Block of Na+ channel conductance by ranolazine displays marked atrial selectivity that is an order of magnitude higher that of other class I antiarrhythmic drugs. Here, we present a Markovian model of the Na+ channel gating, which includes activation-inactivation coupling, aimed at elucidating the mechanisms underlying this potent atrial selectivity of ranolazine. The model incorporates experimentally observed differences between atrial and ventricular Na+ channel gating, including a more negative position of the steady-state inactivation curve in atrial versus ventricular cells. The model assumes that ranolazine requires a hydrophilic access pathway to the channel binding site, which is modulated by both activation and inactivation gates of the channel. Kinetic rate constants were obtained using guarded receptor analysis of the use-dependent block of the fast Na+ current (INa). The model successfully reproduces all experimentally observed phenomena, including the shift of channel availability, the sensitivity of block to holding or diastolic potential, and the preferential block of slow versus fast INa. Using atrial and ventricular action potential-shaped voltage pulses, the model confirms significantly greater use-dependent block of peak INa in atrial versus ventricular cells. The model highlights the importance of action potential prolongation and of a steeper voltage dependence of the time constant of unbinding of ranolazine from the atrial Na+ channel in the development of use-dependent INa block. Our model predictions indicate that differences in channel gating properties as well as action potential morphology between atrial and ventricular cells contribute equally to the atrial selectivity of ranolazine. The model indicates that the steep voltage dependence of ranolazine interaction with the Na+ channel at negative potentials underlies the mechanism of the predominant block of INa in atrial cells by ranolazine.
- antiarrhythmic drugs
- sodium channel
- mathematical models
ranolazine, an antianginal drug with antiarrhythmic properties, has been shown to effectively suppress atrial fibrillation (1–5) by producing potent atrial-selective use-dependent inhibition of Na+ current (INa)-dependent parameters. The results of our experimental study presented in the accompanying paper (19) revealed that ranolazine is an open state blocker that displays almost no interaction with the inactivated state of the channel. However, this conclusion is in apparent disagreement with other results obtained in our voltage-clamp experiments and in other studies. Both tonic and use-dependent block by ranolazine were significantly affected by a moderate increase of the holding potential (Vhold) in the range well below the activation threshold. These effects were much more pronounced in atrial cells compared with ventricular cells, in appearant disagreement with our finding that interaction of ranolazine with binding site on the α-subunit of atrial and ventricular Na+ channels is almost identical. Moreover, the difference in voltage dependence of the block at negative potentials between atrial and ventricular cells was almost identical to the difference in the relative positions of the steady-state inactivation curves between these cell types. In addition, it was shown that ranolazine produced a dose-dependent shift of the steady-state inactivation curves both in atrial and ventricular cells (5). This shift was more pronounced for late INa that for peak INa (15). Therefore, before proceeding to investigate the nature of the atrial-selective block of the Na+ channel by ranolazine, we have to address the mechanism of the state-dependent interaction of ranolazine with the Na+ channel to account for the above experimental observations.
In this study, we developed a Markovian model of Na+ channel gating with the aim to test the hypothesis that exclusive ranolazine interaction with preopen and open states, but not with the inactivated state of the channel, can account for a variety of the above-mentioned experimental observations. To make the main mechanisms easier to grasp, we minimized the number of channel states to a bare minimum and investigated the effects of ranolazine on fast and slow INa using separate models. The results of this modeling study delineate the mechanism of the potent atrial selectivity of ranolazine.
We developed a simplified Markovian model for INa that incorporates the interaction of ranolazine with different states of the channel (Fig. 1A). The main theoretical considerations for the development of the model and model equations are shown in the Supplemental Material.1 The number of closed states was reduced to two: the fully closed state and the preopen nonconducting state (see Supplemental Fig. S1). In addition, we assumed that the inactivation process is coupled with activation and requires at least one of four activation gates to open to proceed (Supplemental Fig. S2). The absence of the closed-inactivated state in the Markovian scheme (Fig. 1A) reflects this fact. In agreement with the experimental data, we shifted the voltage dependence for intrinsic steady-state inactivation in atrial cells to reproduce the observed approximately −10-mV difference in steady-state inactivation curves between ventricular and atrial cells. Due to the corresponding but smaller (−5 mV) difference in the steady-state activation curves, the inactivation time constant in atrial cells was smaller, in agreement with the experimental data (A. C. Zygmunt, unpublished observations).
