Combined hERG channel inhibition and disruption of trafficking in drug-induced long QT syndrome by fluoxetine: a case-study in cardiac safety pharmacology

Abstract

Drug-induced prolongation of the rate-corrected QT interval (QT_CI) on the electrocardiogram occurs as an unwanted effect of diverse clinical and investigational drugs and carries a risk of potentially fatal cardiac arrhythmias. hERG (human ether-à-go-go-related gene) is the gene encoding the α-subunit of channels mediating the rapid delayed rectifier K⁺ current, which plays a vital role in repolarising the ventricles of the heart. Most QT_CI prolonging drugs can inhibit the function of recombinant hERG K⁺ channels, consequently in vitro hERG assays are used widely as front-line screens in cardiac safety-testing of novel chemical entities. In this issue, Rajamani and colleagues report a case of QT_CI prolongation with the antidepressant fluoxetine and correlate this with a dual effect of the drug and of its major metabolite norfluoxetine on hERG channels. Both compounds were found to produce an acute inhibition of the hERG channel by pharmacological blockade, but in addition they also were able to disrupt the normal trafficking of hERG protein to the cell membrane. Mutations to a key component of the drug binding site in the S6 region of the channel greatly attenuated channel block, but did not impair disruption of trafficking; this suggests that channel block and drug effects on trafficking were mediated by different mechanisms. These findings add to growing evidence for disruption of hERG channel trafficking as a mechanism for drug-induced long QT syndrome and raise questions as to possible limitations of acute screening methods in the assessment of QT_cI prolonging liability of drugs in development.

The QT interval on the body surface electrocardiogram represents the time interval between initiation of ventricular depolarization and completion of ventricular repolarization. Prolonged QT intervals typically result from delayed repolarization of cardiac ventricular action potentials and are associated with an increased risk of the potentially fatal polymorphic ventricular tachycardia, torsades de pointes (TdP; Yap and Camm, 2003). QT interval prolongation can result from gene mutations or, more frequently, arises as an unwanted side effect of clinically used drugs. Rate-corrected QT intervals (QT_CI) for males and females of >440 and >460 ms, respectively, are considered to be prolonged, with arrhythmias most commonly associated with QT_CI values of ⩾500 ms (Yap and Camm, 2003). The drugs that exhibit a propensity to prolong the QT_CI are both structurally and therapeutically diverse (e.g. Witchel and Hancox, 2000; Yap and Camm, 2003, and also see www.torsades.org). However, the majority of such drugs have been observed to exert a common effect: pharmacological inhibition of the ‘rapid delayed rectifier' potassium channel current (I_Kr), which is vital for normal ventricular repolarization (Witchel and Hancox, 2000; Sanguinetti and Mitcheson, 2005; Sanguinetti and Tristani-Firouzi, 2006). The human ether-à-go-go-related gene (hERG) encodes functional I_Kr channels and recombinant hERG channels are able to interact with diverse drugs owing to a combination of a large inner cavity and involvement in the drug binding site of aromatic amino-acid residues in the S6 helices of the channel (Sanguinetti and Mitcheson, 2005; Sanguinetti and Tristani-Firouzi, 2006). Although drug-induced torsadogenesis is complex and an area of active debate (e.g. Hoffmann and Warner, 2006), owing to the strong association between QT_CI prolongation and hERG channel blockade, hERG channel inhibition offers a convenient surrogate marker for drug-induced QT_cI prolongation. Recombinant hERG channel assays therefore constitute an important early screen for testing novel chemical entities (NCEs) for their QT_cI-prolonging liability. Moreover, it is hoped that a detailed understanding of the molecular determinants of hERG channel blockade, together with hERG pharmacophore models may ultimately help design out hERG blockade in silico, thereby minimizing attrition of drugs in the design and development process (Recanatini et al., 2005; Sanguinetti and Mitcheson, 2005). An article in this issue of the BJP, dealing with QT_cI prolongation and hERG inhibition by the serotonin-selective-reuptake inhibitor (SSRI) antidepressant, fluoxetine (Prozac) provides pause for thought in this regard (Rajamani et al., 2006). These authors provide evidence for a dual mechanism for attenuated hERG channel function by fluoxetine that incorporates both pharmacological blockade and a disruption of the normal trafficking of the hERG channel protein to the cell membrane.

Although SSRIs are considered to be relatively safe in overdose (Isbister et al., 2004), fluoxetine is occasionally associated with QT_CI prolongation, with published cases in the last 2 years including an elderly individual (Wilting et al., 2006), an individual taking other QT_CI-prolonging drugs (Nykamp et al., 2005), and an infant born to a mother taking fluoxetine (Dubnov et al., 2005). In the present study of Rajamani and co-workers, an adult patient had a markedly prolonged QT_CI of 625 ms, following intentional fluoxetine overdose. The authors hypothesized that this resulted from a reduction in hERG current (I_hERG), which is concordant with a susceptibility of the hERG channel to pharmacological blockade by fluoxetine that had been reported previously (Thomas et al., 2002; Witchel et al., 2002). However, no such data were available before this study on the major, long-lasting metabolite of fluoxetine -norfluoxetine. Using a combination of conventional whole-cell voltage-clamp and ventricular action potential (AP) clamp, Rajamani and co-workers provide convincing evidence that fluoxetine and norfluoxetine are both effective as I_hERG inhibitors, although with fluoxetine slightly the more potent of the two compounds. Thus, levels of both the parent compound and of its major metabolite are likely to be relevant to potential I_Kr/hERG blockade in vivo and, by extension, to possible QT_CI prolongation.

