Quantitative Profiling of the Effects of Vanoxerine on Human Cardiac Ion Channels and its Application to Cardiac Risk

Vanoxerine has been in clinical trials for Parkinsonism, depression and cocaine addiction but lacked efficacy. Although a potent blocker of hERG, it produced no serious adverse events. We attributed the unexpected result to offsetting Multiple Ion Channel Effects (MICE). Vanoxerine’s effects were strongly frequency-dependent and we repositioned it for treatment of atrial fibrillation and flutter. Vanoxerine terminated AF/AFL in an animal model and a dose-ranging clinical trial. Reversion to normal rhythm was associated with QT prolongation yet absent proarrhythmia markers for Torsade de Pointes (TdP). To understand the QT/TdP discordance, we used quantitative profiling and compared vanoxerine with dofetilide, a selective hERG-blocking torsadogen used for intractable AF, verapamil, a non-torsadogenic MICE comparator and bepridil, a torsadogenic MICE comparator. At clinically relevant concentrations, verapamil blocked hCav1.2 and hERG, as did vanoxerine and bepridil both of which also blocked hNav1.5. In acute experiments and simulations, dofetilide produced early after depolarizations (EADs) and arrhythmias, whereas verapamil, vanoxerine and bepridil produced no proarrhythmia markers. Of the MICE drugs only bepridil inhibited hERG trafficking following overnight exposure. The results are consistent with the emphasis on MICE of the CiPA assay. Additionally we propose that trafficking inhibition of hERG be added to CiPA.


Supplementary Figure 2. Concentration response curves for vanoxerine, bepridil, dofetilide and verapamil
Symbols indicate the mean ± sem and the number of associated experiments. The data was obtained using step-ramps except for hNav1.5 late currents in verapamil where the voltage protocol consisted of a 200 ms pulse to -120 mM, a 5 ms pulse to +50 mV followed by a 400 ms pulse to -30 mV.

. Changes in APD90 at steady state as a function of beat period and free drug concentration
Symbols are APD90 changes in simulations where no triggered activity was observed. The simulations show that vanoxerine prolongs APD in a concentration-dependent manner that is increased at slower frequencies. No arrhythmias were observed. The results show the reverse usedependent effect of dofetilide since action potential prolongation was greater at slower pacing rates for the same concentration (e.g. 10% larger at 2s compared to 1 s at 10 nM). The projection of the last point of the Period/Concentration plane shows the functional relationship between triggered activity, beat frequency and drug concentration. Triggered activity occurred after the last plotted concentration and ranged from 45 nM (2 second period) to 80 nM (1 second period). Bluesymbols indicate the last AP in the series showing no arrhythmic markers. Verapamil at concentrations lower than 200-300 nM reduced APD90 but at larger concentrations increased APD90. No arrhythmic activity was observed. Bepridil prolonged APD90 but showed no proarrhythmic markers.

Supplementary Figure 5.
Examples of the effects of vanoxerine, bepridil, verapamil and dofetilide on simulated action potentials at 1 Hz. Line colors are consistent with the experiments shown in Figure 3 in the text.
Supplementary Figure 6. Non-specific tissue binding: an explanation of the limited effects of vanoxerine on the arterially-perfused canine wedge preparation.
S 6.1 summarizes the effects of vanoxerine and dofetilide on APD90 of epicardial action potentials recorded in the canine wedge preparation. Data are from Table 4 in Lacerda et al. 2 .The percent changes from control were fitted to a logistic function with half maximum effect (APD90 half ) at 206 nM for vanoxerine and 18.2 nM for dofetilide. APD prolongation at APD90 half was 3.05% and 11.1% respectively. These values are indicated by vertical lines in the figure.
The maximum responses to vanoxerine and dofetilide were 6.1 and 22.6% respectively. Symbols are mean ± sem.

S 6.1
We hypothesized that the limited effects in situ are due to tissue binding similar to the binding of the drug to plasma proteins in equilibrium dialysis 1 . To estimate the amount of non-specific tissue binding in the wedge experiments we used in silico simulations of the Decker-Rudy model of the canine epicardial action potential 2 .
We investigated the effect of stepwise reductions in the vanoxerine concentration from the simulated APD90 shortening at 206 nM free vanoxerine showing a small attenuated AP not at all like a CAP (blue line in S 6.2c) to the experimentally observed prolongation of 3.05 % at APD90 half . We took as a measure of non-specific (NS) binding the factor that reduces the free concentration from APD90 half to a level that produced 3.05% prolongation. Non-specific tissue binding for vanoxerine was calculated at 99.8% consistent with the strong binding measured with the equilibrium dialysis method. Non-specific binding reduced the total drug concentration from 206 nM to an unbound concentration of 0.5 nM. At that free concentration the shape of the simulated action potential (red line in S 6.2c) is comparable to the action potential recorded in the wedge preparation (S 6.2a).
We used the same approach to assess non-specific binding of dofetilide in the wedge preparation. The model predicts that in the absence of non-specific binding 18.2 nM, the APD90 half for dofetilide, should elicit arrhythmic activity (blue line in S 6.2d). However, 89.4% non-specific binding reduced the free concentration from 18.2 nM to 1.94 nM, a concentration that that produced an 11.1% change in APD90. The shape of the simulated action potential is comparable to the shape of the action potential recorded in the wedge preparation (red line in S 6.2d). The measured 89.4% non-specific binding for dofetilide is close to the 60-70% range that is reported (http://dailymed.nlm.nih.gov/dailymed).  Table 1 in the text. Black lines: no drug; blue lines: simulation based on APD90 half (Note the lack of concordance with the experiments in a and b); red lines: action potential waveforms calculated assuming 99.8% and 89.4% non-specific binding for vanoxerine and dofetilide respectively. Action potentials were simulated using the Decker-Rudy cardiac epicardial dog model 3 .
Simulations used the conductance-block approach where maximal conductances were modified according to the measured IC50s and Hill coefficients ( Table 1). Two thousand action potentials were simulated to ensure attainment of steady state.