The model for slowly inactivating INa was derived by modifying the inactivation rate to obtain an inactivation time constant of ∼200 ms at −30 mV, in agreement with our experimental data. Models for fast and slow (late) INa were investigated separately. We did not make an attempt to combine these models in a more accurate model of cardiac INa.
The description of the ranolazine interaction with the Na+ channel was based on the following considerations. Ranolazine, an open state blocker, gains hydrophilic access to the binding site when two activation gates in domains III and IV are open and the inactivation gate is not yet closed (Fig. 1, C and D). The fraction of the preopen state that permits ranolazine binding and unbinding (δ) is equal to 3/14 (Supplemental Figs. S1 and S3). With this assumption, the blockade becomes significant at voltages well below the activation threshold. Ranolazine is an elongated linear molecule (Fig. 1B), which is different from all typical open state blockers (12). We hypothesized that the unique linear structure of the ranolazine molecule prevents closing of all four activation gates simultaneously, so that it cannot be trapped at the fully closed state (no closed blocked state on the model scheme). Finally, ranolazine does not interact (it neither binds nor unbinds) with the inactivated state, and the corresponding pathways are eliminated from the scheme (Fig. 1A). Previous experimental studies (10, 11) of lidocaine interaction with the Na+ channel have shown that the apparent affinity for lidocaine is increased by partial channel activation, i.e., access is facilitated when channel is in the preopen or intermediate closed states. Furthermore, analysis of the experimental data suggested that the predominant pathway for lidocaine binding to the Na+ channel is provided by the preopen state of the channel (14, 18). The assumption that ranolazine does not unbind from the inactivated state and is effectively trapped in the channel after accessing the binding site via the preopen/open state has previously been shown to be essential for the use-dependent block by flecainide (7).
We considered the hypothesis that interaction of ranolazine with the Na+ channel at negative potentials, when the channel is exclusively in the preopen state(s), provides the basis of its atrial selectivity. As shown in Fig. 1, C and D, the pathway to or from the binding site is available when activation gates in domains III and IV open, which includes not only the fully open (conducting) state but also preopen nonconducting states. According to this hypothesis, the atrial selectivity depends on the ability of ranolazine to unbind from the blocked channel, which is almost exclusively determined by the intrinsic steady-state inactivation curve. Only channels recovered from inactivation provide a hydrophilic pathway for ranolazine to unbind from the channel. The amount of time that the channel spends in the preopen and non-inactivated states during the action potential (AP) significantly affects ranolazine binding and unbinding from the channel. The model equations and further details of the model can be found in the Supplemental Material.
Results of the model simulation of ranolazine interaction with the Na+ channel at negative potentials are shown in Fig. 2. The predominant pathway of ranolazine block below the activation threshold is shown with a thick arrow in the scheme (Fig. 2A). Channel gating proceeds much faster than drug binding/unbinding at all voltages so that the distribution between states connected by double-headed arrows can be considered at steady state. When the membrane potential is stepped moderately positive from Vhold = −140 mV, the fraction of channels in the preopen state increases, providing a hydrophilic pathway to the channel binding site. Simultaneously, a smaller fraction of channels inactivates, closing this pathway. Progressive depolarization increases the fraction of channels in the preopen state, but some of these channels become inactivated. Thus, the accessibility of the binding site will initially increase with voltage until inactivation starts to dominate.
Figure 2B shows simulated steady-state availability curves in the control and in the presence of 15 μM ranolazine in atrial and ventricular models obtained using the standard inactivation protocol (2,000 ms at Vhold = −140 mV followed by a prestep to the indicated voltage for 1,000 ms and a test step to −30 mV for 50 ms). The simulated curves (Fig. 2B, left) were in good agreement with those obtained experimentally under the same conditions (Fig. 2B, right). The shift of the availability curve in the presence of ranolazine is due to additional block of a fraction of the noninactivated channels, which become unavailable upon depolarization. Normalized to the maximal current, the availability curve will shift to more negative potentials compared with the curve in the absence of drug. The difference between these two curves reflects the amount of block accumulated during a 1,000-ms step at each potential.
