Class III antiarrhythmic drugs amiodarone and dronedarone impair KIR2.1 backward trafficking

Abstract Drug‐induced ion channel trafficking disturbance can cause cardiac arrhythmias. The subcellular level at which drugs interfere in trafficking pathways is largely unknown. KIR2.1 inward rectifier channels, largely responsible for the cardiac inward rectifier current (IK 1), are degraded in lysosomes. Amiodarone and dronedarone are class III antiarrhythmics. Chronic use of amiodarone, and to a lesser extent dronedarone, causes serious adverse effects to several organs and tissue types, including the heart. Both drugs have been described to interfere in the late‐endosome/lysosome system. Here we defined the potential interference in KIR2.1 backward trafficking by amiodarone and dronedarone. Both drugs inhibited IK 1 in isolated rabbit ventricular cardiomyocytes at supraclinical doses only. In HK‐KWGF cells, both drugs dose‐ and time‐dependently increased KIR2.1 expression (2.0 ± 0.2‐fold with amiodarone: 10 μM, 24 hrs; 2.3 ± 0.3‐fold with dronedarone: 5 μM, 24 hrs) and late‐endosomal/lysosomal KIR2.1 accumulation. Increased KIR2.1 expression level was also observed in the presence of Nav1.5 co‐expression. Augmented KIR2.1 protein levels and intracellular accumulation were also observed in COS‐7, END‐2, MES‐1 and EPI‐7 cells. Both drugs had no effect on Kv11.1 ion channel protein expression levels. Finally, amiodarone (73.3 ± 10.3% P < 0.05 at −120 mV, 5 μM) enhanced IKIR 2.1 upon 24‐hrs treatment, whereas dronedarone tended to increase IKIR 2.1 and it did not reach significance (43.8 ± 5.5%, P = 0.26 at −120 mV; 2 μM). We conclude that chronic amiodarone, and potentially also dronedarone, treatment can result in enhanced IK 1 by inhibiting KIR2.1 degradation.


Introduction
Proper ion channel expression and function is one of the cornerstones of normal heart function. Unequal ion distribution between the intra-and extracellular compartment in concert with ion specific voltage-sensitive channels in the plasma membrane determines action potential formation. The stable and negative resting membrane potential in between action potentials results from the activity of the inward rectifying ion channels of the KCNJ gene family [1]. In the heart, the K IR 2.1 channel protein, encoded by KCNJ2, is the main contributor to ventricular I K1 . K IR 2.1 loss of function has been associated with Andersen-Tawil syndrome, characterized by action potential prolongation, and thus QT-lengthening on the ECG. Furthermore, patients experience periodic paralysis and mild episodes of cardiac arrhythmia [2]. In contrast, gain-of-function mutations are associated with QT shortening and atrial fibrillation [3]. Besides its important function in the heart, K IR 2.1 proteins also contribute to inward rectifier currents in skeletal and smooth muscle, and several neuronal cell types [4]. In Andersen-Tawil syndrome patients, association with the occurrence of increased U-waves on the ECG has been found [5]. Pharmacological inhibition of KCNJ channels by barium has also been associated with more apparent U-waves [6]. In a study on the presence and amplitudes of U-waves associated with loss-and gain-of-function mutations in KCNJ2 patients at normokalemic conditions, the authors speculate that at least a part of the U-wave is inversely correlated with the amount of I K1 [7]. K IR 2.1 ion channel trafficking is a strictly regulated process that can be divided into forward (anterograde; towards the plasma membrane) and backward (retrograde; from the plasma membrane) trafficking events [8]. K IR 2.1 channels become internalized via a clathrinmediated pathway and subsequently travel towards the lysosome, where the channels ultimately become degraded via an initial discrete cleavage step that removes the N-terminus [9,10]. Interference in lysosomal degradation and upstream trafficking events by specific inhibitors results in increased K IR 2.1 expression levels, and most likely by saturation of the endocytotic machinery, also in increased I K1 densities [9,10]. Also clinical drugs can have significant effects on ion channel trafficking and this can lead to severe adverse effects [8]. Among the variety of affected channel proteins, the K IR 2.1 channel internalization and degradation is sensitive for disturbances by, although old, clinical drugs like chlorpromazine and chloroquine [9][10][11].
