Binding modes of hERG blockers: an unsolved mystery in the drug design arena

1. Introduction

Cardiotoxicity is one serious side effect mainly caused by the off-target interactions of drugs with various voltage-gated ion channels in the heart, particularly the human ether-a-go-go-related gene (hERG) channel [1]. The hERG channel generates the rapidly activating delayed rectifier potassium current (I_Kr) that is required for membrane repolarization and activation. It plays an important role in the regulation of cardiac action potentials in the heart. Unwanted interactions by drugs with this channel could trigger serious cardiac arrhythmia, such as the long QT syndrome that is characterized by the prolongation of the QT interval and Torsades de pointes (TdP) [2,3]. For example, terfenadine, an antihistamine, blocked the hERG channel causing fatal arrhythmias and, therefore, was withdrawn from the markets worldwide [4]. The possible lethal aftermath of drug-induced hERG blockade has made significant impacts on screening strategies, drug development, regulation, and approval procedures [2,3,5]. Recently, FDA has proposed a Comprehensive in-vitro Proarrhythmia Assay initiative to adapt a modern standard for the accurate assessment of potential drug-induced TdP [6]. Both in silico and experimental (electrophysiology and biochemical) screening assays are available to test the possibility of a compound to interrupt the hERG functionality. The advancements in both fields have significantly helped to reduce the adverse drug reaction-related attrition [7]. Nevertheless, we still don’t have the key to develop drugs that lack and/or have only acceptable hERG liability. For example, while these assays can identify potential hERG blockers, they cannot provide answers for much deeper questions, i.e. how a specific drug can bind hERG and what are the chemical features that promote a strong affinity toward the channel. Gaining such information could greatly aid in the development of drugs with an acceptable benefit–risk ratio for a given target and disease. In this context, we would like to discuss how in silico modeling/technologies could be linked, tailored, and used for addressing some of these challenging questions.

2. The puzzle of hERG ‘binding mode’: why is it complex?

hERG, unlike the other ion channels, seems to have an unusual tendency to bind a wide range of small molecules. In order to understand this promiscuity, atomic-level detail of the channel is essential. Until recently, the lack of three-dimensional structure of the hERG channel was itself a big hurdle in solving the puzzle of hERG-binding mode. The previous binding mode hypotheses relied on comparative modeling approach, using homologous, but not highly similar Kv channels [8–12]. Use of different templates to model the hERG channel structure, in part, gave rise to more number of hypotheses on the mode of binding of drugs. Numerous in silico predictive models and tools have been generated to assess the hERG liability of a compound and are being used by the pharmaceutical companies as a part of their screening protocol [13]. These in silico predictors, computational molecular models together with the mutagenesis experiments, greatly enhanced our understanding on the hERG channel and drug binding to it. For instance, most studies agree that majority of the drugs bind in the central cavity of hERG [14]. It was estimated that the central cavity (of hERG) is sufficiently voluminous to accommodate ligands as large as 20 Å × 7 Å [5,15]. Mutagenesis experiments and docking studies confirmed the role of two aromatic residues (Tyr and Phe) in the pore region to play a crucial role in drug binding [16].