Figure 2C shows the predominant pathway for Na+ channel recovery from block at Vhold = −140 mV. As assumed in the model, ranolazine prevents all four activation gates to close simultaneously upon repolarization and the channel cannot enter the fully closed blocked state (absent in the scheme on Fig. 2C). Different from other typical open state blockers, ranolazine is not trapped in the fully closed state and unbinds from the preopen state. However, only a fraction (δ = 3/14) of the preopen state with activation gates in domains III and IV open permit ranolazine unbinding. The rate of recovery from block at −140 mV can be approximated as δ × uo, where uo is the rate of ranolazine unbinding from the open state (0.006 ms−1), as calculated in the accompanying paper (19) and used in this model. This rate corresponds to a time constant of recovery of ≈778 ms, which was in a good agreement with those obtained experimentally in canine myocytes and human embryonic kidney (HEK)-293 cells stably expressing the Nav1.5 channel [see Table 2 in the accompanying paper (19)]. This time constant does not depend on the concentration of ranolazine because at this potential all drug-free channels in the preopen state promptly deactivate eliminating any forward rate.
Figure 2D shows the time course of recovery from inactivation and from inactivation plus block as simulated by the model using the same double-pulse protocol as was used in experimental studies (a single prepulse to −30 mV for 1,000 ms followed by a variable recovery interval at −140 mV and a test pulse of 50 ms to −30 mV). In the control, the recovery process was monoexponential. In the presence of 30 μM ranolazine, a second exponential appeared, with the amplitude corresponding to the block attained during the preceding 1,000-ms prepulse and with the time constant for recovery from block. This simulation reproduced experimental conditions used in HEK-293 cells, and both values (amplitude and time constant) were in good agreement with data obtained experimentally [Fig. 3 in (19)]. The rate of recovery at −140 mV does not depend on cell type and was identical in ventricular and atrial cells. The good agreement between experimental and simulated recovery time constants provides evidence in support of the hypothesis that recovery from block at very negative potentials proceeds via the preopen state and that channels blocked by ranolazine cannot fully close. Conversely, if blocked channels can fully close and trap a drug at the binding site, recovery will depend on infrequent openings of some activation gates and will proceed with the much slower rate typical for all open state blockers.
Figure 3 further illustrates the mechanism of the voltage dependence of ranolazine interaction with the Na+ channel at negative potentials below activation threshold. Figure 3A shows the voltage dependence of the fraction of ventricular and atrial channels blocked in the presence of 15 μM ranolazine. This fraction is larger in ventricular cells because the overlap of the ventricular inactivation and activation curves is larger than that of the atrial curves. Most blocked channels at potentials below activation threshold will accumulate in the preopen inactivated state, which serves as a “sink” for the block process. The rate of block is proportional to the fraction of noninactivated channels in the preopen/open state, which is equal to the product of the preopen/open steady-state curve and the steady-state inactivation curve.
Figure 3B shows the divergence of the time constants of ranolazine interaction with Na+ channels for atrial and ventricular cells at depolarized potentials due to the difference in the fraction of inactivated channels at any given potential between the two cell types. Note that the time constant of the ranolazine interaction with the Na+ channel becomes larger with potentials positive to −140 mV, while the fraction of channels in the preopen state increases. This apparent paradox stems from the fact that the rate of the drug interaction with a binding site is the sum of both forward (binding) and reverse (unbinding) rates. In this potential range, the binding rate is much smaller than the unbinding rate, leading to very small equilibrium (tonic) block. Moderate depolarization results in channel inactivation, which prevents unbinding. This effect has a much larger influence on the total rate of interaction than the increase of the binding rate due to channel activation. Thus, the initial portion of the voltage dependence for the time constant of the ranolazine interaction with the Na+ channel almost coincides with the intrinsic steady-state inactivation curve.