Amiodarone is a class III antiarrhythmic, based on the benzofuran structure used in atrial and ventricular fibrillation therapy [12]. Amiodarone is a multichannel blocker affecting delayed rectifier I Kr , sodium channel and L-type calcium currents. Amiodarone therapy is known for its many adverse effects on the ocular, neurological, dermatological, thyroid, gastrointestinal, pulmonary, cardiac and liver systems [13][14][15]. Some studies demonstrate detrimental effects of amiodarone on cargo trafficking through the late-endosome/lysosome compartments, which could partly explain the plethora of side effects [16][17][18]. Amiodarone has been shown to inhibit the degradation of lung surfactant protein A in vitro and in vivo [16]. Dronedarone is a synthetic analogue of amiodarone developed to preserve antiarrhythmic properties with less adverse effects, especially thyroid and pulmonary toxicity [19]. Compared with amiodarone, dronedarone is less lipophilic and has a much shorter half-life (1-2 versus 30-55 days). Nevertheless, also dronedarone appears to interfere in normal late-endosome/lysosome function [17]. Chronic amiodarone therapy has been associated with the appearance of prominent U-waves [20][21][22], which may allude to a potential disturbance of I K1 . Currently, it is unknown whether amiodarone and dronedarone interfere in the process of K IR 2.1 trafficking, in particular its degradation, which was therefore investigated in the current study.

Rabbit ventricular cardiomyocyte isolation
Animal care and experimental procedures were in accordance with the 'European Directive for the Protection of Vertebrate animals used for Experimental and Scientific Purpose, European Community Directive 2010/63/EU' and were approved by the Committee for Experiments on Animals of the Utrecht University, the Netherlands.
Ventricular rabbit cardiomyocytes were isolated by enzymatic digestion using a Langendorff set-up identical to that described previously [23].
In COS-7 western blot experiments, cells were transfected using linear polyethylenimine (PEI). In short, PEI (Mw 25,000 Polysciences Inc., Eppelheim, Germany) was dissolved in water at 0.323 g/l. PEI solution was subsequently adjusted to pH 8.0, sterilized using filtration and freeze-thawed four times. Aliquots of PEI stock solution were stored at À20°C. For each transfection, 2.5 lg plasmid DNA was added to a 150 mM NaCl solution, total volume 150 ll. 20 ll of PEI stock solution was also added to a 150 mM NaCl solution, total volume 150 ll. Both solutions were mixed, incubated at room temperature for 20 min. and subsequently added to the cells. Medium was replaced at 16 hrs posttransfection. In immunofluorescence microscopy experiments, HEK293t, END-2, MES-1 and EPI-7 cells were transfected with human K IR 2.1 + Rab7-GFP or K IR 2.1 alone using Lipofectamine (Invitrogen, Breda, the Netherlands) according to the manufacturer's protocol.

Electrophysiology
In ventricular rabbit cells, I K1 was measured by patch clamp experiments in whole-cell mode using an Axon amplifier controlled by pClamp9.2 software (Molecular Devices, Sunnyvale, CA, USA). Experiments were performed at 37°C using temperature control (Cell MicroControls, Norfolk, VA, USA). Cardiomyocytes were put in the chamber and superfused with normal Tyrode's solution (mM) (140 NaCl, 5 KCl, 6 HEPES, 6 glucose, 1.8 CaCl 2 , 1 MgCl 2 , pH 7.4 with NaOH). Borosilicate glass pipettes were made with a Sutter P-2000 puller (Sutter Instrument, Novato, CA, USA) and had a pipette resistant of 2-3 MΩ when filled with pipette solution (mM) (110 KCl, 10 EGTA, 10 HEPES, 4 K 2 -ATP, 5.17 CaCl 2 , 1.42 MgCl 2 , pH 7.2 with KOH). The voltage protocol for I K1 measurements was as follows: holding potential was set to À80 mV, and a prepulse at À40 mV for 200 ms was applied to inactivate native sodium current. I K1 was elicited by 1-s step pulses from À120 mV to 30 mV by 10 mV step increments.
Steady-state currents from both cell types were analysed using Clampfit 9.2 software (Molecular Devices) and corrected for membrane capacitance to determine current density.

Statistics
Data were analysed using GraphPad Prism version 6.00 for Windows (GraphPad Software, La Jolla, California USA) or Origin 8 (Microcal Software, Northampton, MA, USA) for rabbit cardiomyocyte measurements. For normally distributed data, Student's t-test or ANOVA for paired samples with Tukey's HSD post hoc or Bonferroni correction for multiple comparisons was used, while nonparametric data were analysed using Wilcoxon rank-sum test and Friedman's test with Dunn's multiple comparison test. Results are presented as mean AE S.E.M. Values of P < 0.05 were considered significant.
