Role of the pH in state-dependent blockade of hERG currents

Mutations that reduce inactivation of the voltage-gated Kv11.1 potassium channel (hERG) reduce binding for a number of blockers. State specific block of the inactivated state of hERG block may increase risks of drug-induced Torsade de pointes. In this study, molecular simulations of dofetilide binding to the previously developed and experimentally validated models of the hERG channel in open and open-inactivated states were combined with voltage-clamp experiments to unravel the mechanism(s) of state-dependent blockade. The computations of the free energy profiles associated with the drug block to its binding pocket in the intra-cavitary site display startling differences in the open and open-inactivated states of the channel. It was also found that drug ionization may play a crucial role in preferential targeting to the open-inactivated state of the pore domain. pH-dependent hERG blockade by dofetilie was studied with patch-clamp recordings. The results show that low pH increases the extent and speed of drug-induced block. Both experimental and computational findings indicate that binding to the open-inactivated state is of key importance to our understanding of the dofetilide’s mode of action.

kinetics of binding to neutral and cationic forms of the blocker. The binding curves obtained from free energy simulations suggest that the cationic form of dofetilide may be a major driver of formation locked-in complex between the inactivated state of the channel and bound drug. The electrophysiological recordings performed with varying intracellular pH provided functional validation of theoretical findings by showing a sharp dependence of the block by intracellular acidity.

Results and Discussions
State-dependent binding of neutral and cationic dofetilide from simulations. The refined structural models of hERG in different conformational states were generated previously [22][23][24][25][26] and have been extensively validated in experimental and theoretical studies since then 6,[27][28][29][30] . A number of predictions made based on these models of open, closed and open-inactivated states have been successfully tested experimentally forming a basis for our current study (Fig. 1a) [31][32][33] . More recently they have been tested with studies of common hERG blockers and mapping of activators sites 27 . Hence, we can assess a state-dependant binding affinity of the drug to this channel in its open, closed and open-inactivated states. As it can be seen in Figures 1b,c the blocker binding site in the intra-cellular cavity (pore-helix and S6 helix) is well captured in different models, which display an RMSD (relative to Eag1 structure) at or below the reported structure resolution (S6 residues are from 635 to 658 and pore-helix residues are from 618 to 629). More importantly, the equilibrium dissociation constants and binding free energies can be readily computed from Potential of Mean Force (PMF) profiles, which are the free energy changes along a defined reaction coordinate. The reaction coordinate defined for modeling the two forms of dofetelide binding to relevant states of hERG is shown in Fig. 2a. The effective (estimated from one-dimensional approximation for the process) equilibrium dissociation constant K D from PMF in the presence of a cylindrical constraint can be expressed as follows 34,35 : where R is the radius of the cylindrical restraint oriented normal to the z-axis and N A Avogadro's number. w(z) was offset to zero for dofetilide in the bulk phase. The binding free energy is calculated then: where c 0 is the standard concentration of dofetilide, 1 M.  Figure 3a shows that the open state of hERG channel displays only one low-affinity site for the cationic form of the blocker located at z = − 16 Å (the location of this binding site is labeled with n in Fig. 3a and shown in Fig. 3b-n).
The simulations for open-inactivated state display a remarkable difference in the binding PMFs for cationic dofetilide. There is a well-defined high-affinity binding site located at Z = − 10 Å, which is corresponding to Fig. 3b-m' . Besides the inner binding site, there is one more local minimum of energy profile located at z = − 16 Å close to the gate as shown in Fig. 3b Table 1. Equilibrium dissociation constants and binding free energies for the four systems.
and flexibility might be coupled with high affinity drug blockade in hERG. As shown in Fig. 3, cationic dofetilide interacts with the hydrophobic residues A653, Y652, and F656, the polar resiudes S624, T623, S649 and S660, and water molecules in the open channel (Figs 3b-m, 3b-n). As shown in Fig. 3b-m' , the bound drug is close to a corner of two subunits. The drug is stabilized by strong hydrophobic and polar interactions with residues Y652 from four subunits, S621-S624 from the bottom of the filter of one monomer, M645, G648, S649, and I655. One head group of dofetilide is stabilized by a hydrogen bond with G648 and a water molecule. It suggests that the binding of cationic dofetilide may help to stabilize the open-inactivated state of hERG. Besides the inner binding site, there is one more local minimum of energy profile located at z = − 16 Å close to the gate (Fig. 3b-n'). Dofetilide is established among four Y652, four F656 and one I655 from distal S6, and one T623 from the bottom of the filter. There is one hydrogen bond between the nitrogen of methanesulfonamide and Y652. Dofetilide also forms bifurcating hydrogen bonds with water molecules around the head groups. The average conformation of cationic dofetelide is remarkable different compared with the neutral form (Fig. 4a). We compared the distances of center of mass of the benzene rings in dofetilide for the open and open-inactivated channels. The benzene rings of cationic dofetilide in the open-inactivated channel are much closer to each other than that in the open channel at z = − 10 Å. The two benzene rings can form π − π stacking interactions to stabilize the ligand. The hydrophilic heads come close to each other forming an intra-molecular interaction illustrated in Fig. 3b-m' . The intra-molecular interactions between two hydrophilic heads of dofetilide result in an increased exposure of hydrophobic part of the drug inside cavity. Combined with an apparent drop in number of water molecules ( Figure S1) in the open-inactivated cavity, this conformation allows optimal stacking and hydrophobic interactions between bound dofetilide and Y652/F656 residues in cavity of hERG. We propose that this may be an essential mechanism for well-documented state-dependency in dofetilide binding. Stabilization of the "closed" conformation of the drug provide natural explanation for the higher  form of the drug is favored, is likely to be responsible for the observed experimental trapping for a number of common hERG blockers. To investigate the extent of the pH-dependence of hERG inhibition, we performed whole-cell patch-clamp experiments at various intracellular pH values using transfected HEK cells. Whole-cell recordings allowed assessment of the effect of pipette pH values on dofetilide-block. According to dofetilide's ionization equilibrium constant, more dofetilide would be protonated when the intracellular pH is decreased. For dofetilide concentration-response relationships, dofetilide was superfused for 10 minutes during constant stimulation (10 pulses/min) with the pulse protocol shown in Fig. 5c. After 3 min, block of the hERG current occurred significantly more rapidly at pH = 6.2 than at pH = 7.2 (Fig. 5a,c,d). Figure 5b compares the mean concentration-dependent block of the hERG at pH6.2 to pH7.2. At intracellular pH7.2, the mean IC50 is 0.041 μ M, Hill's coefficient 2.4 whereas at intracellular pH6.2 the IC50 is 0.015 μ M, Hill's coefficient 4.2. To address use-dependent block, the cell was held constantly at − 80 mV during the first 5 min of dofetilide superfusion (Fig. 6). Thereafter a train of pulses were applied (Fig. 6a-c). The mean time-constant for use-dependent block is shown in Fig. 6d. Dofetilide produce significantly more rapid use-dependent block at intracellular pH 6.2 versus pH 8.0 (Fig. 6d). Thus these experimental results support the computational finding that ionization of the drug is a crucial factor in the process.