Wang and MacKinnon [17] report on the first near-atomic (3.8 Å resolution) structure of hERG channel in the protein data bank (PDB IDs: 5VA1; 5VA2; 5VA3) using the cryo-electron microscopy (cryo-EM) technique. These structures gave us significant insights, especially on the central cavity and pore domain (PD) of the hERG channel, where the high-affinity blockers bind. The cryo-EM structure represents a truncated hERG protein, obtained after deleting most of the cytoplasmic domain regions (residues 141–350; 871–1005). The quaternary architecture of hERG channel is a homotetramer (Figure 1(a,b)), which resembles most of the other voltage-gated ion channels [18]. The four identical α-subunits assemble to form a functional channel with a central pore for the passage of potassium (K⁺) ions. Each subunit is split into four domains: an N-terminal Per–Arnt–Sim domain, a voltage-sensing domain (VSD), a PD, and a nucleotide-binding domain (cNBD) [18]. The VSD (S1–S4) and PD (S5–S6) of each domain forms the transmembrane region (Figure 1(c,d)). The narrow selectivity filter (composed of Ser–Val–Gly–Phe–Gly residues) is located at the extracellular end of the pore and facilitates the selective filtering of K⁺ ions [17]. Lying below the selectivity filter is a polar central cavity, walled by S6 helices from the four subunits [19]. The central cavity of hERG channel is found to be smaller than what was previously envisaged. It also features a stronger electrostatic environment than those found in other similar Kv channels, which probably defines the high affinity of positively charged ligands toward the hERG channel [5,17] (Figure 1(e)). In addition, four extended hydrophobic pockets (of ~8 × 11Å in size) near the central cavity of hERG [17] (Figure 1(e,f)), located between the selectivity filter region and the S6 helices from the tetramer (as shown in Figure 1(e)), were exposed. It is highly likely that these pockets in hERG serve as an extended binding site, particularly, for high-affinity blockers, such as astemizole and dofetilide. Although the absence of a bound drug molecule in the resolved structure of the hERG channel leaves the puzzle of binding mode unsolved, it has definitely provided us a step forward toward answering the fundamental question – ‘how do the drugs bind the hERG channel?’

Figure 1. Structure of the hERG Channel. (a) Schematic representation of the hERG ‘channel pore’ (yellow circle) formed by four hERG subunits; (b) Cartoon representation of the cryo-EM structure of hERG showing the pore at the center formed by the four subunits (c*) Topology diagram of the hERG in lipid bilayer (only two subunits are shown for clarity); (d) Cartoon representation of two subunits of hERG cryo-EM structures; (e^Φ) Electrostatic potential map of the central cavity showing the lateral pockets (extended hydrophobic pockets) on either sides and (f) Zoomed view of the central cavity, selectivity filter (SF) and the two aromatic residues. Full color available online.

*Reprinted from [2] with permission of John Wiley and Sons.

^ΦReprinted from [17] with permission of Elsevier.

Studies that tried to understand the mode of binding of hERG blockers and validate their findings with different experiments did not lead to a consensus finding but provided us with rather contradicting proposals. The two main contradicting hypotheses available on the orientation of drugs in the hERG channel are (1) the parallel – where the drug is proposed to bind parallel to the channel axis, and (2) perpendicular – where the drug binds perpendicular to the channel axis. Interestingly, both the binding modes mostly point to the same residues as playing important roles in stabilizing the binding pose of drugs [2]. Further, mutational experiments on different residues along the central cavity indicate that drugs from same chemical scaffold can still show varied levels of sensitivity to the mutated residues [20,21]. Similarly, several drugs have been reported to show a higher affinity toward the open-inactivated conformation when compared to the open-active conformation [22,23]. This indicates that drugs binding within the central cavity of hERG could adopt different modes of binding and also vary with the functional state of the channel. Thus, in spite of significant efforts, there is still no consensus on the mode of binding of hERG blockers. Resolving the experimental structures of one or more drug-bound complexes of hERG can provide the key to this puzzle. However, it is not always straightforward to perform these experiments. Computational methods like in silico modeling, molecular docking, and dynamic simulation could be particularly valuable in such cases.