We next assessed effects of chronic treatment with amiodarone and dronedarone on K IR 2.1 expression in our previously described model system for K IR 2.1 channel trafficking, HK-KWGF cells [9][10][11]. Both amiodarone and dronedarone resulted in dose-dependent increase in total K IR 2.1 expression as established by Western blotting ( Fig. 2A and B). In these assays, the strongest effects were reached with 20 lM amiodarone (2.9 AE 0.2-fold) and 10 lM dronedarone (6.1 AE 1.5), respectively. No effects on mRNA levels were found by quantitative PCR (1.00 AE 0.02 versus 0.92 AE 0.02 and 1.02 AE 0.01 for control, 10 lM amiodarone and 5 lM dronedarone, respectively). In contrast, amiodarone and dronedarone were unable to increase mature and immature K v 11.1 expression in stably transfected HEK293 cells (Fig. 2C and D).
Finally, we tested whether increased K IR 2.1 expression levels are dependent upon coexpression of Na v 1.5 expression, a cardiac ion channel that has previously been shown to associate with K IR 2.1 and which combined expression demonstrates reciprocal modulation [33]. Ex-293 cells, a HEK293 cell line both expressing K IR 2.1 and Na v 1.5 [28], displayed a dose-dependent increase in K IR 2.1 expression upon treatment with either amiodarone or dronedarone (Fig. 2E and F). Strongest effects were observed with 20 lM amiodarone (3.4 AE 1.2-fold) and 10 lM dronedarone (14.44 AE 6.0).
To determine whether the increased K IR 2.1-GFP expression levels is cell line specific or depends on the GFP tag, experiments were repeated in transiently transfected COS-7 cells. Under these conditions, similar effects were seen for amiodarone and dronedarone on the non-tagged human K IR 2.1 ( Fig. 3A and B). Strongest effects in COS-7 cells were observed with 20 lM amiodarone (3.4 AE 0.6-fold) and 10 lM dronedarone (4.5 AE 0.7).
Significant enhanced expression of K IR 2.1 in HK-KWGF cells was seen from 4 hrs following drug application. Maximal response rates were observed after 4-6 hrs ( Fig. 4A and B).
Immunofluorescence microscopy revealed dose-dependent accumulation of K IR 2.1-GFP (Fig. 5A) in a pattern resembling that of bafilomycin A1 and chloroquine treatment [10]. No intracellular accumulation was seen with 2 lM amiodarone or dronedarone, while relatively small aggregates were seen with 5 lM amiodarone and large aggregates were observed with 10 lM amiodarone or 5 lM dronedarone (Fig. 5A). In order to exclude that K IR 2.1-GFP accumulation in response to amiodarone and dronedarone is cell type specific or depends on the GFP tag, we made use of mouse P19 embryonal carcinoma-derived END-2, MES-1 and EPI-7 cells representing the three different germ layers [34] that were transiently transfected with non-tagged human K IR 2.1. Amiodarone at 10 lM induced clear intracellular aggregates similar as observed in HK-KWGF cells (Fig. 5B). Furthermore, dronedarone at 5 lM induced intracellular K IR 2.1 accumulation in MES-1 cells. In END-2 and EPI-7 cells, dronedarone appeared to induce larger aggregates (Fig. 5B).
We suggested that the intracellular accumulation of K IR 2.1-GFP protein could result in saturation of upstream trafficking pathways which may result in enhanced current levels, as seen before with the lysosomal inhibitor chloroquine [9] and the clathrin-mediated internalization inhibitor dynasore [10]. Cells were treated for 24 hrs with either 2 lM dronedarone or 5 lM amiodarone, and I K1 densities were compared to their non-treated counterparts ( Fig. 7A and B). Chronic dronedarone treatment resulted in a slight trend towards increased I KIR2.1 densities for the inward (43.8 AE 5.5%, P = 0.26 at À120 mV) and a non-significant increase in outward (32.0 AE 7.8%, P = 0.83 at À60 mV) current components. 24-hrs treatment with amiodarone resulted in a significant increase in the inward current component at À120, À110 and À100 mV of 73.3 AE 10.3%, 78.0 AE 10.9% and 84.4 AE 11.5%, respectively, whereas a non-significant increased outward current (75.9 AE 24.9%, P = 0.38 at À60 mV) was observed.