Conclusions
In this study, the binding sites for dofetilide were mapped by the calculation of PMFs. Combining experimental and computational insights, we propose that the state-dependent internal cavity environment and the intracellular pH plays an essential role in the attenuation of hERG current drug blockade by C-type inactivation. We show that, if the different ionization states of dofetilide are considered, the cationic dofetilide is highly stabilized by the C-type inactivation. For the neutral dofetilide, the differences of binding free energy is ~2.   [39][40][41][42] . The topology and parameters of neutral and cationic dofetilide were generated by the CHARMM generalized force field (CGenFF) 43 . The systems were equilibrated for 10 ns using the NAMD2.9 program package 44 . The NPaT ensemble was used for all  Previous studies attempting to buffer pH to target levels have used HEPES at concentrations in the range of 40 mM, similar to concentrations used herein here 50 . Standard patch-clamp methods were used to measure the whole cell currents of hERG1 mutants expressed in HEK 293 cells using the AXOPATCH 200B amplifier (Axon Instruments) 51 . Unless otherwise indicated, the tail currents were recorded when the voltage was returned to − 100 mV from + 50 mV. Transfected HEK cells were patched to record the hERG1 currents 51 . A stock solution of dofetilide was made in DMSO and diluted into the extracellular solution to the requisite concentrations.
To address the impact of changes in intracellular pH on hERG currents in the drug-free state, we compared the conductance-voltage relationship and current densities at intracellular pH 6.2 versus 8.0. Acidification of the intracellular pH produced a small but significant shift in the V1/2 of activation from + 3.7 mV at pH 8.0 to − 2.5 mV at pH 6.2 (p < 0.05; Fig. 7). The mean current densities were not significantly altered by intracellular pH.
Notes added to Proofs. Recently, the full channel structure of the highly homologous Eag1 channel has been resolved through Cryo-EM at 3.78 Å resolution (Ref. 52). The pore models (S6 helix forming intracellular cavity, pore helix and selectivity filter regions) display remarkable agreement to published structure in positions of key residues for drug binding (T623, S624, Y652 and F656). The region that differs the most between models and solved structure is highly mobile S5-pore linker, unique for this family of proteins. While ROSETTA-generated models captured essential elements e.g. amphipathic hellices, their relative packing to the pore domain is different to that seen in Cryo-EM structure. However, located in the extra-cellular millieu, S5-pore linker is unlikely to influence binding profiles reported in this submission. It is also worth-noting, that the recently-solved Eag1 structure has a very small cavity with narrow or no access to the intra-cellular millieu. The pore model that displays lowest RMSD (<2.5 Å) relative to Cryo-EM structure corresponds to the closed conformation of the pore domain. Figure 7. Raw example hERG current traces recorded in the drug-free state at intracellular pH of 6.2 (a) and 8.0 (b) elicited by the pulse protocol shown in the insert. (c) Average g-V relationship of hERG currents at intracellular pH 6.2 versus 8.0. The average V1/2 were − 2.5 + /− 1.9 mV in pH 6.2 n = 5 and 3.7 + /− 2.2 mV at pH 8.0 (n = 4, P < 0.05, t test). The slope factor were 9.2 + /− 0.4 and 8.2 + /− 0.3 respectively. (d) Average current density amplitudes at pH6.2 and pH8.0 (n = 9, 11).