the hyperpolarized direction. Consequently, the blocking process loses its apparent voltage dependence at less depolarized potentials (Fig. 12). As explained below, this phenomenon can also be caused by the binding of a blocker at the shallow site. Blocker binding at the shallow site (not shared with permeant ions) would prevent intracellular ions from accessing the deep site (shared by both permeant ions and blocker) (Fig. 15 B). In this case, refilling the deep site with permeant ions from the intracellular side would no longer be possible. In addition, at very depolarized potentials, the deep site could not be effectively refilled from the extracellular side because depolarization decreases the probability of extracellular ions moving across the selectivity filter to reach the site. Thus, the existence of a shallow blocking site could lower the permeant ion occupancy of the deep site, causing the extrinsic voltage sensitivity to vanish at less depolarized potentials.
In conclusion, given that an ion channel pore has considerable depth, a blocker likely blocks the pore in sequential steps. Experimentally, at appropriate blocker concentrations, association with the shallow site may be the rate-limiting step of current block, even if the steady-state occupancy of that site is negligible. Consequently, a sequential binding mechanism could preclude inferences being drawn regarding channel-gating behavior from the dependence of blocking kinetics on channel open probability. Penetration of blocker into the pore can exhibit voltage sensitivity in two possible ways with experimentally distinguishable characteristics. In one case, the charged blocker itself enters the transmembrane electric field (with and without displacing permeant ions), and the observed voltage sensitivity is an intrinsic property of the blocker. Such voltage sensitivity arising from a direct interaction of the charged blocker with the electric field follows a Boltzmann function. Alternatively, the blocker does not bind within the electric field, and its binding indirectly derives voltage sensitivity solely from the concurrent movement of permeant ions across the electric field. Such acquired voltage sensitivity vanishes at voltages where displaceable permeant ions no longer occupy the blocker binding site, and therefore does not follow a Boltzmann function. This striking feature is the electrophysiological hallmark of the extrinsic type of voltage-dependent ion channel block.
We thank S. Siegelbaum for the CNGA1 channel cDNA clone, C. Armstrong for C10-TEA, and P. De Weer for critical review and discussion of our manuscript.
This study was supported by a grant (GM55560) from the National Institutes of Health. Z. Lu is an investigator of the Howard Hughes Medical Institute.
Christopher Miller served as editor.