Discussion
Amiodarone is known for its hepatic and pulmonary adverse effects in patients. This is associated with the occurrence of lysosomal structural abnormalities such as lamellar lysosomal inclusion bodies [35,36]. Less is known on the effects of amiodarone on muscle cell lysosome morphology and function. Several case reports demonstrate Western blot analysis of K IR 2.1-GFP expression in HK-KWGF cells treated with 2, 5, 10 or 20 lM amiodarone or 2, 5 or 10 lM dronedarone for 24 hrs. C indicates control (untreated) cells. Ponceau staining is used as loading control. Averaged data from eight (amiodarone) and seven (dronedarone) independent experiments, respectively, are depicted in bar graphs in the lower part of both panels.(C and D) Western blot analysis of K v 11.1 expression in HEK-hERG cells treated with 2, 5, 10 or 20 lM amiodarone or 2, 5 or 10 lM dronedarone for 24 hrs. C indicates control (untreated) cells. Ponceau staining is used as loading control. Averaged data from 3 independent experiments are depicted in bar graphs in the lower part of both panels. (E and F) Western blot analysis of K IR 2.1 expression in Ex293 cells treated with 2, 5, 10 or 20 lM amiodarone or 2, 5 or 10 lM dronedarone for 24 hrs. C indicates control (untreated) cells. Ponceau staining is used as loading control. Averaged data from three (amiodarone) and five (dronedarone) independent experiments are depicted in bar graphs in the lower part of both panels. *P < 0.05; ***P < 0.001. the occurrence of skeletal muscle vacuolarization with or without the presence of inclusion bodies upon chronic amiodarone therapy, interpreted as lysosomal defects by the authors [37,38]. In myocardial fibres from the left and right ventricle, and right atrium derived from dogs chronically treated with amiodarone, abnormal lysosomal structures with often dense lamellar inclusion bodies were found [39]. Similar 'autophagic vacuoles' were observed in isolated rat ventricular myocytes chronically treated with amiodarone in vitro [40,41]. Morissette et al. demonstrated that amiodarone application resulted in vacuolar sequestration and evolved towards persistent macroautophagy in macrophages, smooth muscle cells and HEK293 cells [42]. Dronedarone shows strong similarities to amiodarone with respect to induction of the formation of cellular vacuoles containing lamellar bodies (lysosomal structures) as demonstrated in alveolar macrophages [43].
We found that amiodarone and dronedarone treatment increased K IR 2.1 expression and intracellular accumulation, most likely in late endosomes and lysosomes, in several different cell lines. Interestingly, compared with chloroquine treatment that results in lysosomal accumulation of full-length and a discrete N-terminally cleaved K IR 2.1 protein, only accumulation of the full-length product is seen with amiodarone and dronedarone. Therefore, either the majority of the K IR 2.1 accumulates in pre-lysosomal compartments, which is in line with the findings of Picolli et al. [17] who describe that amiodarone and dronedarone do not affect early endosome function, but interferes in the late compartments of the endocytotic pathway, or these compounds interfere in protease function responsible for the N-terminal K IR 2.1 cleavage. The latter explanation is in line with findings of Buratta et al. [44] who describe that specific cathepsins display altered processing in some cell types upon amiodarone treatment. C indicates control (untreated) cells. Ponceau staining is used as loading control. Averaged data from eight (amiodarone) and ten (dronedarone) independent experiments, respectively, are shown in bar graphs in the lower part of both panels. *P < 0.05; **P < 0.01; ***P < 0.001.  Whatever the exact mechanism, our findings for K IR 2.1 are in line with those of Baritussio et al. [16], who demonstrated that amiodarone inhibits surfactant protein A degradation that normally takes place in the lysosomal compartment.
As amiodarone treatment correlates with the induction of autophagocytosis, especially upon longer treatment (>24 hrs), we cannot exclude the possibility that a part of the intracellular K IR 2.1 accumulation occurs in non-functioning, due to the amiodarone and dronedarone acid buffering capacity, macroautophagosomes [42]. This may contribute to the observed colocalization of K IR 2.1 with LAMP-1. Finally, expression level of the K v 11.1 potassium channel protein is not increased by amiodarone or dronedarone, once more demonstrating channel specificity in trafficking pathways and their (patho)physiological regulation [45].
When considering potassium ion channel trafficking with respect to the action of amiodarone and dronedarone, only few data are   [46] found no effect of amiodarone on I Ks channel trafficking in Chinese hamster ovary cells. In the hERG-Lite assay [47], amiodarone inhibits hERG surface expression which may result from impaired forward or enhanced backward trafficking or translation interference. We and others showed that backward trafficking of hERG and K IR 2.1 channels follows different pathways, which makes them react differently to a number of drugs [11]. We showed that amiodarone and dronedarone also affect K IR 2.1 trafficking differently than that for hERG channels. In cardiomyocytes isolated from guinea pigs treated with amiodarone for 7 days, decreased I K1 , I Ks and I Kr densities were found [48]. In contrast, in cardiomyocytes from mice treated with amiodarone for 6 weeks, no differences in I K1 densities, in neither KCNJ2 nor KCNJ12 transcript levels, were observed [49]. For now, it is unclear to what extent and by what mechanisms amiodarone and dronedarone affect potassium ion channel trafficking in vivo which warrants future research.