The time constant of the ranolazine interaction with the binding site increases with voltage (Fig. 3B), so that the accumulated block will be further from equilibrium at more positive potentials (Supplemental Fig. S9), reducing the slope of the availability curve in the presence of the drug. Moreover, the position of the availability curve in the presence of drug will be a function of the inactivating pulse duration, with the drug-induced shift becoming larger and the slope becoming steeper when pulse duration is increased above 1,000 ms (see Supplemental Fig. S10).
When a step potential is applied above the activation threshold, an additional pathway via the fully open (conducting) state transiently opens. Ranolazine interaction with this transient open state explains the substantial block accumulated during steps to these potentials. This block of the fully open state does not depend on the duration of the pulse above a few milliseconds.
Figure 4A shows the simulated effects of various concentrations of ranolazine at Vhold = −140 and −90 mV on the open probability of the fast Na+ channel. Peak open probability elicited by a voltage step to −30 mV for 5 ms was almost unchanged in the presence of 100 μM ranolazine when holding at −140 mV (Fig. 4A, left) but was substantially diminished in the presence of 30 or 100 μM ranolazine when holding at −90 mV (Fig. 4A, right). The concentration-dependent inhibition of peak INa simulated for these two Vhold values is shown in Fig. 4B. Depolarization resulted in a considerable increase of the channel sensitivity to ranolazine. The model predicted the decrease of the apparent dissociation constant (Kd) from 1,800 μM (slope: 0.97) at Vhold = −140 mV to 61 μM (slope: 1.1) at Vhold = −90 mV and to 4.9 μM (slope: 1.37) at Vhold = −70 mV (not shown).
The results of these simulations were in good agreement with the experimental data that were obtained in HEK-293 cells stably expressing different isoforms of rat Na+ channels (16) using a similar voltage protocol (Fig. 4, C and D). Apparent Kd values were found to be 75.0, 65.5, and 59.9 μM for rat rNav1.4, rNav1.5, and rNav1.7, respectively, at Vhold = −70 mV. The 20-mV discrepancy between the model and experimental data for the Vhold value where Kd ≈ 60 μM can be explained by the positions of the control steady-state inactivation curves: −70 mV in the model and −50 mV in rNav1.4 (Fig. 9 in Ref. 17).
The apparent voltage dependence of Kd stems from the differences of voltage dependency for binding and unbinding rates, which, in turn, depend on the fraction of channels in the preopen state and in the inactivated state, correspondingly (Eqs. 18–20 in the Supplemental Material). Figure 5 shows the effects of holding or diastolic potentials on the apparent kinetic rates and Kd. Ranolazine binding requires an activation gate(s) to open, after which the blocked channels promptly inactivate, trapping ranolazine inside the channel. Unbinding requires the inactivation gate to open before ranolazine can unbind. The voltage dependence for access to the binding site is a bell-shaped curve (Fig. 5A, solid curve), reflecting a fraction of preopen, but not inactivated, channels at each voltage. The voltage dependence for the unbinding rate monotonically decreases with voltage, reflecting the voltage dependence of the intrinsic steady-state inactivation (Fig. 5A, dashed curve). These differencies in the voltage dependence for binding and unbinding rates arise from activation-inactivation coupling and from the assumed failure of blocked channels to fully close. Consequently, the appearant Kd defined as a ratio of the apparent unbinding rate to the apparent binding rate is also voltage dependent (Fig. 5B). Note that the voltage dependence of the apparent Kd strongly depends on the particular model of normal Na+ channel gating, which is used to simulate channel blockade.