It may seem curious, therefore, that both fluoxetine and norfluoxetine have been reported to produce action potential shortening rather than lengthening, when applied acutely to isolated cardiac myocytes at concentrations relevant to those reported here to block I_hERG (Pacher et al., 2000; Magyar et al., 2004). The study by Rajamani and co-workers demonstrates that neither fluoxetine nor norfluoxetine suppressed sodium currents carried by heterologously expressed SCN5A Na channels, although native cardiac ion channel currents were not investigated. However, both fluoxetine and norfluoxetine have been previously shown to suppress native cardiac L-type calcium current (Pacher et al., 2000; Magyar et al., 2004), which is a key player in maintaining the ventricular action potential plateau. Thus, the effects on native cardiac tissue of acutely applied fluoxetine and norfluoxetine are unlikely to be limited to a single ion channel type.

What then of the effects of the compounds over a longer time-scale? Macroscopic currents elicited during a cardiac action potential reflect the biophysical properties of the channels and the number of functional channels expressed on the surface membrane. Channel trafficking is integral to functional hERG channel biogenesis and surface expression. Its disruption leads to a reduction in number of functional channels at the cell membrane and is an important mechanism by which a number of hERG mutants cause the LQT2 form of congenital long QT syndrome (Delisle et al., 2004). Recently, reports have emerged of disruption of hERG trafficking by some agents associated with acquired QT_CI prolongation (Ficker et al., 2004; Cordes et al., 2005; Kuryshev et al., 2005; Wible et al., 2005). In the present study, Rajamani et al. (2006) have used Western blotting to demonstrate a decrease in mature hERG protein by prolonged (24 h) exposure to either fluoxetine and norfluoxetine, with similar concentration-dependent profiles to those seen for acute I_hERG blockade. The potent hERG blockers cisapride and E-4031 were used as comparators and had no discernible effect on hERG trafficking at concentrations vastly exceeding those that block I_hERG. Data from laser scanning confocal microscopy were not obtained to confirm a reduction in hERG protein at the cell membrane with fluoxetine or norfluoxetine (cf Cordes et al., 2005), but a loss of functional hERG channels by incubation in fluoxetine was demonstrated electrophysiologically. It seems safe to conclude, therefore, that fluoxetine and norfluoxetine do indeed impair hERG trafficking, particularly when the present data are considered alongside independent findings for fluoxetine made recently using an antibody-based chemiluminescent assay (Wible et al., 2005). The precise mechanism by which these drugs inhibit trafficking remains to be established, but if it involves a direct action of the drugs on the hERG channel protein, the binding site must differ from that which mediates acute I_hERG block. This is because the acute effects of fluoxetine were highly sensitive to mutation of the S6 amino-acid residue Phe-656 (which forms a key part of the binding site for the majority of drugs that block hERG; Sanguinetti and Mitcheson, 2005; Sanguinetti and Tristani-Firouzi, 2006), but disruption of trafficking was unaffected by this mutation. Moreover, a high concentration of the potent hERG blocker E-4031, which binds at Phe-656 (Milnes et al., 2003; Kamiya et al., 2006), was unable to prevent fluoxetine- or norfluoxetine-mediated disruption of trafficking. It is intriguing that although some compounds, including E-4031, can pharmacologically rescue trafficking of some LQT2 mutants (Delisle et al., 2004), for other compounds such as fluoxetine and norfluoxetine, disruption of channel trafficking appears likely to contribute to the genesis of QT_CI prolongation.

The wider relevance of this study relates to the accurate identification during the drug design and discovery process of NCEs that impair hERG channel function. Compounds producing acute hERG channel blockade are likely to be identified by standard hERG assays, but trafficking effects are likely to be missed. In a recent screen for drug-induced suppression of hERG, a surprisingly large proportion (∼40%) of hERG blockers demonstrated trafficking effects in addition to pharmacological block (Wible et al., 2005). Thus, standard hERG assays alone could lead to an underestimation of the net effect on hERG of a drug that combines pharmacological blockade with disruption of trafficking and could overlook entirely an NCE that impaired hERG function due solely to an effect on trafficking. The extent to which this is a real as opposed to theoretical problem remains to be established, although at least six non-blocking hERG trafficking inhibitors are already known to exist (Wible et al., 2005). This adds further to the complexity of screening NCEs for QT_CI-prolonging potential and may underscore the importance both of determining chronic effects of promising candidates on the QT_CI using in vivo models and the value of in vitro assays (e.g Wible et al., 2005) able to detect both hERG channel inhibition and drug effects on trafficking.

Abbreviations

hERG

human ether-à-go-go-related gene

I_hERG

current carried by recombinant hERG channels

I_Kr

current of the rapid delayed rectifier K⁺ channel

NCE

novel chemical entity

QT_CI

rate-corrected QT interval