3. Conclusion

‘How do the high-affinity blockers bind to the hERG ion channel?’ remains a very important question in drug discovery. A recent study reported that ~60% of new compounds that are developed as possible drugs for various targets are showing positive hERG liability [24,25]. However, in silico tools remain to be of great help in filtering out these compounds and, thus, have reduced the incidence of potent hERG blockers in industrial drug discovery programs [26]. Finding the answer to the above question would not only help in refining the existing in silico tools and predictive models but also in the design of drugs with reduced or safe-hERG affinity. The routine computational and/or biological assays, currently used to get a mere ‘Yes/No’ response on hERG liability of compounds, are sufficient to reduce the toxicity-related drug attrition. However, it does not improve the efficiency/success rate in drug discovery and development. Hence, it is becoming increasingly important to gain more structural insights into the high-affinity binding of blockers into the hERG channel. While the studies in the past few years have unraveled significant information about the overall structural properties of the hERG channel, there has been no concrete information on the binding modes adopted by different drugs interacting (off-target) with it. The two predominant modes of binding for hERG blockers proposed until now, such as parallel orientation and perpendicular orientation, have been supported by different mutational experiments and are also convincing in different contexts. The recent cryo-EM structure reported extended hydrophobic pockets near the central cavity, which could possibly be playing a role in stabilizing the hERG blockers in either a complete or a partial perpendicular orientation to that of the channel. It is apparent that only structure determination of one or more drug-bound complexes of hERG channel using experimental techniques could be able to resolve the riddle of drug-binding modes in hERG. However, the complexities involved in capturing these experimental structures remain a critical challenge to achieve this goal. Meanwhile, in silico methods, including molecular modeling, docking, and molecular dynamics (MD) approaches, can be applied efficiently to fill the gap and continue the pace of research. These methods have previously been used to illustrate the dynamic properties, key residues, and the type of interactions involved in binding different classes of drugs. Complementary experimental-computational data could advance our knowledge on drug binding to hERG and thereby help us leap forward toward efficient drug design.

4. Expert opinion

Resolving the experimental structures of the hERG channel represents a significant milestone toward understanding their molecular function and their unusual promiscuity to bind a wide range of drugs. The structural features identified from the cryo-structure spawns several additional questions that should stimulate future research for better understanding of the hERG channels’ drug-binding aspect. Since the structure was resolved without a bound ligand, the prime question of how and where the drug binds is still open. In addition to this, the extended pockets identified in the structures raise questions about the possibility of this region to serve as a binding site. In order to find the suitability of the hydrophobic site for drug binding, it is necessary to know how stable are these pockets? And what happens to this site during repolarization, i.e. when the channel gates are closed?

Long-scale classical MD simulations can help us understand the stability, size, and shape of the extended pockets. However, even with high-end supercomputers, it might not be possible to simulate all of the gating mechanisms of this channel using the traditional MD schemes. In such cases, careful application of advanced MD methods, such as steered MD and accelerated MD, which can speed up these time-limiting events, is needed for understanding the structural changes associated with the gating process. The possibility of a drug to bind the extended pocket should be examined by sampling different conformational and functional states of the channel. Incorporating such receptor flexibility will not only throw light onto the binding modes but also could help in improving the accuracy and predictive power of computational methods to calculate the binding affinity of the ligand–channel complex.

The structures of hERG WT and inactivation-deficient mutant (S631A) did not show much variation except for the slight deviation in the Phe627 residue on the selectivity filter. As the authors of the cryo-structure speculated, if these structures corresponded to the inactivated and activated state of the channel, how different is the binding in these two structures? Again, MD simulation of the drug-bound complexes would help in identifying the key differences in the binding orientations and drug-induced conformational changes between these two states. Further, advanced metadynamics-based simulations might benefit linking the specificity and quantify the affinity for the two different functional states.

Understanding the allosteric effects of auxiliary proteins, which bind the hERG channel, on drug binding, might be quite challenging for the computational modelers. Since most auxiliary proteins bind in the N- and C-terminal regions of the proteins, which were deleted from the hERG construct used for inferring the structure, it might be necessary to build a complete model of the hERG channel with the deleted segments included. As noted in the paper describing the cryo-structure [17], these deleted segments are mostly disordered and, therefore, might not be easy to model them using simple computational modeling methods. Specialized methods that can predict the structural features of disordered proteins are needed. We believe that the future research focusing on these areas will help us improve our knowledge on drug binding and thereby get us closer to solving the puzzle of the drug-binding mode and efficient drug design.

Declaration of Interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties. Peer reviewers on this manuscript have no relevant financial or other relationships to disclose

Additional information

Funding

The authors are funded by the Natural Sciences and Engineering Research Council of Canada and the Li Ka Shing Applied Virology Institute.