Slowly inactivating or late INa is potently blocked by ranolazine. We simulated slow INa by decreasing the inactivation rate of fast INa, as detailed in the Supplemental Material. Fig. 6A shows the simulated open probability for slow INa in the control and in the presence of different concentrations of ranolazine obtained using the complete model (Fig. 1A). The inactivation time constant in the control was 205 ms at −30 mV. This value can be considered a generic one and was set between 10 ms for the R1623Q mutant SCN5A channel expressed in HEK-293 cells (9), 50 ms for the initial inactivation of slow INa in inactivation-deficient Nav1.4 (16), 100 ms obtained in ventricular myocytes using 300-ms pulses (A. C. Zygmunt, unpublished observations), and 500 ms for late INa obtained in failing canine ventricular myocytes using 2-s pulses (15). The simulated traces were in good agreement with experimental recordings (Fig. 6C) obtained using inactivation-deficient Nav1.4 channels (16). The simulated concentration-dependent inhibition of slow INa shown in Fig. 6B demonstrates that slow INa at the end of a 50-ms pulse was much more sensitive to ranolazine (apparent Kd: 5.2 μM, slope: 1.4) than peak INa (apparent Kd: ∼1,300 μM, slope: 0.87), in agreement with experimental data (Fig. 6D). Similar values for IC50 have been reported in other studies: 6.46 μM by Undrovinas et al. (15) and 7.45 ± 0.11 μM by Rajamani et al. (9). Note that the apparent Kd (IC50) for the block of the slowly inactivating current strongly depends on the time into the trace where the block is recorded (see Supplemental Fig. S12) and could be found to be different from the “true” Kd for ranolazine interaction with the channel binding site (≈2 μM in our study).
In addition to these experimental results, our model made a good prediction for the concentration-dependent shift of the steady-state availability curve, as obtained by Undrovinas et al. (15) (see Supplemental Fig. S11). Further discussion of our choice of the experimental data used for comparison with the model predictions can be found in the Supplemental Material.
As expected, the model correctly reproduced the use-dependent inhibition of fast INa by 25 μM ranolazine at 15°C during 20-ms pulse trains to −30 mV from Vhold = −140 mV in ventricular and atrial cells (Fig. 7, A and B). Simulations using the model are illustrated by solid lines, while experimental data are shown as symbols with corresponding error bars (Fig. 7). The magnitude of the steady-state block increases inversely proportional to the interpulse (recovery) interval, while the rate of block (λ; in pulses−1) decreases. The model predicted minimal differences in the use-dependent block between atrial and ventricular myocytes because ranolazine interactions with the Na+ channels in these cells at −140 mV (holding) and −30 mV (step potential) are almost identical. The slightly slower rate and smaller steady-state block in atrial cells compared with ventricular cells are due to faster atrial INa inactivation at −30 mV.
The differences in use-dependent block between atrial and ventricular cells due to a more negative position of the atrial availability curve can be unmasked using moderately depolarized Vhold values. Supplementary Figure S13 shows these differences in tonic block and the rate and steady-state use-dependent block between atrial and ventricular cells simulated using the same pulse trains but varying Vhold between −120 and −100 mV. The results of the simulation were in good agreement with experimental data.
Rates of ranolazine interaction with the Na+ channel were significantly affected by temperature. A train of pulses to −30 mV from Vhold = −120 mV in ventricular and atrial cells produced a use-dependent block that was either too small when the ranolazine concentration was at or below 25 μM or developed too fast at higher concentrations. Parameters of the exponential fit become unreliable for the purpose of guarded receptor analysis if the steady-state block is <10% or if λ (in pulses−1) is >0.5, corresponding to less than two pulses. For this reason, we were unable to directly calculate the kinetic rates of ranolazine interaction with the Na+ channel at 37°C. However, using nonlinear regression analysis, we found that the time course of the use-dependent block of peak INa can be quite accurately reproduced by our model if both kinetic rates are multiplied by a factor of 3 between 15 and 37°C, giving Q10 ≈1.65. This value of Q10 was in the range of temperature sensitivity previously described for lidocaine interaction with the Na+ channel (8). The effect of temperature on normal Na+ channel gating was simulated by increasing the rate of inactivation 2.5-fold at 37°C compared with 15°C (Q10 ≈ 1.5), as described in the Supplemental Material. Figure 7, C and D, shows the experimentally recorded use-dependent block of peak INa in ventricular and atrial cells using a train of 40 pulses in the presence of 50 μM ranolazine (symbols with error bars). The solid lines show the use-dependent block predicted by the model adjusted to 37°C using the same voltage-clamp protocol.
Next, we used our model to predict the inhibition of INa by ranolazine under physiological conditions (Figs. 8⇓–10). Using the INa model adjusted to physiological conditions (37°C and a more positive position of the channel availability curve; see the Supplemental Material), we applied experimentally recorded trains of atrial and ventricular APs as command potentials instead of square voltage pulses. Figure 8A shows the relative block of peak Na+ conductance simulated by applying trains of 20 atrial and ventricular APs experimentally recorded at a basic cycle length (BCL) of 500 ms in the presence of increasing concentrations of ranolazine. The model predictions were in good agreement with experimental data (Fig. 8B). Figure 8C shows the development of additional use-dependent block when the BCL was abbreviated from 500 to 300 ms. The simulations used the same trains of experimentally recorded APs shown in Fig. 8D. The model predicted that in atrial cells the additional block of peak INa was 45.5% when the BCL was abbreviated to 300 ms, whereas the same block of Vmax recorded experimentally was 45.6%. On the other hand, the much smaller additional block of 7.3% was simulated by the model for ventricular APs, in agreement with experimentally recorded 11.4% block. When a similar train of atrial APs recorded in the control was used, the model predicted only 30.3% additional block (see Supplemental Fig. S14). This indicates that the effect of ranolazine on the morphology of the atrial AP is responsible for more than a third of the additional use-dependent block at the faster heart rate. The good agreement between the predictions derived from the model and the experimental data indicates that the model accurately captures the essential features of the interaction of ranolazine with the Na+ channel.
Using the model, we were able to independently investigate the relative contributions of atrioventricular differences in the position of steady-state inactivation curves versus AP morphology. The time course of block development was simulated by applying trains of APs at a BCL of 500 ms in ventricular (Fig. 9A) and atrial (Fig. 9D) cells in the presence of 10 μM ranolazine. The accumulated block after the 10th AP in atrial cells was 8.4 times larger than the block accumulated in ventricular cells.
Next, we simulated conditions in which the atrial AP train was imposed on the ventricular INa model (Fig. 9B) and the ventricular AP train was imposed on the atrial INa model (Fig. 9C). The simulations showed that the shape of the AP is responsible for the 2.0- to 2.4-fold increase in the degree of accumulated block. In contrast, the difference in the position of the steady-state inactivation curve (≈10 mV) is responsible for the 3.5- to 4.1-fold increase of the accumulated block. Therefore, these two prominent features that distinguish atrial from ventricular cells–the longer AP with slow terminal repolarization and the more negative position of the steady-state inactivation curve–are both significant contributors to the atrial-selective block of fast INa by ranolazine.
Finally, the model was used to compare the accumulation of use-dependent block in atrial and ventricular cells in the presence of 10 μM ranolazine, 1 μM propafenone, and 40 μM lidocaine (Fig. 10). Propafenone was assumed to be a pure open state blocker, which also binds to and unbinds from the closed state. Lidocaine was assumed to interact with all three states of the Na+ channel, consistent with published data (14). The schemes for propafenone and lidocaine interactions with Na+ channels and the corresponding kinetic rates are shown in the Supplemental Material (Supplemental Fig. S4 and Supplemental Table S1). The kinetic rates were adopted from published experimental data (6, 13). Supplemental Figure S5 shows the shape of the applied APs together with the time-dependent changes of the fractions of preopen/open and inactivated states. Concentrations were chosen to attain an ∼40% steady-state block at the end of the train in atrial cell. The results of these simulations are shown in Fig. 10. Ranolazine (10 μM) induced considerable use-dependent block of INa in atrial cells but negligible block in the ventricular cell model. The ratio of atrial to ventricular block at the end of the AP trains was 8.6-fold. Propafenone (1 μM) induced potent use-dependent block in both atrial and ventricular cells, yielding a ratio of 1.2, mainly due to a larger tonic block in atrial cells. Lidocaine (40 μM) was somewhat atrial selective, yielding a ratio of 1.6. Only ranolazine was truly atrial selective in its actions to exert use-dependent inhibition of INa.
The mechanism underlying the atrial-selective block of fast INa by ranolazine can be better understood in terms of the voltage dependence of drug interactions with the Na+ channel, as shown in Fig. 10C. According to our model, the ranolazine interaction with the Na+ channel displays a unique decrease of the unbinding rate at moderately depolarized potentials, which is more pronounced in atrial cells compared with ventricular cells. Moderate depolarization results in prompt inactivation of the predominant fraction of the Na+ channels, markedly decreasing the binding/unbinding of ranolazine. On the other hand, lidocaine interacts with the inactivated state of the Na+ channel with faster rates than with the closed state. As a result, the time constant decreases with depolarization similarly in atrial and ventricular cells. Propafenone, a typical open state blocker, is trapped in the fully closed state and displays a very slow rate of unbinding at negative potentials. This rate only increases when channels transition into the preopen state upon depolarization. In this respect, ranolazine is unique because it is not trapped in the fully closed state, likely due to its elongated linear structure.
The present study describes a mathematical model capable of recapitulating the characteristics of interaction of three class I antiarrhythmic agents with the Na+ channel and identifying the mechanisms that contribute to the ability of agents such as ranolazine to selectively inhibit INa in atrial versus ventricular myocytes. Our simulations illustrate how ranolazine interaction with the preopen and open states, but not with the inactivated state, can account for a variety of experimental observations, including the shift of the steady-state availability curve, strong dependence of the block on the resting potential, and the preferential block of late INa compared with peak INa.
The model suggests that the more negative position of the steady-state inactivation curve in atrial cells and the unique morphology of the atrial AP contribute importantly to the atrial-selective effect of ranolazine to inhibit Na+ channel activity. The very gradual phase 3 repolarization in atria causes rapid shortening of the diastolic interval as the heart rate accelerates. Drugs that dissociate slowly from the Na+ channel at positive potentials are able to take advantage of these distinctions because INa block accumulates in atrial cells but not in ventricular cells; the maintenance of a diastolic interval in the latter at fast rates facilitates unblocking of the drug from the channel. The model also highlights the importance of the slower and steeper voltage dependence of unbinding from the Na+ channel and of drug-induced prolongation of the atrial AP duration. Agents that dissociate rapidly from the Na+ channel and also prolong the AP can accentuate the atrioventricular distinctions by facilitating shortening of the diastolic interval in atria, thereby resulting in a more depolarized takeoff potential.
Analysis of the model behavior indicates that one of the main features that distinguishes ranolazine from other class I antiarrhythmic drugs is a much steeper voltage dependence of the unbinding rate, which is exclusively determined by the fraction of blocked channels that have recovered from inactivation at a given voltage. The voltage dependence of the unbinding rate follows the intrinsic steady-state inactivation curve. The more negative position of the steady-state inactivation curve in atrial versus ventricular cells explains the much slower unbinding rate of ranolazine from atrial than ventricular myocytes at diastolic potentials.
The atrial selectivity of ranolazine appears to be due to the fact that it possesses the following properties: it dissociates rapidly from the Na+ channel, causes a preferential prolongation of the atrial AP, and displays a steep voltage dependence of the unbinding rate. In contrast, lidocaine displays much smaller atrial selectivity, owing to the fact that it possesses only one of these properties: rapid dissociation from the Na+ channel. Propafenone, on the other hand, shows little to no atrial selectivity, owing to the fact that it dissociates from the Na+ channel very slowly. None of these drugs, except ranolazine, show deceleration of unbinding at the positive diastolic potentials observed in atrial cells but not in ventricular cells at fast rates. Thus, only ranolazine is able to “take advantage” of the electrophysiological distinctions between atrial and ventricular cells.
A previous study (15) has suggested that ranolazine is an inactivated state blocker based on its effect to cause a leftward shift of the steady-state availability curve. Our model simulations indicate that inactivated state block is not required to reproduce the experimentally observed shift, which, according to the new data presented in the present study and the accompanying paper (19), is due to block that develops at negative potentials via the preopen state of the channel. Ranolazine unbinding from noninactivated channels in the preopen state correctly predicts the rate of recovery from block at very negative potentials, thereby making it unlikely that ranolazine interacts with the closed state of the Na+ channel.
This work was supported by National Heart, Lung, and Blood Institute Grant HL-47678 (to C. Antzelevitch) and by the Masons of New York State and Florida.
C. Antzelevitch received research support and is a consultant to Gilead Sciences, Inc. S. Rajamani and L. Belardinelli are employees of Gilead Sciences, Inc.
↵1 Supplemental Material for this article is available at the American Journal of Physiology-Heart and Circulatory Physiology website.
- Copyright © 2011 the American Physiological Society